Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[deliverable/linux.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
60 void netdev_set_default_ethtool_ops(struct net_device *dev,
61 const struct ethtool_ops *ops);
62
63 /* Backlog congestion levels */
64 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
65 #define NET_RX_DROP 1 /* packet dropped */
66
67 /*
68 * Transmit return codes: transmit return codes originate from three different
69 * namespaces:
70 *
71 * - qdisc return codes
72 * - driver transmit return codes
73 * - errno values
74 *
75 * Drivers are allowed to return any one of those in their hard_start_xmit()
76 * function. Real network devices commonly used with qdiscs should only return
77 * the driver transmit return codes though - when qdiscs are used, the actual
78 * transmission happens asynchronously, so the value is not propagated to
79 * higher layers. Virtual network devices transmit synchronously, in this case
80 * the driver transmit return codes are consumed by dev_queue_xmit(), all
81 * others are propagated to higher layers.
82 */
83
84 /* qdisc ->enqueue() return codes. */
85 #define NET_XMIT_SUCCESS 0x00
86 #define NET_XMIT_DROP 0x01 /* skb dropped */
87 #define NET_XMIT_CN 0x02 /* congestion notification */
88 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
89 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
90
91 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
92 * indicates that the device will soon be dropping packets, or already drops
93 * some packets of the same priority; prompting us to send less aggressively. */
94 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
95 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
96
97 /* Driver transmit return codes */
98 #define NETDEV_TX_MASK 0xf0
99
100 enum netdev_tx {
101 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
102 NETDEV_TX_OK = 0x00, /* driver took care of packet */
103 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
104 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
105 };
106 typedef enum netdev_tx netdev_tx_t;
107
108 /*
109 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
110 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
111 */
112 static inline bool dev_xmit_complete(int rc)
113 {
114 /*
115 * Positive cases with an skb consumed by a driver:
116 * - successful transmission (rc == NETDEV_TX_OK)
117 * - error while transmitting (rc < 0)
118 * - error while queueing to a different device (rc & NET_XMIT_MASK)
119 */
120 if (likely(rc < NET_XMIT_MASK))
121 return true;
122
123 return false;
124 }
125
126 /*
127 * Compute the worst case header length according to the protocols
128 * used.
129 */
130
131 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
132 # if defined(CONFIG_MAC80211_MESH)
133 # define LL_MAX_HEADER 128
134 # else
135 # define LL_MAX_HEADER 96
136 # endif
137 #else
138 # define LL_MAX_HEADER 32
139 #endif
140
141 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
142 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
143 #define MAX_HEADER LL_MAX_HEADER
144 #else
145 #define MAX_HEADER (LL_MAX_HEADER + 48)
146 #endif
147
148 /*
149 * Old network device statistics. Fields are native words
150 * (unsigned long) so they can be read and written atomically.
151 */
152
153 struct net_device_stats {
154 unsigned long rx_packets;
155 unsigned long tx_packets;
156 unsigned long rx_bytes;
157 unsigned long tx_bytes;
158 unsigned long rx_errors;
159 unsigned long tx_errors;
160 unsigned long rx_dropped;
161 unsigned long tx_dropped;
162 unsigned long multicast;
163 unsigned long collisions;
164 unsigned long rx_length_errors;
165 unsigned long rx_over_errors;
166 unsigned long rx_crc_errors;
167 unsigned long rx_frame_errors;
168 unsigned long rx_fifo_errors;
169 unsigned long rx_missed_errors;
170 unsigned long tx_aborted_errors;
171 unsigned long tx_carrier_errors;
172 unsigned long tx_fifo_errors;
173 unsigned long tx_heartbeat_errors;
174 unsigned long tx_window_errors;
175 unsigned long rx_compressed;
176 unsigned long tx_compressed;
177 };
178
179
180 #include <linux/cache.h>
181 #include <linux/skbuff.h>
182
183 #ifdef CONFIG_RPS
184 #include <linux/static_key.h>
185 extern struct static_key rps_needed;
186 #endif
187
188 struct neighbour;
189 struct neigh_parms;
190 struct sk_buff;
191
192 struct netdev_hw_addr {
193 struct list_head list;
194 unsigned char addr[MAX_ADDR_LEN];
195 unsigned char type;
196 #define NETDEV_HW_ADDR_T_LAN 1
197 #define NETDEV_HW_ADDR_T_SAN 2
198 #define NETDEV_HW_ADDR_T_SLAVE 3
199 #define NETDEV_HW_ADDR_T_UNICAST 4
200 #define NETDEV_HW_ADDR_T_MULTICAST 5
201 bool global_use;
202 int sync_cnt;
203 int refcount;
204 int synced;
205 struct rcu_head rcu_head;
206 };
207
208 struct netdev_hw_addr_list {
209 struct list_head list;
210 int count;
211 };
212
213 #define netdev_hw_addr_list_count(l) ((l)->count)
214 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
215 #define netdev_hw_addr_list_for_each(ha, l) \
216 list_for_each_entry(ha, &(l)->list, list)
217
218 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
219 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
220 #define netdev_for_each_uc_addr(ha, dev) \
221 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
222
223 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
224 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
225 #define netdev_for_each_mc_addr(ha, dev) \
226 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
227
228 struct hh_cache {
229 u16 hh_len;
230 u16 __pad;
231 seqlock_t hh_lock;
232
233 /* cached hardware header; allow for machine alignment needs. */
234 #define HH_DATA_MOD 16
235 #define HH_DATA_OFF(__len) \
236 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
237 #define HH_DATA_ALIGN(__len) \
238 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
239 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
240 };
241
242 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
243 * Alternative is:
244 * dev->hard_header_len ? (dev->hard_header_len +
245 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
246 *
247 * We could use other alignment values, but we must maintain the
248 * relationship HH alignment <= LL alignment.
249 */
250 #define LL_RESERVED_SPACE(dev) \
251 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
252 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
253 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
254
255 struct header_ops {
256 int (*create) (struct sk_buff *skb, struct net_device *dev,
257 unsigned short type, const void *daddr,
258 const void *saddr, unsigned int len);
259 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
260 int (*rebuild)(struct sk_buff *skb);
261 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
262 void (*cache_update)(struct hh_cache *hh,
263 const struct net_device *dev,
264 const unsigned char *haddr);
265 };
266
267 /* These flag bits are private to the generic network queueing
268 * layer, they may not be explicitly referenced by any other
269 * code.
270 */
271
272 enum netdev_state_t {
273 __LINK_STATE_START,
274 __LINK_STATE_PRESENT,
275 __LINK_STATE_NOCARRIER,
276 __LINK_STATE_LINKWATCH_PENDING,
277 __LINK_STATE_DORMANT,
278 };
279
280
281 /*
282 * This structure holds at boot time configured netdevice settings. They
283 * are then used in the device probing.
284 */
285 struct netdev_boot_setup {
286 char name[IFNAMSIZ];
287 struct ifmap map;
288 };
289 #define NETDEV_BOOT_SETUP_MAX 8
290
291 int __init netdev_boot_setup(char *str);
292
293 /*
294 * Structure for NAPI scheduling similar to tasklet but with weighting
295 */
296 struct napi_struct {
297 /* The poll_list must only be managed by the entity which
298 * changes the state of the NAPI_STATE_SCHED bit. This means
299 * whoever atomically sets that bit can add this napi_struct
300 * to the per-cpu poll_list, and whoever clears that bit
301 * can remove from the list right before clearing the bit.
302 */
303 struct list_head poll_list;
304
305 unsigned long state;
306 int weight;
307 unsigned int gro_count;
308 int (*poll)(struct napi_struct *, int);
309 #ifdef CONFIG_NETPOLL
310 spinlock_t poll_lock;
311 int poll_owner;
312 #endif
313 struct net_device *dev;
314 struct sk_buff *gro_list;
315 struct sk_buff *skb;
316 struct list_head dev_list;
317 struct hlist_node napi_hash_node;
318 unsigned int napi_id;
319 };
320
321 enum {
322 NAPI_STATE_SCHED, /* Poll is scheduled */
323 NAPI_STATE_DISABLE, /* Disable pending */
324 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
325 NAPI_STATE_HASHED, /* In NAPI hash */
326 };
327
328 enum gro_result {
329 GRO_MERGED,
330 GRO_MERGED_FREE,
331 GRO_HELD,
332 GRO_NORMAL,
333 GRO_DROP,
334 };
335 typedef enum gro_result gro_result_t;
336
337 /*
338 * enum rx_handler_result - Possible return values for rx_handlers.
339 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
340 * further.
341 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
342 * case skb->dev was changed by rx_handler.
343 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
344 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
345 *
346 * rx_handlers are functions called from inside __netif_receive_skb(), to do
347 * special processing of the skb, prior to delivery to protocol handlers.
348 *
349 * Currently, a net_device can only have a single rx_handler registered. Trying
350 * to register a second rx_handler will return -EBUSY.
351 *
352 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
353 * To unregister a rx_handler on a net_device, use
354 * netdev_rx_handler_unregister().
355 *
356 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
357 * do with the skb.
358 *
359 * If the rx_handler consumed to skb in some way, it should return
360 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
361 * the skb to be delivered in some other ways.
362 *
363 * If the rx_handler changed skb->dev, to divert the skb to another
364 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
365 * new device will be called if it exists.
366 *
367 * If the rx_handler consider the skb should be ignored, it should return
368 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
369 * are registered on exact device (ptype->dev == skb->dev).
370 *
371 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
372 * delivered, it should return RX_HANDLER_PASS.
373 *
374 * A device without a registered rx_handler will behave as if rx_handler
375 * returned RX_HANDLER_PASS.
376 */
377
378 enum rx_handler_result {
379 RX_HANDLER_CONSUMED,
380 RX_HANDLER_ANOTHER,
381 RX_HANDLER_EXACT,
382 RX_HANDLER_PASS,
383 };
384 typedef enum rx_handler_result rx_handler_result_t;
385 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
386
387 void __napi_schedule(struct napi_struct *n);
388
389 static inline bool napi_disable_pending(struct napi_struct *n)
390 {
391 return test_bit(NAPI_STATE_DISABLE, &n->state);
392 }
393
394 /**
395 * napi_schedule_prep - check if napi can be scheduled
396 * @n: napi context
397 *
398 * Test if NAPI routine is already running, and if not mark
399 * it as running. This is used as a condition variable
400 * insure only one NAPI poll instance runs. We also make
401 * sure there is no pending NAPI disable.
402 */
403 static inline bool napi_schedule_prep(struct napi_struct *n)
404 {
405 return !napi_disable_pending(n) &&
406 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
407 }
408
409 /**
410 * napi_schedule - schedule NAPI poll
411 * @n: napi context
412 *
413 * Schedule NAPI poll routine to be called if it is not already
414 * running.
415 */
416 static inline void napi_schedule(struct napi_struct *n)
417 {
418 if (napi_schedule_prep(n))
419 __napi_schedule(n);
420 }
421
422 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
423 static inline bool napi_reschedule(struct napi_struct *napi)
424 {
425 if (napi_schedule_prep(napi)) {
426 __napi_schedule(napi);
427 return true;
428 }
429 return false;
430 }
431
432 /**
433 * napi_complete - NAPI processing complete
434 * @n: napi context
435 *
436 * Mark NAPI processing as complete.
437 */
438 void __napi_complete(struct napi_struct *n);
439 void napi_complete(struct napi_struct *n);
440
441 /**
442 * napi_by_id - lookup a NAPI by napi_id
443 * @napi_id: hashed napi_id
444 *
445 * lookup @napi_id in napi_hash table
446 * must be called under rcu_read_lock()
447 */
448 struct napi_struct *napi_by_id(unsigned int napi_id);
449
450 /**
451 * napi_hash_add - add a NAPI to global hashtable
452 * @napi: napi context
453 *
454 * generate a new napi_id and store a @napi under it in napi_hash
455 */
456 void napi_hash_add(struct napi_struct *napi);
457
458 /**
459 * napi_hash_del - remove a NAPI from global table
460 * @napi: napi context
461 *
462 * Warning: caller must observe rcu grace period
463 * before freeing memory containing @napi
464 */
465 void napi_hash_del(struct napi_struct *napi);
466
467 /**
468 * napi_disable - prevent NAPI from scheduling
469 * @n: napi context
470 *
471 * Stop NAPI from being scheduled on this context.
472 * Waits till any outstanding processing completes.
473 */
474 static inline void napi_disable(struct napi_struct *n)
475 {
476 might_sleep();
477 set_bit(NAPI_STATE_DISABLE, &n->state);
478 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
479 msleep(1);
480 clear_bit(NAPI_STATE_DISABLE, &n->state);
481 }
482
483 /**
484 * napi_enable - enable NAPI scheduling
485 * @n: napi context
486 *
487 * Resume NAPI from being scheduled on this context.
488 * Must be paired with napi_disable.
489 */
490 static inline void napi_enable(struct napi_struct *n)
491 {
492 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
493 smp_mb__before_atomic();
494 clear_bit(NAPI_STATE_SCHED, &n->state);
495 }
496
497 #ifdef CONFIG_SMP
498 /**
499 * napi_synchronize - wait until NAPI is not running
500 * @n: napi context
501 *
502 * Wait until NAPI is done being scheduled on this context.
503 * Waits till any outstanding processing completes but
504 * does not disable future activations.
505 */
506 static inline void napi_synchronize(const struct napi_struct *n)
507 {
508 while (test_bit(NAPI_STATE_SCHED, &n->state))
509 msleep(1);
510 }
511 #else
512 # define napi_synchronize(n) barrier()
513 #endif
514
515 enum netdev_queue_state_t {
516 __QUEUE_STATE_DRV_XOFF,
517 __QUEUE_STATE_STACK_XOFF,
518 __QUEUE_STATE_FROZEN,
519 };
520
521 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
522 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
523 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
524
525 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
526 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
527 QUEUE_STATE_FROZEN)
528 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
529 QUEUE_STATE_FROZEN)
530
531 /*
532 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
533 * netif_tx_* functions below are used to manipulate this flag. The
534 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
535 * queue independently. The netif_xmit_*stopped functions below are called
536 * to check if the queue has been stopped by the driver or stack (either
537 * of the XOFF bits are set in the state). Drivers should not need to call
538 * netif_xmit*stopped functions, they should only be using netif_tx_*.
539 */
540
541 struct netdev_queue {
542 /*
543 * read mostly part
544 */
545 struct net_device *dev;
546 struct Qdisc *qdisc;
547 struct Qdisc *qdisc_sleeping;
548 #ifdef CONFIG_SYSFS
549 struct kobject kobj;
550 #endif
551 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
552 int numa_node;
553 #endif
554 /*
555 * write mostly part
556 */
557 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
558 int xmit_lock_owner;
559 /*
560 * please use this field instead of dev->trans_start
561 */
562 unsigned long trans_start;
563
564 /*
565 * Number of TX timeouts for this queue
566 * (/sys/class/net/DEV/Q/trans_timeout)
567 */
568 unsigned long trans_timeout;
569
570 unsigned long state;
571
572 #ifdef CONFIG_BQL
573 struct dql dql;
574 #endif
575 } ____cacheline_aligned_in_smp;
576
577 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
578 {
579 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
580 return q->numa_node;
581 #else
582 return NUMA_NO_NODE;
583 #endif
584 }
585
586 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
587 {
588 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 q->numa_node = node;
590 #endif
591 }
592
593 #ifdef CONFIG_RPS
594 /*
595 * This structure holds an RPS map which can be of variable length. The
596 * map is an array of CPUs.
597 */
598 struct rps_map {
599 unsigned int len;
600 struct rcu_head rcu;
601 u16 cpus[0];
602 };
603 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
604
605 /*
606 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
607 * tail pointer for that CPU's input queue at the time of last enqueue, and
608 * a hardware filter index.
609 */
610 struct rps_dev_flow {
611 u16 cpu;
612 u16 filter;
613 unsigned int last_qtail;
614 };
615 #define RPS_NO_FILTER 0xffff
616
617 /*
618 * The rps_dev_flow_table structure contains a table of flow mappings.
619 */
620 struct rps_dev_flow_table {
621 unsigned int mask;
622 struct rcu_head rcu;
623 struct rps_dev_flow flows[0];
624 };
625 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
626 ((_num) * sizeof(struct rps_dev_flow)))
627
628 /*
629 * The rps_sock_flow_table contains mappings of flows to the last CPU
630 * on which they were processed by the application (set in recvmsg).
631 */
632 struct rps_sock_flow_table {
633 unsigned int mask;
634 u16 ents[0];
635 };
636 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
637 ((_num) * sizeof(u16)))
638
639 #define RPS_NO_CPU 0xffff
640
641 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
642 u32 hash)
643 {
644 if (table && hash) {
645 unsigned int cpu, index = hash & table->mask;
646
647 /* We only give a hint, preemption can change cpu under us */
648 cpu = raw_smp_processor_id();
649
650 if (table->ents[index] != cpu)
651 table->ents[index] = cpu;
652 }
653 }
654
655 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
656 u32 hash)
657 {
658 if (table && hash)
659 table->ents[hash & table->mask] = RPS_NO_CPU;
660 }
661
662 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
663
664 #ifdef CONFIG_RFS_ACCEL
665 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
666 u16 filter_id);
667 #endif
668 #endif /* CONFIG_RPS */
669
670 /* This structure contains an instance of an RX queue. */
671 struct netdev_rx_queue {
672 #ifdef CONFIG_RPS
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
675 #endif
676 struct kobject kobj;
677 struct net_device *dev;
678 } ____cacheline_aligned_in_smp;
679
680 /*
681 * RX queue sysfs structures and functions.
682 */
683 struct rx_queue_attribute {
684 struct attribute attr;
685 ssize_t (*show)(struct netdev_rx_queue *queue,
686 struct rx_queue_attribute *attr, char *buf);
687 ssize_t (*store)(struct netdev_rx_queue *queue,
688 struct rx_queue_attribute *attr, const char *buf, size_t len);
689 };
690
691 #ifdef CONFIG_XPS
692 /*
693 * This structure holds an XPS map which can be of variable length. The
694 * map is an array of queues.
695 */
696 struct xps_map {
697 unsigned int len;
698 unsigned int alloc_len;
699 struct rcu_head rcu;
700 u16 queues[0];
701 };
702 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
703 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
704 / sizeof(u16))
705
706 /*
707 * This structure holds all XPS maps for device. Maps are indexed by CPU.
708 */
709 struct xps_dev_maps {
710 struct rcu_head rcu;
711 struct xps_map __rcu *cpu_map[0];
712 };
713 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
714 (nr_cpu_ids * sizeof(struct xps_map *)))
715 #endif /* CONFIG_XPS */
716
717 #define TC_MAX_QUEUE 16
718 #define TC_BITMASK 15
719 /* HW offloaded queuing disciplines txq count and offset maps */
720 struct netdev_tc_txq {
721 u16 count;
722 u16 offset;
723 };
724
725 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
726 /*
727 * This structure is to hold information about the device
728 * configured to run FCoE protocol stack.
729 */
730 struct netdev_fcoe_hbainfo {
731 char manufacturer[64];
732 char serial_number[64];
733 char hardware_version[64];
734 char driver_version[64];
735 char optionrom_version[64];
736 char firmware_version[64];
737 char model[256];
738 char model_description[256];
739 };
740 #endif
741
742 #define MAX_PHYS_PORT_ID_LEN 32
743
744 /* This structure holds a unique identifier to identify the
745 * physical port used by a netdevice.
746 */
747 struct netdev_phys_port_id {
748 unsigned char id[MAX_PHYS_PORT_ID_LEN];
749 unsigned char id_len;
750 };
751
752 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
753 struct sk_buff *skb);
754
755 /*
756 * This structure defines the management hooks for network devices.
757 * The following hooks can be defined; unless noted otherwise, they are
758 * optional and can be filled with a null pointer.
759 *
760 * int (*ndo_init)(struct net_device *dev);
761 * This function is called once when network device is registered.
762 * The network device can use this to any late stage initializaton
763 * or semantic validattion. It can fail with an error code which will
764 * be propogated back to register_netdev
765 *
766 * void (*ndo_uninit)(struct net_device *dev);
767 * This function is called when device is unregistered or when registration
768 * fails. It is not called if init fails.
769 *
770 * int (*ndo_open)(struct net_device *dev);
771 * This function is called when network device transistions to the up
772 * state.
773 *
774 * int (*ndo_stop)(struct net_device *dev);
775 * This function is called when network device transistions to the down
776 * state.
777 *
778 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
779 * struct net_device *dev);
780 * Called when a packet needs to be transmitted.
781 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
782 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
783 * Required can not be NULL.
784 *
785 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
786 * void *accel_priv, select_queue_fallback_t fallback);
787 * Called to decide which queue to when device supports multiple
788 * transmit queues.
789 *
790 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
791 * This function is called to allow device receiver to make
792 * changes to configuration when multicast or promiscious is enabled.
793 *
794 * void (*ndo_set_rx_mode)(struct net_device *dev);
795 * This function is called device changes address list filtering.
796 * If driver handles unicast address filtering, it should set
797 * IFF_UNICAST_FLT to its priv_flags.
798 *
799 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
800 * This function is called when the Media Access Control address
801 * needs to be changed. If this interface is not defined, the
802 * mac address can not be changed.
803 *
804 * int (*ndo_validate_addr)(struct net_device *dev);
805 * Test if Media Access Control address is valid for the device.
806 *
807 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
808 * Called when a user request an ioctl which can't be handled by
809 * the generic interface code. If not defined ioctl's return
810 * not supported error code.
811 *
812 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
813 * Used to set network devices bus interface parameters. This interface
814 * is retained for legacy reason, new devices should use the bus
815 * interface (PCI) for low level management.
816 *
817 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
818 * Called when a user wants to change the Maximum Transfer Unit
819 * of a device. If not defined, any request to change MTU will
820 * will return an error.
821 *
822 * void (*ndo_tx_timeout)(struct net_device *dev);
823 * Callback uses when the transmitter has not made any progress
824 * for dev->watchdog ticks.
825 *
826 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
827 * struct rtnl_link_stats64 *storage);
828 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
829 * Called when a user wants to get the network device usage
830 * statistics. Drivers must do one of the following:
831 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
832 * rtnl_link_stats64 structure passed by the caller.
833 * 2. Define @ndo_get_stats to update a net_device_stats structure
834 * (which should normally be dev->stats) and return a pointer to
835 * it. The structure may be changed asynchronously only if each
836 * field is written atomically.
837 * 3. Update dev->stats asynchronously and atomically, and define
838 * neither operation.
839 *
840 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
841 * If device support VLAN filtering this function is called when a
842 * VLAN id is registered.
843 *
844 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
845 * If device support VLAN filtering this function is called when a
846 * VLAN id is unregistered.
847 *
848 * void (*ndo_poll_controller)(struct net_device *dev);
849 *
850 * SR-IOV management functions.
851 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
852 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
853 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
854 * int max_tx_rate);
855 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
856 * int (*ndo_get_vf_config)(struct net_device *dev,
857 * int vf, struct ifla_vf_info *ivf);
858 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
859 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
860 * struct nlattr *port[]);
861 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
862 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
863 * Called to setup 'tc' number of traffic classes in the net device. This
864 * is always called from the stack with the rtnl lock held and netif tx
865 * queues stopped. This allows the netdevice to perform queue management
866 * safely.
867 *
868 * Fiber Channel over Ethernet (FCoE) offload functions.
869 * int (*ndo_fcoe_enable)(struct net_device *dev);
870 * Called when the FCoE protocol stack wants to start using LLD for FCoE
871 * so the underlying device can perform whatever needed configuration or
872 * initialization to support acceleration of FCoE traffic.
873 *
874 * int (*ndo_fcoe_disable)(struct net_device *dev);
875 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
876 * so the underlying device can perform whatever needed clean-ups to
877 * stop supporting acceleration of FCoE traffic.
878 *
879 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
880 * struct scatterlist *sgl, unsigned int sgc);
881 * Called when the FCoE Initiator wants to initialize an I/O that
882 * is a possible candidate for Direct Data Placement (DDP). The LLD can
883 * perform necessary setup and returns 1 to indicate the device is set up
884 * successfully to perform DDP on this I/O, otherwise this returns 0.
885 *
886 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
887 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
888 * indicated by the FC exchange id 'xid', so the underlying device can
889 * clean up and reuse resources for later DDP requests.
890 *
891 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
892 * struct scatterlist *sgl, unsigned int sgc);
893 * Called when the FCoE Target wants to initialize an I/O that
894 * is a possible candidate for Direct Data Placement (DDP). The LLD can
895 * perform necessary setup and returns 1 to indicate the device is set up
896 * successfully to perform DDP on this I/O, otherwise this returns 0.
897 *
898 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
899 * struct netdev_fcoe_hbainfo *hbainfo);
900 * Called when the FCoE Protocol stack wants information on the underlying
901 * device. This information is utilized by the FCoE protocol stack to
902 * register attributes with Fiber Channel management service as per the
903 * FC-GS Fabric Device Management Information(FDMI) specification.
904 *
905 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
906 * Called when the underlying device wants to override default World Wide
907 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
908 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
909 * protocol stack to use.
910 *
911 * RFS acceleration.
912 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
913 * u16 rxq_index, u32 flow_id);
914 * Set hardware filter for RFS. rxq_index is the target queue index;
915 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
916 * Return the filter ID on success, or a negative error code.
917 *
918 * Slave management functions (for bridge, bonding, etc).
919 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
920 * Called to make another netdev an underling.
921 *
922 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
923 * Called to release previously enslaved netdev.
924 *
925 * Feature/offload setting functions.
926 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
927 * netdev_features_t features);
928 * Adjusts the requested feature flags according to device-specific
929 * constraints, and returns the resulting flags. Must not modify
930 * the device state.
931 *
932 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
933 * Called to update device configuration to new features. Passed
934 * feature set might be less than what was returned by ndo_fix_features()).
935 * Must return >0 or -errno if it changed dev->features itself.
936 *
937 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
938 * struct net_device *dev,
939 * const unsigned char *addr, u16 flags)
940 * Adds an FDB entry to dev for addr.
941 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
942 * struct net_device *dev,
943 * const unsigned char *addr)
944 * Deletes the FDB entry from dev coresponding to addr.
945 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
946 * struct net_device *dev, int idx)
947 * Used to add FDB entries to dump requests. Implementers should add
948 * entries to skb and update idx with the number of entries.
949 *
950 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
951 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
952 * struct net_device *dev, u32 filter_mask)
953 *
954 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
955 * Called to change device carrier. Soft-devices (like dummy, team, etc)
956 * which do not represent real hardware may define this to allow their
957 * userspace components to manage their virtual carrier state. Devices
958 * that determine carrier state from physical hardware properties (eg
959 * network cables) or protocol-dependent mechanisms (eg
960 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
961 *
962 * int (*ndo_get_phys_port_id)(struct net_device *dev,
963 * struct netdev_phys_port_id *ppid);
964 * Called to get ID of physical port of this device. If driver does
965 * not implement this, it is assumed that the hw is not able to have
966 * multiple net devices on single physical port.
967 *
968 * void (*ndo_add_vxlan_port)(struct net_device *dev,
969 * sa_family_t sa_family, __be16 port);
970 * Called by vxlan to notiy a driver about the UDP port and socket
971 * address family that vxlan is listnening to. It is called only when
972 * a new port starts listening. The operation is protected by the
973 * vxlan_net->sock_lock.
974 *
975 * void (*ndo_del_vxlan_port)(struct net_device *dev,
976 * sa_family_t sa_family, __be16 port);
977 * Called by vxlan to notify the driver about a UDP port and socket
978 * address family that vxlan is not listening to anymore. The operation
979 * is protected by the vxlan_net->sock_lock.
980 *
981 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
982 * struct net_device *dev)
983 * Called by upper layer devices to accelerate switching or other
984 * station functionality into hardware. 'pdev is the lowerdev
985 * to use for the offload and 'dev' is the net device that will
986 * back the offload. Returns a pointer to the private structure
987 * the upper layer will maintain.
988 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
989 * Called by upper layer device to delete the station created
990 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
991 * the station and priv is the structure returned by the add
992 * operation.
993 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
994 * struct net_device *dev,
995 * void *priv);
996 * Callback to use for xmit over the accelerated station. This
997 * is used in place of ndo_start_xmit on accelerated net
998 * devices.
999 */
1000 struct net_device_ops {
1001 int (*ndo_init)(struct net_device *dev);
1002 void (*ndo_uninit)(struct net_device *dev);
1003 int (*ndo_open)(struct net_device *dev);
1004 int (*ndo_stop)(struct net_device *dev);
1005 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1006 struct net_device *dev);
1007 u16 (*ndo_select_queue)(struct net_device *dev,
1008 struct sk_buff *skb,
1009 void *accel_priv,
1010 select_queue_fallback_t fallback);
1011 void (*ndo_change_rx_flags)(struct net_device *dev,
1012 int flags);
1013 void (*ndo_set_rx_mode)(struct net_device *dev);
1014 int (*ndo_set_mac_address)(struct net_device *dev,
1015 void *addr);
1016 int (*ndo_validate_addr)(struct net_device *dev);
1017 int (*ndo_do_ioctl)(struct net_device *dev,
1018 struct ifreq *ifr, int cmd);
1019 int (*ndo_set_config)(struct net_device *dev,
1020 struct ifmap *map);
1021 int (*ndo_change_mtu)(struct net_device *dev,
1022 int new_mtu);
1023 int (*ndo_neigh_setup)(struct net_device *dev,
1024 struct neigh_parms *);
1025 void (*ndo_tx_timeout) (struct net_device *dev);
1026
1027 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1028 struct rtnl_link_stats64 *storage);
1029 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1030
1031 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1032 __be16 proto, u16 vid);
1033 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1034 __be16 proto, u16 vid);
1035 #ifdef CONFIG_NET_POLL_CONTROLLER
1036 void (*ndo_poll_controller)(struct net_device *dev);
1037 int (*ndo_netpoll_setup)(struct net_device *dev,
1038 struct netpoll_info *info);
1039 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1040 #endif
1041 #ifdef CONFIG_NET_RX_BUSY_POLL
1042 int (*ndo_busy_poll)(struct napi_struct *dev);
1043 #endif
1044 int (*ndo_set_vf_mac)(struct net_device *dev,
1045 int queue, u8 *mac);
1046 int (*ndo_set_vf_vlan)(struct net_device *dev,
1047 int queue, u16 vlan, u8 qos);
1048 int (*ndo_set_vf_rate)(struct net_device *dev,
1049 int vf, int min_tx_rate,
1050 int max_tx_rate);
1051 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1052 int vf, bool setting);
1053 int (*ndo_get_vf_config)(struct net_device *dev,
1054 int vf,
1055 struct ifla_vf_info *ivf);
1056 int (*ndo_set_vf_link_state)(struct net_device *dev,
1057 int vf, int link_state);
1058 int (*ndo_set_vf_port)(struct net_device *dev,
1059 int vf,
1060 struct nlattr *port[]);
1061 int (*ndo_get_vf_port)(struct net_device *dev,
1062 int vf, struct sk_buff *skb);
1063 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1064 #if IS_ENABLED(CONFIG_FCOE)
1065 int (*ndo_fcoe_enable)(struct net_device *dev);
1066 int (*ndo_fcoe_disable)(struct net_device *dev);
1067 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1068 u16 xid,
1069 struct scatterlist *sgl,
1070 unsigned int sgc);
1071 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1072 u16 xid);
1073 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1074 u16 xid,
1075 struct scatterlist *sgl,
1076 unsigned int sgc);
1077 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1078 struct netdev_fcoe_hbainfo *hbainfo);
1079 #endif
1080
1081 #if IS_ENABLED(CONFIG_LIBFCOE)
1082 #define NETDEV_FCOE_WWNN 0
1083 #define NETDEV_FCOE_WWPN 1
1084 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1085 u64 *wwn, int type);
1086 #endif
1087
1088 #ifdef CONFIG_RFS_ACCEL
1089 int (*ndo_rx_flow_steer)(struct net_device *dev,
1090 const struct sk_buff *skb,
1091 u16 rxq_index,
1092 u32 flow_id);
1093 #endif
1094 int (*ndo_add_slave)(struct net_device *dev,
1095 struct net_device *slave_dev);
1096 int (*ndo_del_slave)(struct net_device *dev,
1097 struct net_device *slave_dev);
1098 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1099 netdev_features_t features);
1100 int (*ndo_set_features)(struct net_device *dev,
1101 netdev_features_t features);
1102 int (*ndo_neigh_construct)(struct neighbour *n);
1103 void (*ndo_neigh_destroy)(struct neighbour *n);
1104
1105 int (*ndo_fdb_add)(struct ndmsg *ndm,
1106 struct nlattr *tb[],
1107 struct net_device *dev,
1108 const unsigned char *addr,
1109 u16 flags);
1110 int (*ndo_fdb_del)(struct ndmsg *ndm,
1111 struct nlattr *tb[],
1112 struct net_device *dev,
1113 const unsigned char *addr);
1114 int (*ndo_fdb_dump)(struct sk_buff *skb,
1115 struct netlink_callback *cb,
1116 struct net_device *dev,
1117 int idx);
1118
1119 int (*ndo_bridge_setlink)(struct net_device *dev,
1120 struct nlmsghdr *nlh);
1121 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1122 u32 pid, u32 seq,
1123 struct net_device *dev,
1124 u32 filter_mask);
1125 int (*ndo_bridge_dellink)(struct net_device *dev,
1126 struct nlmsghdr *nlh);
1127 int (*ndo_change_carrier)(struct net_device *dev,
1128 bool new_carrier);
1129 int (*ndo_get_phys_port_id)(struct net_device *dev,
1130 struct netdev_phys_port_id *ppid);
1131 void (*ndo_add_vxlan_port)(struct net_device *dev,
1132 sa_family_t sa_family,
1133 __be16 port);
1134 void (*ndo_del_vxlan_port)(struct net_device *dev,
1135 sa_family_t sa_family,
1136 __be16 port);
1137
1138 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1139 struct net_device *dev);
1140 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1141 void *priv);
1142
1143 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1144 struct net_device *dev,
1145 void *priv);
1146 int (*ndo_get_lock_subclass)(struct net_device *dev);
1147 };
1148
1149 /**
1150 * enum net_device_priv_flags - &struct net_device priv_flags
1151 *
1152 * These are the &struct net_device, they are only set internally
1153 * by drivers and used in the kernel. These flags are invisible to
1154 * userspace, this means that the order of these flags can change
1155 * during any kernel release.
1156 *
1157 * You should have a pretty good reason to be extending these flags.
1158 *
1159 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1160 * @IFF_EBRIDGE: Ethernet bridging device
1161 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1162 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1163 * @IFF_MASTER_ALB: bonding master, balance-alb
1164 * @IFF_BONDING: bonding master or slave
1165 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1166 * @IFF_ISATAP: ISATAP interface (RFC4214)
1167 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1168 * @IFF_WAN_HDLC: WAN HDLC device
1169 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1170 * release skb->dst
1171 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1172 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1173 * @IFF_MACVLAN_PORT: device used as macvlan port
1174 * @IFF_BRIDGE_PORT: device used as bridge port
1175 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1176 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1177 * @IFF_UNICAST_FLT: Supports unicast filtering
1178 * @IFF_TEAM_PORT: device used as team port
1179 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1180 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1181 * change when it's running
1182 * @IFF_MACVLAN: Macvlan device
1183 */
1184 enum netdev_priv_flags {
1185 IFF_802_1Q_VLAN = 1<<0,
1186 IFF_EBRIDGE = 1<<1,
1187 IFF_SLAVE_INACTIVE = 1<<2,
1188 IFF_MASTER_8023AD = 1<<3,
1189 IFF_MASTER_ALB = 1<<4,
1190 IFF_BONDING = 1<<5,
1191 IFF_SLAVE_NEEDARP = 1<<6,
1192 IFF_ISATAP = 1<<7,
1193 IFF_MASTER_ARPMON = 1<<8,
1194 IFF_WAN_HDLC = 1<<9,
1195 IFF_XMIT_DST_RELEASE = 1<<10,
1196 IFF_DONT_BRIDGE = 1<<11,
1197 IFF_DISABLE_NETPOLL = 1<<12,
1198 IFF_MACVLAN_PORT = 1<<13,
1199 IFF_BRIDGE_PORT = 1<<14,
1200 IFF_OVS_DATAPATH = 1<<15,
1201 IFF_TX_SKB_SHARING = 1<<16,
1202 IFF_UNICAST_FLT = 1<<17,
1203 IFF_TEAM_PORT = 1<<18,
1204 IFF_SUPP_NOFCS = 1<<19,
1205 IFF_LIVE_ADDR_CHANGE = 1<<20,
1206 IFF_MACVLAN = 1<<21,
1207 };
1208
1209 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1210 #define IFF_EBRIDGE IFF_EBRIDGE
1211 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1212 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1213 #define IFF_MASTER_ALB IFF_MASTER_ALB
1214 #define IFF_BONDING IFF_BONDING
1215 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1216 #define IFF_ISATAP IFF_ISATAP
1217 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1218 #define IFF_WAN_HDLC IFF_WAN_HDLC
1219 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1220 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1221 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1222 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1223 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1224 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1225 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1226 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1227 #define IFF_TEAM_PORT IFF_TEAM_PORT
1228 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1229 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1230 #define IFF_MACVLAN IFF_MACVLAN
1231
1232 /*
1233 * The DEVICE structure.
1234 * Actually, this whole structure is a big mistake. It mixes I/O
1235 * data with strictly "high-level" data, and it has to know about
1236 * almost every data structure used in the INET module.
1237 *
1238 * FIXME: cleanup struct net_device such that network protocol info
1239 * moves out.
1240 */
1241
1242 struct net_device {
1243
1244 /*
1245 * This is the first field of the "visible" part of this structure
1246 * (i.e. as seen by users in the "Space.c" file). It is the name
1247 * of the interface.
1248 */
1249 char name[IFNAMSIZ];
1250
1251 /* device name hash chain, please keep it close to name[] */
1252 struct hlist_node name_hlist;
1253
1254 /* snmp alias */
1255 char *ifalias;
1256
1257 /*
1258 * I/O specific fields
1259 * FIXME: Merge these and struct ifmap into one
1260 */
1261 unsigned long mem_end; /* shared mem end */
1262 unsigned long mem_start; /* shared mem start */
1263 unsigned long base_addr; /* device I/O address */
1264 int irq; /* device IRQ number */
1265
1266 /*
1267 * Some hardware also needs these fields, but they are not
1268 * part of the usual set specified in Space.c.
1269 */
1270
1271 unsigned long state;
1272
1273 struct list_head dev_list;
1274 struct list_head napi_list;
1275 struct list_head unreg_list;
1276 struct list_head close_list;
1277
1278 /* directly linked devices, like slaves for bonding */
1279 struct {
1280 struct list_head upper;
1281 struct list_head lower;
1282 } adj_list;
1283
1284 /* all linked devices, *including* neighbours */
1285 struct {
1286 struct list_head upper;
1287 struct list_head lower;
1288 } all_adj_list;
1289
1290
1291 /* currently active device features */
1292 netdev_features_t features;
1293 /* user-changeable features */
1294 netdev_features_t hw_features;
1295 /* user-requested features */
1296 netdev_features_t wanted_features;
1297 /* mask of features inheritable by VLAN devices */
1298 netdev_features_t vlan_features;
1299 /* mask of features inherited by encapsulating devices
1300 * This field indicates what encapsulation offloads
1301 * the hardware is capable of doing, and drivers will
1302 * need to set them appropriately.
1303 */
1304 netdev_features_t hw_enc_features;
1305 /* mask of fetures inheritable by MPLS */
1306 netdev_features_t mpls_features;
1307
1308 /* Interface index. Unique device identifier */
1309 int ifindex;
1310 int iflink;
1311
1312 struct net_device_stats stats;
1313
1314 /* dropped packets by core network, Do not use this in drivers */
1315 atomic_long_t rx_dropped;
1316 atomic_long_t tx_dropped;
1317
1318 /* Stats to monitor carrier on<->off transitions */
1319 atomic_t carrier_changes;
1320
1321 #ifdef CONFIG_WIRELESS_EXT
1322 /* List of functions to handle Wireless Extensions (instead of ioctl).
1323 * See <net/iw_handler.h> for details. Jean II */
1324 const struct iw_handler_def * wireless_handlers;
1325 /* Instance data managed by the core of Wireless Extensions. */
1326 struct iw_public_data * wireless_data;
1327 #endif
1328 /* Management operations */
1329 const struct net_device_ops *netdev_ops;
1330 const struct ethtool_ops *ethtool_ops;
1331 const struct forwarding_accel_ops *fwd_ops;
1332
1333 /* Hardware header description */
1334 const struct header_ops *header_ops;
1335
1336 unsigned int flags; /* interface flags (a la BSD) */
1337 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1338 * See if.h for definitions. */
1339 unsigned short gflags;
1340 unsigned short padded; /* How much padding added by alloc_netdev() */
1341
1342 unsigned char operstate; /* RFC2863 operstate */
1343 unsigned char link_mode; /* mapping policy to operstate */
1344
1345 unsigned char if_port; /* Selectable AUI, TP,..*/
1346 unsigned char dma; /* DMA channel */
1347
1348 unsigned int mtu; /* interface MTU value */
1349 unsigned short type; /* interface hardware type */
1350 unsigned short hard_header_len; /* hardware hdr length */
1351
1352 /* extra head- and tailroom the hardware may need, but not in all cases
1353 * can this be guaranteed, especially tailroom. Some cases also use
1354 * LL_MAX_HEADER instead to allocate the skb.
1355 */
1356 unsigned short needed_headroom;
1357 unsigned short needed_tailroom;
1358
1359 /* Interface address info. */
1360 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1361 unsigned char addr_assign_type; /* hw address assignment type */
1362 unsigned char addr_len; /* hardware address length */
1363 unsigned short neigh_priv_len;
1364 unsigned short dev_id; /* Used to differentiate devices
1365 * that share the same link
1366 * layer address
1367 */
1368 unsigned short dev_port; /* Used to differentiate
1369 * devices that share the same
1370 * function
1371 */
1372 spinlock_t addr_list_lock;
1373 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1374 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1375 struct netdev_hw_addr_list dev_addrs; /* list of device
1376 * hw addresses
1377 */
1378 #ifdef CONFIG_SYSFS
1379 struct kset *queues_kset;
1380 #endif
1381
1382 bool uc_promisc;
1383 unsigned int promiscuity;
1384 unsigned int allmulti;
1385
1386
1387 /* Protocol specific pointers */
1388
1389 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1390 struct vlan_info __rcu *vlan_info; /* VLAN info */
1391 #endif
1392 #if IS_ENABLED(CONFIG_NET_DSA)
1393 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1394 #endif
1395 #if IS_ENABLED(CONFIG_TIPC)
1396 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1397 #endif
1398 void *atalk_ptr; /* AppleTalk link */
1399 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1400 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1401 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1402 void *ax25_ptr; /* AX.25 specific data */
1403 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1404 assign before registering */
1405
1406 /*
1407 * Cache lines mostly used on receive path (including eth_type_trans())
1408 */
1409 unsigned long last_rx; /* Time of last Rx */
1410
1411 /* Interface address info used in eth_type_trans() */
1412 unsigned char *dev_addr; /* hw address, (before bcast
1413 because most packets are
1414 unicast) */
1415
1416
1417 #ifdef CONFIG_SYSFS
1418 struct netdev_rx_queue *_rx;
1419
1420 /* Number of RX queues allocated at register_netdev() time */
1421 unsigned int num_rx_queues;
1422
1423 /* Number of RX queues currently active in device */
1424 unsigned int real_num_rx_queues;
1425
1426 #endif
1427
1428 rx_handler_func_t __rcu *rx_handler;
1429 void __rcu *rx_handler_data;
1430
1431 struct netdev_queue __rcu *ingress_queue;
1432 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1433
1434
1435 /*
1436 * Cache lines mostly used on transmit path
1437 */
1438 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1439
1440 /* Number of TX queues allocated at alloc_netdev_mq() time */
1441 unsigned int num_tx_queues;
1442
1443 /* Number of TX queues currently active in device */
1444 unsigned int real_num_tx_queues;
1445
1446 /* root qdisc from userspace point of view */
1447 struct Qdisc *qdisc;
1448
1449 unsigned long tx_queue_len; /* Max frames per queue allowed */
1450 spinlock_t tx_global_lock;
1451
1452 #ifdef CONFIG_XPS
1453 struct xps_dev_maps __rcu *xps_maps;
1454 #endif
1455 #ifdef CONFIG_RFS_ACCEL
1456 /* CPU reverse-mapping for RX completion interrupts, indexed
1457 * by RX queue number. Assigned by driver. This must only be
1458 * set if the ndo_rx_flow_steer operation is defined. */
1459 struct cpu_rmap *rx_cpu_rmap;
1460 #endif
1461
1462 /* These may be needed for future network-power-down code. */
1463
1464 /*
1465 * trans_start here is expensive for high speed devices on SMP,
1466 * please use netdev_queue->trans_start instead.
1467 */
1468 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1469
1470 int watchdog_timeo; /* used by dev_watchdog() */
1471 struct timer_list watchdog_timer;
1472
1473 /* Number of references to this device */
1474 int __percpu *pcpu_refcnt;
1475
1476 /* delayed register/unregister */
1477 struct list_head todo_list;
1478 /* device index hash chain */
1479 struct hlist_node index_hlist;
1480
1481 struct list_head link_watch_list;
1482
1483 /* register/unregister state machine */
1484 enum { NETREG_UNINITIALIZED=0,
1485 NETREG_REGISTERED, /* completed register_netdevice */
1486 NETREG_UNREGISTERING, /* called unregister_netdevice */
1487 NETREG_UNREGISTERED, /* completed unregister todo */
1488 NETREG_RELEASED, /* called free_netdev */
1489 NETREG_DUMMY, /* dummy device for NAPI poll */
1490 } reg_state:8;
1491
1492 bool dismantle; /* device is going do be freed */
1493
1494 enum {
1495 RTNL_LINK_INITIALIZED,
1496 RTNL_LINK_INITIALIZING,
1497 } rtnl_link_state:16;
1498
1499 /* Called from unregister, can be used to call free_netdev */
1500 void (*destructor)(struct net_device *dev);
1501
1502 #ifdef CONFIG_NETPOLL
1503 struct netpoll_info __rcu *npinfo;
1504 #endif
1505
1506 #ifdef CONFIG_NET_NS
1507 /* Network namespace this network device is inside */
1508 struct net *nd_net;
1509 #endif
1510
1511 /* mid-layer private */
1512 union {
1513 void *ml_priv;
1514 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1515 struct pcpu_sw_netstats __percpu *tstats;
1516 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1517 struct pcpu_vstats __percpu *vstats; /* veth stats */
1518 };
1519 /* GARP */
1520 struct garp_port __rcu *garp_port;
1521 /* MRP */
1522 struct mrp_port __rcu *mrp_port;
1523
1524 /* class/net/name entry */
1525 struct device dev;
1526 /* space for optional device, statistics, and wireless sysfs groups */
1527 const struct attribute_group *sysfs_groups[4];
1528 /* space for optional per-rx queue attributes */
1529 const struct attribute_group *sysfs_rx_queue_group;
1530
1531 /* rtnetlink link ops */
1532 const struct rtnl_link_ops *rtnl_link_ops;
1533
1534 /* for setting kernel sock attribute on TCP connection setup */
1535 #define GSO_MAX_SIZE 65536
1536 unsigned int gso_max_size;
1537 #define GSO_MAX_SEGS 65535
1538 u16 gso_max_segs;
1539
1540 #ifdef CONFIG_DCB
1541 /* Data Center Bridging netlink ops */
1542 const struct dcbnl_rtnl_ops *dcbnl_ops;
1543 #endif
1544 u8 num_tc;
1545 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1546 u8 prio_tc_map[TC_BITMASK + 1];
1547
1548 #if IS_ENABLED(CONFIG_FCOE)
1549 /* max exchange id for FCoE LRO by ddp */
1550 unsigned int fcoe_ddp_xid;
1551 #endif
1552 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1553 struct netprio_map __rcu *priomap;
1554 #endif
1555 /* phy device may attach itself for hardware timestamping */
1556 struct phy_device *phydev;
1557
1558 struct lock_class_key *qdisc_tx_busylock;
1559
1560 /* group the device belongs to */
1561 int group;
1562
1563 struct pm_qos_request pm_qos_req;
1564 };
1565 #define to_net_dev(d) container_of(d, struct net_device, dev)
1566
1567 #define NETDEV_ALIGN 32
1568
1569 static inline
1570 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1571 {
1572 return dev->prio_tc_map[prio & TC_BITMASK];
1573 }
1574
1575 static inline
1576 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1577 {
1578 if (tc >= dev->num_tc)
1579 return -EINVAL;
1580
1581 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1582 return 0;
1583 }
1584
1585 static inline
1586 void netdev_reset_tc(struct net_device *dev)
1587 {
1588 dev->num_tc = 0;
1589 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1590 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1591 }
1592
1593 static inline
1594 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1595 {
1596 if (tc >= dev->num_tc)
1597 return -EINVAL;
1598
1599 dev->tc_to_txq[tc].count = count;
1600 dev->tc_to_txq[tc].offset = offset;
1601 return 0;
1602 }
1603
1604 static inline
1605 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1606 {
1607 if (num_tc > TC_MAX_QUEUE)
1608 return -EINVAL;
1609
1610 dev->num_tc = num_tc;
1611 return 0;
1612 }
1613
1614 static inline
1615 int netdev_get_num_tc(struct net_device *dev)
1616 {
1617 return dev->num_tc;
1618 }
1619
1620 static inline
1621 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1622 unsigned int index)
1623 {
1624 return &dev->_tx[index];
1625 }
1626
1627 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1628 void (*f)(struct net_device *,
1629 struct netdev_queue *,
1630 void *),
1631 void *arg)
1632 {
1633 unsigned int i;
1634
1635 for (i = 0; i < dev->num_tx_queues; i++)
1636 f(dev, &dev->_tx[i], arg);
1637 }
1638
1639 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1640 struct sk_buff *skb,
1641 void *accel_priv);
1642
1643 /*
1644 * Net namespace inlines
1645 */
1646 static inline
1647 struct net *dev_net(const struct net_device *dev)
1648 {
1649 return read_pnet(&dev->nd_net);
1650 }
1651
1652 static inline
1653 void dev_net_set(struct net_device *dev, struct net *net)
1654 {
1655 #ifdef CONFIG_NET_NS
1656 release_net(dev->nd_net);
1657 dev->nd_net = hold_net(net);
1658 #endif
1659 }
1660
1661 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1662 {
1663 #ifdef CONFIG_NET_DSA_TAG_DSA
1664 if (dev->dsa_ptr != NULL)
1665 return dsa_uses_dsa_tags(dev->dsa_ptr);
1666 #endif
1667
1668 return 0;
1669 }
1670
1671 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1672 {
1673 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1674 if (dev->dsa_ptr != NULL)
1675 return dsa_uses_trailer_tags(dev->dsa_ptr);
1676 #endif
1677
1678 return 0;
1679 }
1680
1681 /**
1682 * netdev_priv - access network device private data
1683 * @dev: network device
1684 *
1685 * Get network device private data
1686 */
1687 static inline void *netdev_priv(const struct net_device *dev)
1688 {
1689 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1690 }
1691
1692 /* Set the sysfs physical device reference for the network logical device
1693 * if set prior to registration will cause a symlink during initialization.
1694 */
1695 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1696
1697 /* Set the sysfs device type for the network logical device to allow
1698 * fine-grained identification of different network device types. For
1699 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1700 */
1701 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1702
1703 /* Default NAPI poll() weight
1704 * Device drivers are strongly advised to not use bigger value
1705 */
1706 #define NAPI_POLL_WEIGHT 64
1707
1708 /**
1709 * netif_napi_add - initialize a napi context
1710 * @dev: network device
1711 * @napi: napi context
1712 * @poll: polling function
1713 * @weight: default weight
1714 *
1715 * netif_napi_add() must be used to initialize a napi context prior to calling
1716 * *any* of the other napi related functions.
1717 */
1718 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1719 int (*poll)(struct napi_struct *, int), int weight);
1720
1721 /**
1722 * netif_napi_del - remove a napi context
1723 * @napi: napi context
1724 *
1725 * netif_napi_del() removes a napi context from the network device napi list
1726 */
1727 void netif_napi_del(struct napi_struct *napi);
1728
1729 struct napi_gro_cb {
1730 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1731 void *frag0;
1732
1733 /* Length of frag0. */
1734 unsigned int frag0_len;
1735
1736 /* This indicates where we are processing relative to skb->data. */
1737 int data_offset;
1738
1739 /* This is non-zero if the packet cannot be merged with the new skb. */
1740 u16 flush;
1741
1742 /* Save the IP ID here and check when we get to the transport layer */
1743 u16 flush_id;
1744
1745 /* Number of segments aggregated. */
1746 u16 count;
1747
1748 /* This is non-zero if the packet may be of the same flow. */
1749 u8 same_flow;
1750
1751 /* Free the skb? */
1752 u8 free;
1753 #define NAPI_GRO_FREE 1
1754 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1755
1756 /* jiffies when first packet was created/queued */
1757 unsigned long age;
1758
1759 /* Used in ipv6_gro_receive() */
1760 u16 proto;
1761
1762 /* Used in udp_gro_receive */
1763 u16 udp_mark;
1764
1765 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1766 __wsum csum;
1767
1768 /* used in skb_gro_receive() slow path */
1769 struct sk_buff *last;
1770 };
1771
1772 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1773
1774 struct packet_type {
1775 __be16 type; /* This is really htons(ether_type). */
1776 struct net_device *dev; /* NULL is wildcarded here */
1777 int (*func) (struct sk_buff *,
1778 struct net_device *,
1779 struct packet_type *,
1780 struct net_device *);
1781 bool (*id_match)(struct packet_type *ptype,
1782 struct sock *sk);
1783 void *af_packet_priv;
1784 struct list_head list;
1785 };
1786
1787 struct offload_callbacks {
1788 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1789 netdev_features_t features);
1790 int (*gso_send_check)(struct sk_buff *skb);
1791 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1792 struct sk_buff *skb);
1793 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1794 };
1795
1796 struct packet_offload {
1797 __be16 type; /* This is really htons(ether_type). */
1798 struct offload_callbacks callbacks;
1799 struct list_head list;
1800 };
1801
1802 struct udp_offload {
1803 __be16 port;
1804 struct offload_callbacks callbacks;
1805 };
1806
1807 /* often modified stats are per cpu, other are shared (netdev->stats) */
1808 struct pcpu_sw_netstats {
1809 u64 rx_packets;
1810 u64 rx_bytes;
1811 u64 tx_packets;
1812 u64 tx_bytes;
1813 struct u64_stats_sync syncp;
1814 };
1815
1816 #define netdev_alloc_pcpu_stats(type) \
1817 ({ \
1818 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1819 if (pcpu_stats) { \
1820 int i; \
1821 for_each_possible_cpu(i) { \
1822 typeof(type) *stat; \
1823 stat = per_cpu_ptr(pcpu_stats, i); \
1824 u64_stats_init(&stat->syncp); \
1825 } \
1826 } \
1827 pcpu_stats; \
1828 })
1829
1830 #include <linux/notifier.h>
1831
1832 /* netdevice notifier chain. Please remember to update the rtnetlink
1833 * notification exclusion list in rtnetlink_event() when adding new
1834 * types.
1835 */
1836 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1837 #define NETDEV_DOWN 0x0002
1838 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1839 detected a hardware crash and restarted
1840 - we can use this eg to kick tcp sessions
1841 once done */
1842 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1843 #define NETDEV_REGISTER 0x0005
1844 #define NETDEV_UNREGISTER 0x0006
1845 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1846 #define NETDEV_CHANGEADDR 0x0008
1847 #define NETDEV_GOING_DOWN 0x0009
1848 #define NETDEV_CHANGENAME 0x000A
1849 #define NETDEV_FEAT_CHANGE 0x000B
1850 #define NETDEV_BONDING_FAILOVER 0x000C
1851 #define NETDEV_PRE_UP 0x000D
1852 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1853 #define NETDEV_POST_TYPE_CHANGE 0x000F
1854 #define NETDEV_POST_INIT 0x0010
1855 #define NETDEV_UNREGISTER_FINAL 0x0011
1856 #define NETDEV_RELEASE 0x0012
1857 #define NETDEV_NOTIFY_PEERS 0x0013
1858 #define NETDEV_JOIN 0x0014
1859 #define NETDEV_CHANGEUPPER 0x0015
1860 #define NETDEV_RESEND_IGMP 0x0016
1861 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1862
1863 int register_netdevice_notifier(struct notifier_block *nb);
1864 int unregister_netdevice_notifier(struct notifier_block *nb);
1865
1866 struct netdev_notifier_info {
1867 struct net_device *dev;
1868 };
1869
1870 struct netdev_notifier_change_info {
1871 struct netdev_notifier_info info; /* must be first */
1872 unsigned int flags_changed;
1873 };
1874
1875 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1876 struct net_device *dev)
1877 {
1878 info->dev = dev;
1879 }
1880
1881 static inline struct net_device *
1882 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1883 {
1884 return info->dev;
1885 }
1886
1887 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1888
1889
1890 extern rwlock_t dev_base_lock; /* Device list lock */
1891
1892 #define for_each_netdev(net, d) \
1893 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1894 #define for_each_netdev_reverse(net, d) \
1895 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1896 #define for_each_netdev_rcu(net, d) \
1897 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1898 #define for_each_netdev_safe(net, d, n) \
1899 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1900 #define for_each_netdev_continue(net, d) \
1901 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1902 #define for_each_netdev_continue_rcu(net, d) \
1903 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1904 #define for_each_netdev_in_bond_rcu(bond, slave) \
1905 for_each_netdev_rcu(&init_net, slave) \
1906 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1907 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1908
1909 static inline struct net_device *next_net_device(struct net_device *dev)
1910 {
1911 struct list_head *lh;
1912 struct net *net;
1913
1914 net = dev_net(dev);
1915 lh = dev->dev_list.next;
1916 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1917 }
1918
1919 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1920 {
1921 struct list_head *lh;
1922 struct net *net;
1923
1924 net = dev_net(dev);
1925 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1926 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1927 }
1928
1929 static inline struct net_device *first_net_device(struct net *net)
1930 {
1931 return list_empty(&net->dev_base_head) ? NULL :
1932 net_device_entry(net->dev_base_head.next);
1933 }
1934
1935 static inline struct net_device *first_net_device_rcu(struct net *net)
1936 {
1937 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1938
1939 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1940 }
1941
1942 int netdev_boot_setup_check(struct net_device *dev);
1943 unsigned long netdev_boot_base(const char *prefix, int unit);
1944 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1945 const char *hwaddr);
1946 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1947 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1948 void dev_add_pack(struct packet_type *pt);
1949 void dev_remove_pack(struct packet_type *pt);
1950 void __dev_remove_pack(struct packet_type *pt);
1951 void dev_add_offload(struct packet_offload *po);
1952 void dev_remove_offload(struct packet_offload *po);
1953
1954 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1955 unsigned short mask);
1956 struct net_device *dev_get_by_name(struct net *net, const char *name);
1957 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1958 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1959 int dev_alloc_name(struct net_device *dev, const char *name);
1960 int dev_open(struct net_device *dev);
1961 int dev_close(struct net_device *dev);
1962 void dev_disable_lro(struct net_device *dev);
1963 int dev_loopback_xmit(struct sk_buff *newskb);
1964 int dev_queue_xmit(struct sk_buff *skb);
1965 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1966 int register_netdevice(struct net_device *dev);
1967 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1968 void unregister_netdevice_many(struct list_head *head);
1969 static inline void unregister_netdevice(struct net_device *dev)
1970 {
1971 unregister_netdevice_queue(dev, NULL);
1972 }
1973
1974 int netdev_refcnt_read(const struct net_device *dev);
1975 void free_netdev(struct net_device *dev);
1976 void netdev_freemem(struct net_device *dev);
1977 void synchronize_net(void);
1978 int init_dummy_netdev(struct net_device *dev);
1979
1980 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1981 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1982 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1983 int netdev_get_name(struct net *net, char *name, int ifindex);
1984 int dev_restart(struct net_device *dev);
1985 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1986
1987 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1988 {
1989 return NAPI_GRO_CB(skb)->data_offset;
1990 }
1991
1992 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1993 {
1994 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1995 }
1996
1997 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1998 {
1999 NAPI_GRO_CB(skb)->data_offset += len;
2000 }
2001
2002 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2003 unsigned int offset)
2004 {
2005 return NAPI_GRO_CB(skb)->frag0 + offset;
2006 }
2007
2008 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2009 {
2010 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2011 }
2012
2013 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2014 unsigned int offset)
2015 {
2016 if (!pskb_may_pull(skb, hlen))
2017 return NULL;
2018
2019 NAPI_GRO_CB(skb)->frag0 = NULL;
2020 NAPI_GRO_CB(skb)->frag0_len = 0;
2021 return skb->data + offset;
2022 }
2023
2024 static inline void *skb_gro_network_header(struct sk_buff *skb)
2025 {
2026 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2027 skb_network_offset(skb);
2028 }
2029
2030 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2031 const void *start, unsigned int len)
2032 {
2033 if (skb->ip_summed == CHECKSUM_COMPLETE)
2034 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2035 csum_partial(start, len, 0));
2036 }
2037
2038 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2039 unsigned short type,
2040 const void *daddr, const void *saddr,
2041 unsigned int len)
2042 {
2043 if (!dev->header_ops || !dev->header_ops->create)
2044 return 0;
2045
2046 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2047 }
2048
2049 static inline int dev_parse_header(const struct sk_buff *skb,
2050 unsigned char *haddr)
2051 {
2052 const struct net_device *dev = skb->dev;
2053
2054 if (!dev->header_ops || !dev->header_ops->parse)
2055 return 0;
2056 return dev->header_ops->parse(skb, haddr);
2057 }
2058
2059 static inline int dev_rebuild_header(struct sk_buff *skb)
2060 {
2061 const struct net_device *dev = skb->dev;
2062
2063 if (!dev->header_ops || !dev->header_ops->rebuild)
2064 return 0;
2065 return dev->header_ops->rebuild(skb);
2066 }
2067
2068 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2069 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2070 static inline int unregister_gifconf(unsigned int family)
2071 {
2072 return register_gifconf(family, NULL);
2073 }
2074
2075 #ifdef CONFIG_NET_FLOW_LIMIT
2076 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2077 struct sd_flow_limit {
2078 u64 count;
2079 unsigned int num_buckets;
2080 unsigned int history_head;
2081 u16 history[FLOW_LIMIT_HISTORY];
2082 u8 buckets[];
2083 };
2084
2085 extern int netdev_flow_limit_table_len;
2086 #endif /* CONFIG_NET_FLOW_LIMIT */
2087
2088 /*
2089 * Incoming packets are placed on per-cpu queues
2090 */
2091 struct softnet_data {
2092 struct Qdisc *output_queue;
2093 struct Qdisc **output_queue_tailp;
2094 struct list_head poll_list;
2095 struct sk_buff *completion_queue;
2096 struct sk_buff_head process_queue;
2097
2098 /* stats */
2099 unsigned int processed;
2100 unsigned int time_squeeze;
2101 unsigned int cpu_collision;
2102 unsigned int received_rps;
2103
2104 #ifdef CONFIG_RPS
2105 struct softnet_data *rps_ipi_list;
2106
2107 /* Elements below can be accessed between CPUs for RPS */
2108 struct call_single_data csd ____cacheline_aligned_in_smp;
2109 struct softnet_data *rps_ipi_next;
2110 unsigned int cpu;
2111 unsigned int input_queue_head;
2112 unsigned int input_queue_tail;
2113 #endif
2114 unsigned int dropped;
2115 struct sk_buff_head input_pkt_queue;
2116 struct napi_struct backlog;
2117
2118 #ifdef CONFIG_NET_FLOW_LIMIT
2119 struct sd_flow_limit __rcu *flow_limit;
2120 #endif
2121 };
2122
2123 static inline void input_queue_head_incr(struct softnet_data *sd)
2124 {
2125 #ifdef CONFIG_RPS
2126 sd->input_queue_head++;
2127 #endif
2128 }
2129
2130 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2131 unsigned int *qtail)
2132 {
2133 #ifdef CONFIG_RPS
2134 *qtail = ++sd->input_queue_tail;
2135 #endif
2136 }
2137
2138 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2139
2140 void __netif_schedule(struct Qdisc *q);
2141
2142 static inline void netif_schedule_queue(struct netdev_queue *txq)
2143 {
2144 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2145 __netif_schedule(txq->qdisc);
2146 }
2147
2148 static inline void netif_tx_schedule_all(struct net_device *dev)
2149 {
2150 unsigned int i;
2151
2152 for (i = 0; i < dev->num_tx_queues; i++)
2153 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2154 }
2155
2156 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2157 {
2158 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2159 }
2160
2161 /**
2162 * netif_start_queue - allow transmit
2163 * @dev: network device
2164 *
2165 * Allow upper layers to call the device hard_start_xmit routine.
2166 */
2167 static inline void netif_start_queue(struct net_device *dev)
2168 {
2169 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2170 }
2171
2172 static inline void netif_tx_start_all_queues(struct net_device *dev)
2173 {
2174 unsigned int i;
2175
2176 for (i = 0; i < dev->num_tx_queues; i++) {
2177 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2178 netif_tx_start_queue(txq);
2179 }
2180 }
2181
2182 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2183 {
2184 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2185 __netif_schedule(dev_queue->qdisc);
2186 }
2187
2188 /**
2189 * netif_wake_queue - restart transmit
2190 * @dev: network device
2191 *
2192 * Allow upper layers to call the device hard_start_xmit routine.
2193 * Used for flow control when transmit resources are available.
2194 */
2195 static inline void netif_wake_queue(struct net_device *dev)
2196 {
2197 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2198 }
2199
2200 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2201 {
2202 unsigned int i;
2203
2204 for (i = 0; i < dev->num_tx_queues; i++) {
2205 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2206 netif_tx_wake_queue(txq);
2207 }
2208 }
2209
2210 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2211 {
2212 if (WARN_ON(!dev_queue)) {
2213 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2214 return;
2215 }
2216 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2217 }
2218
2219 /**
2220 * netif_stop_queue - stop transmitted packets
2221 * @dev: network device
2222 *
2223 * Stop upper layers calling the device hard_start_xmit routine.
2224 * Used for flow control when transmit resources are unavailable.
2225 */
2226 static inline void netif_stop_queue(struct net_device *dev)
2227 {
2228 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2229 }
2230
2231 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2232 {
2233 unsigned int i;
2234
2235 for (i = 0; i < dev->num_tx_queues; i++) {
2236 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2237 netif_tx_stop_queue(txq);
2238 }
2239 }
2240
2241 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2242 {
2243 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2244 }
2245
2246 /**
2247 * netif_queue_stopped - test if transmit queue is flowblocked
2248 * @dev: network device
2249 *
2250 * Test if transmit queue on device is currently unable to send.
2251 */
2252 static inline bool netif_queue_stopped(const struct net_device *dev)
2253 {
2254 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2255 }
2256
2257 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2258 {
2259 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2260 }
2261
2262 static inline bool
2263 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2264 {
2265 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2266 }
2267
2268 static inline bool
2269 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2270 {
2271 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2272 }
2273
2274 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2275 unsigned int bytes)
2276 {
2277 #ifdef CONFIG_BQL
2278 dql_queued(&dev_queue->dql, bytes);
2279
2280 if (likely(dql_avail(&dev_queue->dql) >= 0))
2281 return;
2282
2283 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2284
2285 /*
2286 * The XOFF flag must be set before checking the dql_avail below,
2287 * because in netdev_tx_completed_queue we update the dql_completed
2288 * before checking the XOFF flag.
2289 */
2290 smp_mb();
2291
2292 /* check again in case another CPU has just made room avail */
2293 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2294 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2295 #endif
2296 }
2297
2298 /**
2299 * netdev_sent_queue - report the number of bytes queued to hardware
2300 * @dev: network device
2301 * @bytes: number of bytes queued to the hardware device queue
2302 *
2303 * Report the number of bytes queued for sending/completion to the network
2304 * device hardware queue. @bytes should be a good approximation and should
2305 * exactly match netdev_completed_queue() @bytes
2306 */
2307 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2308 {
2309 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2310 }
2311
2312 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2313 unsigned int pkts, unsigned int bytes)
2314 {
2315 #ifdef CONFIG_BQL
2316 if (unlikely(!bytes))
2317 return;
2318
2319 dql_completed(&dev_queue->dql, bytes);
2320
2321 /*
2322 * Without the memory barrier there is a small possiblity that
2323 * netdev_tx_sent_queue will miss the update and cause the queue to
2324 * be stopped forever
2325 */
2326 smp_mb();
2327
2328 if (dql_avail(&dev_queue->dql) < 0)
2329 return;
2330
2331 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2332 netif_schedule_queue(dev_queue);
2333 #endif
2334 }
2335
2336 /**
2337 * netdev_completed_queue - report bytes and packets completed by device
2338 * @dev: network device
2339 * @pkts: actual number of packets sent over the medium
2340 * @bytes: actual number of bytes sent over the medium
2341 *
2342 * Report the number of bytes and packets transmitted by the network device
2343 * hardware queue over the physical medium, @bytes must exactly match the
2344 * @bytes amount passed to netdev_sent_queue()
2345 */
2346 static inline void netdev_completed_queue(struct net_device *dev,
2347 unsigned int pkts, unsigned int bytes)
2348 {
2349 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2350 }
2351
2352 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2353 {
2354 #ifdef CONFIG_BQL
2355 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2356 dql_reset(&q->dql);
2357 #endif
2358 }
2359
2360 /**
2361 * netdev_reset_queue - reset the packets and bytes count of a network device
2362 * @dev_queue: network device
2363 *
2364 * Reset the bytes and packet count of a network device and clear the
2365 * software flow control OFF bit for this network device
2366 */
2367 static inline void netdev_reset_queue(struct net_device *dev_queue)
2368 {
2369 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2370 }
2371
2372 /**
2373 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2374 * @dev: network device
2375 * @queue_index: given tx queue index
2376 *
2377 * Returns 0 if given tx queue index >= number of device tx queues,
2378 * otherwise returns the originally passed tx queue index.
2379 */
2380 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2381 {
2382 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2383 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2384 dev->name, queue_index,
2385 dev->real_num_tx_queues);
2386 return 0;
2387 }
2388
2389 return queue_index;
2390 }
2391
2392 /**
2393 * netif_running - test if up
2394 * @dev: network device
2395 *
2396 * Test if the device has been brought up.
2397 */
2398 static inline bool netif_running(const struct net_device *dev)
2399 {
2400 return test_bit(__LINK_STATE_START, &dev->state);
2401 }
2402
2403 /*
2404 * Routines to manage the subqueues on a device. We only need start
2405 * stop, and a check if it's stopped. All other device management is
2406 * done at the overall netdevice level.
2407 * Also test the device if we're multiqueue.
2408 */
2409
2410 /**
2411 * netif_start_subqueue - allow sending packets on subqueue
2412 * @dev: network device
2413 * @queue_index: sub queue index
2414 *
2415 * Start individual transmit queue of a device with multiple transmit queues.
2416 */
2417 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2418 {
2419 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2420
2421 netif_tx_start_queue(txq);
2422 }
2423
2424 /**
2425 * netif_stop_subqueue - stop sending packets on subqueue
2426 * @dev: network device
2427 * @queue_index: sub queue index
2428 *
2429 * Stop individual transmit queue of a device with multiple transmit queues.
2430 */
2431 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2432 {
2433 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2434 netif_tx_stop_queue(txq);
2435 }
2436
2437 /**
2438 * netif_subqueue_stopped - test status of subqueue
2439 * @dev: network device
2440 * @queue_index: sub queue index
2441 *
2442 * Check individual transmit queue of a device with multiple transmit queues.
2443 */
2444 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2445 u16 queue_index)
2446 {
2447 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2448
2449 return netif_tx_queue_stopped(txq);
2450 }
2451
2452 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2453 struct sk_buff *skb)
2454 {
2455 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2456 }
2457
2458 /**
2459 * netif_wake_subqueue - allow sending packets on subqueue
2460 * @dev: network device
2461 * @queue_index: sub queue index
2462 *
2463 * Resume individual transmit queue of a device with multiple transmit queues.
2464 */
2465 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2466 {
2467 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2468 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2469 __netif_schedule(txq->qdisc);
2470 }
2471
2472 #ifdef CONFIG_XPS
2473 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2474 u16 index);
2475 #else
2476 static inline int netif_set_xps_queue(struct net_device *dev,
2477 const struct cpumask *mask,
2478 u16 index)
2479 {
2480 return 0;
2481 }
2482 #endif
2483
2484 /*
2485 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2486 * as a distribution range limit for the returned value.
2487 */
2488 static inline u16 skb_tx_hash(const struct net_device *dev,
2489 const struct sk_buff *skb)
2490 {
2491 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2492 }
2493
2494 /**
2495 * netif_is_multiqueue - test if device has multiple transmit queues
2496 * @dev: network device
2497 *
2498 * Check if device has multiple transmit queues
2499 */
2500 static inline bool netif_is_multiqueue(const struct net_device *dev)
2501 {
2502 return dev->num_tx_queues > 1;
2503 }
2504
2505 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2506
2507 #ifdef CONFIG_SYSFS
2508 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2509 #else
2510 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2511 unsigned int rxq)
2512 {
2513 return 0;
2514 }
2515 #endif
2516
2517 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2518 const struct net_device *from_dev)
2519 {
2520 int err;
2521
2522 err = netif_set_real_num_tx_queues(to_dev,
2523 from_dev->real_num_tx_queues);
2524 if (err)
2525 return err;
2526 #ifdef CONFIG_SYSFS
2527 return netif_set_real_num_rx_queues(to_dev,
2528 from_dev->real_num_rx_queues);
2529 #else
2530 return 0;
2531 #endif
2532 }
2533
2534 #ifdef CONFIG_SYSFS
2535 static inline unsigned int get_netdev_rx_queue_index(
2536 struct netdev_rx_queue *queue)
2537 {
2538 struct net_device *dev = queue->dev;
2539 int index = queue - dev->_rx;
2540
2541 BUG_ON(index >= dev->num_rx_queues);
2542 return index;
2543 }
2544 #endif
2545
2546 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2547 int netif_get_num_default_rss_queues(void);
2548
2549 enum skb_free_reason {
2550 SKB_REASON_CONSUMED,
2551 SKB_REASON_DROPPED,
2552 };
2553
2554 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2555 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2556
2557 /*
2558 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2559 * interrupt context or with hardware interrupts being disabled.
2560 * (in_irq() || irqs_disabled())
2561 *
2562 * We provide four helpers that can be used in following contexts :
2563 *
2564 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2565 * replacing kfree_skb(skb)
2566 *
2567 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2568 * Typically used in place of consume_skb(skb) in TX completion path
2569 *
2570 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2571 * replacing kfree_skb(skb)
2572 *
2573 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2574 * and consumed a packet. Used in place of consume_skb(skb)
2575 */
2576 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2577 {
2578 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2579 }
2580
2581 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2582 {
2583 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2584 }
2585
2586 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2587 {
2588 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2589 }
2590
2591 static inline void dev_consume_skb_any(struct sk_buff *skb)
2592 {
2593 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2594 }
2595
2596 int netif_rx(struct sk_buff *skb);
2597 int netif_rx_ni(struct sk_buff *skb);
2598 int netif_receive_skb(struct sk_buff *skb);
2599 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2600 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2601 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2602 gro_result_t napi_gro_frags(struct napi_struct *napi);
2603 struct packet_offload *gro_find_receive_by_type(__be16 type);
2604 struct packet_offload *gro_find_complete_by_type(__be16 type);
2605
2606 static inline void napi_free_frags(struct napi_struct *napi)
2607 {
2608 kfree_skb(napi->skb);
2609 napi->skb = NULL;
2610 }
2611
2612 int netdev_rx_handler_register(struct net_device *dev,
2613 rx_handler_func_t *rx_handler,
2614 void *rx_handler_data);
2615 void netdev_rx_handler_unregister(struct net_device *dev);
2616
2617 bool dev_valid_name(const char *name);
2618 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2619 int dev_ethtool(struct net *net, struct ifreq *);
2620 unsigned int dev_get_flags(const struct net_device *);
2621 int __dev_change_flags(struct net_device *, unsigned int flags);
2622 int dev_change_flags(struct net_device *, unsigned int);
2623 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2624 unsigned int gchanges);
2625 int dev_change_name(struct net_device *, const char *);
2626 int dev_set_alias(struct net_device *, const char *, size_t);
2627 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2628 int dev_set_mtu(struct net_device *, int);
2629 void dev_set_group(struct net_device *, int);
2630 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2631 int dev_change_carrier(struct net_device *, bool new_carrier);
2632 int dev_get_phys_port_id(struct net_device *dev,
2633 struct netdev_phys_port_id *ppid);
2634 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2635 struct netdev_queue *txq);
2636 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2637 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2638 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2639
2640 extern int netdev_budget;
2641
2642 /* Called by rtnetlink.c:rtnl_unlock() */
2643 void netdev_run_todo(void);
2644
2645 /**
2646 * dev_put - release reference to device
2647 * @dev: network device
2648 *
2649 * Release reference to device to allow it to be freed.
2650 */
2651 static inline void dev_put(struct net_device *dev)
2652 {
2653 this_cpu_dec(*dev->pcpu_refcnt);
2654 }
2655
2656 /**
2657 * dev_hold - get reference to device
2658 * @dev: network device
2659 *
2660 * Hold reference to device to keep it from being freed.
2661 */
2662 static inline void dev_hold(struct net_device *dev)
2663 {
2664 this_cpu_inc(*dev->pcpu_refcnt);
2665 }
2666
2667 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2668 * and _off may be called from IRQ context, but it is caller
2669 * who is responsible for serialization of these calls.
2670 *
2671 * The name carrier is inappropriate, these functions should really be
2672 * called netif_lowerlayer_*() because they represent the state of any
2673 * kind of lower layer not just hardware media.
2674 */
2675
2676 void linkwatch_init_dev(struct net_device *dev);
2677 void linkwatch_fire_event(struct net_device *dev);
2678 void linkwatch_forget_dev(struct net_device *dev);
2679
2680 /**
2681 * netif_carrier_ok - test if carrier present
2682 * @dev: network device
2683 *
2684 * Check if carrier is present on device
2685 */
2686 static inline bool netif_carrier_ok(const struct net_device *dev)
2687 {
2688 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2689 }
2690
2691 unsigned long dev_trans_start(struct net_device *dev);
2692
2693 void __netdev_watchdog_up(struct net_device *dev);
2694
2695 void netif_carrier_on(struct net_device *dev);
2696
2697 void netif_carrier_off(struct net_device *dev);
2698
2699 /**
2700 * netif_dormant_on - mark device as dormant.
2701 * @dev: network device
2702 *
2703 * Mark device as dormant (as per RFC2863).
2704 *
2705 * The dormant state indicates that the relevant interface is not
2706 * actually in a condition to pass packets (i.e., it is not 'up') but is
2707 * in a "pending" state, waiting for some external event. For "on-
2708 * demand" interfaces, this new state identifies the situation where the
2709 * interface is waiting for events to place it in the up state.
2710 *
2711 */
2712 static inline void netif_dormant_on(struct net_device *dev)
2713 {
2714 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2715 linkwatch_fire_event(dev);
2716 }
2717
2718 /**
2719 * netif_dormant_off - set device as not dormant.
2720 * @dev: network device
2721 *
2722 * Device is not in dormant state.
2723 */
2724 static inline void netif_dormant_off(struct net_device *dev)
2725 {
2726 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2727 linkwatch_fire_event(dev);
2728 }
2729
2730 /**
2731 * netif_dormant - test if carrier present
2732 * @dev: network device
2733 *
2734 * Check if carrier is present on device
2735 */
2736 static inline bool netif_dormant(const struct net_device *dev)
2737 {
2738 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2739 }
2740
2741
2742 /**
2743 * netif_oper_up - test if device is operational
2744 * @dev: network device
2745 *
2746 * Check if carrier is operational
2747 */
2748 static inline bool netif_oper_up(const struct net_device *dev)
2749 {
2750 return (dev->operstate == IF_OPER_UP ||
2751 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2752 }
2753
2754 /**
2755 * netif_device_present - is device available or removed
2756 * @dev: network device
2757 *
2758 * Check if device has not been removed from system.
2759 */
2760 static inline bool netif_device_present(struct net_device *dev)
2761 {
2762 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2763 }
2764
2765 void netif_device_detach(struct net_device *dev);
2766
2767 void netif_device_attach(struct net_device *dev);
2768
2769 /*
2770 * Network interface message level settings
2771 */
2772
2773 enum {
2774 NETIF_MSG_DRV = 0x0001,
2775 NETIF_MSG_PROBE = 0x0002,
2776 NETIF_MSG_LINK = 0x0004,
2777 NETIF_MSG_TIMER = 0x0008,
2778 NETIF_MSG_IFDOWN = 0x0010,
2779 NETIF_MSG_IFUP = 0x0020,
2780 NETIF_MSG_RX_ERR = 0x0040,
2781 NETIF_MSG_TX_ERR = 0x0080,
2782 NETIF_MSG_TX_QUEUED = 0x0100,
2783 NETIF_MSG_INTR = 0x0200,
2784 NETIF_MSG_TX_DONE = 0x0400,
2785 NETIF_MSG_RX_STATUS = 0x0800,
2786 NETIF_MSG_PKTDATA = 0x1000,
2787 NETIF_MSG_HW = 0x2000,
2788 NETIF_MSG_WOL = 0x4000,
2789 };
2790
2791 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2792 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2793 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2794 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2795 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2796 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2797 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2798 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2799 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2800 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2801 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2802 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2803 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2804 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2805 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2806
2807 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2808 {
2809 /* use default */
2810 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2811 return default_msg_enable_bits;
2812 if (debug_value == 0) /* no output */
2813 return 0;
2814 /* set low N bits */
2815 return (1 << debug_value) - 1;
2816 }
2817
2818 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2819 {
2820 spin_lock(&txq->_xmit_lock);
2821 txq->xmit_lock_owner = cpu;
2822 }
2823
2824 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2825 {
2826 spin_lock_bh(&txq->_xmit_lock);
2827 txq->xmit_lock_owner = smp_processor_id();
2828 }
2829
2830 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2831 {
2832 bool ok = spin_trylock(&txq->_xmit_lock);
2833 if (likely(ok))
2834 txq->xmit_lock_owner = smp_processor_id();
2835 return ok;
2836 }
2837
2838 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2839 {
2840 txq->xmit_lock_owner = -1;
2841 spin_unlock(&txq->_xmit_lock);
2842 }
2843
2844 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2845 {
2846 txq->xmit_lock_owner = -1;
2847 spin_unlock_bh(&txq->_xmit_lock);
2848 }
2849
2850 static inline void txq_trans_update(struct netdev_queue *txq)
2851 {
2852 if (txq->xmit_lock_owner != -1)
2853 txq->trans_start = jiffies;
2854 }
2855
2856 /**
2857 * netif_tx_lock - grab network device transmit lock
2858 * @dev: network device
2859 *
2860 * Get network device transmit lock
2861 */
2862 static inline void netif_tx_lock(struct net_device *dev)
2863 {
2864 unsigned int i;
2865 int cpu;
2866
2867 spin_lock(&dev->tx_global_lock);
2868 cpu = smp_processor_id();
2869 for (i = 0; i < dev->num_tx_queues; i++) {
2870 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2871
2872 /* We are the only thread of execution doing a
2873 * freeze, but we have to grab the _xmit_lock in
2874 * order to synchronize with threads which are in
2875 * the ->hard_start_xmit() handler and already
2876 * checked the frozen bit.
2877 */
2878 __netif_tx_lock(txq, cpu);
2879 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2880 __netif_tx_unlock(txq);
2881 }
2882 }
2883
2884 static inline void netif_tx_lock_bh(struct net_device *dev)
2885 {
2886 local_bh_disable();
2887 netif_tx_lock(dev);
2888 }
2889
2890 static inline void netif_tx_unlock(struct net_device *dev)
2891 {
2892 unsigned int i;
2893
2894 for (i = 0; i < dev->num_tx_queues; i++) {
2895 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2896
2897 /* No need to grab the _xmit_lock here. If the
2898 * queue is not stopped for another reason, we
2899 * force a schedule.
2900 */
2901 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2902 netif_schedule_queue(txq);
2903 }
2904 spin_unlock(&dev->tx_global_lock);
2905 }
2906
2907 static inline void netif_tx_unlock_bh(struct net_device *dev)
2908 {
2909 netif_tx_unlock(dev);
2910 local_bh_enable();
2911 }
2912
2913 #define HARD_TX_LOCK(dev, txq, cpu) { \
2914 if ((dev->features & NETIF_F_LLTX) == 0) { \
2915 __netif_tx_lock(txq, cpu); \
2916 } \
2917 }
2918
2919 #define HARD_TX_TRYLOCK(dev, txq) \
2920 (((dev->features & NETIF_F_LLTX) == 0) ? \
2921 __netif_tx_trylock(txq) : \
2922 true )
2923
2924 #define HARD_TX_UNLOCK(dev, txq) { \
2925 if ((dev->features & NETIF_F_LLTX) == 0) { \
2926 __netif_tx_unlock(txq); \
2927 } \
2928 }
2929
2930 static inline void netif_tx_disable(struct net_device *dev)
2931 {
2932 unsigned int i;
2933 int cpu;
2934
2935 local_bh_disable();
2936 cpu = smp_processor_id();
2937 for (i = 0; i < dev->num_tx_queues; i++) {
2938 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2939
2940 __netif_tx_lock(txq, cpu);
2941 netif_tx_stop_queue(txq);
2942 __netif_tx_unlock(txq);
2943 }
2944 local_bh_enable();
2945 }
2946
2947 static inline void netif_addr_lock(struct net_device *dev)
2948 {
2949 spin_lock(&dev->addr_list_lock);
2950 }
2951
2952 static inline void netif_addr_lock_nested(struct net_device *dev)
2953 {
2954 int subclass = SINGLE_DEPTH_NESTING;
2955
2956 if (dev->netdev_ops->ndo_get_lock_subclass)
2957 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
2958
2959 spin_lock_nested(&dev->addr_list_lock, subclass);
2960 }
2961
2962 static inline void netif_addr_lock_bh(struct net_device *dev)
2963 {
2964 spin_lock_bh(&dev->addr_list_lock);
2965 }
2966
2967 static inline void netif_addr_unlock(struct net_device *dev)
2968 {
2969 spin_unlock(&dev->addr_list_lock);
2970 }
2971
2972 static inline void netif_addr_unlock_bh(struct net_device *dev)
2973 {
2974 spin_unlock_bh(&dev->addr_list_lock);
2975 }
2976
2977 /*
2978 * dev_addrs walker. Should be used only for read access. Call with
2979 * rcu_read_lock held.
2980 */
2981 #define for_each_dev_addr(dev, ha) \
2982 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2983
2984 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2985
2986 void ether_setup(struct net_device *dev);
2987
2988 /* Support for loadable net-drivers */
2989 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2990 void (*setup)(struct net_device *),
2991 unsigned int txqs, unsigned int rxqs);
2992 #define alloc_netdev(sizeof_priv, name, setup) \
2993 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2994
2995 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2996 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2997
2998 int register_netdev(struct net_device *dev);
2999 void unregister_netdev(struct net_device *dev);
3000
3001 /* General hardware address lists handling functions */
3002 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3003 struct netdev_hw_addr_list *from_list, int addr_len);
3004 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3005 struct netdev_hw_addr_list *from_list, int addr_len);
3006 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3007 struct net_device *dev,
3008 int (*sync)(struct net_device *, const unsigned char *),
3009 int (*unsync)(struct net_device *,
3010 const unsigned char *));
3011 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3012 struct net_device *dev,
3013 int (*unsync)(struct net_device *,
3014 const unsigned char *));
3015 void __hw_addr_init(struct netdev_hw_addr_list *list);
3016
3017 /* Functions used for device addresses handling */
3018 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3019 unsigned char addr_type);
3020 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3021 unsigned char addr_type);
3022 void dev_addr_flush(struct net_device *dev);
3023 int dev_addr_init(struct net_device *dev);
3024
3025 /* Functions used for unicast addresses handling */
3026 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3027 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3028 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3029 int dev_uc_sync(struct net_device *to, struct net_device *from);
3030 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3031 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3032 void dev_uc_flush(struct net_device *dev);
3033 void dev_uc_init(struct net_device *dev);
3034
3035 /**
3036 * __dev_uc_sync - Synchonize device's unicast list
3037 * @dev: device to sync
3038 * @sync: function to call if address should be added
3039 * @unsync: function to call if address should be removed
3040 *
3041 * Add newly added addresses to the interface, and release
3042 * addresses that have been deleted.
3043 **/
3044 static inline int __dev_uc_sync(struct net_device *dev,
3045 int (*sync)(struct net_device *,
3046 const unsigned char *),
3047 int (*unsync)(struct net_device *,
3048 const unsigned char *))
3049 {
3050 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3051 }
3052
3053 /**
3054 * __dev_uc_unsync - Remove synchonized addresses from device
3055 * @dev: device to sync
3056 * @unsync: function to call if address should be removed
3057 *
3058 * Remove all addresses that were added to the device by dev_uc_sync().
3059 **/
3060 static inline void __dev_uc_unsync(struct net_device *dev,
3061 int (*unsync)(struct net_device *,
3062 const unsigned char *))
3063 {
3064 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3065 }
3066
3067 /* Functions used for multicast addresses handling */
3068 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3069 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3070 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3071 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3072 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3073 int dev_mc_sync(struct net_device *to, struct net_device *from);
3074 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3075 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3076 void dev_mc_flush(struct net_device *dev);
3077 void dev_mc_init(struct net_device *dev);
3078
3079 /**
3080 * __dev_mc_sync - Synchonize device's multicast list
3081 * @dev: device to sync
3082 * @sync: function to call if address should be added
3083 * @unsync: function to call if address should be removed
3084 *
3085 * Add newly added addresses to the interface, and release
3086 * addresses that have been deleted.
3087 **/
3088 static inline int __dev_mc_sync(struct net_device *dev,
3089 int (*sync)(struct net_device *,
3090 const unsigned char *),
3091 int (*unsync)(struct net_device *,
3092 const unsigned char *))
3093 {
3094 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3095 }
3096
3097 /**
3098 * __dev_mc_unsync - Remove synchonized addresses from device
3099 * @dev: device to sync
3100 * @unsync: function to call if address should be removed
3101 *
3102 * Remove all addresses that were added to the device by dev_mc_sync().
3103 **/
3104 static inline void __dev_mc_unsync(struct net_device *dev,
3105 int (*unsync)(struct net_device *,
3106 const unsigned char *))
3107 {
3108 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3109 }
3110
3111 /* Functions used for secondary unicast and multicast support */
3112 void dev_set_rx_mode(struct net_device *dev);
3113 void __dev_set_rx_mode(struct net_device *dev);
3114 int dev_set_promiscuity(struct net_device *dev, int inc);
3115 int dev_set_allmulti(struct net_device *dev, int inc);
3116 void netdev_state_change(struct net_device *dev);
3117 void netdev_notify_peers(struct net_device *dev);
3118 void netdev_features_change(struct net_device *dev);
3119 /* Load a device via the kmod */
3120 void dev_load(struct net *net, const char *name);
3121 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3122 struct rtnl_link_stats64 *storage);
3123 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3124 const struct net_device_stats *netdev_stats);
3125
3126 extern int netdev_max_backlog;
3127 extern int netdev_tstamp_prequeue;
3128 extern int weight_p;
3129 extern int bpf_jit_enable;
3130
3131 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3132 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3133 struct list_head **iter);
3134 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3135 struct list_head **iter);
3136
3137 /* iterate through upper list, must be called under RCU read lock */
3138 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3139 for (iter = &(dev)->adj_list.upper, \
3140 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3141 updev; \
3142 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3143
3144 /* iterate through upper list, must be called under RCU read lock */
3145 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3146 for (iter = &(dev)->all_adj_list.upper, \
3147 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3148 updev; \
3149 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3150
3151 void *netdev_lower_get_next_private(struct net_device *dev,
3152 struct list_head **iter);
3153 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3154 struct list_head **iter);
3155
3156 #define netdev_for_each_lower_private(dev, priv, iter) \
3157 for (iter = (dev)->adj_list.lower.next, \
3158 priv = netdev_lower_get_next_private(dev, &(iter)); \
3159 priv; \
3160 priv = netdev_lower_get_next_private(dev, &(iter)))
3161
3162 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3163 for (iter = &(dev)->adj_list.lower, \
3164 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3165 priv; \
3166 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3167
3168 void *netdev_lower_get_next(struct net_device *dev,
3169 struct list_head **iter);
3170 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3171 for (iter = &(dev)->adj_list.lower, \
3172 ldev = netdev_lower_get_next(dev, &(iter)); \
3173 ldev; \
3174 ldev = netdev_lower_get_next(dev, &(iter)))
3175
3176 void *netdev_adjacent_get_private(struct list_head *adj_list);
3177 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3178 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3179 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3180 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3181 int netdev_master_upper_dev_link(struct net_device *dev,
3182 struct net_device *upper_dev);
3183 int netdev_master_upper_dev_link_private(struct net_device *dev,
3184 struct net_device *upper_dev,
3185 void *private);
3186 void netdev_upper_dev_unlink(struct net_device *dev,
3187 struct net_device *upper_dev);
3188 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3189 void *netdev_lower_dev_get_private(struct net_device *dev,
3190 struct net_device *lower_dev);
3191 int dev_get_nest_level(struct net_device *dev,
3192 bool (*type_check)(struct net_device *dev));
3193 int skb_checksum_help(struct sk_buff *skb);
3194 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3195 netdev_features_t features, bool tx_path);
3196 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3197 netdev_features_t features);
3198
3199 static inline
3200 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3201 {
3202 return __skb_gso_segment(skb, features, true);
3203 }
3204 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3205
3206 static inline bool can_checksum_protocol(netdev_features_t features,
3207 __be16 protocol)
3208 {
3209 return ((features & NETIF_F_GEN_CSUM) ||
3210 ((features & NETIF_F_V4_CSUM) &&
3211 protocol == htons(ETH_P_IP)) ||
3212 ((features & NETIF_F_V6_CSUM) &&
3213 protocol == htons(ETH_P_IPV6)) ||
3214 ((features & NETIF_F_FCOE_CRC) &&
3215 protocol == htons(ETH_P_FCOE)));
3216 }
3217
3218 #ifdef CONFIG_BUG
3219 void netdev_rx_csum_fault(struct net_device *dev);
3220 #else
3221 static inline void netdev_rx_csum_fault(struct net_device *dev)
3222 {
3223 }
3224 #endif
3225 /* rx skb timestamps */
3226 void net_enable_timestamp(void);
3227 void net_disable_timestamp(void);
3228
3229 #ifdef CONFIG_PROC_FS
3230 int __init dev_proc_init(void);
3231 #else
3232 #define dev_proc_init() 0
3233 #endif
3234
3235 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3236 const void *ns);
3237 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3238 const void *ns);
3239
3240 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3241 {
3242 return netdev_class_create_file_ns(class_attr, NULL);
3243 }
3244
3245 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3246 {
3247 netdev_class_remove_file_ns(class_attr, NULL);
3248 }
3249
3250 extern struct kobj_ns_type_operations net_ns_type_operations;
3251
3252 const char *netdev_drivername(const struct net_device *dev);
3253
3254 void linkwatch_run_queue(void);
3255
3256 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3257 netdev_features_t f2)
3258 {
3259 if (f1 & NETIF_F_GEN_CSUM)
3260 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3261 if (f2 & NETIF_F_GEN_CSUM)
3262 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3263 f1 &= f2;
3264 if (f1 & NETIF_F_GEN_CSUM)
3265 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3266
3267 return f1;
3268 }
3269
3270 static inline netdev_features_t netdev_get_wanted_features(
3271 struct net_device *dev)
3272 {
3273 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3274 }
3275 netdev_features_t netdev_increment_features(netdev_features_t all,
3276 netdev_features_t one, netdev_features_t mask);
3277
3278 /* Allow TSO being used on stacked device :
3279 * Performing the GSO segmentation before last device
3280 * is a performance improvement.
3281 */
3282 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3283 netdev_features_t mask)
3284 {
3285 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3286 }
3287
3288 int __netdev_update_features(struct net_device *dev);
3289 void netdev_update_features(struct net_device *dev);
3290 void netdev_change_features(struct net_device *dev);
3291
3292 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3293 struct net_device *dev);
3294
3295 netdev_features_t netif_skb_features(struct sk_buff *skb);
3296
3297 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3298 {
3299 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3300
3301 /* check flags correspondence */
3302 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3303 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3304 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3305 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3306 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3307 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3308
3309 return (features & feature) == feature;
3310 }
3311
3312 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3313 {
3314 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3315 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3316 }
3317
3318 static inline bool netif_needs_gso(struct sk_buff *skb,
3319 netdev_features_t features)
3320 {
3321 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3322 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3323 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3324 }
3325
3326 static inline void netif_set_gso_max_size(struct net_device *dev,
3327 unsigned int size)
3328 {
3329 dev->gso_max_size = size;
3330 }
3331
3332 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3333 int pulled_hlen, u16 mac_offset,
3334 int mac_len)
3335 {
3336 skb->protocol = protocol;
3337 skb->encapsulation = 1;
3338 skb_push(skb, pulled_hlen);
3339 skb_reset_transport_header(skb);
3340 skb->mac_header = mac_offset;
3341 skb->network_header = skb->mac_header + mac_len;
3342 skb->mac_len = mac_len;
3343 }
3344
3345 static inline bool netif_is_macvlan(struct net_device *dev)
3346 {
3347 return dev->priv_flags & IFF_MACVLAN;
3348 }
3349
3350 static inline bool netif_is_bond_master(struct net_device *dev)
3351 {
3352 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3353 }
3354
3355 static inline bool netif_is_bond_slave(struct net_device *dev)
3356 {
3357 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3358 }
3359
3360 static inline bool netif_supports_nofcs(struct net_device *dev)
3361 {
3362 return dev->priv_flags & IFF_SUPP_NOFCS;
3363 }
3364
3365 extern struct pernet_operations __net_initdata loopback_net_ops;
3366
3367 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3368
3369 /* netdev_printk helpers, similar to dev_printk */
3370
3371 static inline const char *netdev_name(const struct net_device *dev)
3372 {
3373 if (dev->reg_state != NETREG_REGISTERED)
3374 return "(unregistered net_device)";
3375 return dev->name;
3376 }
3377
3378 __printf(3, 4)
3379 int netdev_printk(const char *level, const struct net_device *dev,
3380 const char *format, ...);
3381 __printf(2, 3)
3382 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3383 __printf(2, 3)
3384 int netdev_alert(const struct net_device *dev, const char *format, ...);
3385 __printf(2, 3)
3386 int netdev_crit(const struct net_device *dev, const char *format, ...);
3387 __printf(2, 3)
3388 int netdev_err(const struct net_device *dev, const char *format, ...);
3389 __printf(2, 3)
3390 int netdev_warn(const struct net_device *dev, const char *format, ...);
3391 __printf(2, 3)
3392 int netdev_notice(const struct net_device *dev, const char *format, ...);
3393 __printf(2, 3)
3394 int netdev_info(const struct net_device *dev, const char *format, ...);
3395
3396 #define MODULE_ALIAS_NETDEV(device) \
3397 MODULE_ALIAS("netdev-" device)
3398
3399 #if defined(CONFIG_DYNAMIC_DEBUG)
3400 #define netdev_dbg(__dev, format, args...) \
3401 do { \
3402 dynamic_netdev_dbg(__dev, format, ##args); \
3403 } while (0)
3404 #elif defined(DEBUG)
3405 #define netdev_dbg(__dev, format, args...) \
3406 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3407 #else
3408 #define netdev_dbg(__dev, format, args...) \
3409 ({ \
3410 if (0) \
3411 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3412 0; \
3413 })
3414 #endif
3415
3416 #if defined(VERBOSE_DEBUG)
3417 #define netdev_vdbg netdev_dbg
3418 #else
3419
3420 #define netdev_vdbg(dev, format, args...) \
3421 ({ \
3422 if (0) \
3423 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3424 0; \
3425 })
3426 #endif
3427
3428 /*
3429 * netdev_WARN() acts like dev_printk(), but with the key difference
3430 * of using a WARN/WARN_ON to get the message out, including the
3431 * file/line information and a backtrace.
3432 */
3433 #define netdev_WARN(dev, format, args...) \
3434 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3435
3436 /* netif printk helpers, similar to netdev_printk */
3437
3438 #define netif_printk(priv, type, level, dev, fmt, args...) \
3439 do { \
3440 if (netif_msg_##type(priv)) \
3441 netdev_printk(level, (dev), fmt, ##args); \
3442 } while (0)
3443
3444 #define netif_level(level, priv, type, dev, fmt, args...) \
3445 do { \
3446 if (netif_msg_##type(priv)) \
3447 netdev_##level(dev, fmt, ##args); \
3448 } while (0)
3449
3450 #define netif_emerg(priv, type, dev, fmt, args...) \
3451 netif_level(emerg, priv, type, dev, fmt, ##args)
3452 #define netif_alert(priv, type, dev, fmt, args...) \
3453 netif_level(alert, priv, type, dev, fmt, ##args)
3454 #define netif_crit(priv, type, dev, fmt, args...) \
3455 netif_level(crit, priv, type, dev, fmt, ##args)
3456 #define netif_err(priv, type, dev, fmt, args...) \
3457 netif_level(err, priv, type, dev, fmt, ##args)
3458 #define netif_warn(priv, type, dev, fmt, args...) \
3459 netif_level(warn, priv, type, dev, fmt, ##args)
3460 #define netif_notice(priv, type, dev, fmt, args...) \
3461 netif_level(notice, priv, type, dev, fmt, ##args)
3462 #define netif_info(priv, type, dev, fmt, args...) \
3463 netif_level(info, priv, type, dev, fmt, ##args)
3464
3465 #if defined(CONFIG_DYNAMIC_DEBUG)
3466 #define netif_dbg(priv, type, netdev, format, args...) \
3467 do { \
3468 if (netif_msg_##type(priv)) \
3469 dynamic_netdev_dbg(netdev, format, ##args); \
3470 } while (0)
3471 #elif defined(DEBUG)
3472 #define netif_dbg(priv, type, dev, format, args...) \
3473 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3474 #else
3475 #define netif_dbg(priv, type, dev, format, args...) \
3476 ({ \
3477 if (0) \
3478 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3479 0; \
3480 })
3481 #endif
3482
3483 #if defined(VERBOSE_DEBUG)
3484 #define netif_vdbg netif_dbg
3485 #else
3486 #define netif_vdbg(priv, type, dev, format, args...) \
3487 ({ \
3488 if (0) \
3489 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3490 0; \
3491 })
3492 #endif
3493
3494 /*
3495 * The list of packet types we will receive (as opposed to discard)
3496 * and the routines to invoke.
3497 *
3498 * Why 16. Because with 16 the only overlap we get on a hash of the
3499 * low nibble of the protocol value is RARP/SNAP/X.25.
3500 *
3501 * NOTE: That is no longer true with the addition of VLAN tags. Not
3502 * sure which should go first, but I bet it won't make much
3503 * difference if we are running VLANs. The good news is that
3504 * this protocol won't be in the list unless compiled in, so
3505 * the average user (w/out VLANs) will not be adversely affected.
3506 * --BLG
3507 *
3508 * 0800 IP
3509 * 8100 802.1Q VLAN
3510 * 0001 802.3
3511 * 0002 AX.25
3512 * 0004 802.2
3513 * 8035 RARP
3514 * 0005 SNAP
3515 * 0805 X.25
3516 * 0806 ARP
3517 * 8137 IPX
3518 * 0009 Localtalk
3519 * 86DD IPv6
3520 */
3521 #define PTYPE_HASH_SIZE (16)
3522 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3523
3524 #endif /* _LINUX_NETDEVICE_H */
This page took 0.111143 seconds and 6 git commands to generate.