]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/netdevice.h
net: add netdev_upper_get_next_dev_rcu(dev, iter)
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
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 *
10 * Authors: Ross Biro
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>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
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
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53
54 struct netpoll_info;
55 struct device;
56 struct phy_device;
57 /* 802.11 specific */
58 struct wireless_dev;
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
62
63 extern void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
65
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
72
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
76
77 /*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
92 */
93
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
109
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
115 };
116 typedef enum netdev_tx netdev_tx_t;
117
118 /*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122 static inline bool dev_xmit_complete(int rc)
123 {
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134 }
135
136 /*
137 * Compute the worst case header length according to the protocols
138 * used.
139 */
140
141 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
144 # else
145 # define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
150
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
157
158 /*
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
161 */
162
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187 };
188
189
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
192
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 #endif
197
198 struct neighbour;
199 struct neigh_parms;
200 struct sk_buff;
201
202 struct netdev_hw_addr {
203 struct list_head list;
204 unsigned char addr[MAX_ADDR_LEN];
205 unsigned char type;
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
211 bool global_use;
212 int sync_cnt;
213 int refcount;
214 int synced;
215 struct rcu_head rcu_head;
216 };
217
218 struct netdev_hw_addr_list {
219 struct list_head list;
220 int count;
221 };
222
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
227
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237
238 struct hh_cache {
239 u16 hh_len;
240 u16 __pad;
241 seqlock_t hh_lock;
242
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251
252 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264
265 struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*rebuild)(struct sk_buff *skb);
271 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
272 void (*cache_update)(struct hh_cache *hh,
273 const struct net_device *dev,
274 const unsigned char *haddr);
275 };
276
277 /* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
279 * code.
280 */
281
282 enum netdev_state_t {
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
288 };
289
290
291 /*
292 * This structure holds at boot time configured netdevice settings. They
293 * are then used in the device probing.
294 */
295 struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300
301 extern int __init netdev_boot_setup(char *str);
302
303 /*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306 struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-cpu poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 spinlock_t poll_lock;
321 int poll_owner;
322 #endif
323 struct net_device *dev;
324 struct sk_buff *gro_list;
325 struct sk_buff *skb;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
329 };
330
331 enum {
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash */
336 };
337
338 enum gro_result {
339 GRO_MERGED,
340 GRO_MERGED_FREE,
341 GRO_HELD,
342 GRO_NORMAL,
343 GRO_DROP,
344 };
345 typedef enum gro_result gro_result_t;
346
347 /*
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * further.
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
355 *
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
358 *
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
361 *
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
365 *
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * do with the skb.
368 *
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
372 *
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
376 *
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
380 *
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
383 *
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
386 */
387
388 enum rx_handler_result {
389 RX_HANDLER_CONSUMED,
390 RX_HANDLER_ANOTHER,
391 RX_HANDLER_EXACT,
392 RX_HANDLER_PASS,
393 };
394 typedef enum rx_handler_result rx_handler_result_t;
395 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
396
397 extern void __napi_schedule(struct napi_struct *n);
398
399 static inline bool napi_disable_pending(struct napi_struct *n)
400 {
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402 }
403
404 /**
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
407 *
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
412 */
413 static inline bool napi_schedule_prep(struct napi_struct *n)
414 {
415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
417 }
418
419 /**
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426 static inline void napi_schedule(struct napi_struct *n)
427 {
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430 }
431
432 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
433 static inline bool napi_reschedule(struct napi_struct *napi)
434 {
435 if (napi_schedule_prep(napi)) {
436 __napi_schedule(napi);
437 return true;
438 }
439 return false;
440 }
441
442 /**
443 * napi_complete - NAPI processing complete
444 * @n: napi context
445 *
446 * Mark NAPI processing as complete.
447 */
448 extern void __napi_complete(struct napi_struct *n);
449 extern void napi_complete(struct napi_struct *n);
450
451 /**
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
454 *
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
457 */
458 extern struct napi_struct *napi_by_id(unsigned int napi_id);
459
460 /**
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
463 *
464 * generate a new napi_id and store a @napi under it in napi_hash
465 */
466 extern void napi_hash_add(struct napi_struct *napi);
467
468 /**
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
471 *
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
474 */
475 extern void napi_hash_del(struct napi_struct *napi);
476
477 /**
478 * napi_disable - prevent NAPI from scheduling
479 * @n: napi context
480 *
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
483 */
484 static inline void napi_disable(struct napi_struct *n)
485 {
486 set_bit(NAPI_STATE_DISABLE, &n->state);
487 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
488 msleep(1);
489 clear_bit(NAPI_STATE_DISABLE, &n->state);
490 }
491
492 /**
493 * napi_enable - enable NAPI scheduling
494 * @n: napi context
495 *
496 * Resume NAPI from being scheduled on this context.
497 * Must be paired with napi_disable.
498 */
499 static inline void napi_enable(struct napi_struct *n)
500 {
501 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
502 smp_mb__before_clear_bit();
503 clear_bit(NAPI_STATE_SCHED, &n->state);
504 }
505
506 #ifdef CONFIG_SMP
507 /**
508 * napi_synchronize - wait until NAPI is not running
509 * @n: napi context
510 *
511 * Wait until NAPI is done being scheduled on this context.
512 * Waits till any outstanding processing completes but
513 * does not disable future activations.
514 */
515 static inline void napi_synchronize(const struct napi_struct *n)
516 {
517 while (test_bit(NAPI_STATE_SCHED, &n->state))
518 msleep(1);
519 }
520 #else
521 # define napi_synchronize(n) barrier()
522 #endif
523
524 enum netdev_queue_state_t {
525 __QUEUE_STATE_DRV_XOFF,
526 __QUEUE_STATE_STACK_XOFF,
527 __QUEUE_STATE_FROZEN,
528 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
529 (1 << __QUEUE_STATE_STACK_XOFF))
530 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
531 (1 << __QUEUE_STATE_FROZEN))
532 };
533 /*
534 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
535 * netif_tx_* functions below are used to manipulate this flag. The
536 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
537 * queue independently. The netif_xmit_*stopped functions below are called
538 * to check if the queue has been stopped by the driver or stack (either
539 * of the XOFF bits are set in the state). Drivers should not need to call
540 * netif_xmit*stopped functions, they should only be using netif_tx_*.
541 */
542
543 struct netdev_queue {
544 /*
545 * read mostly part
546 */
547 struct net_device *dev;
548 struct Qdisc *qdisc;
549 struct Qdisc *qdisc_sleeping;
550 #ifdef CONFIG_SYSFS
551 struct kobject kobj;
552 #endif
553 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
554 int numa_node;
555 #endif
556 /*
557 * write mostly part
558 */
559 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
560 int xmit_lock_owner;
561 /*
562 * please use this field instead of dev->trans_start
563 */
564 unsigned long trans_start;
565
566 /*
567 * Number of TX timeouts for this queue
568 * (/sys/class/net/DEV/Q/trans_timeout)
569 */
570 unsigned long trans_timeout;
571
572 unsigned long state;
573
574 #ifdef CONFIG_BQL
575 struct dql dql;
576 #endif
577 } ____cacheline_aligned_in_smp;
578
579 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
580 {
581 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
582 return q->numa_node;
583 #else
584 return NUMA_NO_NODE;
585 #endif
586 }
587
588 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
589 {
590 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
591 q->numa_node = node;
592 #endif
593 }
594
595 #ifdef CONFIG_RPS
596 /*
597 * This structure holds an RPS map which can be of variable length. The
598 * map is an array of CPUs.
599 */
600 struct rps_map {
601 unsigned int len;
602 struct rcu_head rcu;
603 u16 cpus[0];
604 };
605 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
606
607 /*
608 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
609 * tail pointer for that CPU's input queue at the time of last enqueue, and
610 * a hardware filter index.
611 */
612 struct rps_dev_flow {
613 u16 cpu;
614 u16 filter;
615 unsigned int last_qtail;
616 };
617 #define RPS_NO_FILTER 0xffff
618
619 /*
620 * The rps_dev_flow_table structure contains a table of flow mappings.
621 */
622 struct rps_dev_flow_table {
623 unsigned int mask;
624 struct rcu_head rcu;
625 struct rps_dev_flow flows[0];
626 };
627 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
628 ((_num) * sizeof(struct rps_dev_flow)))
629
630 /*
631 * The rps_sock_flow_table contains mappings of flows to the last CPU
632 * on which they were processed by the application (set in recvmsg).
633 */
634 struct rps_sock_flow_table {
635 unsigned int mask;
636 u16 ents[0];
637 };
638 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
639 ((_num) * sizeof(u16)))
640
641 #define RPS_NO_CPU 0xffff
642
643 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
644 u32 hash)
645 {
646 if (table && hash) {
647 unsigned int cpu, index = hash & table->mask;
648
649 /* We only give a hint, preemption can change cpu under us */
650 cpu = raw_smp_processor_id();
651
652 if (table->ents[index] != cpu)
653 table->ents[index] = cpu;
654 }
655 }
656
657 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
658 u32 hash)
659 {
660 if (table && hash)
661 table->ents[hash & table->mask] = RPS_NO_CPU;
662 }
663
664 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
666 #ifdef CONFIG_RFS_ACCEL
667 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
668 u32 flow_id, u16 filter_id);
669 #endif
670
671 /* This structure contains an instance of an RX queue. */
672 struct netdev_rx_queue {
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
675 struct kobject kobj;
676 struct net_device *dev;
677 } ____cacheline_aligned_in_smp;
678 #endif /* CONFIG_RPS */
679
680 #ifdef CONFIG_XPS
681 /*
682 * This structure holds an XPS map which can be of variable length. The
683 * map is an array of queues.
684 */
685 struct xps_map {
686 unsigned int len;
687 unsigned int alloc_len;
688 struct rcu_head rcu;
689 u16 queues[0];
690 };
691 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
692 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
693 / sizeof(u16))
694
695 /*
696 * This structure holds all XPS maps for device. Maps are indexed by CPU.
697 */
698 struct xps_dev_maps {
699 struct rcu_head rcu;
700 struct xps_map __rcu *cpu_map[0];
701 };
702 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
703 (nr_cpu_ids * sizeof(struct xps_map *)))
704 #endif /* CONFIG_XPS */
705
706 #define TC_MAX_QUEUE 16
707 #define TC_BITMASK 15
708 /* HW offloaded queuing disciplines txq count and offset maps */
709 struct netdev_tc_txq {
710 u16 count;
711 u16 offset;
712 };
713
714 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
715 /*
716 * This structure is to hold information about the device
717 * configured to run FCoE protocol stack.
718 */
719 struct netdev_fcoe_hbainfo {
720 char manufacturer[64];
721 char serial_number[64];
722 char hardware_version[64];
723 char driver_version[64];
724 char optionrom_version[64];
725 char firmware_version[64];
726 char model[256];
727 char model_description[256];
728 };
729 #endif
730
731 #define MAX_PHYS_PORT_ID_LEN 32
732
733 /* This structure holds a unique identifier to identify the
734 * physical port used by a netdevice.
735 */
736 struct netdev_phys_port_id {
737 unsigned char id[MAX_PHYS_PORT_ID_LEN];
738 unsigned char id_len;
739 };
740
741 /*
742 * This structure defines the management hooks for network devices.
743 * The following hooks can be defined; unless noted otherwise, they are
744 * optional and can be filled with a null pointer.
745 *
746 * int (*ndo_init)(struct net_device *dev);
747 * This function is called once when network device is registered.
748 * The network device can use this to any late stage initializaton
749 * or semantic validattion. It can fail with an error code which will
750 * be propogated back to register_netdev
751 *
752 * void (*ndo_uninit)(struct net_device *dev);
753 * This function is called when device is unregistered or when registration
754 * fails. It is not called if init fails.
755 *
756 * int (*ndo_open)(struct net_device *dev);
757 * This function is called when network device transistions to the up
758 * state.
759 *
760 * int (*ndo_stop)(struct net_device *dev);
761 * This function is called when network device transistions to the down
762 * state.
763 *
764 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
765 * struct net_device *dev);
766 * Called when a packet needs to be transmitted.
767 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
768 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
769 * Required can not be NULL.
770 *
771 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
772 * Called to decide which queue to when device supports multiple
773 * transmit queues.
774 *
775 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
776 * This function is called to allow device receiver to make
777 * changes to configuration when multicast or promiscious is enabled.
778 *
779 * void (*ndo_set_rx_mode)(struct net_device *dev);
780 * This function is called device changes address list filtering.
781 * If driver handles unicast address filtering, it should set
782 * IFF_UNICAST_FLT to its priv_flags.
783 *
784 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
785 * This function is called when the Media Access Control address
786 * needs to be changed. If this interface is not defined, the
787 * mac address can not be changed.
788 *
789 * int (*ndo_validate_addr)(struct net_device *dev);
790 * Test if Media Access Control address is valid for the device.
791 *
792 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
793 * Called when a user request an ioctl which can't be handled by
794 * the generic interface code. If not defined ioctl's return
795 * not supported error code.
796 *
797 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
798 * Used to set network devices bus interface parameters. This interface
799 * is retained for legacy reason, new devices should use the bus
800 * interface (PCI) for low level management.
801 *
802 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
803 * Called when a user wants to change the Maximum Transfer Unit
804 * of a device. If not defined, any request to change MTU will
805 * will return an error.
806 *
807 * void (*ndo_tx_timeout)(struct net_device *dev);
808 * Callback uses when the transmitter has not made any progress
809 * for dev->watchdog ticks.
810 *
811 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
812 * struct rtnl_link_stats64 *storage);
813 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
814 * Called when a user wants to get the network device usage
815 * statistics. Drivers must do one of the following:
816 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
817 * rtnl_link_stats64 structure passed by the caller.
818 * 2. Define @ndo_get_stats to update a net_device_stats structure
819 * (which should normally be dev->stats) and return a pointer to
820 * it. The structure may be changed asynchronously only if each
821 * field is written atomically.
822 * 3. Update dev->stats asynchronously and atomically, and define
823 * neither operation.
824 *
825 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
826 * If device support VLAN filtering this function is called when a
827 * VLAN id is registered.
828 *
829 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
830 * If device support VLAN filtering this function is called when a
831 * VLAN id is unregistered.
832 *
833 * void (*ndo_poll_controller)(struct net_device *dev);
834 *
835 * SR-IOV management functions.
836 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
837 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
838 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
839 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
840 * int (*ndo_get_vf_config)(struct net_device *dev,
841 * int vf, struct ifla_vf_info *ivf);
842 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
843 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
844 * struct nlattr *port[]);
845 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
846 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
847 * Called to setup 'tc' number of traffic classes in the net device. This
848 * is always called from the stack with the rtnl lock held and netif tx
849 * queues stopped. This allows the netdevice to perform queue management
850 * safely.
851 *
852 * Fiber Channel over Ethernet (FCoE) offload functions.
853 * int (*ndo_fcoe_enable)(struct net_device *dev);
854 * Called when the FCoE protocol stack wants to start using LLD for FCoE
855 * so the underlying device can perform whatever needed configuration or
856 * initialization to support acceleration of FCoE traffic.
857 *
858 * int (*ndo_fcoe_disable)(struct net_device *dev);
859 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
860 * so the underlying device can perform whatever needed clean-ups to
861 * stop supporting acceleration of FCoE traffic.
862 *
863 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
864 * struct scatterlist *sgl, unsigned int sgc);
865 * Called when the FCoE Initiator wants to initialize an I/O that
866 * is a possible candidate for Direct Data Placement (DDP). The LLD can
867 * perform necessary setup and returns 1 to indicate the device is set up
868 * successfully to perform DDP on this I/O, otherwise this returns 0.
869 *
870 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
871 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
872 * indicated by the FC exchange id 'xid', so the underlying device can
873 * clean up and reuse resources for later DDP requests.
874 *
875 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
876 * struct scatterlist *sgl, unsigned int sgc);
877 * Called when the FCoE Target wants to initialize an I/O that
878 * is a possible candidate for Direct Data Placement (DDP). The LLD can
879 * perform necessary setup and returns 1 to indicate the device is set up
880 * successfully to perform DDP on this I/O, otherwise this returns 0.
881 *
882 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
883 * struct netdev_fcoe_hbainfo *hbainfo);
884 * Called when the FCoE Protocol stack wants information on the underlying
885 * device. This information is utilized by the FCoE protocol stack to
886 * register attributes with Fiber Channel management service as per the
887 * FC-GS Fabric Device Management Information(FDMI) specification.
888 *
889 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
890 * Called when the underlying device wants to override default World Wide
891 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
892 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
893 * protocol stack to use.
894 *
895 * RFS acceleration.
896 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
897 * u16 rxq_index, u32 flow_id);
898 * Set hardware filter for RFS. rxq_index is the target queue index;
899 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
900 * Return the filter ID on success, or a negative error code.
901 *
902 * Slave management functions (for bridge, bonding, etc).
903 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
904 * Called to make another netdev an underling.
905 *
906 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
907 * Called to release previously enslaved netdev.
908 *
909 * Feature/offload setting functions.
910 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
911 * netdev_features_t features);
912 * Adjusts the requested feature flags according to device-specific
913 * constraints, and returns the resulting flags. Must not modify
914 * the device state.
915 *
916 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
917 * Called to update device configuration to new features. Passed
918 * feature set might be less than what was returned by ndo_fix_features()).
919 * Must return >0 or -errno if it changed dev->features itself.
920 *
921 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
922 * struct net_device *dev,
923 * const unsigned char *addr, u16 flags)
924 * Adds an FDB entry to dev for addr.
925 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
926 * struct net_device *dev,
927 * const unsigned char *addr)
928 * Deletes the FDB entry from dev coresponding to addr.
929 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
930 * struct net_device *dev, int idx)
931 * Used to add FDB entries to dump requests. Implementers should add
932 * entries to skb and update idx with the number of entries.
933 *
934 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
935 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
936 * struct net_device *dev, u32 filter_mask)
937 *
938 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
939 * Called to change device carrier. Soft-devices (like dummy, team, etc)
940 * which do not represent real hardware may define this to allow their
941 * userspace components to manage their virtual carrier state. Devices
942 * that determine carrier state from physical hardware properties (eg
943 * network cables) or protocol-dependent mechanisms (eg
944 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
945 *
946 * int (*ndo_get_phys_port_id)(struct net_device *dev,
947 * struct netdev_phys_port_id *ppid);
948 * Called to get ID of physical port of this device. If driver does
949 * not implement this, it is assumed that the hw is not able to have
950 * multiple net devices on single physical port.
951 */
952 struct net_device_ops {
953 int (*ndo_init)(struct net_device *dev);
954 void (*ndo_uninit)(struct net_device *dev);
955 int (*ndo_open)(struct net_device *dev);
956 int (*ndo_stop)(struct net_device *dev);
957 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
958 struct net_device *dev);
959 u16 (*ndo_select_queue)(struct net_device *dev,
960 struct sk_buff *skb);
961 void (*ndo_change_rx_flags)(struct net_device *dev,
962 int flags);
963 void (*ndo_set_rx_mode)(struct net_device *dev);
964 int (*ndo_set_mac_address)(struct net_device *dev,
965 void *addr);
966 int (*ndo_validate_addr)(struct net_device *dev);
967 int (*ndo_do_ioctl)(struct net_device *dev,
968 struct ifreq *ifr, int cmd);
969 int (*ndo_set_config)(struct net_device *dev,
970 struct ifmap *map);
971 int (*ndo_change_mtu)(struct net_device *dev,
972 int new_mtu);
973 int (*ndo_neigh_setup)(struct net_device *dev,
974 struct neigh_parms *);
975 void (*ndo_tx_timeout) (struct net_device *dev);
976
977 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
978 struct rtnl_link_stats64 *storage);
979 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
980
981 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
982 __be16 proto, u16 vid);
983 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
984 __be16 proto, u16 vid);
985 #ifdef CONFIG_NET_POLL_CONTROLLER
986 void (*ndo_poll_controller)(struct net_device *dev);
987 int (*ndo_netpoll_setup)(struct net_device *dev,
988 struct netpoll_info *info,
989 gfp_t gfp);
990 void (*ndo_netpoll_cleanup)(struct net_device *dev);
991 #endif
992 #ifdef CONFIG_NET_RX_BUSY_POLL
993 int (*ndo_busy_poll)(struct napi_struct *dev);
994 #endif
995 int (*ndo_set_vf_mac)(struct net_device *dev,
996 int queue, u8 *mac);
997 int (*ndo_set_vf_vlan)(struct net_device *dev,
998 int queue, u16 vlan, u8 qos);
999 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1000 int vf, int rate);
1001 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1002 int vf, bool setting);
1003 int (*ndo_get_vf_config)(struct net_device *dev,
1004 int vf,
1005 struct ifla_vf_info *ivf);
1006 int (*ndo_set_vf_link_state)(struct net_device *dev,
1007 int vf, int link_state);
1008 int (*ndo_set_vf_port)(struct net_device *dev,
1009 int vf,
1010 struct nlattr *port[]);
1011 int (*ndo_get_vf_port)(struct net_device *dev,
1012 int vf, struct sk_buff *skb);
1013 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1014 #if IS_ENABLED(CONFIG_FCOE)
1015 int (*ndo_fcoe_enable)(struct net_device *dev);
1016 int (*ndo_fcoe_disable)(struct net_device *dev);
1017 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1018 u16 xid,
1019 struct scatterlist *sgl,
1020 unsigned int sgc);
1021 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1022 u16 xid);
1023 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1024 u16 xid,
1025 struct scatterlist *sgl,
1026 unsigned int sgc);
1027 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1028 struct netdev_fcoe_hbainfo *hbainfo);
1029 #endif
1030
1031 #if IS_ENABLED(CONFIG_LIBFCOE)
1032 #define NETDEV_FCOE_WWNN 0
1033 #define NETDEV_FCOE_WWPN 1
1034 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1035 u64 *wwn, int type);
1036 #endif
1037
1038 #ifdef CONFIG_RFS_ACCEL
1039 int (*ndo_rx_flow_steer)(struct net_device *dev,
1040 const struct sk_buff *skb,
1041 u16 rxq_index,
1042 u32 flow_id);
1043 #endif
1044 int (*ndo_add_slave)(struct net_device *dev,
1045 struct net_device *slave_dev);
1046 int (*ndo_del_slave)(struct net_device *dev,
1047 struct net_device *slave_dev);
1048 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1049 netdev_features_t features);
1050 int (*ndo_set_features)(struct net_device *dev,
1051 netdev_features_t features);
1052 int (*ndo_neigh_construct)(struct neighbour *n);
1053 void (*ndo_neigh_destroy)(struct neighbour *n);
1054
1055 int (*ndo_fdb_add)(struct ndmsg *ndm,
1056 struct nlattr *tb[],
1057 struct net_device *dev,
1058 const unsigned char *addr,
1059 u16 flags);
1060 int (*ndo_fdb_del)(struct ndmsg *ndm,
1061 struct nlattr *tb[],
1062 struct net_device *dev,
1063 const unsigned char *addr);
1064 int (*ndo_fdb_dump)(struct sk_buff *skb,
1065 struct netlink_callback *cb,
1066 struct net_device *dev,
1067 int idx);
1068
1069 int (*ndo_bridge_setlink)(struct net_device *dev,
1070 struct nlmsghdr *nlh);
1071 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1072 u32 pid, u32 seq,
1073 struct net_device *dev,
1074 u32 filter_mask);
1075 int (*ndo_bridge_dellink)(struct net_device *dev,
1076 struct nlmsghdr *nlh);
1077 int (*ndo_change_carrier)(struct net_device *dev,
1078 bool new_carrier);
1079 int (*ndo_get_phys_port_id)(struct net_device *dev,
1080 struct netdev_phys_port_id *ppid);
1081 };
1082
1083 /*
1084 * The DEVICE structure.
1085 * Actually, this whole structure is a big mistake. It mixes I/O
1086 * data with strictly "high-level" data, and it has to know about
1087 * almost every data structure used in the INET module.
1088 *
1089 * FIXME: cleanup struct net_device such that network protocol info
1090 * moves out.
1091 */
1092
1093 struct net_device {
1094
1095 /*
1096 * This is the first field of the "visible" part of this structure
1097 * (i.e. as seen by users in the "Space.c" file). It is the name
1098 * of the interface.
1099 */
1100 char name[IFNAMSIZ];
1101
1102 /* device name hash chain, please keep it close to name[] */
1103 struct hlist_node name_hlist;
1104
1105 /* snmp alias */
1106 char *ifalias;
1107
1108 /*
1109 * I/O specific fields
1110 * FIXME: Merge these and struct ifmap into one
1111 */
1112 unsigned long mem_end; /* shared mem end */
1113 unsigned long mem_start; /* shared mem start */
1114 unsigned long base_addr; /* device I/O address */
1115 unsigned int irq; /* device IRQ number */
1116
1117 /*
1118 * Some hardware also needs these fields, but they are not
1119 * part of the usual set specified in Space.c.
1120 */
1121
1122 unsigned long state;
1123
1124 struct list_head dev_list;
1125 struct list_head napi_list;
1126 struct list_head unreg_list;
1127 struct list_head upper_dev_list; /* List of upper devices */
1128 struct list_head lower_dev_list;
1129
1130
1131 /* currently active device features */
1132 netdev_features_t features;
1133 /* user-changeable features */
1134 netdev_features_t hw_features;
1135 /* user-requested features */
1136 netdev_features_t wanted_features;
1137 /* mask of features inheritable by VLAN devices */
1138 netdev_features_t vlan_features;
1139 /* mask of features inherited by encapsulating devices
1140 * This field indicates what encapsulation offloads
1141 * the hardware is capable of doing, and drivers will
1142 * need to set them appropriately.
1143 */
1144 netdev_features_t hw_enc_features;
1145 /* mask of fetures inheritable by MPLS */
1146 netdev_features_t mpls_features;
1147
1148 /* Interface index. Unique device identifier */
1149 int ifindex;
1150 int iflink;
1151
1152 struct net_device_stats stats;
1153 atomic_long_t rx_dropped; /* dropped packets by core network
1154 * Do not use this in drivers.
1155 */
1156
1157 #ifdef CONFIG_WIRELESS_EXT
1158 /* List of functions to handle Wireless Extensions (instead of ioctl).
1159 * See <net/iw_handler.h> for details. Jean II */
1160 const struct iw_handler_def * wireless_handlers;
1161 /* Instance data managed by the core of Wireless Extensions. */
1162 struct iw_public_data * wireless_data;
1163 #endif
1164 /* Management operations */
1165 const struct net_device_ops *netdev_ops;
1166 const struct ethtool_ops *ethtool_ops;
1167
1168 /* Hardware header description */
1169 const struct header_ops *header_ops;
1170
1171 unsigned int flags; /* interface flags (a la BSD) */
1172 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1173 * See if.h for definitions. */
1174 unsigned short gflags;
1175 unsigned short padded; /* How much padding added by alloc_netdev() */
1176
1177 unsigned char operstate; /* RFC2863 operstate */
1178 unsigned char link_mode; /* mapping policy to operstate */
1179
1180 unsigned char if_port; /* Selectable AUI, TP,..*/
1181 unsigned char dma; /* DMA channel */
1182
1183 unsigned int mtu; /* interface MTU value */
1184 unsigned short type; /* interface hardware type */
1185 unsigned short hard_header_len; /* hardware hdr length */
1186
1187 /* extra head- and tailroom the hardware may need, but not in all cases
1188 * can this be guaranteed, especially tailroom. Some cases also use
1189 * LL_MAX_HEADER instead to allocate the skb.
1190 */
1191 unsigned short needed_headroom;
1192 unsigned short needed_tailroom;
1193
1194 /* Interface address info. */
1195 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1196 unsigned char addr_assign_type; /* hw address assignment type */
1197 unsigned char addr_len; /* hardware address length */
1198 unsigned char neigh_priv_len;
1199 unsigned short dev_id; /* Used to differentiate devices
1200 * that share the same link
1201 * layer address
1202 */
1203 spinlock_t addr_list_lock;
1204 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1205 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1206 struct netdev_hw_addr_list dev_addrs; /* list of device
1207 * hw addresses
1208 */
1209 #ifdef CONFIG_SYSFS
1210 struct kset *queues_kset;
1211 #endif
1212
1213 bool uc_promisc;
1214 unsigned int promiscuity;
1215 unsigned int allmulti;
1216
1217
1218 /* Protocol specific pointers */
1219
1220 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1221 struct vlan_info __rcu *vlan_info; /* VLAN info */
1222 #endif
1223 #if IS_ENABLED(CONFIG_NET_DSA)
1224 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1225 #endif
1226 void *atalk_ptr; /* AppleTalk link */
1227 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1228 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1229 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1230 void *ax25_ptr; /* AX.25 specific data */
1231 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1232 assign before registering */
1233
1234 /*
1235 * Cache lines mostly used on receive path (including eth_type_trans())
1236 */
1237 unsigned long last_rx; /* Time of last Rx
1238 * This should not be set in
1239 * drivers, unless really needed,
1240 * because network stack (bonding)
1241 * use it if/when necessary, to
1242 * avoid dirtying this cache line.
1243 */
1244
1245 /* Interface address info used in eth_type_trans() */
1246 unsigned char *dev_addr; /* hw address, (before bcast
1247 because most packets are
1248 unicast) */
1249
1250
1251 #ifdef CONFIG_RPS
1252 struct netdev_rx_queue *_rx;
1253
1254 /* Number of RX queues allocated at register_netdev() time */
1255 unsigned int num_rx_queues;
1256
1257 /* Number of RX queues currently active in device */
1258 unsigned int real_num_rx_queues;
1259
1260 #endif
1261
1262 rx_handler_func_t __rcu *rx_handler;
1263 void __rcu *rx_handler_data;
1264
1265 struct netdev_queue __rcu *ingress_queue;
1266 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1267
1268
1269 /*
1270 * Cache lines mostly used on transmit path
1271 */
1272 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1273
1274 /* Number of TX queues allocated at alloc_netdev_mq() time */
1275 unsigned int num_tx_queues;
1276
1277 /* Number of TX queues currently active in device */
1278 unsigned int real_num_tx_queues;
1279
1280 /* root qdisc from userspace point of view */
1281 struct Qdisc *qdisc;
1282
1283 unsigned long tx_queue_len; /* Max frames per queue allowed */
1284 spinlock_t tx_global_lock;
1285
1286 #ifdef CONFIG_XPS
1287 struct xps_dev_maps __rcu *xps_maps;
1288 #endif
1289 #ifdef CONFIG_RFS_ACCEL
1290 /* CPU reverse-mapping for RX completion interrupts, indexed
1291 * by RX queue number. Assigned by driver. This must only be
1292 * set if the ndo_rx_flow_steer operation is defined. */
1293 struct cpu_rmap *rx_cpu_rmap;
1294 #endif
1295
1296 /* These may be needed for future network-power-down code. */
1297
1298 /*
1299 * trans_start here is expensive for high speed devices on SMP,
1300 * please use netdev_queue->trans_start instead.
1301 */
1302 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1303
1304 int watchdog_timeo; /* used by dev_watchdog() */
1305 struct timer_list watchdog_timer;
1306
1307 /* Number of references to this device */
1308 int __percpu *pcpu_refcnt;
1309
1310 /* delayed register/unregister */
1311 struct list_head todo_list;
1312 /* device index hash chain */
1313 struct hlist_node index_hlist;
1314
1315 struct list_head link_watch_list;
1316
1317 /* register/unregister state machine */
1318 enum { NETREG_UNINITIALIZED=0,
1319 NETREG_REGISTERED, /* completed register_netdevice */
1320 NETREG_UNREGISTERING, /* called unregister_netdevice */
1321 NETREG_UNREGISTERED, /* completed unregister todo */
1322 NETREG_RELEASED, /* called free_netdev */
1323 NETREG_DUMMY, /* dummy device for NAPI poll */
1324 } reg_state:8;
1325
1326 bool dismantle; /* device is going do be freed */
1327
1328 enum {
1329 RTNL_LINK_INITIALIZED,
1330 RTNL_LINK_INITIALIZING,
1331 } rtnl_link_state:16;
1332
1333 /* Called from unregister, can be used to call free_netdev */
1334 void (*destructor)(struct net_device *dev);
1335
1336 #ifdef CONFIG_NETPOLL
1337 struct netpoll_info __rcu *npinfo;
1338 #endif
1339
1340 #ifdef CONFIG_NET_NS
1341 /* Network namespace this network device is inside */
1342 struct net *nd_net;
1343 #endif
1344
1345 /* mid-layer private */
1346 union {
1347 void *ml_priv;
1348 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1349 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1350 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1351 struct pcpu_vstats __percpu *vstats; /* veth stats */
1352 };
1353 /* GARP */
1354 struct garp_port __rcu *garp_port;
1355 /* MRP */
1356 struct mrp_port __rcu *mrp_port;
1357
1358 /* class/net/name entry */
1359 struct device dev;
1360 /* space for optional device, statistics, and wireless sysfs groups */
1361 const struct attribute_group *sysfs_groups[4];
1362
1363 /* rtnetlink link ops */
1364 const struct rtnl_link_ops *rtnl_link_ops;
1365
1366 /* for setting kernel sock attribute on TCP connection setup */
1367 #define GSO_MAX_SIZE 65536
1368 unsigned int gso_max_size;
1369 #define GSO_MAX_SEGS 65535
1370 u16 gso_max_segs;
1371
1372 #ifdef CONFIG_DCB
1373 /* Data Center Bridging netlink ops */
1374 const struct dcbnl_rtnl_ops *dcbnl_ops;
1375 #endif
1376 u8 num_tc;
1377 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1378 u8 prio_tc_map[TC_BITMASK + 1];
1379
1380 #if IS_ENABLED(CONFIG_FCOE)
1381 /* max exchange id for FCoE LRO by ddp */
1382 unsigned int fcoe_ddp_xid;
1383 #endif
1384 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1385 struct netprio_map __rcu *priomap;
1386 #endif
1387 /* phy device may attach itself for hardware timestamping */
1388 struct phy_device *phydev;
1389
1390 struct lock_class_key *qdisc_tx_busylock;
1391
1392 /* group the device belongs to */
1393 int group;
1394
1395 struct pm_qos_request pm_qos_req;
1396 };
1397 #define to_net_dev(d) container_of(d, struct net_device, dev)
1398
1399 #define NETDEV_ALIGN 32
1400
1401 static inline
1402 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1403 {
1404 return dev->prio_tc_map[prio & TC_BITMASK];
1405 }
1406
1407 static inline
1408 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1409 {
1410 if (tc >= dev->num_tc)
1411 return -EINVAL;
1412
1413 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1414 return 0;
1415 }
1416
1417 static inline
1418 void netdev_reset_tc(struct net_device *dev)
1419 {
1420 dev->num_tc = 0;
1421 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1422 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1423 }
1424
1425 static inline
1426 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1427 {
1428 if (tc >= dev->num_tc)
1429 return -EINVAL;
1430
1431 dev->tc_to_txq[tc].count = count;
1432 dev->tc_to_txq[tc].offset = offset;
1433 return 0;
1434 }
1435
1436 static inline
1437 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1438 {
1439 if (num_tc > TC_MAX_QUEUE)
1440 return -EINVAL;
1441
1442 dev->num_tc = num_tc;
1443 return 0;
1444 }
1445
1446 static inline
1447 int netdev_get_num_tc(struct net_device *dev)
1448 {
1449 return dev->num_tc;
1450 }
1451
1452 static inline
1453 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1454 unsigned int index)
1455 {
1456 return &dev->_tx[index];
1457 }
1458
1459 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1460 void (*f)(struct net_device *,
1461 struct netdev_queue *,
1462 void *),
1463 void *arg)
1464 {
1465 unsigned int i;
1466
1467 for (i = 0; i < dev->num_tx_queues; i++)
1468 f(dev, &dev->_tx[i], arg);
1469 }
1470
1471 extern struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1472 struct sk_buff *skb);
1473 extern u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb);
1474
1475 /*
1476 * Net namespace inlines
1477 */
1478 static inline
1479 struct net *dev_net(const struct net_device *dev)
1480 {
1481 return read_pnet(&dev->nd_net);
1482 }
1483
1484 static inline
1485 void dev_net_set(struct net_device *dev, struct net *net)
1486 {
1487 #ifdef CONFIG_NET_NS
1488 release_net(dev->nd_net);
1489 dev->nd_net = hold_net(net);
1490 #endif
1491 }
1492
1493 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1494 {
1495 #ifdef CONFIG_NET_DSA_TAG_DSA
1496 if (dev->dsa_ptr != NULL)
1497 return dsa_uses_dsa_tags(dev->dsa_ptr);
1498 #endif
1499
1500 return 0;
1501 }
1502
1503 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1504 {
1505 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1506 if (dev->dsa_ptr != NULL)
1507 return dsa_uses_trailer_tags(dev->dsa_ptr);
1508 #endif
1509
1510 return 0;
1511 }
1512
1513 /**
1514 * netdev_priv - access network device private data
1515 * @dev: network device
1516 *
1517 * Get network device private data
1518 */
1519 static inline void *netdev_priv(const struct net_device *dev)
1520 {
1521 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1522 }
1523
1524 /* Set the sysfs physical device reference for the network logical device
1525 * if set prior to registration will cause a symlink during initialization.
1526 */
1527 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1528
1529 /* Set the sysfs device type for the network logical device to allow
1530 * fin grained indentification of different network device types. For
1531 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1532 */
1533 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1534
1535 /* Default NAPI poll() weight
1536 * Device drivers are strongly advised to not use bigger value
1537 */
1538 #define NAPI_POLL_WEIGHT 64
1539
1540 /**
1541 * netif_napi_add - initialize a napi context
1542 * @dev: network device
1543 * @napi: napi context
1544 * @poll: polling function
1545 * @weight: default weight
1546 *
1547 * netif_napi_add() must be used to initialize a napi context prior to calling
1548 * *any* of the other napi related functions.
1549 */
1550 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1551 int (*poll)(struct napi_struct *, int), int weight);
1552
1553 /**
1554 * netif_napi_del - remove a napi context
1555 * @napi: napi context
1556 *
1557 * netif_napi_del() removes a napi context from the network device napi list
1558 */
1559 void netif_napi_del(struct napi_struct *napi);
1560
1561 struct napi_gro_cb {
1562 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1563 void *frag0;
1564
1565 /* Length of frag0. */
1566 unsigned int frag0_len;
1567
1568 /* This indicates where we are processing relative to skb->data. */
1569 int data_offset;
1570
1571 /* This is non-zero if the packet cannot be merged with the new skb. */
1572 int flush;
1573
1574 /* Number of segments aggregated. */
1575 u16 count;
1576
1577 /* This is non-zero if the packet may be of the same flow. */
1578 u8 same_flow;
1579
1580 /* Free the skb? */
1581 u8 free;
1582 #define NAPI_GRO_FREE 1
1583 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1584
1585 /* jiffies when first packet was created/queued */
1586 unsigned long age;
1587
1588 /* Used in ipv6_gro_receive() */
1589 int proto;
1590
1591 /* used in skb_gro_receive() slow path */
1592 struct sk_buff *last;
1593 };
1594
1595 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1596
1597 struct packet_type {
1598 __be16 type; /* This is really htons(ether_type). */
1599 struct net_device *dev; /* NULL is wildcarded here */
1600 int (*func) (struct sk_buff *,
1601 struct net_device *,
1602 struct packet_type *,
1603 struct net_device *);
1604 bool (*id_match)(struct packet_type *ptype,
1605 struct sock *sk);
1606 void *af_packet_priv;
1607 struct list_head list;
1608 };
1609
1610 struct offload_callbacks {
1611 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1612 netdev_features_t features);
1613 int (*gso_send_check)(struct sk_buff *skb);
1614 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1615 struct sk_buff *skb);
1616 int (*gro_complete)(struct sk_buff *skb);
1617 };
1618
1619 struct packet_offload {
1620 __be16 type; /* This is really htons(ether_type). */
1621 struct offload_callbacks callbacks;
1622 struct list_head list;
1623 };
1624
1625 #include <linux/notifier.h>
1626
1627 /* netdevice notifier chain. Please remember to update the rtnetlink
1628 * notification exclusion list in rtnetlink_event() when adding new
1629 * types.
1630 */
1631 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1632 #define NETDEV_DOWN 0x0002
1633 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1634 detected a hardware crash and restarted
1635 - we can use this eg to kick tcp sessions
1636 once done */
1637 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1638 #define NETDEV_REGISTER 0x0005
1639 #define NETDEV_UNREGISTER 0x0006
1640 #define NETDEV_CHANGEMTU 0x0007
1641 #define NETDEV_CHANGEADDR 0x0008
1642 #define NETDEV_GOING_DOWN 0x0009
1643 #define NETDEV_CHANGENAME 0x000A
1644 #define NETDEV_FEAT_CHANGE 0x000B
1645 #define NETDEV_BONDING_FAILOVER 0x000C
1646 #define NETDEV_PRE_UP 0x000D
1647 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1648 #define NETDEV_POST_TYPE_CHANGE 0x000F
1649 #define NETDEV_POST_INIT 0x0010
1650 #define NETDEV_UNREGISTER_FINAL 0x0011
1651 #define NETDEV_RELEASE 0x0012
1652 #define NETDEV_NOTIFY_PEERS 0x0013
1653 #define NETDEV_JOIN 0x0014
1654 #define NETDEV_CHANGEUPPER 0x0015
1655 #define NETDEV_RESEND_IGMP 0x0016
1656
1657 extern int register_netdevice_notifier(struct notifier_block *nb);
1658 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1659
1660 struct netdev_notifier_info {
1661 struct net_device *dev;
1662 };
1663
1664 struct netdev_notifier_change_info {
1665 struct netdev_notifier_info info; /* must be first */
1666 unsigned int flags_changed;
1667 };
1668
1669 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1670 struct net_device *dev)
1671 {
1672 info->dev = dev;
1673 }
1674
1675 static inline struct net_device *
1676 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1677 {
1678 return info->dev;
1679 }
1680
1681 extern int call_netdevice_notifiers_info(unsigned long val, struct net_device *dev,
1682 struct netdev_notifier_info *info);
1683 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1684
1685
1686 extern rwlock_t dev_base_lock; /* Device list lock */
1687
1688 #define for_each_netdev(net, d) \
1689 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1690 #define for_each_netdev_reverse(net, d) \
1691 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1692 #define for_each_netdev_rcu(net, d) \
1693 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1694 #define for_each_netdev_safe(net, d, n) \
1695 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1696 #define for_each_netdev_continue(net, d) \
1697 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1698 #define for_each_netdev_continue_rcu(net, d) \
1699 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1700 #define for_each_netdev_in_bond_rcu(bond, slave) \
1701 for_each_netdev_rcu(&init_net, slave) \
1702 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1703 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1704
1705 static inline struct net_device *next_net_device(struct net_device *dev)
1706 {
1707 struct list_head *lh;
1708 struct net *net;
1709
1710 net = dev_net(dev);
1711 lh = dev->dev_list.next;
1712 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1713 }
1714
1715 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1716 {
1717 struct list_head *lh;
1718 struct net *net;
1719
1720 net = dev_net(dev);
1721 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1722 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1723 }
1724
1725 static inline struct net_device *first_net_device(struct net *net)
1726 {
1727 return list_empty(&net->dev_base_head) ? NULL :
1728 net_device_entry(net->dev_base_head.next);
1729 }
1730
1731 static inline struct net_device *first_net_device_rcu(struct net *net)
1732 {
1733 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1734
1735 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1736 }
1737
1738 extern int netdev_boot_setup_check(struct net_device *dev);
1739 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1740 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1741 const char *hwaddr);
1742 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1743 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1744 extern void dev_add_pack(struct packet_type *pt);
1745 extern void dev_remove_pack(struct packet_type *pt);
1746 extern void __dev_remove_pack(struct packet_type *pt);
1747 extern void dev_add_offload(struct packet_offload *po);
1748 extern void dev_remove_offload(struct packet_offload *po);
1749 extern void __dev_remove_offload(struct packet_offload *po);
1750
1751 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1752 unsigned short mask);
1753 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1754 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1755 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1756 extern int dev_alloc_name(struct net_device *dev, const char *name);
1757 extern int dev_open(struct net_device *dev);
1758 extern int dev_close(struct net_device *dev);
1759 extern void dev_disable_lro(struct net_device *dev);
1760 extern int dev_loopback_xmit(struct sk_buff *newskb);
1761 extern int dev_queue_xmit(struct sk_buff *skb);
1762 extern int register_netdevice(struct net_device *dev);
1763 extern void unregister_netdevice_queue(struct net_device *dev,
1764 struct list_head *head);
1765 extern void unregister_netdevice_many(struct list_head *head);
1766 static inline void unregister_netdevice(struct net_device *dev)
1767 {
1768 unregister_netdevice_queue(dev, NULL);
1769 }
1770
1771 extern int netdev_refcnt_read(const struct net_device *dev);
1772 extern void free_netdev(struct net_device *dev);
1773 extern void synchronize_net(void);
1774 extern int init_dummy_netdev(struct net_device *dev);
1775
1776 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1777 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1778 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1779 extern int netdev_get_name(struct net *net, char *name, int ifindex);
1780 extern int dev_restart(struct net_device *dev);
1781 #ifdef CONFIG_NETPOLL_TRAP
1782 extern int netpoll_trap(void);
1783 #endif
1784 extern int skb_gro_receive(struct sk_buff **head,
1785 struct sk_buff *skb);
1786
1787 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1788 {
1789 return NAPI_GRO_CB(skb)->data_offset;
1790 }
1791
1792 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1793 {
1794 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1795 }
1796
1797 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1798 {
1799 NAPI_GRO_CB(skb)->data_offset += len;
1800 }
1801
1802 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1803 unsigned int offset)
1804 {
1805 return NAPI_GRO_CB(skb)->frag0 + offset;
1806 }
1807
1808 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1809 {
1810 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1811 }
1812
1813 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1814 unsigned int offset)
1815 {
1816 if (!pskb_may_pull(skb, hlen))
1817 return NULL;
1818
1819 NAPI_GRO_CB(skb)->frag0 = NULL;
1820 NAPI_GRO_CB(skb)->frag0_len = 0;
1821 return skb->data + offset;
1822 }
1823
1824 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1825 {
1826 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1827 }
1828
1829 static inline void *skb_gro_network_header(struct sk_buff *skb)
1830 {
1831 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1832 skb_network_offset(skb);
1833 }
1834
1835 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1836 unsigned short type,
1837 const void *daddr, const void *saddr,
1838 unsigned int len)
1839 {
1840 if (!dev->header_ops || !dev->header_ops->create)
1841 return 0;
1842
1843 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1844 }
1845
1846 static inline int dev_parse_header(const struct sk_buff *skb,
1847 unsigned char *haddr)
1848 {
1849 const struct net_device *dev = skb->dev;
1850
1851 if (!dev->header_ops || !dev->header_ops->parse)
1852 return 0;
1853 return dev->header_ops->parse(skb, haddr);
1854 }
1855
1856 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1857 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1858 static inline int unregister_gifconf(unsigned int family)
1859 {
1860 return register_gifconf(family, NULL);
1861 }
1862
1863 #ifdef CONFIG_NET_FLOW_LIMIT
1864 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
1865 struct sd_flow_limit {
1866 u64 count;
1867 unsigned int num_buckets;
1868 unsigned int history_head;
1869 u16 history[FLOW_LIMIT_HISTORY];
1870 u8 buckets[];
1871 };
1872
1873 extern int netdev_flow_limit_table_len;
1874 #endif /* CONFIG_NET_FLOW_LIMIT */
1875
1876 /*
1877 * Incoming packets are placed on per-cpu queues
1878 */
1879 struct softnet_data {
1880 struct Qdisc *output_queue;
1881 struct Qdisc **output_queue_tailp;
1882 struct list_head poll_list;
1883 struct sk_buff *completion_queue;
1884 struct sk_buff_head process_queue;
1885
1886 /* stats */
1887 unsigned int processed;
1888 unsigned int time_squeeze;
1889 unsigned int cpu_collision;
1890 unsigned int received_rps;
1891
1892 #ifdef CONFIG_RPS
1893 struct softnet_data *rps_ipi_list;
1894
1895 /* Elements below can be accessed between CPUs for RPS */
1896 struct call_single_data csd ____cacheline_aligned_in_smp;
1897 struct softnet_data *rps_ipi_next;
1898 unsigned int cpu;
1899 unsigned int input_queue_head;
1900 unsigned int input_queue_tail;
1901 #endif
1902 unsigned int dropped;
1903 struct sk_buff_head input_pkt_queue;
1904 struct napi_struct backlog;
1905
1906 #ifdef CONFIG_NET_FLOW_LIMIT
1907 struct sd_flow_limit __rcu *flow_limit;
1908 #endif
1909 };
1910
1911 static inline void input_queue_head_incr(struct softnet_data *sd)
1912 {
1913 #ifdef CONFIG_RPS
1914 sd->input_queue_head++;
1915 #endif
1916 }
1917
1918 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1919 unsigned int *qtail)
1920 {
1921 #ifdef CONFIG_RPS
1922 *qtail = ++sd->input_queue_tail;
1923 #endif
1924 }
1925
1926 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1927
1928 extern void __netif_schedule(struct Qdisc *q);
1929
1930 static inline void netif_schedule_queue(struct netdev_queue *txq)
1931 {
1932 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1933 __netif_schedule(txq->qdisc);
1934 }
1935
1936 static inline void netif_tx_schedule_all(struct net_device *dev)
1937 {
1938 unsigned int i;
1939
1940 for (i = 0; i < dev->num_tx_queues; i++)
1941 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1942 }
1943
1944 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1945 {
1946 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1947 }
1948
1949 /**
1950 * netif_start_queue - allow transmit
1951 * @dev: network device
1952 *
1953 * Allow upper layers to call the device hard_start_xmit routine.
1954 */
1955 static inline void netif_start_queue(struct net_device *dev)
1956 {
1957 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1958 }
1959
1960 static inline void netif_tx_start_all_queues(struct net_device *dev)
1961 {
1962 unsigned int i;
1963
1964 for (i = 0; i < dev->num_tx_queues; i++) {
1965 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1966 netif_tx_start_queue(txq);
1967 }
1968 }
1969
1970 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1971 {
1972 #ifdef CONFIG_NETPOLL_TRAP
1973 if (netpoll_trap()) {
1974 netif_tx_start_queue(dev_queue);
1975 return;
1976 }
1977 #endif
1978 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1979 __netif_schedule(dev_queue->qdisc);
1980 }
1981
1982 /**
1983 * netif_wake_queue - restart transmit
1984 * @dev: network device
1985 *
1986 * Allow upper layers to call the device hard_start_xmit routine.
1987 * Used for flow control when transmit resources are available.
1988 */
1989 static inline void netif_wake_queue(struct net_device *dev)
1990 {
1991 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1992 }
1993
1994 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1995 {
1996 unsigned int i;
1997
1998 for (i = 0; i < dev->num_tx_queues; i++) {
1999 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2000 netif_tx_wake_queue(txq);
2001 }
2002 }
2003
2004 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2005 {
2006 if (WARN_ON(!dev_queue)) {
2007 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2008 return;
2009 }
2010 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2011 }
2012
2013 /**
2014 * netif_stop_queue - stop transmitted packets
2015 * @dev: network device
2016 *
2017 * Stop upper layers calling the device hard_start_xmit routine.
2018 * Used for flow control when transmit resources are unavailable.
2019 */
2020 static inline void netif_stop_queue(struct net_device *dev)
2021 {
2022 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2023 }
2024
2025 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2026 {
2027 unsigned int i;
2028
2029 for (i = 0; i < dev->num_tx_queues; i++) {
2030 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2031 netif_tx_stop_queue(txq);
2032 }
2033 }
2034
2035 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2036 {
2037 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2038 }
2039
2040 /**
2041 * netif_queue_stopped - test if transmit queue is flowblocked
2042 * @dev: network device
2043 *
2044 * Test if transmit queue on device is currently unable to send.
2045 */
2046 static inline bool netif_queue_stopped(const struct net_device *dev)
2047 {
2048 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2049 }
2050
2051 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2052 {
2053 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2054 }
2055
2056 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2057 {
2058 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2059 }
2060
2061 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2062 unsigned int bytes)
2063 {
2064 #ifdef CONFIG_BQL
2065 dql_queued(&dev_queue->dql, bytes);
2066
2067 if (likely(dql_avail(&dev_queue->dql) >= 0))
2068 return;
2069
2070 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2071
2072 /*
2073 * The XOFF flag must be set before checking the dql_avail below,
2074 * because in netdev_tx_completed_queue we update the dql_completed
2075 * before checking the XOFF flag.
2076 */
2077 smp_mb();
2078
2079 /* check again in case another CPU has just made room avail */
2080 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2081 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2082 #endif
2083 }
2084
2085 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2086 {
2087 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2088 }
2089
2090 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2091 unsigned int pkts, unsigned int bytes)
2092 {
2093 #ifdef CONFIG_BQL
2094 if (unlikely(!bytes))
2095 return;
2096
2097 dql_completed(&dev_queue->dql, bytes);
2098
2099 /*
2100 * Without the memory barrier there is a small possiblity that
2101 * netdev_tx_sent_queue will miss the update and cause the queue to
2102 * be stopped forever
2103 */
2104 smp_mb();
2105
2106 if (dql_avail(&dev_queue->dql) < 0)
2107 return;
2108
2109 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2110 netif_schedule_queue(dev_queue);
2111 #endif
2112 }
2113
2114 static inline void netdev_completed_queue(struct net_device *dev,
2115 unsigned int pkts, unsigned int bytes)
2116 {
2117 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2118 }
2119
2120 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2121 {
2122 #ifdef CONFIG_BQL
2123 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2124 dql_reset(&q->dql);
2125 #endif
2126 }
2127
2128 static inline void netdev_reset_queue(struct net_device *dev_queue)
2129 {
2130 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2131 }
2132
2133 /**
2134 * netif_running - test if up
2135 * @dev: network device
2136 *
2137 * Test if the device has been brought up.
2138 */
2139 static inline bool netif_running(const struct net_device *dev)
2140 {
2141 return test_bit(__LINK_STATE_START, &dev->state);
2142 }
2143
2144 /*
2145 * Routines to manage the subqueues on a device. We only need start
2146 * stop, and a check if it's stopped. All other device management is
2147 * done at the overall netdevice level.
2148 * Also test the device if we're multiqueue.
2149 */
2150
2151 /**
2152 * netif_start_subqueue - allow sending packets on subqueue
2153 * @dev: network device
2154 * @queue_index: sub queue index
2155 *
2156 * Start individual transmit queue of a device with multiple transmit queues.
2157 */
2158 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2159 {
2160 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2161
2162 netif_tx_start_queue(txq);
2163 }
2164
2165 /**
2166 * netif_stop_subqueue - stop sending packets on subqueue
2167 * @dev: network device
2168 * @queue_index: sub queue index
2169 *
2170 * Stop individual transmit queue of a device with multiple transmit queues.
2171 */
2172 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2173 {
2174 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2175 #ifdef CONFIG_NETPOLL_TRAP
2176 if (netpoll_trap())
2177 return;
2178 #endif
2179 netif_tx_stop_queue(txq);
2180 }
2181
2182 /**
2183 * netif_subqueue_stopped - test status of subqueue
2184 * @dev: network device
2185 * @queue_index: sub queue index
2186 *
2187 * Check individual transmit queue of a device with multiple transmit queues.
2188 */
2189 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2190 u16 queue_index)
2191 {
2192 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2193
2194 return netif_tx_queue_stopped(txq);
2195 }
2196
2197 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2198 struct sk_buff *skb)
2199 {
2200 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2201 }
2202
2203 /**
2204 * netif_wake_subqueue - allow sending packets on subqueue
2205 * @dev: network device
2206 * @queue_index: sub queue index
2207 *
2208 * Resume individual transmit queue of a device with multiple transmit queues.
2209 */
2210 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2211 {
2212 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2213 #ifdef CONFIG_NETPOLL_TRAP
2214 if (netpoll_trap())
2215 return;
2216 #endif
2217 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2218 __netif_schedule(txq->qdisc);
2219 }
2220
2221 #ifdef CONFIG_XPS
2222 extern int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask,
2223 u16 index);
2224 #else
2225 static inline int netif_set_xps_queue(struct net_device *dev,
2226 struct cpumask *mask,
2227 u16 index)
2228 {
2229 return 0;
2230 }
2231 #endif
2232
2233 /*
2234 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2235 * as a distribution range limit for the returned value.
2236 */
2237 static inline u16 skb_tx_hash(const struct net_device *dev,
2238 const struct sk_buff *skb)
2239 {
2240 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2241 }
2242
2243 /**
2244 * netif_is_multiqueue - test if device has multiple transmit queues
2245 * @dev: network device
2246 *
2247 * Check if device has multiple transmit queues
2248 */
2249 static inline bool netif_is_multiqueue(const struct net_device *dev)
2250 {
2251 return dev->num_tx_queues > 1;
2252 }
2253
2254 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2255 unsigned int txq);
2256
2257 #ifdef CONFIG_RPS
2258 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2259 unsigned int rxq);
2260 #else
2261 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2262 unsigned int rxq)
2263 {
2264 return 0;
2265 }
2266 #endif
2267
2268 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2269 const struct net_device *from_dev)
2270 {
2271 int err;
2272
2273 err = netif_set_real_num_tx_queues(to_dev,
2274 from_dev->real_num_tx_queues);
2275 if (err)
2276 return err;
2277 #ifdef CONFIG_RPS
2278 return netif_set_real_num_rx_queues(to_dev,
2279 from_dev->real_num_rx_queues);
2280 #else
2281 return 0;
2282 #endif
2283 }
2284
2285 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2286 extern int netif_get_num_default_rss_queues(void);
2287
2288 /* Use this variant when it is known for sure that it
2289 * is executing from hardware interrupt context or with hardware interrupts
2290 * disabled.
2291 */
2292 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2293
2294 /* Use this variant in places where it could be invoked
2295 * from either hardware interrupt or other context, with hardware interrupts
2296 * either disabled or enabled.
2297 */
2298 extern void dev_kfree_skb_any(struct sk_buff *skb);
2299
2300 extern int netif_rx(struct sk_buff *skb);
2301 extern int netif_rx_ni(struct sk_buff *skb);
2302 extern int netif_receive_skb(struct sk_buff *skb);
2303 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2304 struct sk_buff *skb);
2305 extern void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2306 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2307 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2308
2309 static inline void napi_free_frags(struct napi_struct *napi)
2310 {
2311 kfree_skb(napi->skb);
2312 napi->skb = NULL;
2313 }
2314
2315 extern int netdev_rx_handler_register(struct net_device *dev,
2316 rx_handler_func_t *rx_handler,
2317 void *rx_handler_data);
2318 extern void netdev_rx_handler_unregister(struct net_device *dev);
2319
2320 extern bool dev_valid_name(const char *name);
2321 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2322 extern int dev_ethtool(struct net *net, struct ifreq *);
2323 extern unsigned int dev_get_flags(const struct net_device *);
2324 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2325 extern int dev_change_flags(struct net_device *, unsigned int);
2326 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2327 extern int dev_change_name(struct net_device *, const char *);
2328 extern int dev_set_alias(struct net_device *, const char *, size_t);
2329 extern int dev_change_net_namespace(struct net_device *,
2330 struct net *, const char *);
2331 extern int dev_set_mtu(struct net_device *, int);
2332 extern void dev_set_group(struct net_device *, int);
2333 extern int dev_set_mac_address(struct net_device *,
2334 struct sockaddr *);
2335 extern int dev_change_carrier(struct net_device *,
2336 bool new_carrier);
2337 extern int dev_get_phys_port_id(struct net_device *dev,
2338 struct netdev_phys_port_id *ppid);
2339 extern int dev_hard_start_xmit(struct sk_buff *skb,
2340 struct net_device *dev,
2341 struct netdev_queue *txq);
2342 extern int dev_forward_skb(struct net_device *dev,
2343 struct sk_buff *skb);
2344
2345 extern int netdev_budget;
2346
2347 /* Called by rtnetlink.c:rtnl_unlock() */
2348 extern void netdev_run_todo(void);
2349
2350 /**
2351 * dev_put - release reference to device
2352 * @dev: network device
2353 *
2354 * Release reference to device to allow it to be freed.
2355 */
2356 static inline void dev_put(struct net_device *dev)
2357 {
2358 this_cpu_dec(*dev->pcpu_refcnt);
2359 }
2360
2361 /**
2362 * dev_hold - get reference to device
2363 * @dev: network device
2364 *
2365 * Hold reference to device to keep it from being freed.
2366 */
2367 static inline void dev_hold(struct net_device *dev)
2368 {
2369 this_cpu_inc(*dev->pcpu_refcnt);
2370 }
2371
2372 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2373 * and _off may be called from IRQ context, but it is caller
2374 * who is responsible for serialization of these calls.
2375 *
2376 * The name carrier is inappropriate, these functions should really be
2377 * called netif_lowerlayer_*() because they represent the state of any
2378 * kind of lower layer not just hardware media.
2379 */
2380
2381 extern void linkwatch_init_dev(struct net_device *dev);
2382 extern void linkwatch_fire_event(struct net_device *dev);
2383 extern void linkwatch_forget_dev(struct net_device *dev);
2384
2385 /**
2386 * netif_carrier_ok - test if carrier present
2387 * @dev: network device
2388 *
2389 * Check if carrier is present on device
2390 */
2391 static inline bool netif_carrier_ok(const struct net_device *dev)
2392 {
2393 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2394 }
2395
2396 extern unsigned long dev_trans_start(struct net_device *dev);
2397
2398 extern void __netdev_watchdog_up(struct net_device *dev);
2399
2400 extern void netif_carrier_on(struct net_device *dev);
2401
2402 extern void netif_carrier_off(struct net_device *dev);
2403
2404 /**
2405 * netif_dormant_on - mark device as dormant.
2406 * @dev: network device
2407 *
2408 * Mark device as dormant (as per RFC2863).
2409 *
2410 * The dormant state indicates that the relevant interface is not
2411 * actually in a condition to pass packets (i.e., it is not 'up') but is
2412 * in a "pending" state, waiting for some external event. For "on-
2413 * demand" interfaces, this new state identifies the situation where the
2414 * interface is waiting for events to place it in the up state.
2415 *
2416 */
2417 static inline void netif_dormant_on(struct net_device *dev)
2418 {
2419 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2420 linkwatch_fire_event(dev);
2421 }
2422
2423 /**
2424 * netif_dormant_off - set device as not dormant.
2425 * @dev: network device
2426 *
2427 * Device is not in dormant state.
2428 */
2429 static inline void netif_dormant_off(struct net_device *dev)
2430 {
2431 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2432 linkwatch_fire_event(dev);
2433 }
2434
2435 /**
2436 * netif_dormant - test if carrier present
2437 * @dev: network device
2438 *
2439 * Check if carrier is present on device
2440 */
2441 static inline bool netif_dormant(const struct net_device *dev)
2442 {
2443 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2444 }
2445
2446
2447 /**
2448 * netif_oper_up - test if device is operational
2449 * @dev: network device
2450 *
2451 * Check if carrier is operational
2452 */
2453 static inline bool netif_oper_up(const struct net_device *dev)
2454 {
2455 return (dev->operstate == IF_OPER_UP ||
2456 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2457 }
2458
2459 /**
2460 * netif_device_present - is device available or removed
2461 * @dev: network device
2462 *
2463 * Check if device has not been removed from system.
2464 */
2465 static inline bool netif_device_present(struct net_device *dev)
2466 {
2467 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2468 }
2469
2470 extern void netif_device_detach(struct net_device *dev);
2471
2472 extern void netif_device_attach(struct net_device *dev);
2473
2474 /*
2475 * Network interface message level settings
2476 */
2477
2478 enum {
2479 NETIF_MSG_DRV = 0x0001,
2480 NETIF_MSG_PROBE = 0x0002,
2481 NETIF_MSG_LINK = 0x0004,
2482 NETIF_MSG_TIMER = 0x0008,
2483 NETIF_MSG_IFDOWN = 0x0010,
2484 NETIF_MSG_IFUP = 0x0020,
2485 NETIF_MSG_RX_ERR = 0x0040,
2486 NETIF_MSG_TX_ERR = 0x0080,
2487 NETIF_MSG_TX_QUEUED = 0x0100,
2488 NETIF_MSG_INTR = 0x0200,
2489 NETIF_MSG_TX_DONE = 0x0400,
2490 NETIF_MSG_RX_STATUS = 0x0800,
2491 NETIF_MSG_PKTDATA = 0x1000,
2492 NETIF_MSG_HW = 0x2000,
2493 NETIF_MSG_WOL = 0x4000,
2494 };
2495
2496 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2497 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2498 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2499 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2500 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2501 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2502 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2503 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2504 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2505 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2506 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2507 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2508 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2509 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2510 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2511
2512 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2513 {
2514 /* use default */
2515 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2516 return default_msg_enable_bits;
2517 if (debug_value == 0) /* no output */
2518 return 0;
2519 /* set low N bits */
2520 return (1 << debug_value) - 1;
2521 }
2522
2523 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2524 {
2525 spin_lock(&txq->_xmit_lock);
2526 txq->xmit_lock_owner = cpu;
2527 }
2528
2529 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2530 {
2531 spin_lock_bh(&txq->_xmit_lock);
2532 txq->xmit_lock_owner = smp_processor_id();
2533 }
2534
2535 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2536 {
2537 bool ok = spin_trylock(&txq->_xmit_lock);
2538 if (likely(ok))
2539 txq->xmit_lock_owner = smp_processor_id();
2540 return ok;
2541 }
2542
2543 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2544 {
2545 txq->xmit_lock_owner = -1;
2546 spin_unlock(&txq->_xmit_lock);
2547 }
2548
2549 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2550 {
2551 txq->xmit_lock_owner = -1;
2552 spin_unlock_bh(&txq->_xmit_lock);
2553 }
2554
2555 static inline void txq_trans_update(struct netdev_queue *txq)
2556 {
2557 if (txq->xmit_lock_owner != -1)
2558 txq->trans_start = jiffies;
2559 }
2560
2561 /**
2562 * netif_tx_lock - grab network device transmit lock
2563 * @dev: network device
2564 *
2565 * Get network device transmit lock
2566 */
2567 static inline void netif_tx_lock(struct net_device *dev)
2568 {
2569 unsigned int i;
2570 int cpu;
2571
2572 spin_lock(&dev->tx_global_lock);
2573 cpu = smp_processor_id();
2574 for (i = 0; i < dev->num_tx_queues; i++) {
2575 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2576
2577 /* We are the only thread of execution doing a
2578 * freeze, but we have to grab the _xmit_lock in
2579 * order to synchronize with threads which are in
2580 * the ->hard_start_xmit() handler and already
2581 * checked the frozen bit.
2582 */
2583 __netif_tx_lock(txq, cpu);
2584 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2585 __netif_tx_unlock(txq);
2586 }
2587 }
2588
2589 static inline void netif_tx_lock_bh(struct net_device *dev)
2590 {
2591 local_bh_disable();
2592 netif_tx_lock(dev);
2593 }
2594
2595 static inline void netif_tx_unlock(struct net_device *dev)
2596 {
2597 unsigned int i;
2598
2599 for (i = 0; i < dev->num_tx_queues; i++) {
2600 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2601
2602 /* No need to grab the _xmit_lock here. If the
2603 * queue is not stopped for another reason, we
2604 * force a schedule.
2605 */
2606 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2607 netif_schedule_queue(txq);
2608 }
2609 spin_unlock(&dev->tx_global_lock);
2610 }
2611
2612 static inline void netif_tx_unlock_bh(struct net_device *dev)
2613 {
2614 netif_tx_unlock(dev);
2615 local_bh_enable();
2616 }
2617
2618 #define HARD_TX_LOCK(dev, txq, cpu) { \
2619 if ((dev->features & NETIF_F_LLTX) == 0) { \
2620 __netif_tx_lock(txq, cpu); \
2621 } \
2622 }
2623
2624 #define HARD_TX_UNLOCK(dev, txq) { \
2625 if ((dev->features & NETIF_F_LLTX) == 0) { \
2626 __netif_tx_unlock(txq); \
2627 } \
2628 }
2629
2630 static inline void netif_tx_disable(struct net_device *dev)
2631 {
2632 unsigned int i;
2633 int cpu;
2634
2635 local_bh_disable();
2636 cpu = smp_processor_id();
2637 for (i = 0; i < dev->num_tx_queues; i++) {
2638 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2639
2640 __netif_tx_lock(txq, cpu);
2641 netif_tx_stop_queue(txq);
2642 __netif_tx_unlock(txq);
2643 }
2644 local_bh_enable();
2645 }
2646
2647 static inline void netif_addr_lock(struct net_device *dev)
2648 {
2649 spin_lock(&dev->addr_list_lock);
2650 }
2651
2652 static inline void netif_addr_lock_nested(struct net_device *dev)
2653 {
2654 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2655 }
2656
2657 static inline void netif_addr_lock_bh(struct net_device *dev)
2658 {
2659 spin_lock_bh(&dev->addr_list_lock);
2660 }
2661
2662 static inline void netif_addr_unlock(struct net_device *dev)
2663 {
2664 spin_unlock(&dev->addr_list_lock);
2665 }
2666
2667 static inline void netif_addr_unlock_bh(struct net_device *dev)
2668 {
2669 spin_unlock_bh(&dev->addr_list_lock);
2670 }
2671
2672 /*
2673 * dev_addrs walker. Should be used only for read access. Call with
2674 * rcu_read_lock held.
2675 */
2676 #define for_each_dev_addr(dev, ha) \
2677 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2678
2679 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2680
2681 extern void ether_setup(struct net_device *dev);
2682
2683 /* Support for loadable net-drivers */
2684 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2685 void (*setup)(struct net_device *),
2686 unsigned int txqs, unsigned int rxqs);
2687 #define alloc_netdev(sizeof_priv, name, setup) \
2688 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2689
2690 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2691 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2692
2693 extern int register_netdev(struct net_device *dev);
2694 extern void unregister_netdev(struct net_device *dev);
2695
2696 /* General hardware address lists handling functions */
2697 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2698 struct netdev_hw_addr_list *from_list,
2699 int addr_len, unsigned char addr_type);
2700 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2701 struct netdev_hw_addr_list *from_list,
2702 int addr_len, unsigned char addr_type);
2703 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2704 struct netdev_hw_addr_list *from_list,
2705 int addr_len);
2706 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2707 struct netdev_hw_addr_list *from_list,
2708 int addr_len);
2709 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2710 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2711
2712 /* Functions used for device addresses handling */
2713 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2714 unsigned char addr_type);
2715 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2716 unsigned char addr_type);
2717 extern int dev_addr_add_multiple(struct net_device *to_dev,
2718 struct net_device *from_dev,
2719 unsigned char addr_type);
2720 extern int dev_addr_del_multiple(struct net_device *to_dev,
2721 struct net_device *from_dev,
2722 unsigned char addr_type);
2723 extern void dev_addr_flush(struct net_device *dev);
2724 extern int dev_addr_init(struct net_device *dev);
2725
2726 /* Functions used for unicast addresses handling */
2727 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2728 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2729 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2730 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2731 extern int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
2732 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2733 extern void dev_uc_flush(struct net_device *dev);
2734 extern void dev_uc_init(struct net_device *dev);
2735
2736 /* Functions used for multicast addresses handling */
2737 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2738 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2739 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2740 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2741 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2742 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2743 extern int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
2744 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2745 extern void dev_mc_flush(struct net_device *dev);
2746 extern void dev_mc_init(struct net_device *dev);
2747
2748 /* Functions used for secondary unicast and multicast support */
2749 extern void dev_set_rx_mode(struct net_device *dev);
2750 extern void __dev_set_rx_mode(struct net_device *dev);
2751 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2752 extern int dev_set_allmulti(struct net_device *dev, int inc);
2753 extern void netdev_state_change(struct net_device *dev);
2754 extern void netdev_notify_peers(struct net_device *dev);
2755 extern void netdev_features_change(struct net_device *dev);
2756 /* Load a device via the kmod */
2757 extern void dev_load(struct net *net, const char *name);
2758 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2759 struct rtnl_link_stats64 *storage);
2760 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2761 const struct net_device_stats *netdev_stats);
2762
2763 extern int netdev_max_backlog;
2764 extern int netdev_tstamp_prequeue;
2765 extern int weight_p;
2766 extern int bpf_jit_enable;
2767
2768 extern bool netdev_has_upper_dev(struct net_device *dev,
2769 struct net_device *upper_dev);
2770 extern bool netdev_has_any_upper_dev(struct net_device *dev);
2771 extern struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
2772 extern struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
2773 extern int netdev_upper_dev_link(struct net_device *dev,
2774 struct net_device *upper_dev);
2775 extern int netdev_master_upper_dev_link(struct net_device *dev,
2776 struct net_device *upper_dev);
2777 extern void netdev_upper_dev_unlink(struct net_device *dev,
2778 struct net_device *upper_dev);
2779 extern int skb_checksum_help(struct sk_buff *skb);
2780 extern struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
2781 netdev_features_t features, bool tx_path);
2782 extern struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
2783 netdev_features_t features);
2784
2785 static inline
2786 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
2787 {
2788 return __skb_gso_segment(skb, features, true);
2789 }
2790 __be16 skb_network_protocol(struct sk_buff *skb);
2791
2792 static inline bool can_checksum_protocol(netdev_features_t features,
2793 __be16 protocol)
2794 {
2795 return ((features & NETIF_F_GEN_CSUM) ||
2796 ((features & NETIF_F_V4_CSUM) &&
2797 protocol == htons(ETH_P_IP)) ||
2798 ((features & NETIF_F_V6_CSUM) &&
2799 protocol == htons(ETH_P_IPV6)) ||
2800 ((features & NETIF_F_FCOE_CRC) &&
2801 protocol == htons(ETH_P_FCOE)));
2802 }
2803
2804 #ifdef CONFIG_BUG
2805 extern void netdev_rx_csum_fault(struct net_device *dev);
2806 #else
2807 static inline void netdev_rx_csum_fault(struct net_device *dev)
2808 {
2809 }
2810 #endif
2811 /* rx skb timestamps */
2812 extern void net_enable_timestamp(void);
2813 extern void net_disable_timestamp(void);
2814
2815 #ifdef CONFIG_PROC_FS
2816 extern int __init dev_proc_init(void);
2817 #else
2818 #define dev_proc_init() 0
2819 #endif
2820
2821 extern int netdev_class_create_file(struct class_attribute *class_attr);
2822 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2823
2824 extern struct kobj_ns_type_operations net_ns_type_operations;
2825
2826 extern const char *netdev_drivername(const struct net_device *dev);
2827
2828 extern void linkwatch_run_queue(void);
2829
2830 static inline netdev_features_t netdev_get_wanted_features(
2831 struct net_device *dev)
2832 {
2833 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2834 }
2835 netdev_features_t netdev_increment_features(netdev_features_t all,
2836 netdev_features_t one, netdev_features_t mask);
2837
2838 /* Allow TSO being used on stacked device :
2839 * Performing the GSO segmentation before last device
2840 * is a performance improvement.
2841 */
2842 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
2843 netdev_features_t mask)
2844 {
2845 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
2846 }
2847
2848 int __netdev_update_features(struct net_device *dev);
2849 void netdev_update_features(struct net_device *dev);
2850 void netdev_change_features(struct net_device *dev);
2851
2852 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2853 struct net_device *dev);
2854
2855 netdev_features_t netif_skb_features(struct sk_buff *skb);
2856
2857 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2858 {
2859 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2860
2861 /* check flags correspondence */
2862 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2863 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2864 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2865 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2866 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2867 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2868
2869 return (features & feature) == feature;
2870 }
2871
2872 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2873 {
2874 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2875 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2876 }
2877
2878 static inline bool netif_needs_gso(struct sk_buff *skb,
2879 netdev_features_t features)
2880 {
2881 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2882 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2883 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2884 }
2885
2886 static inline void netif_set_gso_max_size(struct net_device *dev,
2887 unsigned int size)
2888 {
2889 dev->gso_max_size = size;
2890 }
2891
2892 static inline bool netif_is_bond_master(struct net_device *dev)
2893 {
2894 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
2895 }
2896
2897 static inline bool netif_is_bond_slave(struct net_device *dev)
2898 {
2899 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2900 }
2901
2902 static inline bool netif_supports_nofcs(struct net_device *dev)
2903 {
2904 return dev->priv_flags & IFF_SUPP_NOFCS;
2905 }
2906
2907 extern struct pernet_operations __net_initdata loopback_net_ops;
2908
2909 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2910
2911 /* netdev_printk helpers, similar to dev_printk */
2912
2913 static inline const char *netdev_name(const struct net_device *dev)
2914 {
2915 if (dev->reg_state != NETREG_REGISTERED)
2916 return "(unregistered net_device)";
2917 return dev->name;
2918 }
2919
2920 extern __printf(3, 4)
2921 int netdev_printk(const char *level, const struct net_device *dev,
2922 const char *format, ...);
2923 extern __printf(2, 3)
2924 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2925 extern __printf(2, 3)
2926 int netdev_alert(const struct net_device *dev, const char *format, ...);
2927 extern __printf(2, 3)
2928 int netdev_crit(const struct net_device *dev, const char *format, ...);
2929 extern __printf(2, 3)
2930 int netdev_err(const struct net_device *dev, const char *format, ...);
2931 extern __printf(2, 3)
2932 int netdev_warn(const struct net_device *dev, const char *format, ...);
2933 extern __printf(2, 3)
2934 int netdev_notice(const struct net_device *dev, const char *format, ...);
2935 extern __printf(2, 3)
2936 int netdev_info(const struct net_device *dev, const char *format, ...);
2937
2938 #define MODULE_ALIAS_NETDEV(device) \
2939 MODULE_ALIAS("netdev-" device)
2940
2941 #if defined(CONFIG_DYNAMIC_DEBUG)
2942 #define netdev_dbg(__dev, format, args...) \
2943 do { \
2944 dynamic_netdev_dbg(__dev, format, ##args); \
2945 } while (0)
2946 #elif defined(DEBUG)
2947 #define netdev_dbg(__dev, format, args...) \
2948 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2949 #else
2950 #define netdev_dbg(__dev, format, args...) \
2951 ({ \
2952 if (0) \
2953 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2954 0; \
2955 })
2956 #endif
2957
2958 #if defined(VERBOSE_DEBUG)
2959 #define netdev_vdbg netdev_dbg
2960 #else
2961
2962 #define netdev_vdbg(dev, format, args...) \
2963 ({ \
2964 if (0) \
2965 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2966 0; \
2967 })
2968 #endif
2969
2970 /*
2971 * netdev_WARN() acts like dev_printk(), but with the key difference
2972 * of using a WARN/WARN_ON to get the message out, including the
2973 * file/line information and a backtrace.
2974 */
2975 #define netdev_WARN(dev, format, args...) \
2976 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2977
2978 /* netif printk helpers, similar to netdev_printk */
2979
2980 #define netif_printk(priv, type, level, dev, fmt, args...) \
2981 do { \
2982 if (netif_msg_##type(priv)) \
2983 netdev_printk(level, (dev), fmt, ##args); \
2984 } while (0)
2985
2986 #define netif_level(level, priv, type, dev, fmt, args...) \
2987 do { \
2988 if (netif_msg_##type(priv)) \
2989 netdev_##level(dev, fmt, ##args); \
2990 } while (0)
2991
2992 #define netif_emerg(priv, type, dev, fmt, args...) \
2993 netif_level(emerg, priv, type, dev, fmt, ##args)
2994 #define netif_alert(priv, type, dev, fmt, args...) \
2995 netif_level(alert, priv, type, dev, fmt, ##args)
2996 #define netif_crit(priv, type, dev, fmt, args...) \
2997 netif_level(crit, priv, type, dev, fmt, ##args)
2998 #define netif_err(priv, type, dev, fmt, args...) \
2999 netif_level(err, priv, type, dev, fmt, ##args)
3000 #define netif_warn(priv, type, dev, fmt, args...) \
3001 netif_level(warn, priv, type, dev, fmt, ##args)
3002 #define netif_notice(priv, type, dev, fmt, args...) \
3003 netif_level(notice, priv, type, dev, fmt, ##args)
3004 #define netif_info(priv, type, dev, fmt, args...) \
3005 netif_level(info, priv, type, dev, fmt, ##args)
3006
3007 #if defined(CONFIG_DYNAMIC_DEBUG)
3008 #define netif_dbg(priv, type, netdev, format, args...) \
3009 do { \
3010 if (netif_msg_##type(priv)) \
3011 dynamic_netdev_dbg(netdev, format, ##args); \
3012 } while (0)
3013 #elif defined(DEBUG)
3014 #define netif_dbg(priv, type, dev, format, args...) \
3015 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3016 #else
3017 #define netif_dbg(priv, type, dev, format, args...) \
3018 ({ \
3019 if (0) \
3020 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3021 0; \
3022 })
3023 #endif
3024
3025 #if defined(VERBOSE_DEBUG)
3026 #define netif_vdbg netif_dbg
3027 #else
3028 #define netif_vdbg(priv, type, dev, format, args...) \
3029 ({ \
3030 if (0) \
3031 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3032 0; \
3033 })
3034 #endif
3035
3036 /*
3037 * The list of packet types we will receive (as opposed to discard)
3038 * and the routines to invoke.
3039 *
3040 * Why 16. Because with 16 the only overlap we get on a hash of the
3041 * low nibble of the protocol value is RARP/SNAP/X.25.
3042 *
3043 * NOTE: That is no longer true with the addition of VLAN tags. Not
3044 * sure which should go first, but I bet it won't make much
3045 * difference if we are running VLANs. The good news is that
3046 * this protocol won't be in the list unless compiled in, so
3047 * the average user (w/out VLANs) will not be adversely affected.
3048 * --BLG
3049 *
3050 * 0800 IP
3051 * 8100 802.1Q VLAN
3052 * 0001 802.3
3053 * 0002 AX.25
3054 * 0004 802.2
3055 * 8035 RARP
3056 * 0005 SNAP
3057 * 0805 X.25
3058 * 0806 ARP
3059 * 8137 IPX
3060 * 0009 Localtalk
3061 * 86DD IPv6
3062 */
3063 #define PTYPE_HASH_SIZE (16)
3064 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3065
3066 #endif /* _LINUX_NETDEVICE_H */