net: provide generic busy polling to all NAPI drivers
[deliverable/linux.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.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 #include <uapi/linux/if_bonding.h>
54
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
70
71 /*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
103
104 enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 };
110 typedef enum netdev_tx netdev_tx_t;
111
112 /*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116 static inline bool dev_xmit_complete(int rc)
117 {
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128 }
129
130 /*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
138 # else
139 # define LL_MAX_HEADER 96
140 # endif
141 #else
142 # define LL_MAX_HEADER 32
143 #endif
144
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
148 #else
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
150 #endif
151
152 /*
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
155 */
156
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181 };
182
183
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
186
187 #ifdef CONFIG_RPS
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
190 #endif
191
192 struct neighbour;
193 struct neigh_parms;
194 struct sk_buff;
195
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
205 bool global_use;
206 int sync_cnt;
207 int refcount;
208 int synced;
209 struct rcu_head rcu_head;
210 };
211
212 struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215 };
216
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
231
232 struct hh_cache {
233 u16 hh_len;
234 u16 __pad;
235 seqlock_t hh_lock;
236
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244 };
245
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258
259 struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268 };
269
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
275 enum netdev_state_t {
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
281 };
282
283
284 /*
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
287 */
288 struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291 };
292 #define NETDEV_BOOT_SETUP_MAX 8
293
294 int __init netdev_boot_setup(char *str);
295
296 /*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299 struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
315 #endif
316 struct net_device *dev;
317 struct sk_buff *gro_list;
318 struct sk_buff *skb;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
323 };
324
325 enum {
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
330 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
331 };
332
333 enum gro_result {
334 GRO_MERGED,
335 GRO_MERGED_FREE,
336 GRO_HELD,
337 GRO_NORMAL,
338 GRO_DROP,
339 };
340 typedef enum gro_result gro_result_t;
341
342 /*
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * further.
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 *
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
353 *
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
356 *
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
360 *
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * do with the skb.
363 *
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
367 *
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
371 *
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
374 * are registered on exact device (ptype->dev == skb->dev).
375 *
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
378 *
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
381 */
382
383 enum rx_handler_result {
384 RX_HANDLER_CONSUMED,
385 RX_HANDLER_ANOTHER,
386 RX_HANDLER_EXACT,
387 RX_HANDLER_PASS,
388 };
389 typedef enum rx_handler_result rx_handler_result_t;
390 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
391
392 void __napi_schedule(struct napi_struct *n);
393 void __napi_schedule_irqoff(struct napi_struct *n);
394
395 static inline bool napi_disable_pending(struct napi_struct *n)
396 {
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
398 }
399
400 /**
401 * napi_schedule_prep - check if napi can be scheduled
402 * @n: napi context
403 *
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
408 */
409 static inline bool napi_schedule_prep(struct napi_struct *n)
410 {
411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
413 }
414
415 /**
416 * napi_schedule - schedule NAPI poll
417 * @n: napi context
418 *
419 * Schedule NAPI poll routine to be called if it is not already
420 * running.
421 */
422 static inline void napi_schedule(struct napi_struct *n)
423 {
424 if (napi_schedule_prep(n))
425 __napi_schedule(n);
426 }
427
428 /**
429 * napi_schedule_irqoff - schedule NAPI poll
430 * @n: napi context
431 *
432 * Variant of napi_schedule(), assuming hard irqs are masked.
433 */
434 static inline void napi_schedule_irqoff(struct napi_struct *n)
435 {
436 if (napi_schedule_prep(n))
437 __napi_schedule_irqoff(n);
438 }
439
440 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
441 static inline bool napi_reschedule(struct napi_struct *napi)
442 {
443 if (napi_schedule_prep(napi)) {
444 __napi_schedule(napi);
445 return true;
446 }
447 return false;
448 }
449
450 void __napi_complete(struct napi_struct *n);
451 void napi_complete_done(struct napi_struct *n, int work_done);
452 /**
453 * napi_complete - NAPI processing complete
454 * @n: napi context
455 *
456 * Mark NAPI processing as complete.
457 * Consider using napi_complete_done() instead.
458 */
459 static inline void napi_complete(struct napi_struct *n)
460 {
461 return napi_complete_done(n, 0);
462 }
463
464 /**
465 * napi_hash_add - add a NAPI to global hashtable
466 * @napi: napi context
467 *
468 * generate a new napi_id and store a @napi under it in napi_hash
469 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
470 * Note: This is normally automatically done from netif_napi_add(),
471 * so might disappear in a future linux version.
472 */
473 void napi_hash_add(struct napi_struct *napi);
474
475 /**
476 * napi_hash_del - remove a NAPI from global table
477 * @napi: napi context
478 *
479 * Warning: caller must observe rcu grace period
480 * before freeing memory containing @napi, if
481 * this function returns true.
482 * Note: core networking stack automatically calls it
483 * from netif_napi_del()
484 * Drivers might want to call this helper to combine all
485 * the needed rcu grace periods into a single one.
486 */
487 bool napi_hash_del(struct napi_struct *napi);
488
489 /**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
496 void napi_disable(struct napi_struct *n);
497
498 /**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505 static inline void napi_enable(struct napi_struct *n)
506 {
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510 clear_bit(NAPI_STATE_NPSVC, &n->state);
511 }
512
513 #ifdef CONFIG_SMP
514 /**
515 * napi_synchronize - wait until NAPI is not running
516 * @n: napi context
517 *
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
521 */
522 static inline void napi_synchronize(const struct napi_struct *n)
523 {
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 msleep(1);
526 }
527 #else
528 # define napi_synchronize(n) barrier()
529 #endif
530
531 enum netdev_queue_state_t {
532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
534 __QUEUE_STATE_FROZEN,
535 };
536
537 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540
541 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 QUEUE_STATE_FROZEN)
544 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 QUEUE_STATE_FROZEN)
546
547 /*
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
555 */
556
557 struct netdev_queue {
558 /*
559 * read mostly part
560 */
561 struct net_device *dev;
562 struct Qdisc __rcu *qdisc;
563 struct Qdisc *qdisc_sleeping;
564 #ifdef CONFIG_SYSFS
565 struct kobject kobj;
566 #endif
567 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
568 int numa_node;
569 #endif
570 /*
571 * write mostly part
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
575 /*
576 * please use this field instead of dev->trans_start
577 */
578 unsigned long trans_start;
579
580 /*
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
583 */
584 unsigned long trans_timeout;
585
586 unsigned long state;
587
588 #ifdef CONFIG_BQL
589 struct dql dql;
590 #endif
591 unsigned long tx_maxrate;
592 } ____cacheline_aligned_in_smp;
593
594 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
595 {
596 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 return q->numa_node;
598 #else
599 return NUMA_NO_NODE;
600 #endif
601 }
602
603 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
604 {
605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 q->numa_node = node;
607 #endif
608 }
609
610 #ifdef CONFIG_RPS
611 /*
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
614 */
615 struct rps_map {
616 unsigned int len;
617 struct rcu_head rcu;
618 u16 cpus[0];
619 };
620 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
621
622 /*
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
626 */
627 struct rps_dev_flow {
628 u16 cpu;
629 u16 filter;
630 unsigned int last_qtail;
631 };
632 #define RPS_NO_FILTER 0xffff
633
634 /*
635 * The rps_dev_flow_table structure contains a table of flow mappings.
636 */
637 struct rps_dev_flow_table {
638 unsigned int mask;
639 struct rcu_head rcu;
640 struct rps_dev_flow flows[0];
641 };
642 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
643 ((_num) * sizeof(struct rps_dev_flow)))
644
645 /*
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
648 * Each entry is a 32bit value. Upper part is the high order bits
649 * of flow hash, lower part is cpu number.
650 * rps_cpu_mask is used to partition the space, depending on number of
651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
653 * meaning we use 32-6=26 bits for the hash.
654 */
655 struct rps_sock_flow_table {
656 u32 mask;
657
658 u32 ents[0] ____cacheline_aligned_in_smp;
659 };
660 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
661
662 #define RPS_NO_CPU 0xffff
663
664 extern u32 rps_cpu_mask;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
666
667 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
668 u32 hash)
669 {
670 if (table && hash) {
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
673
674 /* We only give a hint, preemption can change cpu under us */
675 val |= raw_smp_processor_id();
676
677 if (table->ents[index] != val)
678 table->ents[index] = val;
679 }
680 }
681
682 #ifdef CONFIG_RFS_ACCEL
683 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 u16 filter_id);
685 #endif
686 #endif /* CONFIG_RPS */
687
688 /* This structure contains an instance of an RX queue. */
689 struct netdev_rx_queue {
690 #ifdef CONFIG_RPS
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
693 #endif
694 struct kobject kobj;
695 struct net_device *dev;
696 } ____cacheline_aligned_in_smp;
697
698 /*
699 * RX queue sysfs structures and functions.
700 */
701 struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
707 };
708
709 #ifdef CONFIG_XPS
710 /*
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
713 */
714 struct xps_map {
715 unsigned int len;
716 unsigned int alloc_len;
717 struct rcu_head rcu;
718 u16 queues[0];
719 };
720 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
721 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
722 - sizeof(struct xps_map)) / sizeof(u16))
723
724 /*
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 */
727 struct xps_dev_maps {
728 struct rcu_head rcu;
729 struct xps_map __rcu *cpu_map[0];
730 };
731 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733 #endif /* CONFIG_XPS */
734
735 #define TC_MAX_QUEUE 16
736 #define TC_BITMASK 15
737 /* HW offloaded queuing disciplines txq count and offset maps */
738 struct netdev_tc_txq {
739 u16 count;
740 u16 offset;
741 };
742
743 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744 /*
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
747 */
748 struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
755 char model[256];
756 char model_description[256];
757 };
758 #endif
759
760 #define MAX_PHYS_ITEM_ID_LEN 32
761
762 /* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
764 */
765 struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
767 unsigned char id_len;
768 };
769
770 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
771 struct netdev_phys_item_id *b)
772 {
773 return a->id_len == b->id_len &&
774 memcmp(a->id, b->id, a->id_len) == 0;
775 }
776
777 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
778 struct sk_buff *skb);
779
780 /*
781 * This structure defines the management hooks for network devices.
782 * The following hooks can be defined; unless noted otherwise, they are
783 * optional and can be filled with a null pointer.
784 *
785 * int (*ndo_init)(struct net_device *dev);
786 * This function is called once when network device is registered.
787 * The network device can use this to any late stage initializaton
788 * or semantic validattion. It can fail with an error code which will
789 * be propogated back to register_netdev
790 *
791 * void (*ndo_uninit)(struct net_device *dev);
792 * This function is called when device is unregistered or when registration
793 * fails. It is not called if init fails.
794 *
795 * int (*ndo_open)(struct net_device *dev);
796 * This function is called when network device transistions to the up
797 * state.
798 *
799 * int (*ndo_stop)(struct net_device *dev);
800 * This function is called when network device transistions to the down
801 * state.
802 *
803 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
804 * struct net_device *dev);
805 * Called when a packet needs to be transmitted.
806 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
807 * the queue before that can happen; it's for obsolete devices and weird
808 * corner cases, but the stack really does a non-trivial amount
809 * of useless work if you return NETDEV_TX_BUSY.
810 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
811 * Required can not be NULL.
812 *
813 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
814 * void *accel_priv, select_queue_fallback_t fallback);
815 * Called to decide which queue to when device supports multiple
816 * transmit queues.
817 *
818 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
819 * This function is called to allow device receiver to make
820 * changes to configuration when multicast or promiscious is enabled.
821 *
822 * void (*ndo_set_rx_mode)(struct net_device *dev);
823 * This function is called device changes address list filtering.
824 * If driver handles unicast address filtering, it should set
825 * IFF_UNICAST_FLT to its priv_flags.
826 *
827 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
828 * This function is called when the Media Access Control address
829 * needs to be changed. If this interface is not defined, the
830 * mac address can not be changed.
831 *
832 * int (*ndo_validate_addr)(struct net_device *dev);
833 * Test if Media Access Control address is valid for the device.
834 *
835 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
836 * Called when a user request an ioctl which can't be handled by
837 * the generic interface code. If not defined ioctl's return
838 * not supported error code.
839 *
840 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
841 * Used to set network devices bus interface parameters. This interface
842 * is retained for legacy reason, new devices should use the bus
843 * interface (PCI) for low level management.
844 *
845 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
846 * Called when a user wants to change the Maximum Transfer Unit
847 * of a device. If not defined, any request to change MTU will
848 * will return an error.
849 *
850 * void (*ndo_tx_timeout)(struct net_device *dev);
851 * Callback uses when the transmitter has not made any progress
852 * for dev->watchdog ticks.
853 *
854 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
855 * struct rtnl_link_stats64 *storage);
856 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
857 * Called when a user wants to get the network device usage
858 * statistics. Drivers must do one of the following:
859 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
860 * rtnl_link_stats64 structure passed by the caller.
861 * 2. Define @ndo_get_stats to update a net_device_stats structure
862 * (which should normally be dev->stats) and return a pointer to
863 * it. The structure may be changed asynchronously only if each
864 * field is written atomically.
865 * 3. Update dev->stats asynchronously and atomically, and define
866 * neither operation.
867 *
868 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
869 * If device support VLAN filtering this function is called when a
870 * VLAN id is registered.
871 *
872 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
873 * If device support VLAN filtering this function is called when a
874 * VLAN id is unregistered.
875 *
876 * void (*ndo_poll_controller)(struct net_device *dev);
877 *
878 * SR-IOV management functions.
879 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
880 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
881 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
882 * int max_tx_rate);
883 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
884 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
885 * int (*ndo_get_vf_config)(struct net_device *dev,
886 * int vf, struct ifla_vf_info *ivf);
887 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
888 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
889 * struct nlattr *port[]);
890 *
891 * Enable or disable the VF ability to query its RSS Redirection Table and
892 * Hash Key. This is needed since on some devices VF share this information
893 * with PF and querying it may adduce a theoretical security risk.
894 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
895 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
896 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
897 * Called to setup 'tc' number of traffic classes in the net device. This
898 * is always called from the stack with the rtnl lock held and netif tx
899 * queues stopped. This allows the netdevice to perform queue management
900 * safely.
901 *
902 * Fiber Channel over Ethernet (FCoE) offload functions.
903 * int (*ndo_fcoe_enable)(struct net_device *dev);
904 * Called when the FCoE protocol stack wants to start using LLD for FCoE
905 * so the underlying device can perform whatever needed configuration or
906 * initialization to support acceleration of FCoE traffic.
907 *
908 * int (*ndo_fcoe_disable)(struct net_device *dev);
909 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
910 * so the underlying device can perform whatever needed clean-ups to
911 * stop supporting acceleration of FCoE traffic.
912 *
913 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
914 * struct scatterlist *sgl, unsigned int sgc);
915 * Called when the FCoE Initiator wants to initialize an I/O that
916 * is a possible candidate for Direct Data Placement (DDP). The LLD can
917 * perform necessary setup and returns 1 to indicate the device is set up
918 * successfully to perform DDP on this I/O, otherwise this returns 0.
919 *
920 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
921 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
922 * indicated by the FC exchange id 'xid', so the underlying device can
923 * clean up and reuse resources for later DDP requests.
924 *
925 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
926 * struct scatterlist *sgl, unsigned int sgc);
927 * Called when the FCoE Target wants to initialize an I/O that
928 * is a possible candidate for Direct Data Placement (DDP). The LLD can
929 * perform necessary setup and returns 1 to indicate the device is set up
930 * successfully to perform DDP on this I/O, otherwise this returns 0.
931 *
932 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
933 * struct netdev_fcoe_hbainfo *hbainfo);
934 * Called when the FCoE Protocol stack wants information on the underlying
935 * device. This information is utilized by the FCoE protocol stack to
936 * register attributes with Fiber Channel management service as per the
937 * FC-GS Fabric Device Management Information(FDMI) specification.
938 *
939 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
940 * Called when the underlying device wants to override default World Wide
941 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
942 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
943 * protocol stack to use.
944 *
945 * RFS acceleration.
946 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
947 * u16 rxq_index, u32 flow_id);
948 * Set hardware filter for RFS. rxq_index is the target queue index;
949 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
950 * Return the filter ID on success, or a negative error code.
951 *
952 * Slave management functions (for bridge, bonding, etc).
953 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
954 * Called to make another netdev an underling.
955 *
956 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
957 * Called to release previously enslaved netdev.
958 *
959 * Feature/offload setting functions.
960 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
961 * netdev_features_t features);
962 * Adjusts the requested feature flags according to device-specific
963 * constraints, and returns the resulting flags. Must not modify
964 * the device state.
965 *
966 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
967 * Called to update device configuration to new features. Passed
968 * feature set might be less than what was returned by ndo_fix_features()).
969 * Must return >0 or -errno if it changed dev->features itself.
970 *
971 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
972 * struct net_device *dev,
973 * const unsigned char *addr, u16 vid, u16 flags)
974 * Adds an FDB entry to dev for addr.
975 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
976 * struct net_device *dev,
977 * const unsigned char *addr, u16 vid)
978 * Deletes the FDB entry from dev coresponding to addr.
979 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
980 * struct net_device *dev, struct net_device *filter_dev,
981 * int idx)
982 * Used to add FDB entries to dump requests. Implementers should add
983 * entries to skb and update idx with the number of entries.
984 *
985 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
986 * u16 flags)
987 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
988 * struct net_device *dev, u32 filter_mask,
989 * int nlflags)
990 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
991 * u16 flags);
992 *
993 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
994 * Called to change device carrier. Soft-devices (like dummy, team, etc)
995 * which do not represent real hardware may define this to allow their
996 * userspace components to manage their virtual carrier state. Devices
997 * that determine carrier state from physical hardware properties (eg
998 * network cables) or protocol-dependent mechanisms (eg
999 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1000 *
1001 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1002 * struct netdev_phys_item_id *ppid);
1003 * Called to get ID of physical port of this device. If driver does
1004 * not implement this, it is assumed that the hw is not able to have
1005 * multiple net devices on single physical port.
1006 *
1007 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1008 * sa_family_t sa_family, __be16 port);
1009 * Called by vxlan to notiy a driver about the UDP port and socket
1010 * address family that vxlan is listnening to. It is called only when
1011 * a new port starts listening. The operation is protected by the
1012 * vxlan_net->sock_lock.
1013 *
1014 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1015 * sa_family_t sa_family, __be16 port);
1016 * Called by vxlan to notify the driver about a UDP port and socket
1017 * address family that vxlan is not listening to anymore. The operation
1018 * is protected by the vxlan_net->sock_lock.
1019 *
1020 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1021 * struct net_device *dev)
1022 * Called by upper layer devices to accelerate switching or other
1023 * station functionality into hardware. 'pdev is the lowerdev
1024 * to use for the offload and 'dev' is the net device that will
1025 * back the offload. Returns a pointer to the private structure
1026 * the upper layer will maintain.
1027 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1028 * Called by upper layer device to delete the station created
1029 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1030 * the station and priv is the structure returned by the add
1031 * operation.
1032 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1033 * struct net_device *dev,
1034 * void *priv);
1035 * Callback to use for xmit over the accelerated station. This
1036 * is used in place of ndo_start_xmit on accelerated net
1037 * devices.
1038 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1039 * struct net_device *dev
1040 * netdev_features_t features);
1041 * Called by core transmit path to determine if device is capable of
1042 * performing offload operations on a given packet. This is to give
1043 * the device an opportunity to implement any restrictions that cannot
1044 * be otherwise expressed by feature flags. The check is called with
1045 * the set of features that the stack has calculated and it returns
1046 * those the driver believes to be appropriate.
1047 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1048 * int queue_index, u32 maxrate);
1049 * Called when a user wants to set a max-rate limitation of specific
1050 * TX queue.
1051 * int (*ndo_get_iflink)(const struct net_device *dev);
1052 * Called to get the iflink value of this device.
1053 * void (*ndo_change_proto_down)(struct net_device *dev,
1054 * bool proto_down);
1055 * This function is used to pass protocol port error state information
1056 * to the switch driver. The switch driver can react to the proto_down
1057 * by doing a phys down on the associated switch port.
1058 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1059 * This function is used to get egress tunnel information for given skb.
1060 * This is useful for retrieving outer tunnel header parameters while
1061 * sampling packet.
1062 *
1063 */
1064 struct net_device_ops {
1065 int (*ndo_init)(struct net_device *dev);
1066 void (*ndo_uninit)(struct net_device *dev);
1067 int (*ndo_open)(struct net_device *dev);
1068 int (*ndo_stop)(struct net_device *dev);
1069 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1070 struct net_device *dev);
1071 u16 (*ndo_select_queue)(struct net_device *dev,
1072 struct sk_buff *skb,
1073 void *accel_priv,
1074 select_queue_fallback_t fallback);
1075 void (*ndo_change_rx_flags)(struct net_device *dev,
1076 int flags);
1077 void (*ndo_set_rx_mode)(struct net_device *dev);
1078 int (*ndo_set_mac_address)(struct net_device *dev,
1079 void *addr);
1080 int (*ndo_validate_addr)(struct net_device *dev);
1081 int (*ndo_do_ioctl)(struct net_device *dev,
1082 struct ifreq *ifr, int cmd);
1083 int (*ndo_set_config)(struct net_device *dev,
1084 struct ifmap *map);
1085 int (*ndo_change_mtu)(struct net_device *dev,
1086 int new_mtu);
1087 int (*ndo_neigh_setup)(struct net_device *dev,
1088 struct neigh_parms *);
1089 void (*ndo_tx_timeout) (struct net_device *dev);
1090
1091 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1092 struct rtnl_link_stats64 *storage);
1093 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1094
1095 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1096 __be16 proto, u16 vid);
1097 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1098 __be16 proto, u16 vid);
1099 #ifdef CONFIG_NET_POLL_CONTROLLER
1100 void (*ndo_poll_controller)(struct net_device *dev);
1101 int (*ndo_netpoll_setup)(struct net_device *dev,
1102 struct netpoll_info *info);
1103 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1104 #endif
1105 #ifdef CONFIG_NET_RX_BUSY_POLL
1106 int (*ndo_busy_poll)(struct napi_struct *dev);
1107 #endif
1108 int (*ndo_set_vf_mac)(struct net_device *dev,
1109 int queue, u8 *mac);
1110 int (*ndo_set_vf_vlan)(struct net_device *dev,
1111 int queue, u16 vlan, u8 qos);
1112 int (*ndo_set_vf_rate)(struct net_device *dev,
1113 int vf, int min_tx_rate,
1114 int max_tx_rate);
1115 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1116 int vf, bool setting);
1117 int (*ndo_set_vf_trust)(struct net_device *dev,
1118 int vf, bool setting);
1119 int (*ndo_get_vf_config)(struct net_device *dev,
1120 int vf,
1121 struct ifla_vf_info *ivf);
1122 int (*ndo_set_vf_link_state)(struct net_device *dev,
1123 int vf, int link_state);
1124 int (*ndo_get_vf_stats)(struct net_device *dev,
1125 int vf,
1126 struct ifla_vf_stats
1127 *vf_stats);
1128 int (*ndo_set_vf_port)(struct net_device *dev,
1129 int vf,
1130 struct nlattr *port[]);
1131 int (*ndo_get_vf_port)(struct net_device *dev,
1132 int vf, struct sk_buff *skb);
1133 int (*ndo_set_vf_rss_query_en)(
1134 struct net_device *dev,
1135 int vf, bool setting);
1136 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1137 #if IS_ENABLED(CONFIG_FCOE)
1138 int (*ndo_fcoe_enable)(struct net_device *dev);
1139 int (*ndo_fcoe_disable)(struct net_device *dev);
1140 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1141 u16 xid,
1142 struct scatterlist *sgl,
1143 unsigned int sgc);
1144 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1145 u16 xid);
1146 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1147 u16 xid,
1148 struct scatterlist *sgl,
1149 unsigned int sgc);
1150 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1151 struct netdev_fcoe_hbainfo *hbainfo);
1152 #endif
1153
1154 #if IS_ENABLED(CONFIG_LIBFCOE)
1155 #define NETDEV_FCOE_WWNN 0
1156 #define NETDEV_FCOE_WWPN 1
1157 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1158 u64 *wwn, int type);
1159 #endif
1160
1161 #ifdef CONFIG_RFS_ACCEL
1162 int (*ndo_rx_flow_steer)(struct net_device *dev,
1163 const struct sk_buff *skb,
1164 u16 rxq_index,
1165 u32 flow_id);
1166 #endif
1167 int (*ndo_add_slave)(struct net_device *dev,
1168 struct net_device *slave_dev);
1169 int (*ndo_del_slave)(struct net_device *dev,
1170 struct net_device *slave_dev);
1171 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1172 netdev_features_t features);
1173 int (*ndo_set_features)(struct net_device *dev,
1174 netdev_features_t features);
1175 int (*ndo_neigh_construct)(struct neighbour *n);
1176 void (*ndo_neigh_destroy)(struct neighbour *n);
1177
1178 int (*ndo_fdb_add)(struct ndmsg *ndm,
1179 struct nlattr *tb[],
1180 struct net_device *dev,
1181 const unsigned char *addr,
1182 u16 vid,
1183 u16 flags);
1184 int (*ndo_fdb_del)(struct ndmsg *ndm,
1185 struct nlattr *tb[],
1186 struct net_device *dev,
1187 const unsigned char *addr,
1188 u16 vid);
1189 int (*ndo_fdb_dump)(struct sk_buff *skb,
1190 struct netlink_callback *cb,
1191 struct net_device *dev,
1192 struct net_device *filter_dev,
1193 int idx);
1194
1195 int (*ndo_bridge_setlink)(struct net_device *dev,
1196 struct nlmsghdr *nlh,
1197 u16 flags);
1198 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1199 u32 pid, u32 seq,
1200 struct net_device *dev,
1201 u32 filter_mask,
1202 int nlflags);
1203 int (*ndo_bridge_dellink)(struct net_device *dev,
1204 struct nlmsghdr *nlh,
1205 u16 flags);
1206 int (*ndo_change_carrier)(struct net_device *dev,
1207 bool new_carrier);
1208 int (*ndo_get_phys_port_id)(struct net_device *dev,
1209 struct netdev_phys_item_id *ppid);
1210 int (*ndo_get_phys_port_name)(struct net_device *dev,
1211 char *name, size_t len);
1212 void (*ndo_add_vxlan_port)(struct net_device *dev,
1213 sa_family_t sa_family,
1214 __be16 port);
1215 void (*ndo_del_vxlan_port)(struct net_device *dev,
1216 sa_family_t sa_family,
1217 __be16 port);
1218
1219 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1220 struct net_device *dev);
1221 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1222 void *priv);
1223
1224 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1225 struct net_device *dev,
1226 void *priv);
1227 int (*ndo_get_lock_subclass)(struct net_device *dev);
1228 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1229 struct net_device *dev,
1230 netdev_features_t features);
1231 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1232 int queue_index,
1233 u32 maxrate);
1234 int (*ndo_get_iflink)(const struct net_device *dev);
1235 int (*ndo_change_proto_down)(struct net_device *dev,
1236 bool proto_down);
1237 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1238 struct sk_buff *skb);
1239 };
1240
1241 /**
1242 * enum net_device_priv_flags - &struct net_device priv_flags
1243 *
1244 * These are the &struct net_device, they are only set internally
1245 * by drivers and used in the kernel. These flags are invisible to
1246 * userspace, this means that the order of these flags can change
1247 * during any kernel release.
1248 *
1249 * You should have a pretty good reason to be extending these flags.
1250 *
1251 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1252 * @IFF_EBRIDGE: Ethernet bridging device
1253 * @IFF_BONDING: bonding master or slave
1254 * @IFF_ISATAP: ISATAP interface (RFC4214)
1255 * @IFF_WAN_HDLC: WAN HDLC device
1256 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1257 * release skb->dst
1258 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1259 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1260 * @IFF_MACVLAN_PORT: device used as macvlan port
1261 * @IFF_BRIDGE_PORT: device used as bridge port
1262 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1263 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1264 * @IFF_UNICAST_FLT: Supports unicast filtering
1265 * @IFF_TEAM_PORT: device used as team port
1266 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1267 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1268 * change when it's running
1269 * @IFF_MACVLAN: Macvlan device
1270 * @IFF_L3MDEV_MASTER: device is an L3 master device
1271 * @IFF_NO_QUEUE: device can run without qdisc attached
1272 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1273 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1274 */
1275 enum netdev_priv_flags {
1276 IFF_802_1Q_VLAN = 1<<0,
1277 IFF_EBRIDGE = 1<<1,
1278 IFF_BONDING = 1<<2,
1279 IFF_ISATAP = 1<<3,
1280 IFF_WAN_HDLC = 1<<4,
1281 IFF_XMIT_DST_RELEASE = 1<<5,
1282 IFF_DONT_BRIDGE = 1<<6,
1283 IFF_DISABLE_NETPOLL = 1<<7,
1284 IFF_MACVLAN_PORT = 1<<8,
1285 IFF_BRIDGE_PORT = 1<<9,
1286 IFF_OVS_DATAPATH = 1<<10,
1287 IFF_TX_SKB_SHARING = 1<<11,
1288 IFF_UNICAST_FLT = 1<<12,
1289 IFF_TEAM_PORT = 1<<13,
1290 IFF_SUPP_NOFCS = 1<<14,
1291 IFF_LIVE_ADDR_CHANGE = 1<<15,
1292 IFF_MACVLAN = 1<<16,
1293 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1294 IFF_IPVLAN_MASTER = 1<<18,
1295 IFF_IPVLAN_SLAVE = 1<<19,
1296 IFF_L3MDEV_MASTER = 1<<20,
1297 IFF_NO_QUEUE = 1<<21,
1298 IFF_OPENVSWITCH = 1<<22,
1299 IFF_L3MDEV_SLAVE = 1<<23,
1300 };
1301
1302 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1303 #define IFF_EBRIDGE IFF_EBRIDGE
1304 #define IFF_BONDING IFF_BONDING
1305 #define IFF_ISATAP IFF_ISATAP
1306 #define IFF_WAN_HDLC IFF_WAN_HDLC
1307 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1308 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1309 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1310 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1311 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1312 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1313 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1314 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1315 #define IFF_TEAM_PORT IFF_TEAM_PORT
1316 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1317 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1318 #define IFF_MACVLAN IFF_MACVLAN
1319 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1320 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1321 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1322 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1323 #define IFF_NO_QUEUE IFF_NO_QUEUE
1324 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1325 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1326
1327 /**
1328 * struct net_device - The DEVICE structure.
1329 * Actually, this whole structure is a big mistake. It mixes I/O
1330 * data with strictly "high-level" data, and it has to know about
1331 * almost every data structure used in the INET module.
1332 *
1333 * @name: This is the first field of the "visible" part of this structure
1334 * (i.e. as seen by users in the "Space.c" file). It is the name
1335 * of the interface.
1336 *
1337 * @name_hlist: Device name hash chain, please keep it close to name[]
1338 * @ifalias: SNMP alias
1339 * @mem_end: Shared memory end
1340 * @mem_start: Shared memory start
1341 * @base_addr: Device I/O address
1342 * @irq: Device IRQ number
1343 *
1344 * @carrier_changes: Stats to monitor carrier on<->off transitions
1345 *
1346 * @state: Generic network queuing layer state, see netdev_state_t
1347 * @dev_list: The global list of network devices
1348 * @napi_list: List entry, that is used for polling napi devices
1349 * @unreg_list: List entry, that is used, when we are unregistering the
1350 * device, see the function unregister_netdev
1351 * @close_list: List entry, that is used, when we are closing the device
1352 *
1353 * @adj_list: Directly linked devices, like slaves for bonding
1354 * @all_adj_list: All linked devices, *including* neighbours
1355 * @features: Currently active device features
1356 * @hw_features: User-changeable features
1357 *
1358 * @wanted_features: User-requested features
1359 * @vlan_features: Mask of features inheritable by VLAN devices
1360 *
1361 * @hw_enc_features: Mask of features inherited by encapsulating devices
1362 * This field indicates what encapsulation
1363 * offloads the hardware is capable of doing,
1364 * and drivers will need to set them appropriately.
1365 *
1366 * @mpls_features: Mask of features inheritable by MPLS
1367 *
1368 * @ifindex: interface index
1369 * @group: The group, that the device belongs to
1370 *
1371 * @stats: Statistics struct, which was left as a legacy, use
1372 * rtnl_link_stats64 instead
1373 *
1374 * @rx_dropped: Dropped packets by core network,
1375 * do not use this in drivers
1376 * @tx_dropped: Dropped packets by core network,
1377 * do not use this in drivers
1378 *
1379 * @wireless_handlers: List of functions to handle Wireless Extensions,
1380 * instead of ioctl,
1381 * see <net/iw_handler.h> for details.
1382 * @wireless_data: Instance data managed by the core of wireless extensions
1383 *
1384 * @netdev_ops: Includes several pointers to callbacks,
1385 * if one wants to override the ndo_*() functions
1386 * @ethtool_ops: Management operations
1387 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1388 * of Layer 2 headers.
1389 *
1390 * @flags: Interface flags (a la BSD)
1391 * @priv_flags: Like 'flags' but invisible to userspace,
1392 * see if.h for the definitions
1393 * @gflags: Global flags ( kept as legacy )
1394 * @padded: How much padding added by alloc_netdev()
1395 * @operstate: RFC2863 operstate
1396 * @link_mode: Mapping policy to operstate
1397 * @if_port: Selectable AUI, TP, ...
1398 * @dma: DMA channel
1399 * @mtu: Interface MTU value
1400 * @type: Interface hardware type
1401 * @hard_header_len: Hardware header length
1402 *
1403 * @needed_headroom: Extra headroom the hardware may need, but not in all
1404 * cases can this be guaranteed
1405 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1406 * cases can this be guaranteed. Some cases also use
1407 * LL_MAX_HEADER instead to allocate the skb
1408 *
1409 * interface address info:
1410 *
1411 * @perm_addr: Permanent hw address
1412 * @addr_assign_type: Hw address assignment type
1413 * @addr_len: Hardware address length
1414 * @neigh_priv_len; Used in neigh_alloc(),
1415 * initialized only in atm/clip.c
1416 * @dev_id: Used to differentiate devices that share
1417 * the same link layer address
1418 * @dev_port: Used to differentiate devices that share
1419 * the same function
1420 * @addr_list_lock: XXX: need comments on this one
1421 * @uc_promisc: Counter, that indicates, that promiscuous mode
1422 * has been enabled due to the need to listen to
1423 * additional unicast addresses in a device that
1424 * does not implement ndo_set_rx_mode()
1425 * @uc: unicast mac addresses
1426 * @mc: multicast mac addresses
1427 * @dev_addrs: list of device hw addresses
1428 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1429 * @promiscuity: Number of times, the NIC is told to work in
1430 * Promiscuous mode, if it becomes 0 the NIC will
1431 * exit from working in Promiscuous mode
1432 * @allmulti: Counter, enables or disables allmulticast mode
1433 *
1434 * @vlan_info: VLAN info
1435 * @dsa_ptr: dsa specific data
1436 * @tipc_ptr: TIPC specific data
1437 * @atalk_ptr: AppleTalk link
1438 * @ip_ptr: IPv4 specific data
1439 * @dn_ptr: DECnet specific data
1440 * @ip6_ptr: IPv6 specific data
1441 * @ax25_ptr: AX.25 specific data
1442 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1443 *
1444 * @last_rx: Time of last Rx
1445 * @dev_addr: Hw address (before bcast,
1446 * because most packets are unicast)
1447 *
1448 * @_rx: Array of RX queues
1449 * @num_rx_queues: Number of RX queues
1450 * allocated at register_netdev() time
1451 * @real_num_rx_queues: Number of RX queues currently active in device
1452 *
1453 * @rx_handler: handler for received packets
1454 * @rx_handler_data: XXX: need comments on this one
1455 * @ingress_queue: XXX: need comments on this one
1456 * @broadcast: hw bcast address
1457 *
1458 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1459 * indexed by RX queue number. Assigned by driver.
1460 * This must only be set if the ndo_rx_flow_steer
1461 * operation is defined
1462 * @index_hlist: Device index hash chain
1463 *
1464 * @_tx: Array of TX queues
1465 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1466 * @real_num_tx_queues: Number of TX queues currently active in device
1467 * @qdisc: Root qdisc from userspace point of view
1468 * @tx_queue_len: Max frames per queue allowed
1469 * @tx_global_lock: XXX: need comments on this one
1470 *
1471 * @xps_maps: XXX: need comments on this one
1472 *
1473 * @offload_fwd_mark: Offload device fwding mark
1474 *
1475 * @trans_start: Time (in jiffies) of last Tx
1476 * @watchdog_timeo: Represents the timeout that is used by
1477 * the watchdog ( see dev_watchdog() )
1478 * @watchdog_timer: List of timers
1479 *
1480 * @pcpu_refcnt: Number of references to this device
1481 * @todo_list: Delayed register/unregister
1482 * @link_watch_list: XXX: need comments on this one
1483 *
1484 * @reg_state: Register/unregister state machine
1485 * @dismantle: Device is going to be freed
1486 * @rtnl_link_state: This enum represents the phases of creating
1487 * a new link
1488 *
1489 * @destructor: Called from unregister,
1490 * can be used to call free_netdev
1491 * @npinfo: XXX: need comments on this one
1492 * @nd_net: Network namespace this network device is inside
1493 *
1494 * @ml_priv: Mid-layer private
1495 * @lstats: Loopback statistics
1496 * @tstats: Tunnel statistics
1497 * @dstats: Dummy statistics
1498 * @vstats: Virtual ethernet statistics
1499 *
1500 * @garp_port: GARP
1501 * @mrp_port: MRP
1502 *
1503 * @dev: Class/net/name entry
1504 * @sysfs_groups: Space for optional device, statistics and wireless
1505 * sysfs groups
1506 *
1507 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1508 * @rtnl_link_ops: Rtnl_link_ops
1509 *
1510 * @gso_max_size: Maximum size of generic segmentation offload
1511 * @gso_max_segs: Maximum number of segments that can be passed to the
1512 * NIC for GSO
1513 * @gso_min_segs: Minimum number of segments that can be passed to the
1514 * NIC for GSO
1515 *
1516 * @dcbnl_ops: Data Center Bridging netlink ops
1517 * @num_tc: Number of traffic classes in the net device
1518 * @tc_to_txq: XXX: need comments on this one
1519 * @prio_tc_map XXX: need comments on this one
1520 *
1521 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1522 *
1523 * @priomap: XXX: need comments on this one
1524 * @phydev: Physical device may attach itself
1525 * for hardware timestamping
1526 *
1527 * @qdisc_tx_busylock: XXX: need comments on this one
1528 *
1529 * @proto_down: protocol port state information can be sent to the
1530 * switch driver and used to set the phys state of the
1531 * switch port.
1532 *
1533 * FIXME: cleanup struct net_device such that network protocol info
1534 * moves out.
1535 */
1536
1537 struct net_device {
1538 char name[IFNAMSIZ];
1539 struct hlist_node name_hlist;
1540 char *ifalias;
1541 /*
1542 * I/O specific fields
1543 * FIXME: Merge these and struct ifmap into one
1544 */
1545 unsigned long mem_end;
1546 unsigned long mem_start;
1547 unsigned long base_addr;
1548 int irq;
1549
1550 atomic_t carrier_changes;
1551
1552 /*
1553 * Some hardware also needs these fields (state,dev_list,
1554 * napi_list,unreg_list,close_list) but they are not
1555 * part of the usual set specified in Space.c.
1556 */
1557
1558 unsigned long state;
1559
1560 struct list_head dev_list;
1561 struct list_head napi_list;
1562 struct list_head unreg_list;
1563 struct list_head close_list;
1564 struct list_head ptype_all;
1565 struct list_head ptype_specific;
1566
1567 struct {
1568 struct list_head upper;
1569 struct list_head lower;
1570 } adj_list;
1571
1572 struct {
1573 struct list_head upper;
1574 struct list_head lower;
1575 } all_adj_list;
1576
1577 netdev_features_t features;
1578 netdev_features_t hw_features;
1579 netdev_features_t wanted_features;
1580 netdev_features_t vlan_features;
1581 netdev_features_t hw_enc_features;
1582 netdev_features_t mpls_features;
1583
1584 int ifindex;
1585 int group;
1586
1587 struct net_device_stats stats;
1588
1589 atomic_long_t rx_dropped;
1590 atomic_long_t tx_dropped;
1591
1592 #ifdef CONFIG_WIRELESS_EXT
1593 const struct iw_handler_def * wireless_handlers;
1594 struct iw_public_data * wireless_data;
1595 #endif
1596 const struct net_device_ops *netdev_ops;
1597 const struct ethtool_ops *ethtool_ops;
1598 #ifdef CONFIG_NET_SWITCHDEV
1599 const struct switchdev_ops *switchdev_ops;
1600 #endif
1601 #ifdef CONFIG_NET_L3_MASTER_DEV
1602 const struct l3mdev_ops *l3mdev_ops;
1603 #endif
1604
1605 const struct header_ops *header_ops;
1606
1607 unsigned int flags;
1608 unsigned int priv_flags;
1609
1610 unsigned short gflags;
1611 unsigned short padded;
1612
1613 unsigned char operstate;
1614 unsigned char link_mode;
1615
1616 unsigned char if_port;
1617 unsigned char dma;
1618
1619 unsigned int mtu;
1620 unsigned short type;
1621 unsigned short hard_header_len;
1622
1623 unsigned short needed_headroom;
1624 unsigned short needed_tailroom;
1625
1626 /* Interface address info. */
1627 unsigned char perm_addr[MAX_ADDR_LEN];
1628 unsigned char addr_assign_type;
1629 unsigned char addr_len;
1630 unsigned short neigh_priv_len;
1631 unsigned short dev_id;
1632 unsigned short dev_port;
1633 spinlock_t addr_list_lock;
1634 unsigned char name_assign_type;
1635 bool uc_promisc;
1636 struct netdev_hw_addr_list uc;
1637 struct netdev_hw_addr_list mc;
1638 struct netdev_hw_addr_list dev_addrs;
1639
1640 #ifdef CONFIG_SYSFS
1641 struct kset *queues_kset;
1642 #endif
1643 unsigned int promiscuity;
1644 unsigned int allmulti;
1645
1646
1647 /* Protocol specific pointers */
1648
1649 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1650 struct vlan_info __rcu *vlan_info;
1651 #endif
1652 #if IS_ENABLED(CONFIG_NET_DSA)
1653 struct dsa_switch_tree *dsa_ptr;
1654 #endif
1655 #if IS_ENABLED(CONFIG_TIPC)
1656 struct tipc_bearer __rcu *tipc_ptr;
1657 #endif
1658 void *atalk_ptr;
1659 struct in_device __rcu *ip_ptr;
1660 struct dn_dev __rcu *dn_ptr;
1661 struct inet6_dev __rcu *ip6_ptr;
1662 void *ax25_ptr;
1663 struct wireless_dev *ieee80211_ptr;
1664 struct wpan_dev *ieee802154_ptr;
1665 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1666 struct mpls_dev __rcu *mpls_ptr;
1667 #endif
1668
1669 /*
1670 * Cache lines mostly used on receive path (including eth_type_trans())
1671 */
1672 unsigned long last_rx;
1673
1674 /* Interface address info used in eth_type_trans() */
1675 unsigned char *dev_addr;
1676
1677
1678 #ifdef CONFIG_SYSFS
1679 struct netdev_rx_queue *_rx;
1680
1681 unsigned int num_rx_queues;
1682 unsigned int real_num_rx_queues;
1683
1684 #endif
1685
1686 unsigned long gro_flush_timeout;
1687 rx_handler_func_t __rcu *rx_handler;
1688 void __rcu *rx_handler_data;
1689
1690 #ifdef CONFIG_NET_CLS_ACT
1691 struct tcf_proto __rcu *ingress_cl_list;
1692 #endif
1693 struct netdev_queue __rcu *ingress_queue;
1694 #ifdef CONFIG_NETFILTER_INGRESS
1695 struct list_head nf_hooks_ingress;
1696 #endif
1697
1698 unsigned char broadcast[MAX_ADDR_LEN];
1699 #ifdef CONFIG_RFS_ACCEL
1700 struct cpu_rmap *rx_cpu_rmap;
1701 #endif
1702 struct hlist_node index_hlist;
1703
1704 /*
1705 * Cache lines mostly used on transmit path
1706 */
1707 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1708 unsigned int num_tx_queues;
1709 unsigned int real_num_tx_queues;
1710 struct Qdisc *qdisc;
1711 unsigned long tx_queue_len;
1712 spinlock_t tx_global_lock;
1713 int watchdog_timeo;
1714
1715 #ifdef CONFIG_XPS
1716 struct xps_dev_maps __rcu *xps_maps;
1717 #endif
1718
1719 #ifdef CONFIG_NET_SWITCHDEV
1720 u32 offload_fwd_mark;
1721 #endif
1722
1723 /* These may be needed for future network-power-down code. */
1724
1725 /*
1726 * trans_start here is expensive for high speed devices on SMP,
1727 * please use netdev_queue->trans_start instead.
1728 */
1729 unsigned long trans_start;
1730
1731 struct timer_list watchdog_timer;
1732
1733 int __percpu *pcpu_refcnt;
1734 struct list_head todo_list;
1735
1736 struct list_head link_watch_list;
1737
1738 enum { NETREG_UNINITIALIZED=0,
1739 NETREG_REGISTERED, /* completed register_netdevice */
1740 NETREG_UNREGISTERING, /* called unregister_netdevice */
1741 NETREG_UNREGISTERED, /* completed unregister todo */
1742 NETREG_RELEASED, /* called free_netdev */
1743 NETREG_DUMMY, /* dummy device for NAPI poll */
1744 } reg_state:8;
1745
1746 bool dismantle;
1747
1748 enum {
1749 RTNL_LINK_INITIALIZED,
1750 RTNL_LINK_INITIALIZING,
1751 } rtnl_link_state:16;
1752
1753 void (*destructor)(struct net_device *dev);
1754
1755 #ifdef CONFIG_NETPOLL
1756 struct netpoll_info __rcu *npinfo;
1757 #endif
1758
1759 possible_net_t nd_net;
1760
1761 /* mid-layer private */
1762 union {
1763 void *ml_priv;
1764 struct pcpu_lstats __percpu *lstats;
1765 struct pcpu_sw_netstats __percpu *tstats;
1766 struct pcpu_dstats __percpu *dstats;
1767 struct pcpu_vstats __percpu *vstats;
1768 };
1769
1770 struct garp_port __rcu *garp_port;
1771 struct mrp_port __rcu *mrp_port;
1772
1773 struct device dev;
1774 const struct attribute_group *sysfs_groups[4];
1775 const struct attribute_group *sysfs_rx_queue_group;
1776
1777 const struct rtnl_link_ops *rtnl_link_ops;
1778
1779 /* for setting kernel sock attribute on TCP connection setup */
1780 #define GSO_MAX_SIZE 65536
1781 unsigned int gso_max_size;
1782 #define GSO_MAX_SEGS 65535
1783 u16 gso_max_segs;
1784 u16 gso_min_segs;
1785 #ifdef CONFIG_DCB
1786 const struct dcbnl_rtnl_ops *dcbnl_ops;
1787 #endif
1788 u8 num_tc;
1789 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1790 u8 prio_tc_map[TC_BITMASK + 1];
1791
1792 #if IS_ENABLED(CONFIG_FCOE)
1793 unsigned int fcoe_ddp_xid;
1794 #endif
1795 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1796 struct netprio_map __rcu *priomap;
1797 #endif
1798 struct phy_device *phydev;
1799 struct lock_class_key *qdisc_tx_busylock;
1800 bool proto_down;
1801 };
1802 #define to_net_dev(d) container_of(d, struct net_device, dev)
1803
1804 #define NETDEV_ALIGN 32
1805
1806 static inline
1807 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1808 {
1809 return dev->prio_tc_map[prio & TC_BITMASK];
1810 }
1811
1812 static inline
1813 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1814 {
1815 if (tc >= dev->num_tc)
1816 return -EINVAL;
1817
1818 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1819 return 0;
1820 }
1821
1822 static inline
1823 void netdev_reset_tc(struct net_device *dev)
1824 {
1825 dev->num_tc = 0;
1826 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1827 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1828 }
1829
1830 static inline
1831 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1832 {
1833 if (tc >= dev->num_tc)
1834 return -EINVAL;
1835
1836 dev->tc_to_txq[tc].count = count;
1837 dev->tc_to_txq[tc].offset = offset;
1838 return 0;
1839 }
1840
1841 static inline
1842 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1843 {
1844 if (num_tc > TC_MAX_QUEUE)
1845 return -EINVAL;
1846
1847 dev->num_tc = num_tc;
1848 return 0;
1849 }
1850
1851 static inline
1852 int netdev_get_num_tc(struct net_device *dev)
1853 {
1854 return dev->num_tc;
1855 }
1856
1857 static inline
1858 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1859 unsigned int index)
1860 {
1861 return &dev->_tx[index];
1862 }
1863
1864 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1865 const struct sk_buff *skb)
1866 {
1867 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1868 }
1869
1870 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1871 void (*f)(struct net_device *,
1872 struct netdev_queue *,
1873 void *),
1874 void *arg)
1875 {
1876 unsigned int i;
1877
1878 for (i = 0; i < dev->num_tx_queues; i++)
1879 f(dev, &dev->_tx[i], arg);
1880 }
1881
1882 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1883 struct sk_buff *skb,
1884 void *accel_priv);
1885
1886 /*
1887 * Net namespace inlines
1888 */
1889 static inline
1890 struct net *dev_net(const struct net_device *dev)
1891 {
1892 return read_pnet(&dev->nd_net);
1893 }
1894
1895 static inline
1896 void dev_net_set(struct net_device *dev, struct net *net)
1897 {
1898 write_pnet(&dev->nd_net, net);
1899 }
1900
1901 static inline bool netdev_uses_dsa(struct net_device *dev)
1902 {
1903 #if IS_ENABLED(CONFIG_NET_DSA)
1904 if (dev->dsa_ptr != NULL)
1905 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1906 #endif
1907 return false;
1908 }
1909
1910 /**
1911 * netdev_priv - access network device private data
1912 * @dev: network device
1913 *
1914 * Get network device private data
1915 */
1916 static inline void *netdev_priv(const struct net_device *dev)
1917 {
1918 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1919 }
1920
1921 /* Set the sysfs physical device reference for the network logical device
1922 * if set prior to registration will cause a symlink during initialization.
1923 */
1924 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1925
1926 /* Set the sysfs device type for the network logical device to allow
1927 * fine-grained identification of different network device types. For
1928 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1929 */
1930 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1931
1932 /* Default NAPI poll() weight
1933 * Device drivers are strongly advised to not use bigger value
1934 */
1935 #define NAPI_POLL_WEIGHT 64
1936
1937 /**
1938 * netif_napi_add - initialize a napi context
1939 * @dev: network device
1940 * @napi: napi context
1941 * @poll: polling function
1942 * @weight: default weight
1943 *
1944 * netif_napi_add() must be used to initialize a napi context prior to calling
1945 * *any* of the other napi related functions.
1946 */
1947 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1948 int (*poll)(struct napi_struct *, int), int weight);
1949
1950 /**
1951 * netif_tx_napi_add - initialize a napi context
1952 * @dev: network device
1953 * @napi: napi context
1954 * @poll: polling function
1955 * @weight: default weight
1956 *
1957 * This variant of netif_napi_add() should be used from drivers using NAPI
1958 * to exclusively poll a TX queue.
1959 * This will avoid we add it into napi_hash[], thus polluting this hash table.
1960 */
1961 static inline void netif_tx_napi_add(struct net_device *dev,
1962 struct napi_struct *napi,
1963 int (*poll)(struct napi_struct *, int),
1964 int weight)
1965 {
1966 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
1967 netif_napi_add(dev, napi, poll, weight);
1968 }
1969
1970 /**
1971 * netif_napi_del - remove a napi context
1972 * @napi: napi context
1973 *
1974 * netif_napi_del() removes a napi context from the network device napi list
1975 */
1976 void netif_napi_del(struct napi_struct *napi);
1977
1978 struct napi_gro_cb {
1979 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1980 void *frag0;
1981
1982 /* Length of frag0. */
1983 unsigned int frag0_len;
1984
1985 /* This indicates where we are processing relative to skb->data. */
1986 int data_offset;
1987
1988 /* This is non-zero if the packet cannot be merged with the new skb. */
1989 u16 flush;
1990
1991 /* Save the IP ID here and check when we get to the transport layer */
1992 u16 flush_id;
1993
1994 /* Number of segments aggregated. */
1995 u16 count;
1996
1997 /* Start offset for remote checksum offload */
1998 u16 gro_remcsum_start;
1999
2000 /* jiffies when first packet was created/queued */
2001 unsigned long age;
2002
2003 /* Used in ipv6_gro_receive() and foo-over-udp */
2004 u16 proto;
2005
2006 /* This is non-zero if the packet may be of the same flow. */
2007 u8 same_flow:1;
2008
2009 /* Used in udp_gro_receive */
2010 u8 udp_mark:1;
2011
2012 /* GRO checksum is valid */
2013 u8 csum_valid:1;
2014
2015 /* Number of checksums via CHECKSUM_UNNECESSARY */
2016 u8 csum_cnt:3;
2017
2018 /* Free the skb? */
2019 u8 free:2;
2020 #define NAPI_GRO_FREE 1
2021 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2022
2023 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2024 u8 is_ipv6:1;
2025
2026 /* 7 bit hole */
2027
2028 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2029 __wsum csum;
2030
2031 /* used in skb_gro_receive() slow path */
2032 struct sk_buff *last;
2033 };
2034
2035 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2036
2037 struct packet_type {
2038 __be16 type; /* This is really htons(ether_type). */
2039 struct net_device *dev; /* NULL is wildcarded here */
2040 int (*func) (struct sk_buff *,
2041 struct net_device *,
2042 struct packet_type *,
2043 struct net_device *);
2044 bool (*id_match)(struct packet_type *ptype,
2045 struct sock *sk);
2046 void *af_packet_priv;
2047 struct list_head list;
2048 };
2049
2050 struct offload_callbacks {
2051 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2052 netdev_features_t features);
2053 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2054 struct sk_buff *skb);
2055 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2056 };
2057
2058 struct packet_offload {
2059 __be16 type; /* This is really htons(ether_type). */
2060 u16 priority;
2061 struct offload_callbacks callbacks;
2062 struct list_head list;
2063 };
2064
2065 struct udp_offload;
2066
2067 struct udp_offload_callbacks {
2068 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2069 struct sk_buff *skb,
2070 struct udp_offload *uoff);
2071 int (*gro_complete)(struct sk_buff *skb,
2072 int nhoff,
2073 struct udp_offload *uoff);
2074 };
2075
2076 struct udp_offload {
2077 __be16 port;
2078 u8 ipproto;
2079 struct udp_offload_callbacks callbacks;
2080 };
2081
2082 /* often modified stats are per cpu, other are shared (netdev->stats) */
2083 struct pcpu_sw_netstats {
2084 u64 rx_packets;
2085 u64 rx_bytes;
2086 u64 tx_packets;
2087 u64 tx_bytes;
2088 struct u64_stats_sync syncp;
2089 };
2090
2091 #define __netdev_alloc_pcpu_stats(type, gfp) \
2092 ({ \
2093 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2094 if (pcpu_stats) { \
2095 int __cpu; \
2096 for_each_possible_cpu(__cpu) { \
2097 typeof(type) *stat; \
2098 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2099 u64_stats_init(&stat->syncp); \
2100 } \
2101 } \
2102 pcpu_stats; \
2103 })
2104
2105 #define netdev_alloc_pcpu_stats(type) \
2106 __netdev_alloc_pcpu_stats(type, GFP_KERNEL);
2107
2108 #include <linux/notifier.h>
2109
2110 /* netdevice notifier chain. Please remember to update the rtnetlink
2111 * notification exclusion list in rtnetlink_event() when adding new
2112 * types.
2113 */
2114 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2115 #define NETDEV_DOWN 0x0002
2116 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2117 detected a hardware crash and restarted
2118 - we can use this eg to kick tcp sessions
2119 once done */
2120 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2121 #define NETDEV_REGISTER 0x0005
2122 #define NETDEV_UNREGISTER 0x0006
2123 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2124 #define NETDEV_CHANGEADDR 0x0008
2125 #define NETDEV_GOING_DOWN 0x0009
2126 #define NETDEV_CHANGENAME 0x000A
2127 #define NETDEV_FEAT_CHANGE 0x000B
2128 #define NETDEV_BONDING_FAILOVER 0x000C
2129 #define NETDEV_PRE_UP 0x000D
2130 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2131 #define NETDEV_POST_TYPE_CHANGE 0x000F
2132 #define NETDEV_POST_INIT 0x0010
2133 #define NETDEV_UNREGISTER_FINAL 0x0011
2134 #define NETDEV_RELEASE 0x0012
2135 #define NETDEV_NOTIFY_PEERS 0x0013
2136 #define NETDEV_JOIN 0x0014
2137 #define NETDEV_CHANGEUPPER 0x0015
2138 #define NETDEV_RESEND_IGMP 0x0016
2139 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2140 #define NETDEV_CHANGEINFODATA 0x0018
2141 #define NETDEV_BONDING_INFO 0x0019
2142 #define NETDEV_PRECHANGEUPPER 0x001A
2143
2144 int register_netdevice_notifier(struct notifier_block *nb);
2145 int unregister_netdevice_notifier(struct notifier_block *nb);
2146
2147 struct netdev_notifier_info {
2148 struct net_device *dev;
2149 };
2150
2151 struct netdev_notifier_change_info {
2152 struct netdev_notifier_info info; /* must be first */
2153 unsigned int flags_changed;
2154 };
2155
2156 struct netdev_notifier_changeupper_info {
2157 struct netdev_notifier_info info; /* must be first */
2158 struct net_device *upper_dev; /* new upper dev */
2159 bool master; /* is upper dev master */
2160 bool linking; /* is the nofication for link or unlink */
2161 };
2162
2163 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2164 struct net_device *dev)
2165 {
2166 info->dev = dev;
2167 }
2168
2169 static inline struct net_device *
2170 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2171 {
2172 return info->dev;
2173 }
2174
2175 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2176
2177
2178 extern rwlock_t dev_base_lock; /* Device list lock */
2179
2180 #define for_each_netdev(net, d) \
2181 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2182 #define for_each_netdev_reverse(net, d) \
2183 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2184 #define for_each_netdev_rcu(net, d) \
2185 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2186 #define for_each_netdev_safe(net, d, n) \
2187 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2188 #define for_each_netdev_continue(net, d) \
2189 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2190 #define for_each_netdev_continue_rcu(net, d) \
2191 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2192 #define for_each_netdev_in_bond_rcu(bond, slave) \
2193 for_each_netdev_rcu(&init_net, slave) \
2194 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2195 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2196
2197 static inline struct net_device *next_net_device(struct net_device *dev)
2198 {
2199 struct list_head *lh;
2200 struct net *net;
2201
2202 net = dev_net(dev);
2203 lh = dev->dev_list.next;
2204 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2205 }
2206
2207 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2208 {
2209 struct list_head *lh;
2210 struct net *net;
2211
2212 net = dev_net(dev);
2213 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2214 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2215 }
2216
2217 static inline struct net_device *first_net_device(struct net *net)
2218 {
2219 return list_empty(&net->dev_base_head) ? NULL :
2220 net_device_entry(net->dev_base_head.next);
2221 }
2222
2223 static inline struct net_device *first_net_device_rcu(struct net *net)
2224 {
2225 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2226
2227 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2228 }
2229
2230 int netdev_boot_setup_check(struct net_device *dev);
2231 unsigned long netdev_boot_base(const char *prefix, int unit);
2232 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2233 const char *hwaddr);
2234 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2235 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2236 void dev_add_pack(struct packet_type *pt);
2237 void dev_remove_pack(struct packet_type *pt);
2238 void __dev_remove_pack(struct packet_type *pt);
2239 void dev_add_offload(struct packet_offload *po);
2240 void dev_remove_offload(struct packet_offload *po);
2241
2242 int dev_get_iflink(const struct net_device *dev);
2243 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2244 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2245 unsigned short mask);
2246 struct net_device *dev_get_by_name(struct net *net, const char *name);
2247 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2248 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2249 int dev_alloc_name(struct net_device *dev, const char *name);
2250 int dev_open(struct net_device *dev);
2251 int dev_close(struct net_device *dev);
2252 int dev_close_many(struct list_head *head, bool unlink);
2253 void dev_disable_lro(struct net_device *dev);
2254 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2255 int dev_queue_xmit(struct sk_buff *skb);
2256 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2257 int register_netdevice(struct net_device *dev);
2258 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2259 void unregister_netdevice_many(struct list_head *head);
2260 static inline void unregister_netdevice(struct net_device *dev)
2261 {
2262 unregister_netdevice_queue(dev, NULL);
2263 }
2264
2265 int netdev_refcnt_read(const struct net_device *dev);
2266 void free_netdev(struct net_device *dev);
2267 void netdev_freemem(struct net_device *dev);
2268 void synchronize_net(void);
2269 int init_dummy_netdev(struct net_device *dev);
2270
2271 DECLARE_PER_CPU(int, xmit_recursion);
2272 static inline int dev_recursion_level(void)
2273 {
2274 return this_cpu_read(xmit_recursion);
2275 }
2276
2277 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2278 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2279 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2280 int netdev_get_name(struct net *net, char *name, int ifindex);
2281 int dev_restart(struct net_device *dev);
2282 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2283
2284 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2285 {
2286 return NAPI_GRO_CB(skb)->data_offset;
2287 }
2288
2289 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2290 {
2291 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2292 }
2293
2294 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2295 {
2296 NAPI_GRO_CB(skb)->data_offset += len;
2297 }
2298
2299 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2300 unsigned int offset)
2301 {
2302 return NAPI_GRO_CB(skb)->frag0 + offset;
2303 }
2304
2305 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2306 {
2307 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2308 }
2309
2310 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2311 unsigned int offset)
2312 {
2313 if (!pskb_may_pull(skb, hlen))
2314 return NULL;
2315
2316 NAPI_GRO_CB(skb)->frag0 = NULL;
2317 NAPI_GRO_CB(skb)->frag0_len = 0;
2318 return skb->data + offset;
2319 }
2320
2321 static inline void *skb_gro_network_header(struct sk_buff *skb)
2322 {
2323 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2324 skb_network_offset(skb);
2325 }
2326
2327 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2328 const void *start, unsigned int len)
2329 {
2330 if (NAPI_GRO_CB(skb)->csum_valid)
2331 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2332 csum_partial(start, len, 0));
2333 }
2334
2335 /* GRO checksum functions. These are logical equivalents of the normal
2336 * checksum functions (in skbuff.h) except that they operate on the GRO
2337 * offsets and fields in sk_buff.
2338 */
2339
2340 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2341
2342 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2343 {
2344 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2345 }
2346
2347 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2348 bool zero_okay,
2349 __sum16 check)
2350 {
2351 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2352 skb_checksum_start_offset(skb) <
2353 skb_gro_offset(skb)) &&
2354 !skb_at_gro_remcsum_start(skb) &&
2355 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2356 (!zero_okay || check));
2357 }
2358
2359 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2360 __wsum psum)
2361 {
2362 if (NAPI_GRO_CB(skb)->csum_valid &&
2363 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2364 return 0;
2365
2366 NAPI_GRO_CB(skb)->csum = psum;
2367
2368 return __skb_gro_checksum_complete(skb);
2369 }
2370
2371 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2372 {
2373 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2374 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2375 NAPI_GRO_CB(skb)->csum_cnt--;
2376 } else {
2377 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2378 * verified a new top level checksum or an encapsulated one
2379 * during GRO. This saves work if we fallback to normal path.
2380 */
2381 __skb_incr_checksum_unnecessary(skb);
2382 }
2383 }
2384
2385 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2386 compute_pseudo) \
2387 ({ \
2388 __sum16 __ret = 0; \
2389 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2390 __ret = __skb_gro_checksum_validate_complete(skb, \
2391 compute_pseudo(skb, proto)); \
2392 if (__ret) \
2393 __skb_mark_checksum_bad(skb); \
2394 else \
2395 skb_gro_incr_csum_unnecessary(skb); \
2396 __ret; \
2397 })
2398
2399 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2400 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2401
2402 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2403 compute_pseudo) \
2404 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2405
2406 #define skb_gro_checksum_simple_validate(skb) \
2407 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2408
2409 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2410 {
2411 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2412 !NAPI_GRO_CB(skb)->csum_valid);
2413 }
2414
2415 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2416 __sum16 check, __wsum pseudo)
2417 {
2418 NAPI_GRO_CB(skb)->csum = ~pseudo;
2419 NAPI_GRO_CB(skb)->csum_valid = 1;
2420 }
2421
2422 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2423 do { \
2424 if (__skb_gro_checksum_convert_check(skb)) \
2425 __skb_gro_checksum_convert(skb, check, \
2426 compute_pseudo(skb, proto)); \
2427 } while (0)
2428
2429 struct gro_remcsum {
2430 int offset;
2431 __wsum delta;
2432 };
2433
2434 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2435 {
2436 grc->offset = 0;
2437 grc->delta = 0;
2438 }
2439
2440 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2441 unsigned int off, size_t hdrlen,
2442 int start, int offset,
2443 struct gro_remcsum *grc,
2444 bool nopartial)
2445 {
2446 __wsum delta;
2447 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2448
2449 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2450
2451 if (!nopartial) {
2452 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2453 return ptr;
2454 }
2455
2456 ptr = skb_gro_header_fast(skb, off);
2457 if (skb_gro_header_hard(skb, off + plen)) {
2458 ptr = skb_gro_header_slow(skb, off + plen, off);
2459 if (!ptr)
2460 return NULL;
2461 }
2462
2463 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2464 start, offset);
2465
2466 /* Adjust skb->csum since we changed the packet */
2467 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2468
2469 grc->offset = off + hdrlen + offset;
2470 grc->delta = delta;
2471
2472 return ptr;
2473 }
2474
2475 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2476 struct gro_remcsum *grc)
2477 {
2478 void *ptr;
2479 size_t plen = grc->offset + sizeof(u16);
2480
2481 if (!grc->delta)
2482 return;
2483
2484 ptr = skb_gro_header_fast(skb, grc->offset);
2485 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2486 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2487 if (!ptr)
2488 return;
2489 }
2490
2491 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2492 }
2493
2494 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2495 unsigned short type,
2496 const void *daddr, const void *saddr,
2497 unsigned int len)
2498 {
2499 if (!dev->header_ops || !dev->header_ops->create)
2500 return 0;
2501
2502 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2503 }
2504
2505 static inline int dev_parse_header(const struct sk_buff *skb,
2506 unsigned char *haddr)
2507 {
2508 const struct net_device *dev = skb->dev;
2509
2510 if (!dev->header_ops || !dev->header_ops->parse)
2511 return 0;
2512 return dev->header_ops->parse(skb, haddr);
2513 }
2514
2515 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2516 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2517 static inline int unregister_gifconf(unsigned int family)
2518 {
2519 return register_gifconf(family, NULL);
2520 }
2521
2522 #ifdef CONFIG_NET_FLOW_LIMIT
2523 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2524 struct sd_flow_limit {
2525 u64 count;
2526 unsigned int num_buckets;
2527 unsigned int history_head;
2528 u16 history[FLOW_LIMIT_HISTORY];
2529 u8 buckets[];
2530 };
2531
2532 extern int netdev_flow_limit_table_len;
2533 #endif /* CONFIG_NET_FLOW_LIMIT */
2534
2535 /*
2536 * Incoming packets are placed on per-cpu queues
2537 */
2538 struct softnet_data {
2539 struct list_head poll_list;
2540 struct sk_buff_head process_queue;
2541
2542 /* stats */
2543 unsigned int processed;
2544 unsigned int time_squeeze;
2545 unsigned int cpu_collision;
2546 unsigned int received_rps;
2547 #ifdef CONFIG_RPS
2548 struct softnet_data *rps_ipi_list;
2549 #endif
2550 #ifdef CONFIG_NET_FLOW_LIMIT
2551 struct sd_flow_limit __rcu *flow_limit;
2552 #endif
2553 struct Qdisc *output_queue;
2554 struct Qdisc **output_queue_tailp;
2555 struct sk_buff *completion_queue;
2556
2557 #ifdef CONFIG_RPS
2558 /* Elements below can be accessed between CPUs for RPS */
2559 struct call_single_data csd ____cacheline_aligned_in_smp;
2560 struct softnet_data *rps_ipi_next;
2561 unsigned int cpu;
2562 unsigned int input_queue_head;
2563 unsigned int input_queue_tail;
2564 #endif
2565 unsigned int dropped;
2566 struct sk_buff_head input_pkt_queue;
2567 struct napi_struct backlog;
2568
2569 };
2570
2571 static inline void input_queue_head_incr(struct softnet_data *sd)
2572 {
2573 #ifdef CONFIG_RPS
2574 sd->input_queue_head++;
2575 #endif
2576 }
2577
2578 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2579 unsigned int *qtail)
2580 {
2581 #ifdef CONFIG_RPS
2582 *qtail = ++sd->input_queue_tail;
2583 #endif
2584 }
2585
2586 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2587
2588 void __netif_schedule(struct Qdisc *q);
2589 void netif_schedule_queue(struct netdev_queue *txq);
2590
2591 static inline void netif_tx_schedule_all(struct net_device *dev)
2592 {
2593 unsigned int i;
2594
2595 for (i = 0; i < dev->num_tx_queues; i++)
2596 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2597 }
2598
2599 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2600 {
2601 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2602 }
2603
2604 /**
2605 * netif_start_queue - allow transmit
2606 * @dev: network device
2607 *
2608 * Allow upper layers to call the device hard_start_xmit routine.
2609 */
2610 static inline void netif_start_queue(struct net_device *dev)
2611 {
2612 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2613 }
2614
2615 static inline void netif_tx_start_all_queues(struct net_device *dev)
2616 {
2617 unsigned int i;
2618
2619 for (i = 0; i < dev->num_tx_queues; i++) {
2620 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2621 netif_tx_start_queue(txq);
2622 }
2623 }
2624
2625 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2626
2627 /**
2628 * netif_wake_queue - restart transmit
2629 * @dev: network device
2630 *
2631 * Allow upper layers to call the device hard_start_xmit routine.
2632 * Used for flow control when transmit resources are available.
2633 */
2634 static inline void netif_wake_queue(struct net_device *dev)
2635 {
2636 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2637 }
2638
2639 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2640 {
2641 unsigned int i;
2642
2643 for (i = 0; i < dev->num_tx_queues; i++) {
2644 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2645 netif_tx_wake_queue(txq);
2646 }
2647 }
2648
2649 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2650 {
2651 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2652 }
2653
2654 /**
2655 * netif_stop_queue - stop transmitted packets
2656 * @dev: network device
2657 *
2658 * Stop upper layers calling the device hard_start_xmit routine.
2659 * Used for flow control when transmit resources are unavailable.
2660 */
2661 static inline void netif_stop_queue(struct net_device *dev)
2662 {
2663 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2664 }
2665
2666 void netif_tx_stop_all_queues(struct net_device *dev);
2667
2668 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2669 {
2670 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2671 }
2672
2673 /**
2674 * netif_queue_stopped - test if transmit queue is flowblocked
2675 * @dev: network device
2676 *
2677 * Test if transmit queue on device is currently unable to send.
2678 */
2679 static inline bool netif_queue_stopped(const struct net_device *dev)
2680 {
2681 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2682 }
2683
2684 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2685 {
2686 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2687 }
2688
2689 static inline bool
2690 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2691 {
2692 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2693 }
2694
2695 static inline bool
2696 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2697 {
2698 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2699 }
2700
2701 /**
2702 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2703 * @dev_queue: pointer to transmit queue
2704 *
2705 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2706 * to give appropriate hint to the cpu.
2707 */
2708 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2709 {
2710 #ifdef CONFIG_BQL
2711 prefetchw(&dev_queue->dql.num_queued);
2712 #endif
2713 }
2714
2715 /**
2716 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2717 * @dev_queue: pointer to transmit queue
2718 *
2719 * BQL enabled drivers might use this helper in their TX completion path,
2720 * to give appropriate hint to the cpu.
2721 */
2722 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2723 {
2724 #ifdef CONFIG_BQL
2725 prefetchw(&dev_queue->dql.limit);
2726 #endif
2727 }
2728
2729 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2730 unsigned int bytes)
2731 {
2732 #ifdef CONFIG_BQL
2733 dql_queued(&dev_queue->dql, bytes);
2734
2735 if (likely(dql_avail(&dev_queue->dql) >= 0))
2736 return;
2737
2738 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2739
2740 /*
2741 * The XOFF flag must be set before checking the dql_avail below,
2742 * because in netdev_tx_completed_queue we update the dql_completed
2743 * before checking the XOFF flag.
2744 */
2745 smp_mb();
2746
2747 /* check again in case another CPU has just made room avail */
2748 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2749 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2750 #endif
2751 }
2752
2753 /**
2754 * netdev_sent_queue - report the number of bytes queued to hardware
2755 * @dev: network device
2756 * @bytes: number of bytes queued to the hardware device queue
2757 *
2758 * Report the number of bytes queued for sending/completion to the network
2759 * device hardware queue. @bytes should be a good approximation and should
2760 * exactly match netdev_completed_queue() @bytes
2761 */
2762 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2763 {
2764 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2765 }
2766
2767 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2768 unsigned int pkts, unsigned int bytes)
2769 {
2770 #ifdef CONFIG_BQL
2771 if (unlikely(!bytes))
2772 return;
2773
2774 dql_completed(&dev_queue->dql, bytes);
2775
2776 /*
2777 * Without the memory barrier there is a small possiblity that
2778 * netdev_tx_sent_queue will miss the update and cause the queue to
2779 * be stopped forever
2780 */
2781 smp_mb();
2782
2783 if (dql_avail(&dev_queue->dql) < 0)
2784 return;
2785
2786 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2787 netif_schedule_queue(dev_queue);
2788 #endif
2789 }
2790
2791 /**
2792 * netdev_completed_queue - report bytes and packets completed by device
2793 * @dev: network device
2794 * @pkts: actual number of packets sent over the medium
2795 * @bytes: actual number of bytes sent over the medium
2796 *
2797 * Report the number of bytes and packets transmitted by the network device
2798 * hardware queue over the physical medium, @bytes must exactly match the
2799 * @bytes amount passed to netdev_sent_queue()
2800 */
2801 static inline void netdev_completed_queue(struct net_device *dev,
2802 unsigned int pkts, unsigned int bytes)
2803 {
2804 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2805 }
2806
2807 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2808 {
2809 #ifdef CONFIG_BQL
2810 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2811 dql_reset(&q->dql);
2812 #endif
2813 }
2814
2815 /**
2816 * netdev_reset_queue - reset the packets and bytes count of a network device
2817 * @dev_queue: network device
2818 *
2819 * Reset the bytes and packet count of a network device and clear the
2820 * software flow control OFF bit for this network device
2821 */
2822 static inline void netdev_reset_queue(struct net_device *dev_queue)
2823 {
2824 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2825 }
2826
2827 /**
2828 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2829 * @dev: network device
2830 * @queue_index: given tx queue index
2831 *
2832 * Returns 0 if given tx queue index >= number of device tx queues,
2833 * otherwise returns the originally passed tx queue index.
2834 */
2835 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2836 {
2837 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2838 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2839 dev->name, queue_index,
2840 dev->real_num_tx_queues);
2841 return 0;
2842 }
2843
2844 return queue_index;
2845 }
2846
2847 /**
2848 * netif_running - test if up
2849 * @dev: network device
2850 *
2851 * Test if the device has been brought up.
2852 */
2853 static inline bool netif_running(const struct net_device *dev)
2854 {
2855 return test_bit(__LINK_STATE_START, &dev->state);
2856 }
2857
2858 /*
2859 * Routines to manage the subqueues on a device. We only need start
2860 * stop, and a check if it's stopped. All other device management is
2861 * done at the overall netdevice level.
2862 * Also test the device if we're multiqueue.
2863 */
2864
2865 /**
2866 * netif_start_subqueue - allow sending packets on subqueue
2867 * @dev: network device
2868 * @queue_index: sub queue index
2869 *
2870 * Start individual transmit queue of a device with multiple transmit queues.
2871 */
2872 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2873 {
2874 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2875
2876 netif_tx_start_queue(txq);
2877 }
2878
2879 /**
2880 * netif_stop_subqueue - stop sending packets on subqueue
2881 * @dev: network device
2882 * @queue_index: sub queue index
2883 *
2884 * Stop individual transmit queue of a device with multiple transmit queues.
2885 */
2886 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2887 {
2888 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2889 netif_tx_stop_queue(txq);
2890 }
2891
2892 /**
2893 * netif_subqueue_stopped - test status of subqueue
2894 * @dev: network device
2895 * @queue_index: sub queue index
2896 *
2897 * Check individual transmit queue of a device with multiple transmit queues.
2898 */
2899 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2900 u16 queue_index)
2901 {
2902 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2903
2904 return netif_tx_queue_stopped(txq);
2905 }
2906
2907 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2908 struct sk_buff *skb)
2909 {
2910 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2911 }
2912
2913 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2914
2915 #ifdef CONFIG_XPS
2916 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2917 u16 index);
2918 #else
2919 static inline int netif_set_xps_queue(struct net_device *dev,
2920 const struct cpumask *mask,
2921 u16 index)
2922 {
2923 return 0;
2924 }
2925 #endif
2926
2927 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2928 unsigned int num_tx_queues);
2929
2930 /*
2931 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2932 * as a distribution range limit for the returned value.
2933 */
2934 static inline u16 skb_tx_hash(const struct net_device *dev,
2935 struct sk_buff *skb)
2936 {
2937 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2938 }
2939
2940 /**
2941 * netif_is_multiqueue - test if device has multiple transmit queues
2942 * @dev: network device
2943 *
2944 * Check if device has multiple transmit queues
2945 */
2946 static inline bool netif_is_multiqueue(const struct net_device *dev)
2947 {
2948 return dev->num_tx_queues > 1;
2949 }
2950
2951 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2952
2953 #ifdef CONFIG_SYSFS
2954 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2955 #else
2956 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2957 unsigned int rxq)
2958 {
2959 return 0;
2960 }
2961 #endif
2962
2963 #ifdef CONFIG_SYSFS
2964 static inline unsigned int get_netdev_rx_queue_index(
2965 struct netdev_rx_queue *queue)
2966 {
2967 struct net_device *dev = queue->dev;
2968 int index = queue - dev->_rx;
2969
2970 BUG_ON(index >= dev->num_rx_queues);
2971 return index;
2972 }
2973 #endif
2974
2975 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2976 int netif_get_num_default_rss_queues(void);
2977
2978 enum skb_free_reason {
2979 SKB_REASON_CONSUMED,
2980 SKB_REASON_DROPPED,
2981 };
2982
2983 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2984 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2985
2986 /*
2987 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2988 * interrupt context or with hardware interrupts being disabled.
2989 * (in_irq() || irqs_disabled())
2990 *
2991 * We provide four helpers that can be used in following contexts :
2992 *
2993 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2994 * replacing kfree_skb(skb)
2995 *
2996 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2997 * Typically used in place of consume_skb(skb) in TX completion path
2998 *
2999 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3000 * replacing kfree_skb(skb)
3001 *
3002 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3003 * and consumed a packet. Used in place of consume_skb(skb)
3004 */
3005 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3006 {
3007 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3008 }
3009
3010 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3011 {
3012 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3013 }
3014
3015 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3016 {
3017 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3018 }
3019
3020 static inline void dev_consume_skb_any(struct sk_buff *skb)
3021 {
3022 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3023 }
3024
3025 int netif_rx(struct sk_buff *skb);
3026 int netif_rx_ni(struct sk_buff *skb);
3027 int netif_receive_skb(struct sk_buff *skb);
3028 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3029 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3030 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3031 gro_result_t napi_gro_frags(struct napi_struct *napi);
3032 struct packet_offload *gro_find_receive_by_type(__be16 type);
3033 struct packet_offload *gro_find_complete_by_type(__be16 type);
3034
3035 static inline void napi_free_frags(struct napi_struct *napi)
3036 {
3037 kfree_skb(napi->skb);
3038 napi->skb = NULL;
3039 }
3040
3041 int netdev_rx_handler_register(struct net_device *dev,
3042 rx_handler_func_t *rx_handler,
3043 void *rx_handler_data);
3044 void netdev_rx_handler_unregister(struct net_device *dev);
3045
3046 bool dev_valid_name(const char *name);
3047 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3048 int dev_ethtool(struct net *net, struct ifreq *);
3049 unsigned int dev_get_flags(const struct net_device *);
3050 int __dev_change_flags(struct net_device *, unsigned int flags);
3051 int dev_change_flags(struct net_device *, unsigned int);
3052 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3053 unsigned int gchanges);
3054 int dev_change_name(struct net_device *, const char *);
3055 int dev_set_alias(struct net_device *, const char *, size_t);
3056 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3057 int dev_set_mtu(struct net_device *, int);
3058 void dev_set_group(struct net_device *, int);
3059 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3060 int dev_change_carrier(struct net_device *, bool new_carrier);
3061 int dev_get_phys_port_id(struct net_device *dev,
3062 struct netdev_phys_item_id *ppid);
3063 int dev_get_phys_port_name(struct net_device *dev,
3064 char *name, size_t len);
3065 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3066 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3067 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3068 struct netdev_queue *txq, int *ret);
3069 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3070 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3071 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3072
3073 extern int netdev_budget;
3074
3075 /* Called by rtnetlink.c:rtnl_unlock() */
3076 void netdev_run_todo(void);
3077
3078 /**
3079 * dev_put - release reference to device
3080 * @dev: network device
3081 *
3082 * Release reference to device to allow it to be freed.
3083 */
3084 static inline void dev_put(struct net_device *dev)
3085 {
3086 this_cpu_dec(*dev->pcpu_refcnt);
3087 }
3088
3089 /**
3090 * dev_hold - get reference to device
3091 * @dev: network device
3092 *
3093 * Hold reference to device to keep it from being freed.
3094 */
3095 static inline void dev_hold(struct net_device *dev)
3096 {
3097 this_cpu_inc(*dev->pcpu_refcnt);
3098 }
3099
3100 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3101 * and _off may be called from IRQ context, but it is caller
3102 * who is responsible for serialization of these calls.
3103 *
3104 * The name carrier is inappropriate, these functions should really be
3105 * called netif_lowerlayer_*() because they represent the state of any
3106 * kind of lower layer not just hardware media.
3107 */
3108
3109 void linkwatch_init_dev(struct net_device *dev);
3110 void linkwatch_fire_event(struct net_device *dev);
3111 void linkwatch_forget_dev(struct net_device *dev);
3112
3113 /**
3114 * netif_carrier_ok - test if carrier present
3115 * @dev: network device
3116 *
3117 * Check if carrier is present on device
3118 */
3119 static inline bool netif_carrier_ok(const struct net_device *dev)
3120 {
3121 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3122 }
3123
3124 unsigned long dev_trans_start(struct net_device *dev);
3125
3126 void __netdev_watchdog_up(struct net_device *dev);
3127
3128 void netif_carrier_on(struct net_device *dev);
3129
3130 void netif_carrier_off(struct net_device *dev);
3131
3132 /**
3133 * netif_dormant_on - mark device as dormant.
3134 * @dev: network device
3135 *
3136 * Mark device as dormant (as per RFC2863).
3137 *
3138 * The dormant state indicates that the relevant interface is not
3139 * actually in a condition to pass packets (i.e., it is not 'up') but is
3140 * in a "pending" state, waiting for some external event. For "on-
3141 * demand" interfaces, this new state identifies the situation where the
3142 * interface is waiting for events to place it in the up state.
3143 *
3144 */
3145 static inline void netif_dormant_on(struct net_device *dev)
3146 {
3147 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3148 linkwatch_fire_event(dev);
3149 }
3150
3151 /**
3152 * netif_dormant_off - set device as not dormant.
3153 * @dev: network device
3154 *
3155 * Device is not in dormant state.
3156 */
3157 static inline void netif_dormant_off(struct net_device *dev)
3158 {
3159 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3160 linkwatch_fire_event(dev);
3161 }
3162
3163 /**
3164 * netif_dormant - test if carrier present
3165 * @dev: network device
3166 *
3167 * Check if carrier is present on device
3168 */
3169 static inline bool netif_dormant(const struct net_device *dev)
3170 {
3171 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3172 }
3173
3174
3175 /**
3176 * netif_oper_up - test if device is operational
3177 * @dev: network device
3178 *
3179 * Check if carrier is operational
3180 */
3181 static inline bool netif_oper_up(const struct net_device *dev)
3182 {
3183 return (dev->operstate == IF_OPER_UP ||
3184 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3185 }
3186
3187 /**
3188 * netif_device_present - is device available or removed
3189 * @dev: network device
3190 *
3191 * Check if device has not been removed from system.
3192 */
3193 static inline bool netif_device_present(struct net_device *dev)
3194 {
3195 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3196 }
3197
3198 void netif_device_detach(struct net_device *dev);
3199
3200 void netif_device_attach(struct net_device *dev);
3201
3202 /*
3203 * Network interface message level settings
3204 */
3205
3206 enum {
3207 NETIF_MSG_DRV = 0x0001,
3208 NETIF_MSG_PROBE = 0x0002,
3209 NETIF_MSG_LINK = 0x0004,
3210 NETIF_MSG_TIMER = 0x0008,
3211 NETIF_MSG_IFDOWN = 0x0010,
3212 NETIF_MSG_IFUP = 0x0020,
3213 NETIF_MSG_RX_ERR = 0x0040,
3214 NETIF_MSG_TX_ERR = 0x0080,
3215 NETIF_MSG_TX_QUEUED = 0x0100,
3216 NETIF_MSG_INTR = 0x0200,
3217 NETIF_MSG_TX_DONE = 0x0400,
3218 NETIF_MSG_RX_STATUS = 0x0800,
3219 NETIF_MSG_PKTDATA = 0x1000,
3220 NETIF_MSG_HW = 0x2000,
3221 NETIF_MSG_WOL = 0x4000,
3222 };
3223
3224 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3225 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3226 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3227 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3228 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3229 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3230 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3231 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3232 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3233 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3234 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3235 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3236 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3237 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3238 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3239
3240 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3241 {
3242 /* use default */
3243 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3244 return default_msg_enable_bits;
3245 if (debug_value == 0) /* no output */
3246 return 0;
3247 /* set low N bits */
3248 return (1 << debug_value) - 1;
3249 }
3250
3251 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3252 {
3253 spin_lock(&txq->_xmit_lock);
3254 txq->xmit_lock_owner = cpu;
3255 }
3256
3257 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3258 {
3259 spin_lock_bh(&txq->_xmit_lock);
3260 txq->xmit_lock_owner = smp_processor_id();
3261 }
3262
3263 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3264 {
3265 bool ok = spin_trylock(&txq->_xmit_lock);
3266 if (likely(ok))
3267 txq->xmit_lock_owner = smp_processor_id();
3268 return ok;
3269 }
3270
3271 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3272 {
3273 txq->xmit_lock_owner = -1;
3274 spin_unlock(&txq->_xmit_lock);
3275 }
3276
3277 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3278 {
3279 txq->xmit_lock_owner = -1;
3280 spin_unlock_bh(&txq->_xmit_lock);
3281 }
3282
3283 static inline void txq_trans_update(struct netdev_queue *txq)
3284 {
3285 if (txq->xmit_lock_owner != -1)
3286 txq->trans_start = jiffies;
3287 }
3288
3289 /**
3290 * netif_tx_lock - grab network device transmit lock
3291 * @dev: network device
3292 *
3293 * Get network device transmit lock
3294 */
3295 static inline void netif_tx_lock(struct net_device *dev)
3296 {
3297 unsigned int i;
3298 int cpu;
3299
3300 spin_lock(&dev->tx_global_lock);
3301 cpu = smp_processor_id();
3302 for (i = 0; i < dev->num_tx_queues; i++) {
3303 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3304
3305 /* We are the only thread of execution doing a
3306 * freeze, but we have to grab the _xmit_lock in
3307 * order to synchronize with threads which are in
3308 * the ->hard_start_xmit() handler and already
3309 * checked the frozen bit.
3310 */
3311 __netif_tx_lock(txq, cpu);
3312 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3313 __netif_tx_unlock(txq);
3314 }
3315 }
3316
3317 static inline void netif_tx_lock_bh(struct net_device *dev)
3318 {
3319 local_bh_disable();
3320 netif_tx_lock(dev);
3321 }
3322
3323 static inline void netif_tx_unlock(struct net_device *dev)
3324 {
3325 unsigned int i;
3326
3327 for (i = 0; i < dev->num_tx_queues; i++) {
3328 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3329
3330 /* No need to grab the _xmit_lock here. If the
3331 * queue is not stopped for another reason, we
3332 * force a schedule.
3333 */
3334 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3335 netif_schedule_queue(txq);
3336 }
3337 spin_unlock(&dev->tx_global_lock);
3338 }
3339
3340 static inline void netif_tx_unlock_bh(struct net_device *dev)
3341 {
3342 netif_tx_unlock(dev);
3343 local_bh_enable();
3344 }
3345
3346 #define HARD_TX_LOCK(dev, txq, cpu) { \
3347 if ((dev->features & NETIF_F_LLTX) == 0) { \
3348 __netif_tx_lock(txq, cpu); \
3349 } \
3350 }
3351
3352 #define HARD_TX_TRYLOCK(dev, txq) \
3353 (((dev->features & NETIF_F_LLTX) == 0) ? \
3354 __netif_tx_trylock(txq) : \
3355 true )
3356
3357 #define HARD_TX_UNLOCK(dev, txq) { \
3358 if ((dev->features & NETIF_F_LLTX) == 0) { \
3359 __netif_tx_unlock(txq); \
3360 } \
3361 }
3362
3363 static inline void netif_tx_disable(struct net_device *dev)
3364 {
3365 unsigned int i;
3366 int cpu;
3367
3368 local_bh_disable();
3369 cpu = smp_processor_id();
3370 for (i = 0; i < dev->num_tx_queues; i++) {
3371 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3372
3373 __netif_tx_lock(txq, cpu);
3374 netif_tx_stop_queue(txq);
3375 __netif_tx_unlock(txq);
3376 }
3377 local_bh_enable();
3378 }
3379
3380 static inline void netif_addr_lock(struct net_device *dev)
3381 {
3382 spin_lock(&dev->addr_list_lock);
3383 }
3384
3385 static inline void netif_addr_lock_nested(struct net_device *dev)
3386 {
3387 int subclass = SINGLE_DEPTH_NESTING;
3388
3389 if (dev->netdev_ops->ndo_get_lock_subclass)
3390 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3391
3392 spin_lock_nested(&dev->addr_list_lock, subclass);
3393 }
3394
3395 static inline void netif_addr_lock_bh(struct net_device *dev)
3396 {
3397 spin_lock_bh(&dev->addr_list_lock);
3398 }
3399
3400 static inline void netif_addr_unlock(struct net_device *dev)
3401 {
3402 spin_unlock(&dev->addr_list_lock);
3403 }
3404
3405 static inline void netif_addr_unlock_bh(struct net_device *dev)
3406 {
3407 spin_unlock_bh(&dev->addr_list_lock);
3408 }
3409
3410 /*
3411 * dev_addrs walker. Should be used only for read access. Call with
3412 * rcu_read_lock held.
3413 */
3414 #define for_each_dev_addr(dev, ha) \
3415 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3416
3417 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3418
3419 void ether_setup(struct net_device *dev);
3420
3421 /* Support for loadable net-drivers */
3422 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3423 unsigned char name_assign_type,
3424 void (*setup)(struct net_device *),
3425 unsigned int txqs, unsigned int rxqs);
3426 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3427 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3428
3429 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3430 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3431 count)
3432
3433 int register_netdev(struct net_device *dev);
3434 void unregister_netdev(struct net_device *dev);
3435
3436 /* General hardware address lists handling functions */
3437 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3438 struct netdev_hw_addr_list *from_list, int addr_len);
3439 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3440 struct netdev_hw_addr_list *from_list, int addr_len);
3441 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3442 struct net_device *dev,
3443 int (*sync)(struct net_device *, const unsigned char *),
3444 int (*unsync)(struct net_device *,
3445 const unsigned char *));
3446 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3447 struct net_device *dev,
3448 int (*unsync)(struct net_device *,
3449 const unsigned char *));
3450 void __hw_addr_init(struct netdev_hw_addr_list *list);
3451
3452 /* Functions used for device addresses handling */
3453 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3454 unsigned char addr_type);
3455 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3456 unsigned char addr_type);
3457 void dev_addr_flush(struct net_device *dev);
3458 int dev_addr_init(struct net_device *dev);
3459
3460 /* Functions used for unicast addresses handling */
3461 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3462 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3463 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3464 int dev_uc_sync(struct net_device *to, struct net_device *from);
3465 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3466 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3467 void dev_uc_flush(struct net_device *dev);
3468 void dev_uc_init(struct net_device *dev);
3469
3470 /**
3471 * __dev_uc_sync - Synchonize device's unicast list
3472 * @dev: device to sync
3473 * @sync: function to call if address should be added
3474 * @unsync: function to call if address should be removed
3475 *
3476 * Add newly added addresses to the interface, and release
3477 * addresses that have been deleted.
3478 **/
3479 static inline int __dev_uc_sync(struct net_device *dev,
3480 int (*sync)(struct net_device *,
3481 const unsigned char *),
3482 int (*unsync)(struct net_device *,
3483 const unsigned char *))
3484 {
3485 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3486 }
3487
3488 /**
3489 * __dev_uc_unsync - Remove synchronized addresses from device
3490 * @dev: device to sync
3491 * @unsync: function to call if address should be removed
3492 *
3493 * Remove all addresses that were added to the device by dev_uc_sync().
3494 **/
3495 static inline void __dev_uc_unsync(struct net_device *dev,
3496 int (*unsync)(struct net_device *,
3497 const unsigned char *))
3498 {
3499 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3500 }
3501
3502 /* Functions used for multicast addresses handling */
3503 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3504 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3505 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3506 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3507 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3508 int dev_mc_sync(struct net_device *to, struct net_device *from);
3509 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3510 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3511 void dev_mc_flush(struct net_device *dev);
3512 void dev_mc_init(struct net_device *dev);
3513
3514 /**
3515 * __dev_mc_sync - Synchonize device's multicast list
3516 * @dev: device to sync
3517 * @sync: function to call if address should be added
3518 * @unsync: function to call if address should be removed
3519 *
3520 * Add newly added addresses to the interface, and release
3521 * addresses that have been deleted.
3522 **/
3523 static inline int __dev_mc_sync(struct net_device *dev,
3524 int (*sync)(struct net_device *,
3525 const unsigned char *),
3526 int (*unsync)(struct net_device *,
3527 const unsigned char *))
3528 {
3529 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3530 }
3531
3532 /**
3533 * __dev_mc_unsync - Remove synchronized addresses from device
3534 * @dev: device to sync
3535 * @unsync: function to call if address should be removed
3536 *
3537 * Remove all addresses that were added to the device by dev_mc_sync().
3538 **/
3539 static inline void __dev_mc_unsync(struct net_device *dev,
3540 int (*unsync)(struct net_device *,
3541 const unsigned char *))
3542 {
3543 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3544 }
3545
3546 /* Functions used for secondary unicast and multicast support */
3547 void dev_set_rx_mode(struct net_device *dev);
3548 void __dev_set_rx_mode(struct net_device *dev);
3549 int dev_set_promiscuity(struct net_device *dev, int inc);
3550 int dev_set_allmulti(struct net_device *dev, int inc);
3551 void netdev_state_change(struct net_device *dev);
3552 void netdev_notify_peers(struct net_device *dev);
3553 void netdev_features_change(struct net_device *dev);
3554 /* Load a device via the kmod */
3555 void dev_load(struct net *net, const char *name);
3556 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3557 struct rtnl_link_stats64 *storage);
3558 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3559 const struct net_device_stats *netdev_stats);
3560
3561 extern int netdev_max_backlog;
3562 extern int netdev_tstamp_prequeue;
3563 extern int weight_p;
3564 extern int bpf_jit_enable;
3565
3566 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3567 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3568 struct list_head **iter);
3569 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3570 struct list_head **iter);
3571
3572 /* iterate through upper list, must be called under RCU read lock */
3573 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3574 for (iter = &(dev)->adj_list.upper, \
3575 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3576 updev; \
3577 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3578
3579 /* iterate through upper list, must be called under RCU read lock */
3580 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3581 for (iter = &(dev)->all_adj_list.upper, \
3582 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3583 updev; \
3584 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3585
3586 void *netdev_lower_get_next_private(struct net_device *dev,
3587 struct list_head **iter);
3588 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3589 struct list_head **iter);
3590
3591 #define netdev_for_each_lower_private(dev, priv, iter) \
3592 for (iter = (dev)->adj_list.lower.next, \
3593 priv = netdev_lower_get_next_private(dev, &(iter)); \
3594 priv; \
3595 priv = netdev_lower_get_next_private(dev, &(iter)))
3596
3597 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3598 for (iter = &(dev)->adj_list.lower, \
3599 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3600 priv; \
3601 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3602
3603 void *netdev_lower_get_next(struct net_device *dev,
3604 struct list_head **iter);
3605 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3606 for (iter = &(dev)->adj_list.lower, \
3607 ldev = netdev_lower_get_next(dev, &(iter)); \
3608 ldev; \
3609 ldev = netdev_lower_get_next(dev, &(iter)))
3610
3611 void *netdev_adjacent_get_private(struct list_head *adj_list);
3612 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3613 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3614 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3615 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3616 int netdev_master_upper_dev_link(struct net_device *dev,
3617 struct net_device *upper_dev);
3618 int netdev_master_upper_dev_link_private(struct net_device *dev,
3619 struct net_device *upper_dev,
3620 void *private);
3621 void netdev_upper_dev_unlink(struct net_device *dev,
3622 struct net_device *upper_dev);
3623 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3624 void *netdev_lower_dev_get_private(struct net_device *dev,
3625 struct net_device *lower_dev);
3626
3627 /* RSS keys are 40 or 52 bytes long */
3628 #define NETDEV_RSS_KEY_LEN 52
3629 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3630 void netdev_rss_key_fill(void *buffer, size_t len);
3631
3632 int dev_get_nest_level(struct net_device *dev,
3633 bool (*type_check)(struct net_device *dev));
3634 int skb_checksum_help(struct sk_buff *skb);
3635 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3636 netdev_features_t features, bool tx_path);
3637 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3638 netdev_features_t features);
3639
3640 struct netdev_bonding_info {
3641 ifslave slave;
3642 ifbond master;
3643 };
3644
3645 struct netdev_notifier_bonding_info {
3646 struct netdev_notifier_info info; /* must be first */
3647 struct netdev_bonding_info bonding_info;
3648 };
3649
3650 void netdev_bonding_info_change(struct net_device *dev,
3651 struct netdev_bonding_info *bonding_info);
3652
3653 static inline
3654 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3655 {
3656 return __skb_gso_segment(skb, features, true);
3657 }
3658 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3659
3660 static inline bool can_checksum_protocol(netdev_features_t features,
3661 __be16 protocol)
3662 {
3663 return ((features & NETIF_F_GEN_CSUM) ||
3664 ((features & NETIF_F_V4_CSUM) &&
3665 protocol == htons(ETH_P_IP)) ||
3666 ((features & NETIF_F_V6_CSUM) &&
3667 protocol == htons(ETH_P_IPV6)) ||
3668 ((features & NETIF_F_FCOE_CRC) &&
3669 protocol == htons(ETH_P_FCOE)));
3670 }
3671
3672 #ifdef CONFIG_BUG
3673 void netdev_rx_csum_fault(struct net_device *dev);
3674 #else
3675 static inline void netdev_rx_csum_fault(struct net_device *dev)
3676 {
3677 }
3678 #endif
3679 /* rx skb timestamps */
3680 void net_enable_timestamp(void);
3681 void net_disable_timestamp(void);
3682
3683 #ifdef CONFIG_PROC_FS
3684 int __init dev_proc_init(void);
3685 #else
3686 #define dev_proc_init() 0
3687 #endif
3688
3689 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3690 struct sk_buff *skb, struct net_device *dev,
3691 bool more)
3692 {
3693 skb->xmit_more = more ? 1 : 0;
3694 return ops->ndo_start_xmit(skb, dev);
3695 }
3696
3697 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3698 struct netdev_queue *txq, bool more)
3699 {
3700 const struct net_device_ops *ops = dev->netdev_ops;
3701 int rc;
3702
3703 rc = __netdev_start_xmit(ops, skb, dev, more);
3704 if (rc == NETDEV_TX_OK)
3705 txq_trans_update(txq);
3706
3707 return rc;
3708 }
3709
3710 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3711 const void *ns);
3712 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3713 const void *ns);
3714
3715 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3716 {
3717 return netdev_class_create_file_ns(class_attr, NULL);
3718 }
3719
3720 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3721 {
3722 netdev_class_remove_file_ns(class_attr, NULL);
3723 }
3724
3725 extern struct kobj_ns_type_operations net_ns_type_operations;
3726
3727 const char *netdev_drivername(const struct net_device *dev);
3728
3729 void linkwatch_run_queue(void);
3730
3731 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3732 netdev_features_t f2)
3733 {
3734 if (f1 & NETIF_F_GEN_CSUM)
3735 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3736 if (f2 & NETIF_F_GEN_CSUM)
3737 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3738 f1 &= f2;
3739 if (f1 & NETIF_F_GEN_CSUM)
3740 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3741
3742 return f1;
3743 }
3744
3745 static inline netdev_features_t netdev_get_wanted_features(
3746 struct net_device *dev)
3747 {
3748 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3749 }
3750 netdev_features_t netdev_increment_features(netdev_features_t all,
3751 netdev_features_t one, netdev_features_t mask);
3752
3753 /* Allow TSO being used on stacked device :
3754 * Performing the GSO segmentation before last device
3755 * is a performance improvement.
3756 */
3757 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3758 netdev_features_t mask)
3759 {
3760 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3761 }
3762
3763 int __netdev_update_features(struct net_device *dev);
3764 void netdev_update_features(struct net_device *dev);
3765 void netdev_change_features(struct net_device *dev);
3766
3767 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3768 struct net_device *dev);
3769
3770 netdev_features_t passthru_features_check(struct sk_buff *skb,
3771 struct net_device *dev,
3772 netdev_features_t features);
3773 netdev_features_t netif_skb_features(struct sk_buff *skb);
3774
3775 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3776 {
3777 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3778
3779 /* check flags correspondence */
3780 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3781 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3782 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3783 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3784 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3785 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3786 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3787 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3788 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3789 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3790 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3791 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3792 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3793
3794 return (features & feature) == feature;
3795 }
3796
3797 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3798 {
3799 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3800 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3801 }
3802
3803 static inline bool netif_needs_gso(struct sk_buff *skb,
3804 netdev_features_t features)
3805 {
3806 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3807 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3808 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3809 }
3810
3811 static inline void netif_set_gso_max_size(struct net_device *dev,
3812 unsigned int size)
3813 {
3814 dev->gso_max_size = size;
3815 }
3816
3817 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3818 int pulled_hlen, u16 mac_offset,
3819 int mac_len)
3820 {
3821 skb->protocol = protocol;
3822 skb->encapsulation = 1;
3823 skb_push(skb, pulled_hlen);
3824 skb_reset_transport_header(skb);
3825 skb->mac_header = mac_offset;
3826 skb->network_header = skb->mac_header + mac_len;
3827 skb->mac_len = mac_len;
3828 }
3829
3830 static inline bool netif_is_macvlan(struct net_device *dev)
3831 {
3832 return dev->priv_flags & IFF_MACVLAN;
3833 }
3834
3835 static inline bool netif_is_macvlan_port(struct net_device *dev)
3836 {
3837 return dev->priv_flags & IFF_MACVLAN_PORT;
3838 }
3839
3840 static inline bool netif_is_ipvlan(struct net_device *dev)
3841 {
3842 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3843 }
3844
3845 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3846 {
3847 return dev->priv_flags & IFF_IPVLAN_MASTER;
3848 }
3849
3850 static inline bool netif_is_bond_master(struct net_device *dev)
3851 {
3852 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3853 }
3854
3855 static inline bool netif_is_bond_slave(struct net_device *dev)
3856 {
3857 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3858 }
3859
3860 static inline bool netif_supports_nofcs(struct net_device *dev)
3861 {
3862 return dev->priv_flags & IFF_SUPP_NOFCS;
3863 }
3864
3865 static inline bool netif_is_l3_master(const struct net_device *dev)
3866 {
3867 return dev->priv_flags & IFF_L3MDEV_MASTER;
3868 }
3869
3870 static inline bool netif_is_l3_slave(const struct net_device *dev)
3871 {
3872 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3873 }
3874
3875 static inline bool netif_is_bridge_master(const struct net_device *dev)
3876 {
3877 return dev->priv_flags & IFF_EBRIDGE;
3878 }
3879
3880 static inline bool netif_is_bridge_port(const struct net_device *dev)
3881 {
3882 return dev->priv_flags & IFF_BRIDGE_PORT;
3883 }
3884
3885 static inline bool netif_is_ovs_master(const struct net_device *dev)
3886 {
3887 return dev->priv_flags & IFF_OPENVSWITCH;
3888 }
3889
3890 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3891 static inline void netif_keep_dst(struct net_device *dev)
3892 {
3893 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3894 }
3895
3896 extern struct pernet_operations __net_initdata loopback_net_ops;
3897
3898 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3899
3900 /* netdev_printk helpers, similar to dev_printk */
3901
3902 static inline const char *netdev_name(const struct net_device *dev)
3903 {
3904 if (!dev->name[0] || strchr(dev->name, '%'))
3905 return "(unnamed net_device)";
3906 return dev->name;
3907 }
3908
3909 static inline const char *netdev_reg_state(const struct net_device *dev)
3910 {
3911 switch (dev->reg_state) {
3912 case NETREG_UNINITIALIZED: return " (uninitialized)";
3913 case NETREG_REGISTERED: return "";
3914 case NETREG_UNREGISTERING: return " (unregistering)";
3915 case NETREG_UNREGISTERED: return " (unregistered)";
3916 case NETREG_RELEASED: return " (released)";
3917 case NETREG_DUMMY: return " (dummy)";
3918 }
3919
3920 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3921 return " (unknown)";
3922 }
3923
3924 __printf(3, 4)
3925 void netdev_printk(const char *level, const struct net_device *dev,
3926 const char *format, ...);
3927 __printf(2, 3)
3928 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3929 __printf(2, 3)
3930 void netdev_alert(const struct net_device *dev, const char *format, ...);
3931 __printf(2, 3)
3932 void netdev_crit(const struct net_device *dev, const char *format, ...);
3933 __printf(2, 3)
3934 void netdev_err(const struct net_device *dev, const char *format, ...);
3935 __printf(2, 3)
3936 void netdev_warn(const struct net_device *dev, const char *format, ...);
3937 __printf(2, 3)
3938 void netdev_notice(const struct net_device *dev, const char *format, ...);
3939 __printf(2, 3)
3940 void netdev_info(const struct net_device *dev, const char *format, ...);
3941
3942 #define MODULE_ALIAS_NETDEV(device) \
3943 MODULE_ALIAS("netdev-" device)
3944
3945 #if defined(CONFIG_DYNAMIC_DEBUG)
3946 #define netdev_dbg(__dev, format, args...) \
3947 do { \
3948 dynamic_netdev_dbg(__dev, format, ##args); \
3949 } while (0)
3950 #elif defined(DEBUG)
3951 #define netdev_dbg(__dev, format, args...) \
3952 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3953 #else
3954 #define netdev_dbg(__dev, format, args...) \
3955 ({ \
3956 if (0) \
3957 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3958 })
3959 #endif
3960
3961 #if defined(VERBOSE_DEBUG)
3962 #define netdev_vdbg netdev_dbg
3963 #else
3964
3965 #define netdev_vdbg(dev, format, args...) \
3966 ({ \
3967 if (0) \
3968 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3969 0; \
3970 })
3971 #endif
3972
3973 /*
3974 * netdev_WARN() acts like dev_printk(), but with the key difference
3975 * of using a WARN/WARN_ON to get the message out, including the
3976 * file/line information and a backtrace.
3977 */
3978 #define netdev_WARN(dev, format, args...) \
3979 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3980 netdev_reg_state(dev), ##args)
3981
3982 /* netif printk helpers, similar to netdev_printk */
3983
3984 #define netif_printk(priv, type, level, dev, fmt, args...) \
3985 do { \
3986 if (netif_msg_##type(priv)) \
3987 netdev_printk(level, (dev), fmt, ##args); \
3988 } while (0)
3989
3990 #define netif_level(level, priv, type, dev, fmt, args...) \
3991 do { \
3992 if (netif_msg_##type(priv)) \
3993 netdev_##level(dev, fmt, ##args); \
3994 } while (0)
3995
3996 #define netif_emerg(priv, type, dev, fmt, args...) \
3997 netif_level(emerg, priv, type, dev, fmt, ##args)
3998 #define netif_alert(priv, type, dev, fmt, args...) \
3999 netif_level(alert, priv, type, dev, fmt, ##args)
4000 #define netif_crit(priv, type, dev, fmt, args...) \
4001 netif_level(crit, priv, type, dev, fmt, ##args)
4002 #define netif_err(priv, type, dev, fmt, args...) \
4003 netif_level(err, priv, type, dev, fmt, ##args)
4004 #define netif_warn(priv, type, dev, fmt, args...) \
4005 netif_level(warn, priv, type, dev, fmt, ##args)
4006 #define netif_notice(priv, type, dev, fmt, args...) \
4007 netif_level(notice, priv, type, dev, fmt, ##args)
4008 #define netif_info(priv, type, dev, fmt, args...) \
4009 netif_level(info, priv, type, dev, fmt, ##args)
4010
4011 #if defined(CONFIG_DYNAMIC_DEBUG)
4012 #define netif_dbg(priv, type, netdev, format, args...) \
4013 do { \
4014 if (netif_msg_##type(priv)) \
4015 dynamic_netdev_dbg(netdev, format, ##args); \
4016 } while (0)
4017 #elif defined(DEBUG)
4018 #define netif_dbg(priv, type, dev, format, args...) \
4019 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4020 #else
4021 #define netif_dbg(priv, type, dev, format, args...) \
4022 ({ \
4023 if (0) \
4024 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4025 0; \
4026 })
4027 #endif
4028
4029 #if defined(VERBOSE_DEBUG)
4030 #define netif_vdbg netif_dbg
4031 #else
4032 #define netif_vdbg(priv, type, dev, format, args...) \
4033 ({ \
4034 if (0) \
4035 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4036 0; \
4037 })
4038 #endif
4039
4040 /*
4041 * The list of packet types we will receive (as opposed to discard)
4042 * and the routines to invoke.
4043 *
4044 * Why 16. Because with 16 the only overlap we get on a hash of the
4045 * low nibble of the protocol value is RARP/SNAP/X.25.
4046 *
4047 * NOTE: That is no longer true with the addition of VLAN tags. Not
4048 * sure which should go first, but I bet it won't make much
4049 * difference if we are running VLANs. The good news is that
4050 * this protocol won't be in the list unless compiled in, so
4051 * the average user (w/out VLANs) will not be adversely affected.
4052 * --BLG
4053 *
4054 * 0800 IP
4055 * 8100 802.1Q VLAN
4056 * 0001 802.3
4057 * 0002 AX.25
4058 * 0004 802.2
4059 * 8035 RARP
4060 * 0005 SNAP
4061 * 0805 X.25
4062 * 0806 ARP
4063 * 8137 IPX
4064 * 0009 Localtalk
4065 * 86DD IPv6
4066 */
4067 #define PTYPE_HASH_SIZE (16)
4068 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4069
4070 #endif /* _LINUX_NETDEVICE_H */
This page took 0.127189 seconds and 5 git commands to generate.