]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/netdevice.h
hyperv: Add hash value into RNDIS Per-packet info
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53
54 struct netpoll_info;
55 struct device;
56 struct phy_device;
57 /* 802.11 specific */
58 struct wireless_dev;
59
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_clear_bit();
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_tx_rate)(struct net_device *dev, int vf, int rate);
854 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
855 * int (*ndo_get_vf_config)(struct net_device *dev,
856 * int vf, struct ifla_vf_info *ivf);
857 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
858 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
859 * struct nlattr *port[]);
860 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
861 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
862 * Called to setup 'tc' number of traffic classes in the net device. This
863 * is always called from the stack with the rtnl lock held and netif tx
864 * queues stopped. This allows the netdevice to perform queue management
865 * safely.
866 *
867 * Fiber Channel over Ethernet (FCoE) offload functions.
868 * int (*ndo_fcoe_enable)(struct net_device *dev);
869 * Called when the FCoE protocol stack wants to start using LLD for FCoE
870 * so the underlying device can perform whatever needed configuration or
871 * initialization to support acceleration of FCoE traffic.
872 *
873 * int (*ndo_fcoe_disable)(struct net_device *dev);
874 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
875 * so the underlying device can perform whatever needed clean-ups to
876 * stop supporting acceleration of FCoE traffic.
877 *
878 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
879 * struct scatterlist *sgl, unsigned int sgc);
880 * Called when the FCoE Initiator wants to initialize an I/O that
881 * is a possible candidate for Direct Data Placement (DDP). The LLD can
882 * perform necessary setup and returns 1 to indicate the device is set up
883 * successfully to perform DDP on this I/O, otherwise this returns 0.
884 *
885 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
886 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
887 * indicated by the FC exchange id 'xid', so the underlying device can
888 * clean up and reuse resources for later DDP requests.
889 *
890 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
891 * struct scatterlist *sgl, unsigned int sgc);
892 * Called when the FCoE Target wants to initialize an I/O that
893 * is a possible candidate for Direct Data Placement (DDP). The LLD can
894 * perform necessary setup and returns 1 to indicate the device is set up
895 * successfully to perform DDP on this I/O, otherwise this returns 0.
896 *
897 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
898 * struct netdev_fcoe_hbainfo *hbainfo);
899 * Called when the FCoE Protocol stack wants information on the underlying
900 * device. This information is utilized by the FCoE protocol stack to
901 * register attributes with Fiber Channel management service as per the
902 * FC-GS Fabric Device Management Information(FDMI) specification.
903 *
904 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
905 * Called when the underlying device wants to override default World Wide
906 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
907 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
908 * protocol stack to use.
909 *
910 * RFS acceleration.
911 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
912 * u16 rxq_index, u32 flow_id);
913 * Set hardware filter for RFS. rxq_index is the target queue index;
914 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
915 * Return the filter ID on success, or a negative error code.
916 *
917 * Slave management functions (for bridge, bonding, etc).
918 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
919 * Called to make another netdev an underling.
920 *
921 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
922 * Called to release previously enslaved netdev.
923 *
924 * Feature/offload setting functions.
925 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
926 * netdev_features_t features);
927 * Adjusts the requested feature flags according to device-specific
928 * constraints, and returns the resulting flags. Must not modify
929 * the device state.
930 *
931 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
932 * Called to update device configuration to new features. Passed
933 * feature set might be less than what was returned by ndo_fix_features()).
934 * Must return >0 or -errno if it changed dev->features itself.
935 *
936 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
937 * struct net_device *dev,
938 * const unsigned char *addr, u16 flags)
939 * Adds an FDB entry to dev for addr.
940 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
941 * struct net_device *dev,
942 * const unsigned char *addr)
943 * Deletes the FDB entry from dev coresponding to addr.
944 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
945 * struct net_device *dev, int idx)
946 * Used to add FDB entries to dump requests. Implementers should add
947 * entries to skb and update idx with the number of entries.
948 *
949 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
950 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
951 * struct net_device *dev, u32 filter_mask)
952 *
953 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
954 * Called to change device carrier. Soft-devices (like dummy, team, etc)
955 * which do not represent real hardware may define this to allow their
956 * userspace components to manage their virtual carrier state. Devices
957 * that determine carrier state from physical hardware properties (eg
958 * network cables) or protocol-dependent mechanisms (eg
959 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
960 *
961 * int (*ndo_get_phys_port_id)(struct net_device *dev,
962 * struct netdev_phys_port_id *ppid);
963 * Called to get ID of physical port of this device. If driver does
964 * not implement this, it is assumed that the hw is not able to have
965 * multiple net devices on single physical port.
966 *
967 * void (*ndo_add_vxlan_port)(struct net_device *dev,
968 * sa_family_t sa_family, __be16 port);
969 * Called by vxlan to notiy a driver about the UDP port and socket
970 * address family that vxlan is listnening to. It is called only when
971 * a new port starts listening. The operation is protected by the
972 * vxlan_net->sock_lock.
973 *
974 * void (*ndo_del_vxlan_port)(struct net_device *dev,
975 * sa_family_t sa_family, __be16 port);
976 * Called by vxlan to notify the driver about a UDP port and socket
977 * address family that vxlan is not listening to anymore. The operation
978 * is protected by the vxlan_net->sock_lock.
979 *
980 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
981 * struct net_device *dev)
982 * Called by upper layer devices to accelerate switching or other
983 * station functionality into hardware. 'pdev is the lowerdev
984 * to use for the offload and 'dev' is the net device that will
985 * back the offload. Returns a pointer to the private structure
986 * the upper layer will maintain.
987 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
988 * Called by upper layer device to delete the station created
989 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
990 * the station and priv is the structure returned by the add
991 * operation.
992 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
993 * struct net_device *dev,
994 * void *priv);
995 * Callback to use for xmit over the accelerated station. This
996 * is used in place of ndo_start_xmit on accelerated net
997 * devices.
998 */
999 struct net_device_ops {
1000 int (*ndo_init)(struct net_device *dev);
1001 void (*ndo_uninit)(struct net_device *dev);
1002 int (*ndo_open)(struct net_device *dev);
1003 int (*ndo_stop)(struct net_device *dev);
1004 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1005 struct net_device *dev);
1006 u16 (*ndo_select_queue)(struct net_device *dev,
1007 struct sk_buff *skb,
1008 void *accel_priv,
1009 select_queue_fallback_t fallback);
1010 void (*ndo_change_rx_flags)(struct net_device *dev,
1011 int flags);
1012 void (*ndo_set_rx_mode)(struct net_device *dev);
1013 int (*ndo_set_mac_address)(struct net_device *dev,
1014 void *addr);
1015 int (*ndo_validate_addr)(struct net_device *dev);
1016 int (*ndo_do_ioctl)(struct net_device *dev,
1017 struct ifreq *ifr, int cmd);
1018 int (*ndo_set_config)(struct net_device *dev,
1019 struct ifmap *map);
1020 int (*ndo_change_mtu)(struct net_device *dev,
1021 int new_mtu);
1022 int (*ndo_neigh_setup)(struct net_device *dev,
1023 struct neigh_parms *);
1024 void (*ndo_tx_timeout) (struct net_device *dev);
1025
1026 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1027 struct rtnl_link_stats64 *storage);
1028 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1029
1030 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1031 __be16 proto, u16 vid);
1032 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1033 __be16 proto, u16 vid);
1034 #ifdef CONFIG_NET_POLL_CONTROLLER
1035 void (*ndo_poll_controller)(struct net_device *dev);
1036 int (*ndo_netpoll_setup)(struct net_device *dev,
1037 struct netpoll_info *info);
1038 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1039 #endif
1040 #ifdef CONFIG_NET_RX_BUSY_POLL
1041 int (*ndo_busy_poll)(struct napi_struct *dev);
1042 #endif
1043 int (*ndo_set_vf_mac)(struct net_device *dev,
1044 int queue, u8 *mac);
1045 int (*ndo_set_vf_vlan)(struct net_device *dev,
1046 int queue, u16 vlan, u8 qos);
1047 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
1048 int vf, int rate);
1049 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1050 int vf, bool setting);
1051 int (*ndo_get_vf_config)(struct net_device *dev,
1052 int vf,
1053 struct ifla_vf_info *ivf);
1054 int (*ndo_set_vf_link_state)(struct net_device *dev,
1055 int vf, int link_state);
1056 int (*ndo_set_vf_port)(struct net_device *dev,
1057 int vf,
1058 struct nlattr *port[]);
1059 int (*ndo_get_vf_port)(struct net_device *dev,
1060 int vf, struct sk_buff *skb);
1061 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1062 #if IS_ENABLED(CONFIG_FCOE)
1063 int (*ndo_fcoe_enable)(struct net_device *dev);
1064 int (*ndo_fcoe_disable)(struct net_device *dev);
1065 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1066 u16 xid,
1067 struct scatterlist *sgl,
1068 unsigned int sgc);
1069 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1070 u16 xid);
1071 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1072 u16 xid,
1073 struct scatterlist *sgl,
1074 unsigned int sgc);
1075 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1076 struct netdev_fcoe_hbainfo *hbainfo);
1077 #endif
1078
1079 #if IS_ENABLED(CONFIG_LIBFCOE)
1080 #define NETDEV_FCOE_WWNN 0
1081 #define NETDEV_FCOE_WWPN 1
1082 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1083 u64 *wwn, int type);
1084 #endif
1085
1086 #ifdef CONFIG_RFS_ACCEL
1087 int (*ndo_rx_flow_steer)(struct net_device *dev,
1088 const struct sk_buff *skb,
1089 u16 rxq_index,
1090 u32 flow_id);
1091 #endif
1092 int (*ndo_add_slave)(struct net_device *dev,
1093 struct net_device *slave_dev);
1094 int (*ndo_del_slave)(struct net_device *dev,
1095 struct net_device *slave_dev);
1096 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1097 netdev_features_t features);
1098 int (*ndo_set_features)(struct net_device *dev,
1099 netdev_features_t features);
1100 int (*ndo_neigh_construct)(struct neighbour *n);
1101 void (*ndo_neigh_destroy)(struct neighbour *n);
1102
1103 int (*ndo_fdb_add)(struct ndmsg *ndm,
1104 struct nlattr *tb[],
1105 struct net_device *dev,
1106 const unsigned char *addr,
1107 u16 flags);
1108 int (*ndo_fdb_del)(struct ndmsg *ndm,
1109 struct nlattr *tb[],
1110 struct net_device *dev,
1111 const unsigned char *addr);
1112 int (*ndo_fdb_dump)(struct sk_buff *skb,
1113 struct netlink_callback *cb,
1114 struct net_device *dev,
1115 int idx);
1116
1117 int (*ndo_bridge_setlink)(struct net_device *dev,
1118 struct nlmsghdr *nlh);
1119 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1120 u32 pid, u32 seq,
1121 struct net_device *dev,
1122 u32 filter_mask);
1123 int (*ndo_bridge_dellink)(struct net_device *dev,
1124 struct nlmsghdr *nlh);
1125 int (*ndo_change_carrier)(struct net_device *dev,
1126 bool new_carrier);
1127 int (*ndo_get_phys_port_id)(struct net_device *dev,
1128 struct netdev_phys_port_id *ppid);
1129 void (*ndo_add_vxlan_port)(struct net_device *dev,
1130 sa_family_t sa_family,
1131 __be16 port);
1132 void (*ndo_del_vxlan_port)(struct net_device *dev,
1133 sa_family_t sa_family,
1134 __be16 port);
1135
1136 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1137 struct net_device *dev);
1138 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1139 void *priv);
1140
1141 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1142 struct net_device *dev,
1143 void *priv);
1144 };
1145
1146 /**
1147 * enum net_device_priv_flags - &struct net_device priv_flags
1148 *
1149 * These are the &struct net_device, they are only set internally
1150 * by drivers and used in the kernel. These flags are invisible to
1151 * userspace, this means that the order of these flags can change
1152 * during any kernel release.
1153 *
1154 * You should have a pretty good reason to be extending these flags.
1155 *
1156 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1157 * @IFF_EBRIDGE: Ethernet bridging device
1158 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1159 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1160 * @IFF_MASTER_ALB: bonding master, balance-alb
1161 * @IFF_BONDING: bonding master or slave
1162 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1163 * @IFF_ISATAP: ISATAP interface (RFC4214)
1164 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1165 * @IFF_WAN_HDLC: WAN HDLC device
1166 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1167 * release skb->dst
1168 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1169 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1170 * @IFF_MACVLAN_PORT: device used as macvlan port
1171 * @IFF_BRIDGE_PORT: device used as bridge port
1172 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1173 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1174 * @IFF_UNICAST_FLT: Supports unicast filtering
1175 * @IFF_TEAM_PORT: device used as team port
1176 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1177 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1178 * change when it's running
1179 * @IFF_MACVLAN: Macvlan device
1180 */
1181 enum netdev_priv_flags {
1182 IFF_802_1Q_VLAN = 1<<0,
1183 IFF_EBRIDGE = 1<<1,
1184 IFF_SLAVE_INACTIVE = 1<<2,
1185 IFF_MASTER_8023AD = 1<<3,
1186 IFF_MASTER_ALB = 1<<4,
1187 IFF_BONDING = 1<<5,
1188 IFF_SLAVE_NEEDARP = 1<<6,
1189 IFF_ISATAP = 1<<7,
1190 IFF_MASTER_ARPMON = 1<<8,
1191 IFF_WAN_HDLC = 1<<9,
1192 IFF_XMIT_DST_RELEASE = 1<<10,
1193 IFF_DONT_BRIDGE = 1<<11,
1194 IFF_DISABLE_NETPOLL = 1<<12,
1195 IFF_MACVLAN_PORT = 1<<13,
1196 IFF_BRIDGE_PORT = 1<<14,
1197 IFF_OVS_DATAPATH = 1<<15,
1198 IFF_TX_SKB_SHARING = 1<<16,
1199 IFF_UNICAST_FLT = 1<<17,
1200 IFF_TEAM_PORT = 1<<18,
1201 IFF_SUPP_NOFCS = 1<<19,
1202 IFF_LIVE_ADDR_CHANGE = 1<<20,
1203 IFF_MACVLAN = 1<<21,
1204 };
1205
1206 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1207 #define IFF_EBRIDGE IFF_EBRIDGE
1208 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1209 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1210 #define IFF_MASTER_ALB IFF_MASTER_ALB
1211 #define IFF_BONDING IFF_BONDING
1212 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1213 #define IFF_ISATAP IFF_ISATAP
1214 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1215 #define IFF_WAN_HDLC IFF_WAN_HDLC
1216 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1217 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1218 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1219 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1220 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1221 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1222 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1223 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1224 #define IFF_TEAM_PORT IFF_TEAM_PORT
1225 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1226 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1227 #define IFF_MACVLAN IFF_MACVLAN
1228
1229 /*
1230 * The DEVICE structure.
1231 * Actually, this whole structure is a big mistake. It mixes I/O
1232 * data with strictly "high-level" data, and it has to know about
1233 * almost every data structure used in the INET module.
1234 *
1235 * FIXME: cleanup struct net_device such that network protocol info
1236 * moves out.
1237 */
1238
1239 struct net_device {
1240
1241 /*
1242 * This is the first field of the "visible" part of this structure
1243 * (i.e. as seen by users in the "Space.c" file). It is the name
1244 * of the interface.
1245 */
1246 char name[IFNAMSIZ];
1247
1248 /* device name hash chain, please keep it close to name[] */
1249 struct hlist_node name_hlist;
1250
1251 /* snmp alias */
1252 char *ifalias;
1253
1254 /*
1255 * I/O specific fields
1256 * FIXME: Merge these and struct ifmap into one
1257 */
1258 unsigned long mem_end; /* shared mem end */
1259 unsigned long mem_start; /* shared mem start */
1260 unsigned long base_addr; /* device I/O address */
1261 int irq; /* device IRQ number */
1262
1263 /*
1264 * Some hardware also needs these fields, but they are not
1265 * part of the usual set specified in Space.c.
1266 */
1267
1268 unsigned long state;
1269
1270 struct list_head dev_list;
1271 struct list_head napi_list;
1272 struct list_head unreg_list;
1273 struct list_head close_list;
1274
1275 /* directly linked devices, like slaves for bonding */
1276 struct {
1277 struct list_head upper;
1278 struct list_head lower;
1279 } adj_list;
1280
1281 /* all linked devices, *including* neighbours */
1282 struct {
1283 struct list_head upper;
1284 struct list_head lower;
1285 } all_adj_list;
1286
1287
1288 /* currently active device features */
1289 netdev_features_t features;
1290 /* user-changeable features */
1291 netdev_features_t hw_features;
1292 /* user-requested features */
1293 netdev_features_t wanted_features;
1294 /* mask of features inheritable by VLAN devices */
1295 netdev_features_t vlan_features;
1296 /* mask of features inherited by encapsulating devices
1297 * This field indicates what encapsulation offloads
1298 * the hardware is capable of doing, and drivers will
1299 * need to set them appropriately.
1300 */
1301 netdev_features_t hw_enc_features;
1302 /* mask of fetures inheritable by MPLS */
1303 netdev_features_t mpls_features;
1304
1305 /* Interface index. Unique device identifier */
1306 int ifindex;
1307 int iflink;
1308
1309 struct net_device_stats stats;
1310
1311 /* dropped packets by core network, Do not use this in drivers */
1312 atomic_long_t rx_dropped;
1313 atomic_long_t tx_dropped;
1314
1315 /* Stats to monitor carrier on<->off transitions */
1316 atomic_t carrier_changes;
1317
1318 #ifdef CONFIG_WIRELESS_EXT
1319 /* List of functions to handle Wireless Extensions (instead of ioctl).
1320 * See <net/iw_handler.h> for details. Jean II */
1321 const struct iw_handler_def * wireless_handlers;
1322 /* Instance data managed by the core of Wireless Extensions. */
1323 struct iw_public_data * wireless_data;
1324 #endif
1325 /* Management operations */
1326 const struct net_device_ops *netdev_ops;
1327 const struct ethtool_ops *ethtool_ops;
1328 const struct forwarding_accel_ops *fwd_ops;
1329
1330 /* Hardware header description */
1331 const struct header_ops *header_ops;
1332
1333 unsigned int flags; /* interface flags (a la BSD) */
1334 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1335 * See if.h for definitions. */
1336 unsigned short gflags;
1337 unsigned short padded; /* How much padding added by alloc_netdev() */
1338
1339 unsigned char operstate; /* RFC2863 operstate */
1340 unsigned char link_mode; /* mapping policy to operstate */
1341
1342 unsigned char if_port; /* Selectable AUI, TP,..*/
1343 unsigned char dma; /* DMA channel */
1344
1345 unsigned int mtu; /* interface MTU value */
1346 unsigned short type; /* interface hardware type */
1347 unsigned short hard_header_len; /* hardware hdr length */
1348
1349 /* extra head- and tailroom the hardware may need, but not in all cases
1350 * can this be guaranteed, especially tailroom. Some cases also use
1351 * LL_MAX_HEADER instead to allocate the skb.
1352 */
1353 unsigned short needed_headroom;
1354 unsigned short needed_tailroom;
1355
1356 /* Interface address info. */
1357 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1358 unsigned char addr_assign_type; /* hw address assignment type */
1359 unsigned char addr_len; /* hardware address length */
1360 unsigned short neigh_priv_len;
1361 unsigned short dev_id; /* Used to differentiate devices
1362 * that share the same link
1363 * layer address
1364 */
1365 unsigned short dev_port; /* Used to differentiate
1366 * devices that share the same
1367 * function
1368 */
1369 spinlock_t addr_list_lock;
1370 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1371 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1372 struct netdev_hw_addr_list dev_addrs; /* list of device
1373 * hw addresses
1374 */
1375 #ifdef CONFIG_SYSFS
1376 struct kset *queues_kset;
1377 #endif
1378
1379 bool uc_promisc;
1380 unsigned int promiscuity;
1381 unsigned int allmulti;
1382
1383
1384 /* Protocol specific pointers */
1385
1386 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1387 struct vlan_info __rcu *vlan_info; /* VLAN info */
1388 #endif
1389 #if IS_ENABLED(CONFIG_NET_DSA)
1390 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1391 #endif
1392 #if IS_ENABLED(CONFIG_TIPC)
1393 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1394 #endif
1395 void *atalk_ptr; /* AppleTalk link */
1396 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1397 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1398 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1399 void *ax25_ptr; /* AX.25 specific data */
1400 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1401 assign before registering */
1402
1403 /*
1404 * Cache lines mostly used on receive path (including eth_type_trans())
1405 */
1406 unsigned long last_rx; /* Time of last Rx */
1407
1408 /* Interface address info used in eth_type_trans() */
1409 unsigned char *dev_addr; /* hw address, (before bcast
1410 because most packets are
1411 unicast) */
1412
1413
1414 #ifdef CONFIG_SYSFS
1415 struct netdev_rx_queue *_rx;
1416
1417 /* Number of RX queues allocated at register_netdev() time */
1418 unsigned int num_rx_queues;
1419
1420 /* Number of RX queues currently active in device */
1421 unsigned int real_num_rx_queues;
1422
1423 #endif
1424
1425 rx_handler_func_t __rcu *rx_handler;
1426 void __rcu *rx_handler_data;
1427
1428 struct netdev_queue __rcu *ingress_queue;
1429 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1430
1431
1432 /*
1433 * Cache lines mostly used on transmit path
1434 */
1435 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1436
1437 /* Number of TX queues allocated at alloc_netdev_mq() time */
1438 unsigned int num_tx_queues;
1439
1440 /* Number of TX queues currently active in device */
1441 unsigned int real_num_tx_queues;
1442
1443 /* root qdisc from userspace point of view */
1444 struct Qdisc *qdisc;
1445
1446 unsigned long tx_queue_len; /* Max frames per queue allowed */
1447 spinlock_t tx_global_lock;
1448
1449 #ifdef CONFIG_XPS
1450 struct xps_dev_maps __rcu *xps_maps;
1451 #endif
1452 #ifdef CONFIG_RFS_ACCEL
1453 /* CPU reverse-mapping for RX completion interrupts, indexed
1454 * by RX queue number. Assigned by driver. This must only be
1455 * set if the ndo_rx_flow_steer operation is defined. */
1456 struct cpu_rmap *rx_cpu_rmap;
1457 #endif
1458
1459 /* These may be needed for future network-power-down code. */
1460
1461 /*
1462 * trans_start here is expensive for high speed devices on SMP,
1463 * please use netdev_queue->trans_start instead.
1464 */
1465 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1466
1467 int watchdog_timeo; /* used by dev_watchdog() */
1468 struct timer_list watchdog_timer;
1469
1470 /* Number of references to this device */
1471 int __percpu *pcpu_refcnt;
1472
1473 /* delayed register/unregister */
1474 struct list_head todo_list;
1475 /* device index hash chain */
1476 struct hlist_node index_hlist;
1477
1478 struct list_head link_watch_list;
1479
1480 /* register/unregister state machine */
1481 enum { NETREG_UNINITIALIZED=0,
1482 NETREG_REGISTERED, /* completed register_netdevice */
1483 NETREG_UNREGISTERING, /* called unregister_netdevice */
1484 NETREG_UNREGISTERED, /* completed unregister todo */
1485 NETREG_RELEASED, /* called free_netdev */
1486 NETREG_DUMMY, /* dummy device for NAPI poll */
1487 } reg_state:8;
1488
1489 bool dismantle; /* device is going do be freed */
1490
1491 enum {
1492 RTNL_LINK_INITIALIZED,
1493 RTNL_LINK_INITIALIZING,
1494 } rtnl_link_state:16;
1495
1496 /* Called from unregister, can be used to call free_netdev */
1497 void (*destructor)(struct net_device *dev);
1498
1499 #ifdef CONFIG_NETPOLL
1500 struct netpoll_info __rcu *npinfo;
1501 #endif
1502
1503 #ifdef CONFIG_NET_NS
1504 /* Network namespace this network device is inside */
1505 struct net *nd_net;
1506 #endif
1507
1508 /* mid-layer private */
1509 union {
1510 void *ml_priv;
1511 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1512 struct pcpu_sw_netstats __percpu *tstats;
1513 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1514 struct pcpu_vstats __percpu *vstats; /* veth stats */
1515 };
1516 /* GARP */
1517 struct garp_port __rcu *garp_port;
1518 /* MRP */
1519 struct mrp_port __rcu *mrp_port;
1520
1521 /* class/net/name entry */
1522 struct device dev;
1523 /* space for optional device, statistics, and wireless sysfs groups */
1524 const struct attribute_group *sysfs_groups[4];
1525 /* space for optional per-rx queue attributes */
1526 const struct attribute_group *sysfs_rx_queue_group;
1527
1528 /* rtnetlink link ops */
1529 const struct rtnl_link_ops *rtnl_link_ops;
1530
1531 /* for setting kernel sock attribute on TCP connection setup */
1532 #define GSO_MAX_SIZE 65536
1533 unsigned int gso_max_size;
1534 #define GSO_MAX_SEGS 65535
1535 u16 gso_max_segs;
1536
1537 #ifdef CONFIG_DCB
1538 /* Data Center Bridging netlink ops */
1539 const struct dcbnl_rtnl_ops *dcbnl_ops;
1540 #endif
1541 u8 num_tc;
1542 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1543 u8 prio_tc_map[TC_BITMASK + 1];
1544
1545 #if IS_ENABLED(CONFIG_FCOE)
1546 /* max exchange id for FCoE LRO by ddp */
1547 unsigned int fcoe_ddp_xid;
1548 #endif
1549 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1550 struct netprio_map __rcu *priomap;
1551 #endif
1552 /* phy device may attach itself for hardware timestamping */
1553 struct phy_device *phydev;
1554
1555 struct lock_class_key *qdisc_tx_busylock;
1556
1557 /* group the device belongs to */
1558 int group;
1559
1560 struct pm_qos_request pm_qos_req;
1561 };
1562 #define to_net_dev(d) container_of(d, struct net_device, dev)
1563
1564 #define NETDEV_ALIGN 32
1565
1566 static inline
1567 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1568 {
1569 return dev->prio_tc_map[prio & TC_BITMASK];
1570 }
1571
1572 static inline
1573 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1574 {
1575 if (tc >= dev->num_tc)
1576 return -EINVAL;
1577
1578 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1579 return 0;
1580 }
1581
1582 static inline
1583 void netdev_reset_tc(struct net_device *dev)
1584 {
1585 dev->num_tc = 0;
1586 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1587 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1588 }
1589
1590 static inline
1591 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1592 {
1593 if (tc >= dev->num_tc)
1594 return -EINVAL;
1595
1596 dev->tc_to_txq[tc].count = count;
1597 dev->tc_to_txq[tc].offset = offset;
1598 return 0;
1599 }
1600
1601 static inline
1602 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1603 {
1604 if (num_tc > TC_MAX_QUEUE)
1605 return -EINVAL;
1606
1607 dev->num_tc = num_tc;
1608 return 0;
1609 }
1610
1611 static inline
1612 int netdev_get_num_tc(struct net_device *dev)
1613 {
1614 return dev->num_tc;
1615 }
1616
1617 static inline
1618 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1619 unsigned int index)
1620 {
1621 return &dev->_tx[index];
1622 }
1623
1624 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1625 void (*f)(struct net_device *,
1626 struct netdev_queue *,
1627 void *),
1628 void *arg)
1629 {
1630 unsigned int i;
1631
1632 for (i = 0; i < dev->num_tx_queues; i++)
1633 f(dev, &dev->_tx[i], arg);
1634 }
1635
1636 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1637 struct sk_buff *skb,
1638 void *accel_priv);
1639
1640 /*
1641 * Net namespace inlines
1642 */
1643 static inline
1644 struct net *dev_net(const struct net_device *dev)
1645 {
1646 return read_pnet(&dev->nd_net);
1647 }
1648
1649 static inline
1650 void dev_net_set(struct net_device *dev, struct net *net)
1651 {
1652 #ifdef CONFIG_NET_NS
1653 release_net(dev->nd_net);
1654 dev->nd_net = hold_net(net);
1655 #endif
1656 }
1657
1658 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1659 {
1660 #ifdef CONFIG_NET_DSA_TAG_DSA
1661 if (dev->dsa_ptr != NULL)
1662 return dsa_uses_dsa_tags(dev->dsa_ptr);
1663 #endif
1664
1665 return 0;
1666 }
1667
1668 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1669 {
1670 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1671 if (dev->dsa_ptr != NULL)
1672 return dsa_uses_trailer_tags(dev->dsa_ptr);
1673 #endif
1674
1675 return 0;
1676 }
1677
1678 /**
1679 * netdev_priv - access network device private data
1680 * @dev: network device
1681 *
1682 * Get network device private data
1683 */
1684 static inline void *netdev_priv(const struct net_device *dev)
1685 {
1686 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1687 }
1688
1689 /* Set the sysfs physical device reference for the network logical device
1690 * if set prior to registration will cause a symlink during initialization.
1691 */
1692 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1693
1694 /* Set the sysfs device type for the network logical device to allow
1695 * fine-grained identification of different network device types. For
1696 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1697 */
1698 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1699
1700 /* Default NAPI poll() weight
1701 * Device drivers are strongly advised to not use bigger value
1702 */
1703 #define NAPI_POLL_WEIGHT 64
1704
1705 /**
1706 * netif_napi_add - initialize a napi context
1707 * @dev: network device
1708 * @napi: napi context
1709 * @poll: polling function
1710 * @weight: default weight
1711 *
1712 * netif_napi_add() must be used to initialize a napi context prior to calling
1713 * *any* of the other napi related functions.
1714 */
1715 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1716 int (*poll)(struct napi_struct *, int), int weight);
1717
1718 /**
1719 * netif_napi_del - remove a napi context
1720 * @napi: napi context
1721 *
1722 * netif_napi_del() removes a napi context from the network device napi list
1723 */
1724 void netif_napi_del(struct napi_struct *napi);
1725
1726 struct napi_gro_cb {
1727 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1728 void *frag0;
1729
1730 /* Length of frag0. */
1731 unsigned int frag0_len;
1732
1733 /* This indicates where we are processing relative to skb->data. */
1734 int data_offset;
1735
1736 /* This is non-zero if the packet cannot be merged with the new skb. */
1737 u16 flush;
1738
1739 /* Save the IP ID here and check when we get to the transport layer */
1740 u16 flush_id;
1741
1742 /* Number of segments aggregated. */
1743 u16 count;
1744
1745 /* This is non-zero if the packet may be of the same flow. */
1746 u8 same_flow;
1747
1748 /* Free the skb? */
1749 u8 free;
1750 #define NAPI_GRO_FREE 1
1751 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1752
1753 /* jiffies when first packet was created/queued */
1754 unsigned long age;
1755
1756 /* Used in ipv6_gro_receive() */
1757 u16 proto;
1758
1759 /* Used in udp_gro_receive */
1760 u16 udp_mark;
1761
1762 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1763 __wsum csum;
1764
1765 /* used in skb_gro_receive() slow path */
1766 struct sk_buff *last;
1767 };
1768
1769 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1770
1771 struct packet_type {
1772 __be16 type; /* This is really htons(ether_type). */
1773 struct net_device *dev; /* NULL is wildcarded here */
1774 int (*func) (struct sk_buff *,
1775 struct net_device *,
1776 struct packet_type *,
1777 struct net_device *);
1778 bool (*id_match)(struct packet_type *ptype,
1779 struct sock *sk);
1780 void *af_packet_priv;
1781 struct list_head list;
1782 };
1783
1784 struct offload_callbacks {
1785 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1786 netdev_features_t features);
1787 int (*gso_send_check)(struct sk_buff *skb);
1788 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1789 struct sk_buff *skb);
1790 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1791 };
1792
1793 struct packet_offload {
1794 __be16 type; /* This is really htons(ether_type). */
1795 struct offload_callbacks callbacks;
1796 struct list_head list;
1797 };
1798
1799 struct udp_offload {
1800 __be16 port;
1801 struct offload_callbacks callbacks;
1802 };
1803
1804 /* often modified stats are per cpu, other are shared (netdev->stats) */
1805 struct pcpu_sw_netstats {
1806 u64 rx_packets;
1807 u64 rx_bytes;
1808 u64 tx_packets;
1809 u64 tx_bytes;
1810 struct u64_stats_sync syncp;
1811 };
1812
1813 #define netdev_alloc_pcpu_stats(type) \
1814 ({ \
1815 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1816 if (pcpu_stats) { \
1817 int i; \
1818 for_each_possible_cpu(i) { \
1819 typeof(type) *stat; \
1820 stat = per_cpu_ptr(pcpu_stats, i); \
1821 u64_stats_init(&stat->syncp); \
1822 } \
1823 } \
1824 pcpu_stats; \
1825 })
1826
1827 #include <linux/notifier.h>
1828
1829 /* netdevice notifier chain. Please remember to update the rtnetlink
1830 * notification exclusion list in rtnetlink_event() when adding new
1831 * types.
1832 */
1833 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1834 #define NETDEV_DOWN 0x0002
1835 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1836 detected a hardware crash and restarted
1837 - we can use this eg to kick tcp sessions
1838 once done */
1839 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1840 #define NETDEV_REGISTER 0x0005
1841 #define NETDEV_UNREGISTER 0x0006
1842 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1843 #define NETDEV_CHANGEADDR 0x0008
1844 #define NETDEV_GOING_DOWN 0x0009
1845 #define NETDEV_CHANGENAME 0x000A
1846 #define NETDEV_FEAT_CHANGE 0x000B
1847 #define NETDEV_BONDING_FAILOVER 0x000C
1848 #define NETDEV_PRE_UP 0x000D
1849 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1850 #define NETDEV_POST_TYPE_CHANGE 0x000F
1851 #define NETDEV_POST_INIT 0x0010
1852 #define NETDEV_UNREGISTER_FINAL 0x0011
1853 #define NETDEV_RELEASE 0x0012
1854 #define NETDEV_NOTIFY_PEERS 0x0013
1855 #define NETDEV_JOIN 0x0014
1856 #define NETDEV_CHANGEUPPER 0x0015
1857 #define NETDEV_RESEND_IGMP 0x0016
1858 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1859
1860 int register_netdevice_notifier(struct notifier_block *nb);
1861 int unregister_netdevice_notifier(struct notifier_block *nb);
1862
1863 struct netdev_notifier_info {
1864 struct net_device *dev;
1865 };
1866
1867 struct netdev_notifier_change_info {
1868 struct netdev_notifier_info info; /* must be first */
1869 unsigned int flags_changed;
1870 };
1871
1872 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1873 struct net_device *dev)
1874 {
1875 info->dev = dev;
1876 }
1877
1878 static inline struct net_device *
1879 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1880 {
1881 return info->dev;
1882 }
1883
1884 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1885
1886
1887 extern rwlock_t dev_base_lock; /* Device list lock */
1888
1889 #define for_each_netdev(net, d) \
1890 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1891 #define for_each_netdev_reverse(net, d) \
1892 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1893 #define for_each_netdev_rcu(net, d) \
1894 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1895 #define for_each_netdev_safe(net, d, n) \
1896 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1897 #define for_each_netdev_continue(net, d) \
1898 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1899 #define for_each_netdev_continue_rcu(net, d) \
1900 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1901 #define for_each_netdev_in_bond_rcu(bond, slave) \
1902 for_each_netdev_rcu(&init_net, slave) \
1903 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1904 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1905
1906 static inline struct net_device *next_net_device(struct net_device *dev)
1907 {
1908 struct list_head *lh;
1909 struct net *net;
1910
1911 net = dev_net(dev);
1912 lh = dev->dev_list.next;
1913 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1914 }
1915
1916 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1917 {
1918 struct list_head *lh;
1919 struct net *net;
1920
1921 net = dev_net(dev);
1922 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1923 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1924 }
1925
1926 static inline struct net_device *first_net_device(struct net *net)
1927 {
1928 return list_empty(&net->dev_base_head) ? NULL :
1929 net_device_entry(net->dev_base_head.next);
1930 }
1931
1932 static inline struct net_device *first_net_device_rcu(struct net *net)
1933 {
1934 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1935
1936 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1937 }
1938
1939 int netdev_boot_setup_check(struct net_device *dev);
1940 unsigned long netdev_boot_base(const char *prefix, int unit);
1941 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1942 const char *hwaddr);
1943 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1944 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1945 void dev_add_pack(struct packet_type *pt);
1946 void dev_remove_pack(struct packet_type *pt);
1947 void __dev_remove_pack(struct packet_type *pt);
1948 void dev_add_offload(struct packet_offload *po);
1949 void dev_remove_offload(struct packet_offload *po);
1950
1951 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1952 unsigned short mask);
1953 struct net_device *dev_get_by_name(struct net *net, const char *name);
1954 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1955 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1956 int dev_alloc_name(struct net_device *dev, const char *name);
1957 int dev_open(struct net_device *dev);
1958 int dev_close(struct net_device *dev);
1959 void dev_disable_lro(struct net_device *dev);
1960 int dev_loopback_xmit(struct sk_buff *newskb);
1961 int dev_queue_xmit(struct sk_buff *skb);
1962 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1963 int register_netdevice(struct net_device *dev);
1964 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1965 void unregister_netdevice_many(struct list_head *head);
1966 static inline void unregister_netdevice(struct net_device *dev)
1967 {
1968 unregister_netdevice_queue(dev, NULL);
1969 }
1970
1971 int netdev_refcnt_read(const struct net_device *dev);
1972 void free_netdev(struct net_device *dev);
1973 void netdev_freemem(struct net_device *dev);
1974 void synchronize_net(void);
1975 int init_dummy_netdev(struct net_device *dev);
1976
1977 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1978 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1979 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1980 int netdev_get_name(struct net *net, char *name, int ifindex);
1981 int dev_restart(struct net_device *dev);
1982 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1983
1984 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1985 {
1986 return NAPI_GRO_CB(skb)->data_offset;
1987 }
1988
1989 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1990 {
1991 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1992 }
1993
1994 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1995 {
1996 NAPI_GRO_CB(skb)->data_offset += len;
1997 }
1998
1999 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2000 unsigned int offset)
2001 {
2002 return NAPI_GRO_CB(skb)->frag0 + offset;
2003 }
2004
2005 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2006 {
2007 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2008 }
2009
2010 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2011 unsigned int offset)
2012 {
2013 if (!pskb_may_pull(skb, hlen))
2014 return NULL;
2015
2016 NAPI_GRO_CB(skb)->frag0 = NULL;
2017 NAPI_GRO_CB(skb)->frag0_len = 0;
2018 return skb->data + offset;
2019 }
2020
2021 static inline void *skb_gro_network_header(struct sk_buff *skb)
2022 {
2023 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2024 skb_network_offset(skb);
2025 }
2026
2027 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2028 const void *start, unsigned int len)
2029 {
2030 if (skb->ip_summed == CHECKSUM_COMPLETE)
2031 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2032 csum_partial(start, len, 0));
2033 }
2034
2035 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2036 unsigned short type,
2037 const void *daddr, const void *saddr,
2038 unsigned int len)
2039 {
2040 if (!dev->header_ops || !dev->header_ops->create)
2041 return 0;
2042
2043 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2044 }
2045
2046 static inline int dev_parse_header(const struct sk_buff *skb,
2047 unsigned char *haddr)
2048 {
2049 const struct net_device *dev = skb->dev;
2050
2051 if (!dev->header_ops || !dev->header_ops->parse)
2052 return 0;
2053 return dev->header_ops->parse(skb, haddr);
2054 }
2055
2056 static inline int dev_rebuild_header(struct sk_buff *skb)
2057 {
2058 const struct net_device *dev = skb->dev;
2059
2060 if (!dev->header_ops || !dev->header_ops->rebuild)
2061 return 0;
2062 return dev->header_ops->rebuild(skb);
2063 }
2064
2065 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2066 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2067 static inline int unregister_gifconf(unsigned int family)
2068 {
2069 return register_gifconf(family, NULL);
2070 }
2071
2072 #ifdef CONFIG_NET_FLOW_LIMIT
2073 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2074 struct sd_flow_limit {
2075 u64 count;
2076 unsigned int num_buckets;
2077 unsigned int history_head;
2078 u16 history[FLOW_LIMIT_HISTORY];
2079 u8 buckets[];
2080 };
2081
2082 extern int netdev_flow_limit_table_len;
2083 #endif /* CONFIG_NET_FLOW_LIMIT */
2084
2085 /*
2086 * Incoming packets are placed on per-cpu queues
2087 */
2088 struct softnet_data {
2089 struct Qdisc *output_queue;
2090 struct Qdisc **output_queue_tailp;
2091 struct list_head poll_list;
2092 struct sk_buff *completion_queue;
2093 struct sk_buff_head process_queue;
2094
2095 /* stats */
2096 unsigned int processed;
2097 unsigned int time_squeeze;
2098 unsigned int cpu_collision;
2099 unsigned int received_rps;
2100
2101 #ifdef CONFIG_RPS
2102 struct softnet_data *rps_ipi_list;
2103
2104 /* Elements below can be accessed between CPUs for RPS */
2105 struct call_single_data csd ____cacheline_aligned_in_smp;
2106 struct softnet_data *rps_ipi_next;
2107 unsigned int cpu;
2108 unsigned int input_queue_head;
2109 unsigned int input_queue_tail;
2110 #endif
2111 unsigned int dropped;
2112 struct sk_buff_head input_pkt_queue;
2113 struct napi_struct backlog;
2114
2115 #ifdef CONFIG_NET_FLOW_LIMIT
2116 struct sd_flow_limit __rcu *flow_limit;
2117 #endif
2118 };
2119
2120 static inline void input_queue_head_incr(struct softnet_data *sd)
2121 {
2122 #ifdef CONFIG_RPS
2123 sd->input_queue_head++;
2124 #endif
2125 }
2126
2127 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2128 unsigned int *qtail)
2129 {
2130 #ifdef CONFIG_RPS
2131 *qtail = ++sd->input_queue_tail;
2132 #endif
2133 }
2134
2135 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2136
2137 void __netif_schedule(struct Qdisc *q);
2138
2139 static inline void netif_schedule_queue(struct netdev_queue *txq)
2140 {
2141 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2142 __netif_schedule(txq->qdisc);
2143 }
2144
2145 static inline void netif_tx_schedule_all(struct net_device *dev)
2146 {
2147 unsigned int i;
2148
2149 for (i = 0; i < dev->num_tx_queues; i++)
2150 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2151 }
2152
2153 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2154 {
2155 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2156 }
2157
2158 /**
2159 * netif_start_queue - allow transmit
2160 * @dev: network device
2161 *
2162 * Allow upper layers to call the device hard_start_xmit routine.
2163 */
2164 static inline void netif_start_queue(struct net_device *dev)
2165 {
2166 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2167 }
2168
2169 static inline void netif_tx_start_all_queues(struct net_device *dev)
2170 {
2171 unsigned int i;
2172
2173 for (i = 0; i < dev->num_tx_queues; i++) {
2174 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2175 netif_tx_start_queue(txq);
2176 }
2177 }
2178
2179 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2180 {
2181 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2182 __netif_schedule(dev_queue->qdisc);
2183 }
2184
2185 /**
2186 * netif_wake_queue - restart transmit
2187 * @dev: network device
2188 *
2189 * Allow upper layers to call the device hard_start_xmit routine.
2190 * Used for flow control when transmit resources are available.
2191 */
2192 static inline void netif_wake_queue(struct net_device *dev)
2193 {
2194 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2195 }
2196
2197 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2198 {
2199 unsigned int i;
2200
2201 for (i = 0; i < dev->num_tx_queues; i++) {
2202 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2203 netif_tx_wake_queue(txq);
2204 }
2205 }
2206
2207 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2208 {
2209 if (WARN_ON(!dev_queue)) {
2210 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2211 return;
2212 }
2213 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2214 }
2215
2216 /**
2217 * netif_stop_queue - stop transmitted packets
2218 * @dev: network device
2219 *
2220 * Stop upper layers calling the device hard_start_xmit routine.
2221 * Used for flow control when transmit resources are unavailable.
2222 */
2223 static inline void netif_stop_queue(struct net_device *dev)
2224 {
2225 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2226 }
2227
2228 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2229 {
2230 unsigned int i;
2231
2232 for (i = 0; i < dev->num_tx_queues; i++) {
2233 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2234 netif_tx_stop_queue(txq);
2235 }
2236 }
2237
2238 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2239 {
2240 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2241 }
2242
2243 /**
2244 * netif_queue_stopped - test if transmit queue is flowblocked
2245 * @dev: network device
2246 *
2247 * Test if transmit queue on device is currently unable to send.
2248 */
2249 static inline bool netif_queue_stopped(const struct net_device *dev)
2250 {
2251 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2252 }
2253
2254 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2255 {
2256 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2257 }
2258
2259 static inline bool
2260 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2261 {
2262 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2263 }
2264
2265 static inline bool
2266 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2267 {
2268 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2269 }
2270
2271 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2272 unsigned int bytes)
2273 {
2274 #ifdef CONFIG_BQL
2275 dql_queued(&dev_queue->dql, bytes);
2276
2277 if (likely(dql_avail(&dev_queue->dql) >= 0))
2278 return;
2279
2280 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2281
2282 /*
2283 * The XOFF flag must be set before checking the dql_avail below,
2284 * because in netdev_tx_completed_queue we update the dql_completed
2285 * before checking the XOFF flag.
2286 */
2287 smp_mb();
2288
2289 /* check again in case another CPU has just made room avail */
2290 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2291 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2292 #endif
2293 }
2294
2295 /**
2296 * netdev_sent_queue - report the number of bytes queued to hardware
2297 * @dev: network device
2298 * @bytes: number of bytes queued to the hardware device queue
2299 *
2300 * Report the number of bytes queued for sending/completion to the network
2301 * device hardware queue. @bytes should be a good approximation and should
2302 * exactly match netdev_completed_queue() @bytes
2303 */
2304 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2305 {
2306 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2307 }
2308
2309 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2310 unsigned int pkts, unsigned int bytes)
2311 {
2312 #ifdef CONFIG_BQL
2313 if (unlikely(!bytes))
2314 return;
2315
2316 dql_completed(&dev_queue->dql, bytes);
2317
2318 /*
2319 * Without the memory barrier there is a small possiblity that
2320 * netdev_tx_sent_queue will miss the update and cause the queue to
2321 * be stopped forever
2322 */
2323 smp_mb();
2324
2325 if (dql_avail(&dev_queue->dql) < 0)
2326 return;
2327
2328 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2329 netif_schedule_queue(dev_queue);
2330 #endif
2331 }
2332
2333 /**
2334 * netdev_completed_queue - report bytes and packets completed by device
2335 * @dev: network device
2336 * @pkts: actual number of packets sent over the medium
2337 * @bytes: actual number of bytes sent over the medium
2338 *
2339 * Report the number of bytes and packets transmitted by the network device
2340 * hardware queue over the physical medium, @bytes must exactly match the
2341 * @bytes amount passed to netdev_sent_queue()
2342 */
2343 static inline void netdev_completed_queue(struct net_device *dev,
2344 unsigned int pkts, unsigned int bytes)
2345 {
2346 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2347 }
2348
2349 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2350 {
2351 #ifdef CONFIG_BQL
2352 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2353 dql_reset(&q->dql);
2354 #endif
2355 }
2356
2357 /**
2358 * netdev_reset_queue - reset the packets and bytes count of a network device
2359 * @dev_queue: network device
2360 *
2361 * Reset the bytes and packet count of a network device and clear the
2362 * software flow control OFF bit for this network device
2363 */
2364 static inline void netdev_reset_queue(struct net_device *dev_queue)
2365 {
2366 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2367 }
2368
2369 /**
2370 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2371 * @dev: network device
2372 * @queue_index: given tx queue index
2373 *
2374 * Returns 0 if given tx queue index >= number of device tx queues,
2375 * otherwise returns the originally passed tx queue index.
2376 */
2377 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2378 {
2379 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2380 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2381 dev->name, queue_index,
2382 dev->real_num_tx_queues);
2383 return 0;
2384 }
2385
2386 return queue_index;
2387 }
2388
2389 /**
2390 * netif_running - test if up
2391 * @dev: network device
2392 *
2393 * Test if the device has been brought up.
2394 */
2395 static inline bool netif_running(const struct net_device *dev)
2396 {
2397 return test_bit(__LINK_STATE_START, &dev->state);
2398 }
2399
2400 /*
2401 * Routines to manage the subqueues on a device. We only need start
2402 * stop, and a check if it's stopped. All other device management is
2403 * done at the overall netdevice level.
2404 * Also test the device if we're multiqueue.
2405 */
2406
2407 /**
2408 * netif_start_subqueue - allow sending packets on subqueue
2409 * @dev: network device
2410 * @queue_index: sub queue index
2411 *
2412 * Start individual transmit queue of a device with multiple transmit queues.
2413 */
2414 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2415 {
2416 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2417
2418 netif_tx_start_queue(txq);
2419 }
2420
2421 /**
2422 * netif_stop_subqueue - stop sending packets on subqueue
2423 * @dev: network device
2424 * @queue_index: sub queue index
2425 *
2426 * Stop individual transmit queue of a device with multiple transmit queues.
2427 */
2428 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2429 {
2430 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2431 netif_tx_stop_queue(txq);
2432 }
2433
2434 /**
2435 * netif_subqueue_stopped - test status of subqueue
2436 * @dev: network device
2437 * @queue_index: sub queue index
2438 *
2439 * Check individual transmit queue of a device with multiple transmit queues.
2440 */
2441 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2442 u16 queue_index)
2443 {
2444 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2445
2446 return netif_tx_queue_stopped(txq);
2447 }
2448
2449 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2450 struct sk_buff *skb)
2451 {
2452 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2453 }
2454
2455 /**
2456 * netif_wake_subqueue - allow sending packets on subqueue
2457 * @dev: network device
2458 * @queue_index: sub queue index
2459 *
2460 * Resume individual transmit queue of a device with multiple transmit queues.
2461 */
2462 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2463 {
2464 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2465 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2466 __netif_schedule(txq->qdisc);
2467 }
2468
2469 #ifdef CONFIG_XPS
2470 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2471 u16 index);
2472 #else
2473 static inline int netif_set_xps_queue(struct net_device *dev,
2474 const struct cpumask *mask,
2475 u16 index)
2476 {
2477 return 0;
2478 }
2479 #endif
2480
2481 /*
2482 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2483 * as a distribution range limit for the returned value.
2484 */
2485 static inline u16 skb_tx_hash(const struct net_device *dev,
2486 const struct sk_buff *skb)
2487 {
2488 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2489 }
2490
2491 /**
2492 * netif_is_multiqueue - test if device has multiple transmit queues
2493 * @dev: network device
2494 *
2495 * Check if device has multiple transmit queues
2496 */
2497 static inline bool netif_is_multiqueue(const struct net_device *dev)
2498 {
2499 return dev->num_tx_queues > 1;
2500 }
2501
2502 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2503
2504 #ifdef CONFIG_SYSFS
2505 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2506 #else
2507 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2508 unsigned int rxq)
2509 {
2510 return 0;
2511 }
2512 #endif
2513
2514 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2515 const struct net_device *from_dev)
2516 {
2517 int err;
2518
2519 err = netif_set_real_num_tx_queues(to_dev,
2520 from_dev->real_num_tx_queues);
2521 if (err)
2522 return err;
2523 #ifdef CONFIG_SYSFS
2524 return netif_set_real_num_rx_queues(to_dev,
2525 from_dev->real_num_rx_queues);
2526 #else
2527 return 0;
2528 #endif
2529 }
2530
2531 #ifdef CONFIG_SYSFS
2532 static inline unsigned int get_netdev_rx_queue_index(
2533 struct netdev_rx_queue *queue)
2534 {
2535 struct net_device *dev = queue->dev;
2536 int index = queue - dev->_rx;
2537
2538 BUG_ON(index >= dev->num_rx_queues);
2539 return index;
2540 }
2541 #endif
2542
2543 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2544 int netif_get_num_default_rss_queues(void);
2545
2546 enum skb_free_reason {
2547 SKB_REASON_CONSUMED,
2548 SKB_REASON_DROPPED,
2549 };
2550
2551 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2552 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2553
2554 /*
2555 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2556 * interrupt context or with hardware interrupts being disabled.
2557 * (in_irq() || irqs_disabled())
2558 *
2559 * We provide four helpers that can be used in following contexts :
2560 *
2561 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2562 * replacing kfree_skb(skb)
2563 *
2564 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2565 * Typically used in place of consume_skb(skb) in TX completion path
2566 *
2567 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2568 * replacing kfree_skb(skb)
2569 *
2570 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2571 * and consumed a packet. Used in place of consume_skb(skb)
2572 */
2573 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2574 {
2575 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2576 }
2577
2578 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2579 {
2580 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2581 }
2582
2583 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2584 {
2585 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2586 }
2587
2588 static inline void dev_consume_skb_any(struct sk_buff *skb)
2589 {
2590 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2591 }
2592
2593 int netif_rx(struct sk_buff *skb);
2594 int netif_rx_ni(struct sk_buff *skb);
2595 int netif_receive_skb(struct sk_buff *skb);
2596 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2597 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2598 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2599 gro_result_t napi_gro_frags(struct napi_struct *napi);
2600 struct packet_offload *gro_find_receive_by_type(__be16 type);
2601 struct packet_offload *gro_find_complete_by_type(__be16 type);
2602
2603 static inline void napi_free_frags(struct napi_struct *napi)
2604 {
2605 kfree_skb(napi->skb);
2606 napi->skb = NULL;
2607 }
2608
2609 int netdev_rx_handler_register(struct net_device *dev,
2610 rx_handler_func_t *rx_handler,
2611 void *rx_handler_data);
2612 void netdev_rx_handler_unregister(struct net_device *dev);
2613
2614 bool dev_valid_name(const char *name);
2615 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2616 int dev_ethtool(struct net *net, struct ifreq *);
2617 unsigned int dev_get_flags(const struct net_device *);
2618 int __dev_change_flags(struct net_device *, unsigned int flags);
2619 int dev_change_flags(struct net_device *, unsigned int);
2620 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2621 unsigned int gchanges);
2622 int dev_change_name(struct net_device *, const char *);
2623 int dev_set_alias(struct net_device *, const char *, size_t);
2624 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2625 int dev_set_mtu(struct net_device *, int);
2626 void dev_set_group(struct net_device *, int);
2627 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2628 int dev_change_carrier(struct net_device *, bool new_carrier);
2629 int dev_get_phys_port_id(struct net_device *dev,
2630 struct netdev_phys_port_id *ppid);
2631 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2632 struct netdev_queue *txq);
2633 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2634 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2635 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2636
2637 extern int netdev_budget;
2638
2639 /* Called by rtnetlink.c:rtnl_unlock() */
2640 void netdev_run_todo(void);
2641
2642 /**
2643 * dev_put - release reference to device
2644 * @dev: network device
2645 *
2646 * Release reference to device to allow it to be freed.
2647 */
2648 static inline void dev_put(struct net_device *dev)
2649 {
2650 this_cpu_dec(*dev->pcpu_refcnt);
2651 }
2652
2653 /**
2654 * dev_hold - get reference to device
2655 * @dev: network device
2656 *
2657 * Hold reference to device to keep it from being freed.
2658 */
2659 static inline void dev_hold(struct net_device *dev)
2660 {
2661 this_cpu_inc(*dev->pcpu_refcnt);
2662 }
2663
2664 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2665 * and _off may be called from IRQ context, but it is caller
2666 * who is responsible for serialization of these calls.
2667 *
2668 * The name carrier is inappropriate, these functions should really be
2669 * called netif_lowerlayer_*() because they represent the state of any
2670 * kind of lower layer not just hardware media.
2671 */
2672
2673 void linkwatch_init_dev(struct net_device *dev);
2674 void linkwatch_fire_event(struct net_device *dev);
2675 void linkwatch_forget_dev(struct net_device *dev);
2676
2677 /**
2678 * netif_carrier_ok - test if carrier present
2679 * @dev: network device
2680 *
2681 * Check if carrier is present on device
2682 */
2683 static inline bool netif_carrier_ok(const struct net_device *dev)
2684 {
2685 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2686 }
2687
2688 unsigned long dev_trans_start(struct net_device *dev);
2689
2690 void __netdev_watchdog_up(struct net_device *dev);
2691
2692 void netif_carrier_on(struct net_device *dev);
2693
2694 void netif_carrier_off(struct net_device *dev);
2695
2696 /**
2697 * netif_dormant_on - mark device as dormant.
2698 * @dev: network device
2699 *
2700 * Mark device as dormant (as per RFC2863).
2701 *
2702 * The dormant state indicates that the relevant interface is not
2703 * actually in a condition to pass packets (i.e., it is not 'up') but is
2704 * in a "pending" state, waiting for some external event. For "on-
2705 * demand" interfaces, this new state identifies the situation where the
2706 * interface is waiting for events to place it in the up state.
2707 *
2708 */
2709 static inline void netif_dormant_on(struct net_device *dev)
2710 {
2711 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2712 linkwatch_fire_event(dev);
2713 }
2714
2715 /**
2716 * netif_dormant_off - set device as not dormant.
2717 * @dev: network device
2718 *
2719 * Device is not in dormant state.
2720 */
2721 static inline void netif_dormant_off(struct net_device *dev)
2722 {
2723 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2724 linkwatch_fire_event(dev);
2725 }
2726
2727 /**
2728 * netif_dormant - test if carrier present
2729 * @dev: network device
2730 *
2731 * Check if carrier is present on device
2732 */
2733 static inline bool netif_dormant(const struct net_device *dev)
2734 {
2735 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2736 }
2737
2738
2739 /**
2740 * netif_oper_up - test if device is operational
2741 * @dev: network device
2742 *
2743 * Check if carrier is operational
2744 */
2745 static inline bool netif_oper_up(const struct net_device *dev)
2746 {
2747 return (dev->operstate == IF_OPER_UP ||
2748 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2749 }
2750
2751 /**
2752 * netif_device_present - is device available or removed
2753 * @dev: network device
2754 *
2755 * Check if device has not been removed from system.
2756 */
2757 static inline bool netif_device_present(struct net_device *dev)
2758 {
2759 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2760 }
2761
2762 void netif_device_detach(struct net_device *dev);
2763
2764 void netif_device_attach(struct net_device *dev);
2765
2766 /*
2767 * Network interface message level settings
2768 */
2769
2770 enum {
2771 NETIF_MSG_DRV = 0x0001,
2772 NETIF_MSG_PROBE = 0x0002,
2773 NETIF_MSG_LINK = 0x0004,
2774 NETIF_MSG_TIMER = 0x0008,
2775 NETIF_MSG_IFDOWN = 0x0010,
2776 NETIF_MSG_IFUP = 0x0020,
2777 NETIF_MSG_RX_ERR = 0x0040,
2778 NETIF_MSG_TX_ERR = 0x0080,
2779 NETIF_MSG_TX_QUEUED = 0x0100,
2780 NETIF_MSG_INTR = 0x0200,
2781 NETIF_MSG_TX_DONE = 0x0400,
2782 NETIF_MSG_RX_STATUS = 0x0800,
2783 NETIF_MSG_PKTDATA = 0x1000,
2784 NETIF_MSG_HW = 0x2000,
2785 NETIF_MSG_WOL = 0x4000,
2786 };
2787
2788 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2789 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2790 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2791 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2792 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2793 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2794 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2795 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2796 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2797 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2798 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2799 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2800 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2801 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2802 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2803
2804 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2805 {
2806 /* use default */
2807 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2808 return default_msg_enable_bits;
2809 if (debug_value == 0) /* no output */
2810 return 0;
2811 /* set low N bits */
2812 return (1 << debug_value) - 1;
2813 }
2814
2815 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2816 {
2817 spin_lock(&txq->_xmit_lock);
2818 txq->xmit_lock_owner = cpu;
2819 }
2820
2821 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2822 {
2823 spin_lock_bh(&txq->_xmit_lock);
2824 txq->xmit_lock_owner = smp_processor_id();
2825 }
2826
2827 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2828 {
2829 bool ok = spin_trylock(&txq->_xmit_lock);
2830 if (likely(ok))
2831 txq->xmit_lock_owner = smp_processor_id();
2832 return ok;
2833 }
2834
2835 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2836 {
2837 txq->xmit_lock_owner = -1;
2838 spin_unlock(&txq->_xmit_lock);
2839 }
2840
2841 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2842 {
2843 txq->xmit_lock_owner = -1;
2844 spin_unlock_bh(&txq->_xmit_lock);
2845 }
2846
2847 static inline void txq_trans_update(struct netdev_queue *txq)
2848 {
2849 if (txq->xmit_lock_owner != -1)
2850 txq->trans_start = jiffies;
2851 }
2852
2853 /**
2854 * netif_tx_lock - grab network device transmit lock
2855 * @dev: network device
2856 *
2857 * Get network device transmit lock
2858 */
2859 static inline void netif_tx_lock(struct net_device *dev)
2860 {
2861 unsigned int i;
2862 int cpu;
2863
2864 spin_lock(&dev->tx_global_lock);
2865 cpu = smp_processor_id();
2866 for (i = 0; i < dev->num_tx_queues; i++) {
2867 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2868
2869 /* We are the only thread of execution doing a
2870 * freeze, but we have to grab the _xmit_lock in
2871 * order to synchronize with threads which are in
2872 * the ->hard_start_xmit() handler and already
2873 * checked the frozen bit.
2874 */
2875 __netif_tx_lock(txq, cpu);
2876 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2877 __netif_tx_unlock(txq);
2878 }
2879 }
2880
2881 static inline void netif_tx_lock_bh(struct net_device *dev)
2882 {
2883 local_bh_disable();
2884 netif_tx_lock(dev);
2885 }
2886
2887 static inline void netif_tx_unlock(struct net_device *dev)
2888 {
2889 unsigned int i;
2890
2891 for (i = 0; i < dev->num_tx_queues; i++) {
2892 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2893
2894 /* No need to grab the _xmit_lock here. If the
2895 * queue is not stopped for another reason, we
2896 * force a schedule.
2897 */
2898 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2899 netif_schedule_queue(txq);
2900 }
2901 spin_unlock(&dev->tx_global_lock);
2902 }
2903
2904 static inline void netif_tx_unlock_bh(struct net_device *dev)
2905 {
2906 netif_tx_unlock(dev);
2907 local_bh_enable();
2908 }
2909
2910 #define HARD_TX_LOCK(dev, txq, cpu) { \
2911 if ((dev->features & NETIF_F_LLTX) == 0) { \
2912 __netif_tx_lock(txq, cpu); \
2913 } \
2914 }
2915
2916 #define HARD_TX_TRYLOCK(dev, txq) \
2917 (((dev->features & NETIF_F_LLTX) == 0) ? \
2918 __netif_tx_trylock(txq) : \
2919 true )
2920
2921 #define HARD_TX_UNLOCK(dev, txq) { \
2922 if ((dev->features & NETIF_F_LLTX) == 0) { \
2923 __netif_tx_unlock(txq); \
2924 } \
2925 }
2926
2927 static inline void netif_tx_disable(struct net_device *dev)
2928 {
2929 unsigned int i;
2930 int cpu;
2931
2932 local_bh_disable();
2933 cpu = smp_processor_id();
2934 for (i = 0; i < dev->num_tx_queues; i++) {
2935 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2936
2937 __netif_tx_lock(txq, cpu);
2938 netif_tx_stop_queue(txq);
2939 __netif_tx_unlock(txq);
2940 }
2941 local_bh_enable();
2942 }
2943
2944 static inline void netif_addr_lock(struct net_device *dev)
2945 {
2946 spin_lock(&dev->addr_list_lock);
2947 }
2948
2949 static inline void netif_addr_lock_nested(struct net_device *dev)
2950 {
2951 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2952 }
2953
2954 static inline void netif_addr_lock_bh(struct net_device *dev)
2955 {
2956 spin_lock_bh(&dev->addr_list_lock);
2957 }
2958
2959 static inline void netif_addr_unlock(struct net_device *dev)
2960 {
2961 spin_unlock(&dev->addr_list_lock);
2962 }
2963
2964 static inline void netif_addr_unlock_bh(struct net_device *dev)
2965 {
2966 spin_unlock_bh(&dev->addr_list_lock);
2967 }
2968
2969 /*
2970 * dev_addrs walker. Should be used only for read access. Call with
2971 * rcu_read_lock held.
2972 */
2973 #define for_each_dev_addr(dev, ha) \
2974 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2975
2976 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2977
2978 void ether_setup(struct net_device *dev);
2979
2980 /* Support for loadable net-drivers */
2981 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2982 void (*setup)(struct net_device *),
2983 unsigned int txqs, unsigned int rxqs);
2984 #define alloc_netdev(sizeof_priv, name, setup) \
2985 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2986
2987 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2988 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2989
2990 int register_netdev(struct net_device *dev);
2991 void unregister_netdev(struct net_device *dev);
2992
2993 /* General hardware address lists handling functions */
2994 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2995 struct netdev_hw_addr_list *from_list, int addr_len);
2996 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2997 struct netdev_hw_addr_list *from_list, int addr_len);
2998 void __hw_addr_init(struct netdev_hw_addr_list *list);
2999
3000 /* Functions used for device addresses handling */
3001 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3002 unsigned char addr_type);
3003 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3004 unsigned char addr_type);
3005 void dev_addr_flush(struct net_device *dev);
3006 int dev_addr_init(struct net_device *dev);
3007
3008 /* Functions used for unicast addresses handling */
3009 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3010 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3011 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3012 int dev_uc_sync(struct net_device *to, struct net_device *from);
3013 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3014 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3015 void dev_uc_flush(struct net_device *dev);
3016 void dev_uc_init(struct net_device *dev);
3017
3018 /* Functions used for multicast addresses handling */
3019 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3020 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3021 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3022 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3023 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3024 int dev_mc_sync(struct net_device *to, struct net_device *from);
3025 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3026 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3027 void dev_mc_flush(struct net_device *dev);
3028 void dev_mc_init(struct net_device *dev);
3029
3030 /* Functions used for secondary unicast and multicast support */
3031 void dev_set_rx_mode(struct net_device *dev);
3032 void __dev_set_rx_mode(struct net_device *dev);
3033 int dev_set_promiscuity(struct net_device *dev, int inc);
3034 int dev_set_allmulti(struct net_device *dev, int inc);
3035 void netdev_state_change(struct net_device *dev);
3036 void netdev_notify_peers(struct net_device *dev);
3037 void netdev_features_change(struct net_device *dev);
3038 /* Load a device via the kmod */
3039 void dev_load(struct net *net, const char *name);
3040 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3041 struct rtnl_link_stats64 *storage);
3042 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3043 const struct net_device_stats *netdev_stats);
3044
3045 extern int netdev_max_backlog;
3046 extern int netdev_tstamp_prequeue;
3047 extern int weight_p;
3048 extern int bpf_jit_enable;
3049
3050 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3051 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3052 struct list_head **iter);
3053
3054 /* iterate through upper list, must be called under RCU read lock */
3055 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3056 for (iter = &(dev)->all_adj_list.upper, \
3057 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3058 updev; \
3059 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3060
3061 void *netdev_lower_get_next_private(struct net_device *dev,
3062 struct list_head **iter);
3063 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3064 struct list_head **iter);
3065
3066 #define netdev_for_each_lower_private(dev, priv, iter) \
3067 for (iter = (dev)->adj_list.lower.next, \
3068 priv = netdev_lower_get_next_private(dev, &(iter)); \
3069 priv; \
3070 priv = netdev_lower_get_next_private(dev, &(iter)))
3071
3072 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3073 for (iter = &(dev)->adj_list.lower, \
3074 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3075 priv; \
3076 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3077
3078 void *netdev_adjacent_get_private(struct list_head *adj_list);
3079 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3080 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3081 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3082 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3083 int netdev_master_upper_dev_link(struct net_device *dev,
3084 struct net_device *upper_dev);
3085 int netdev_master_upper_dev_link_private(struct net_device *dev,
3086 struct net_device *upper_dev,
3087 void *private);
3088 void netdev_upper_dev_unlink(struct net_device *dev,
3089 struct net_device *upper_dev);
3090 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3091 void *netdev_lower_dev_get_private(struct net_device *dev,
3092 struct net_device *lower_dev);
3093 int skb_checksum_help(struct sk_buff *skb);
3094 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3095 netdev_features_t features, bool tx_path);
3096 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3097 netdev_features_t features);
3098
3099 static inline
3100 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3101 {
3102 return __skb_gso_segment(skb, features, true);
3103 }
3104 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3105
3106 static inline bool can_checksum_protocol(netdev_features_t features,
3107 __be16 protocol)
3108 {
3109 return ((features & NETIF_F_GEN_CSUM) ||
3110 ((features & NETIF_F_V4_CSUM) &&
3111 protocol == htons(ETH_P_IP)) ||
3112 ((features & NETIF_F_V6_CSUM) &&
3113 protocol == htons(ETH_P_IPV6)) ||
3114 ((features & NETIF_F_FCOE_CRC) &&
3115 protocol == htons(ETH_P_FCOE)));
3116 }
3117
3118 #ifdef CONFIG_BUG
3119 void netdev_rx_csum_fault(struct net_device *dev);
3120 #else
3121 static inline void netdev_rx_csum_fault(struct net_device *dev)
3122 {
3123 }
3124 #endif
3125 /* rx skb timestamps */
3126 void net_enable_timestamp(void);
3127 void net_disable_timestamp(void);
3128
3129 #ifdef CONFIG_PROC_FS
3130 int __init dev_proc_init(void);
3131 #else
3132 #define dev_proc_init() 0
3133 #endif
3134
3135 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3136 const void *ns);
3137 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3138 const void *ns);
3139
3140 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3141 {
3142 return netdev_class_create_file_ns(class_attr, NULL);
3143 }
3144
3145 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3146 {
3147 netdev_class_remove_file_ns(class_attr, NULL);
3148 }
3149
3150 extern struct kobj_ns_type_operations net_ns_type_operations;
3151
3152 const char *netdev_drivername(const struct net_device *dev);
3153
3154 void linkwatch_run_queue(void);
3155
3156 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3157 netdev_features_t f2)
3158 {
3159 if (f1 & NETIF_F_GEN_CSUM)
3160 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3161 if (f2 & NETIF_F_GEN_CSUM)
3162 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3163 f1 &= f2;
3164 if (f1 & NETIF_F_GEN_CSUM)
3165 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3166
3167 return f1;
3168 }
3169
3170 static inline netdev_features_t netdev_get_wanted_features(
3171 struct net_device *dev)
3172 {
3173 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3174 }
3175 netdev_features_t netdev_increment_features(netdev_features_t all,
3176 netdev_features_t one, netdev_features_t mask);
3177
3178 /* Allow TSO being used on stacked device :
3179 * Performing the GSO segmentation before last device
3180 * is a performance improvement.
3181 */
3182 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3183 netdev_features_t mask)
3184 {
3185 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3186 }
3187
3188 int __netdev_update_features(struct net_device *dev);
3189 void netdev_update_features(struct net_device *dev);
3190 void netdev_change_features(struct net_device *dev);
3191
3192 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3193 struct net_device *dev);
3194
3195 netdev_features_t netif_skb_features(struct sk_buff *skb);
3196
3197 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3198 {
3199 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3200
3201 /* check flags correspondence */
3202 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3203 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3204 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3205 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3206 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3207 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3208
3209 return (features & feature) == feature;
3210 }
3211
3212 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3213 {
3214 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3215 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3216 }
3217
3218 static inline bool netif_needs_gso(struct sk_buff *skb,
3219 netdev_features_t features)
3220 {
3221 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3222 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3223 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3224 }
3225
3226 static inline void netif_set_gso_max_size(struct net_device *dev,
3227 unsigned int size)
3228 {
3229 dev->gso_max_size = size;
3230 }
3231
3232 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3233 int pulled_hlen, u16 mac_offset,
3234 int mac_len)
3235 {
3236 skb->protocol = protocol;
3237 skb->encapsulation = 1;
3238 skb_push(skb, pulled_hlen);
3239 skb_reset_transport_header(skb);
3240 skb->mac_header = mac_offset;
3241 skb->network_header = skb->mac_header + mac_len;
3242 skb->mac_len = mac_len;
3243 }
3244
3245 static inline bool netif_is_macvlan(struct net_device *dev)
3246 {
3247 return dev->priv_flags & IFF_MACVLAN;
3248 }
3249
3250 static inline bool netif_is_bond_master(struct net_device *dev)
3251 {
3252 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3253 }
3254
3255 static inline bool netif_is_bond_slave(struct net_device *dev)
3256 {
3257 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3258 }
3259
3260 static inline bool netif_supports_nofcs(struct net_device *dev)
3261 {
3262 return dev->priv_flags & IFF_SUPP_NOFCS;
3263 }
3264
3265 extern struct pernet_operations __net_initdata loopback_net_ops;
3266
3267 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3268
3269 /* netdev_printk helpers, similar to dev_printk */
3270
3271 static inline const char *netdev_name(const struct net_device *dev)
3272 {
3273 if (dev->reg_state != NETREG_REGISTERED)
3274 return "(unregistered net_device)";
3275 return dev->name;
3276 }
3277
3278 __printf(3, 4)
3279 int netdev_printk(const char *level, const struct net_device *dev,
3280 const char *format, ...);
3281 __printf(2, 3)
3282 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3283 __printf(2, 3)
3284 int netdev_alert(const struct net_device *dev, const char *format, ...);
3285 __printf(2, 3)
3286 int netdev_crit(const struct net_device *dev, const char *format, ...);
3287 __printf(2, 3)
3288 int netdev_err(const struct net_device *dev, const char *format, ...);
3289 __printf(2, 3)
3290 int netdev_warn(const struct net_device *dev, const char *format, ...);
3291 __printf(2, 3)
3292 int netdev_notice(const struct net_device *dev, const char *format, ...);
3293 __printf(2, 3)
3294 int netdev_info(const struct net_device *dev, const char *format, ...);
3295
3296 #define MODULE_ALIAS_NETDEV(device) \
3297 MODULE_ALIAS("netdev-" device)
3298
3299 #if defined(CONFIG_DYNAMIC_DEBUG)
3300 #define netdev_dbg(__dev, format, args...) \
3301 do { \
3302 dynamic_netdev_dbg(__dev, format, ##args); \
3303 } while (0)
3304 #elif defined(DEBUG)
3305 #define netdev_dbg(__dev, format, args...) \
3306 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3307 #else
3308 #define netdev_dbg(__dev, format, args...) \
3309 ({ \
3310 if (0) \
3311 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3312 0; \
3313 })
3314 #endif
3315
3316 #if defined(VERBOSE_DEBUG)
3317 #define netdev_vdbg netdev_dbg
3318 #else
3319
3320 #define netdev_vdbg(dev, format, args...) \
3321 ({ \
3322 if (0) \
3323 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3324 0; \
3325 })
3326 #endif
3327
3328 /*
3329 * netdev_WARN() acts like dev_printk(), but with the key difference
3330 * of using a WARN/WARN_ON to get the message out, including the
3331 * file/line information and a backtrace.
3332 */
3333 #define netdev_WARN(dev, format, args...) \
3334 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3335
3336 /* netif printk helpers, similar to netdev_printk */
3337
3338 #define netif_printk(priv, type, level, dev, fmt, args...) \
3339 do { \
3340 if (netif_msg_##type(priv)) \
3341 netdev_printk(level, (dev), fmt, ##args); \
3342 } while (0)
3343
3344 #define netif_level(level, priv, type, dev, fmt, args...) \
3345 do { \
3346 if (netif_msg_##type(priv)) \
3347 netdev_##level(dev, fmt, ##args); \
3348 } while (0)
3349
3350 #define netif_emerg(priv, type, dev, fmt, args...) \
3351 netif_level(emerg, priv, type, dev, fmt, ##args)
3352 #define netif_alert(priv, type, dev, fmt, args...) \
3353 netif_level(alert, priv, type, dev, fmt, ##args)
3354 #define netif_crit(priv, type, dev, fmt, args...) \
3355 netif_level(crit, priv, type, dev, fmt, ##args)
3356 #define netif_err(priv, type, dev, fmt, args...) \
3357 netif_level(err, priv, type, dev, fmt, ##args)
3358 #define netif_warn(priv, type, dev, fmt, args...) \
3359 netif_level(warn, priv, type, dev, fmt, ##args)
3360 #define netif_notice(priv, type, dev, fmt, args...) \
3361 netif_level(notice, priv, type, dev, fmt, ##args)
3362 #define netif_info(priv, type, dev, fmt, args...) \
3363 netif_level(info, priv, type, dev, fmt, ##args)
3364
3365 #if defined(CONFIG_DYNAMIC_DEBUG)
3366 #define netif_dbg(priv, type, netdev, format, args...) \
3367 do { \
3368 if (netif_msg_##type(priv)) \
3369 dynamic_netdev_dbg(netdev, format, ##args); \
3370 } while (0)
3371 #elif defined(DEBUG)
3372 #define netif_dbg(priv, type, dev, format, args...) \
3373 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3374 #else
3375 #define netif_dbg(priv, type, dev, format, args...) \
3376 ({ \
3377 if (0) \
3378 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3379 0; \
3380 })
3381 #endif
3382
3383 #if defined(VERBOSE_DEBUG)
3384 #define netif_vdbg netif_dbg
3385 #else
3386 #define netif_vdbg(priv, type, dev, format, args...) \
3387 ({ \
3388 if (0) \
3389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3390 0; \
3391 })
3392 #endif
3393
3394 /*
3395 * The list of packet types we will receive (as opposed to discard)
3396 * and the routines to invoke.
3397 *
3398 * Why 16. Because with 16 the only overlap we get on a hash of the
3399 * low nibble of the protocol value is RARP/SNAP/X.25.
3400 *
3401 * NOTE: That is no longer true with the addition of VLAN tags. Not
3402 * sure which should go first, but I bet it won't make much
3403 * difference if we are running VLANs. The good news is that
3404 * this protocol won't be in the list unless compiled in, so
3405 * the average user (w/out VLANs) will not be adversely affected.
3406 * --BLG
3407 *
3408 * 0800 IP
3409 * 8100 802.1Q VLAN
3410 * 0001 802.3
3411 * 0002 AX.25
3412 * 0004 802.2
3413 * 8035 RARP
3414 * 0005 SNAP
3415 * 0805 X.25
3416 * 0806 ARP
3417 * 8137 IPX
3418 * 0009 Localtalk
3419 * 86DD IPv6
3420 */
3421 #define PTYPE_HASH_SIZE (16)
3422 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3423
3424 #endif /* _LINUX_NETDEVICE_H */