net: Rename IFF_VRF_MASTER to IFF_L3MDEV_MASTER
[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 */
330 };
331
332 enum gro_result {
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338 };
339 typedef enum gro_result gro_result_t;
340
341 /*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382 enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387 };
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
393
394 static inline bool napi_disable_pending(struct napi_struct *n)
395 {
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397 }
398
399 /**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
407 */
408 static inline bool napi_schedule_prep(struct napi_struct *n)
409 {
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
412 }
413
414 /**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421 static inline void napi_schedule(struct napi_struct *n)
422 {
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425 }
426
427 /**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
434 {
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437 }
438
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
441 {
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
444 return true;
445 }
446 return false;
447 }
448
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
451 /**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
457 */
458 static inline void napi_complete(struct napi_struct *n)
459 {
460 return napi_complete_done(n, 0);
461 }
462
463 /**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
470 struct napi_struct *napi_by_id(unsigned int napi_id);
471
472 /**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
478 void napi_hash_add(struct napi_struct *napi);
479
480 /**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
487 void 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_BYTES - sizeof(struct xps_map)) \
722 / 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_get_vf_config)(struct net_device *dev,
885 * int vf, struct ifla_vf_info *ivf);
886 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
887 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
888 * struct nlattr *port[]);
889 *
890 * Enable or disable the VF ability to query its RSS Redirection Table and
891 * Hash Key. This is needed since on some devices VF share this information
892 * with PF and querying it may adduce a theoretical security risk.
893 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
894 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
895 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
896 * Called to setup 'tc' number of traffic classes in the net device. This
897 * is always called from the stack with the rtnl lock held and netif tx
898 * queues stopped. This allows the netdevice to perform queue management
899 * safely.
900 *
901 * Fiber Channel over Ethernet (FCoE) offload functions.
902 * int (*ndo_fcoe_enable)(struct net_device *dev);
903 * Called when the FCoE protocol stack wants to start using LLD for FCoE
904 * so the underlying device can perform whatever needed configuration or
905 * initialization to support acceleration of FCoE traffic.
906 *
907 * int (*ndo_fcoe_disable)(struct net_device *dev);
908 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
909 * so the underlying device can perform whatever needed clean-ups to
910 * stop supporting acceleration of FCoE traffic.
911 *
912 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
913 * struct scatterlist *sgl, unsigned int sgc);
914 * Called when the FCoE Initiator wants to initialize an I/O that
915 * is a possible candidate for Direct Data Placement (DDP). The LLD can
916 * perform necessary setup and returns 1 to indicate the device is set up
917 * successfully to perform DDP on this I/O, otherwise this returns 0.
918 *
919 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
920 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
921 * indicated by the FC exchange id 'xid', so the underlying device can
922 * clean up and reuse resources for later DDP requests.
923 *
924 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
925 * struct scatterlist *sgl, unsigned int sgc);
926 * Called when the FCoE Target wants to initialize an I/O that
927 * is a possible candidate for Direct Data Placement (DDP). The LLD can
928 * perform necessary setup and returns 1 to indicate the device is set up
929 * successfully to perform DDP on this I/O, otherwise this returns 0.
930 *
931 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
932 * struct netdev_fcoe_hbainfo *hbainfo);
933 * Called when the FCoE Protocol stack wants information on the underlying
934 * device. This information is utilized by the FCoE protocol stack to
935 * register attributes with Fiber Channel management service as per the
936 * FC-GS Fabric Device Management Information(FDMI) specification.
937 *
938 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
939 * Called when the underlying device wants to override default World Wide
940 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
941 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
942 * protocol stack to use.
943 *
944 * RFS acceleration.
945 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
946 * u16 rxq_index, u32 flow_id);
947 * Set hardware filter for RFS. rxq_index is the target queue index;
948 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
949 * Return the filter ID on success, or a negative error code.
950 *
951 * Slave management functions (for bridge, bonding, etc).
952 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
953 * Called to make another netdev an underling.
954 *
955 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
956 * Called to release previously enslaved netdev.
957 *
958 * Feature/offload setting functions.
959 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
960 * netdev_features_t features);
961 * Adjusts the requested feature flags according to device-specific
962 * constraints, and returns the resulting flags. Must not modify
963 * the device state.
964 *
965 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
966 * Called to update device configuration to new features. Passed
967 * feature set might be less than what was returned by ndo_fix_features()).
968 * Must return >0 or -errno if it changed dev->features itself.
969 *
970 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
971 * struct net_device *dev,
972 * const unsigned char *addr, u16 vid, u16 flags)
973 * Adds an FDB entry to dev for addr.
974 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
975 * struct net_device *dev,
976 * const unsigned char *addr, u16 vid)
977 * Deletes the FDB entry from dev coresponding to addr.
978 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
979 * struct net_device *dev, struct net_device *filter_dev,
980 * int idx)
981 * Used to add FDB entries to dump requests. Implementers should add
982 * entries to skb and update idx with the number of entries.
983 *
984 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
985 * u16 flags)
986 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
987 * struct net_device *dev, u32 filter_mask,
988 * int nlflags)
989 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
990 * u16 flags);
991 *
992 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
993 * Called to change device carrier. Soft-devices (like dummy, team, etc)
994 * which do not represent real hardware may define this to allow their
995 * userspace components to manage their virtual carrier state. Devices
996 * that determine carrier state from physical hardware properties (eg
997 * network cables) or protocol-dependent mechanisms (eg
998 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
999 *
1000 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1001 * struct netdev_phys_item_id *ppid);
1002 * Called to get ID of physical port of this device. If driver does
1003 * not implement this, it is assumed that the hw is not able to have
1004 * multiple net devices on single physical port.
1005 *
1006 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1007 * sa_family_t sa_family, __be16 port);
1008 * Called by vxlan to notiy a driver about the UDP port and socket
1009 * address family that vxlan is listnening to. It is called only when
1010 * a new port starts listening. The operation is protected by the
1011 * vxlan_net->sock_lock.
1012 *
1013 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1014 * sa_family_t sa_family, __be16 port);
1015 * Called by vxlan to notify the driver about a UDP port and socket
1016 * address family that vxlan is not listening to anymore. The operation
1017 * is protected by the vxlan_net->sock_lock.
1018 *
1019 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1020 * struct net_device *dev)
1021 * Called by upper layer devices to accelerate switching or other
1022 * station functionality into hardware. 'pdev is the lowerdev
1023 * to use for the offload and 'dev' is the net device that will
1024 * back the offload. Returns a pointer to the private structure
1025 * the upper layer will maintain.
1026 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1027 * Called by upper layer device to delete the station created
1028 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1029 * the station and priv is the structure returned by the add
1030 * operation.
1031 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1032 * struct net_device *dev,
1033 * void *priv);
1034 * Callback to use for xmit over the accelerated station. This
1035 * is used in place of ndo_start_xmit on accelerated net
1036 * devices.
1037 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1038 * struct net_device *dev
1039 * netdev_features_t features);
1040 * Called by core transmit path to determine if device is capable of
1041 * performing offload operations on a given packet. This is to give
1042 * the device an opportunity to implement any restrictions that cannot
1043 * be otherwise expressed by feature flags. The check is called with
1044 * the set of features that the stack has calculated and it returns
1045 * those the driver believes to be appropriate.
1046 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1047 * int queue_index, u32 maxrate);
1048 * Called when a user wants to set a max-rate limitation of specific
1049 * TX queue.
1050 * int (*ndo_get_iflink)(const struct net_device *dev);
1051 * Called to get the iflink value of this device.
1052 * void (*ndo_change_proto_down)(struct net_device *dev,
1053 * bool proto_down);
1054 * This function is used to pass protocol port error state information
1055 * to the switch driver. The switch driver can react to the proto_down
1056 * by doing a phys down on the associated switch port.
1057 *
1058 */
1059 struct net_device_ops {
1060 int (*ndo_init)(struct net_device *dev);
1061 void (*ndo_uninit)(struct net_device *dev);
1062 int (*ndo_open)(struct net_device *dev);
1063 int (*ndo_stop)(struct net_device *dev);
1064 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1065 struct net_device *dev);
1066 u16 (*ndo_select_queue)(struct net_device *dev,
1067 struct sk_buff *skb,
1068 void *accel_priv,
1069 select_queue_fallback_t fallback);
1070 void (*ndo_change_rx_flags)(struct net_device *dev,
1071 int flags);
1072 void (*ndo_set_rx_mode)(struct net_device *dev);
1073 int (*ndo_set_mac_address)(struct net_device *dev,
1074 void *addr);
1075 int (*ndo_validate_addr)(struct net_device *dev);
1076 int (*ndo_do_ioctl)(struct net_device *dev,
1077 struct ifreq *ifr, int cmd);
1078 int (*ndo_set_config)(struct net_device *dev,
1079 struct ifmap *map);
1080 int (*ndo_change_mtu)(struct net_device *dev,
1081 int new_mtu);
1082 int (*ndo_neigh_setup)(struct net_device *dev,
1083 struct neigh_parms *);
1084 void (*ndo_tx_timeout) (struct net_device *dev);
1085
1086 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1087 struct rtnl_link_stats64 *storage);
1088 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1089
1090 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1091 __be16 proto, u16 vid);
1092 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1093 __be16 proto, u16 vid);
1094 #ifdef CONFIG_NET_POLL_CONTROLLER
1095 void (*ndo_poll_controller)(struct net_device *dev);
1096 int (*ndo_netpoll_setup)(struct net_device *dev,
1097 struct netpoll_info *info);
1098 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1099 #endif
1100 #ifdef CONFIG_NET_RX_BUSY_POLL
1101 int (*ndo_busy_poll)(struct napi_struct *dev);
1102 #endif
1103 int (*ndo_set_vf_mac)(struct net_device *dev,
1104 int queue, u8 *mac);
1105 int (*ndo_set_vf_vlan)(struct net_device *dev,
1106 int queue, u16 vlan, u8 qos);
1107 int (*ndo_set_vf_rate)(struct net_device *dev,
1108 int vf, int min_tx_rate,
1109 int max_tx_rate);
1110 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1111 int vf, bool setting);
1112 int (*ndo_get_vf_config)(struct net_device *dev,
1113 int vf,
1114 struct ifla_vf_info *ivf);
1115 int (*ndo_set_vf_link_state)(struct net_device *dev,
1116 int vf, int link_state);
1117 int (*ndo_get_vf_stats)(struct net_device *dev,
1118 int vf,
1119 struct ifla_vf_stats
1120 *vf_stats);
1121 int (*ndo_set_vf_port)(struct net_device *dev,
1122 int vf,
1123 struct nlattr *port[]);
1124 int (*ndo_get_vf_port)(struct net_device *dev,
1125 int vf, struct sk_buff *skb);
1126 int (*ndo_set_vf_rss_query_en)(
1127 struct net_device *dev,
1128 int vf, bool setting);
1129 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1130 #if IS_ENABLED(CONFIG_FCOE)
1131 int (*ndo_fcoe_enable)(struct net_device *dev);
1132 int (*ndo_fcoe_disable)(struct net_device *dev);
1133 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1134 u16 xid,
1135 struct scatterlist *sgl,
1136 unsigned int sgc);
1137 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1138 u16 xid);
1139 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1140 u16 xid,
1141 struct scatterlist *sgl,
1142 unsigned int sgc);
1143 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1144 struct netdev_fcoe_hbainfo *hbainfo);
1145 #endif
1146
1147 #if IS_ENABLED(CONFIG_LIBFCOE)
1148 #define NETDEV_FCOE_WWNN 0
1149 #define NETDEV_FCOE_WWPN 1
1150 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1151 u64 *wwn, int type);
1152 #endif
1153
1154 #ifdef CONFIG_RFS_ACCEL
1155 int (*ndo_rx_flow_steer)(struct net_device *dev,
1156 const struct sk_buff *skb,
1157 u16 rxq_index,
1158 u32 flow_id);
1159 #endif
1160 int (*ndo_add_slave)(struct net_device *dev,
1161 struct net_device *slave_dev);
1162 int (*ndo_del_slave)(struct net_device *dev,
1163 struct net_device *slave_dev);
1164 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1165 netdev_features_t features);
1166 int (*ndo_set_features)(struct net_device *dev,
1167 netdev_features_t features);
1168 int (*ndo_neigh_construct)(struct neighbour *n);
1169 void (*ndo_neigh_destroy)(struct neighbour *n);
1170
1171 int (*ndo_fdb_add)(struct ndmsg *ndm,
1172 struct nlattr *tb[],
1173 struct net_device *dev,
1174 const unsigned char *addr,
1175 u16 vid,
1176 u16 flags);
1177 int (*ndo_fdb_del)(struct ndmsg *ndm,
1178 struct nlattr *tb[],
1179 struct net_device *dev,
1180 const unsigned char *addr,
1181 u16 vid);
1182 int (*ndo_fdb_dump)(struct sk_buff *skb,
1183 struct netlink_callback *cb,
1184 struct net_device *dev,
1185 struct net_device *filter_dev,
1186 int idx);
1187
1188 int (*ndo_bridge_setlink)(struct net_device *dev,
1189 struct nlmsghdr *nlh,
1190 u16 flags);
1191 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1192 u32 pid, u32 seq,
1193 struct net_device *dev,
1194 u32 filter_mask,
1195 int nlflags);
1196 int (*ndo_bridge_dellink)(struct net_device *dev,
1197 struct nlmsghdr *nlh,
1198 u16 flags);
1199 int (*ndo_change_carrier)(struct net_device *dev,
1200 bool new_carrier);
1201 int (*ndo_get_phys_port_id)(struct net_device *dev,
1202 struct netdev_phys_item_id *ppid);
1203 int (*ndo_get_phys_port_name)(struct net_device *dev,
1204 char *name, size_t len);
1205 void (*ndo_add_vxlan_port)(struct net_device *dev,
1206 sa_family_t sa_family,
1207 __be16 port);
1208 void (*ndo_del_vxlan_port)(struct net_device *dev,
1209 sa_family_t sa_family,
1210 __be16 port);
1211
1212 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1213 struct net_device *dev);
1214 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1215 void *priv);
1216
1217 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1218 struct net_device *dev,
1219 void *priv);
1220 int (*ndo_get_lock_subclass)(struct net_device *dev);
1221 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1222 struct net_device *dev,
1223 netdev_features_t features);
1224 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1225 int queue_index,
1226 u32 maxrate);
1227 int (*ndo_get_iflink)(const struct net_device *dev);
1228 int (*ndo_change_proto_down)(struct net_device *dev,
1229 bool proto_down);
1230 };
1231
1232 /**
1233 * enum net_device_priv_flags - &struct net_device priv_flags
1234 *
1235 * These are the &struct net_device, they are only set internally
1236 * by drivers and used in the kernel. These flags are invisible to
1237 * userspace, this means that the order of these flags can change
1238 * during any kernel release.
1239 *
1240 * You should have a pretty good reason to be extending these flags.
1241 *
1242 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1243 * @IFF_EBRIDGE: Ethernet bridging device
1244 * @IFF_BONDING: bonding master or slave
1245 * @IFF_ISATAP: ISATAP interface (RFC4214)
1246 * @IFF_WAN_HDLC: WAN HDLC device
1247 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1248 * release skb->dst
1249 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1250 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1251 * @IFF_MACVLAN_PORT: device used as macvlan port
1252 * @IFF_BRIDGE_PORT: device used as bridge port
1253 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1254 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1255 * @IFF_UNICAST_FLT: Supports unicast filtering
1256 * @IFF_TEAM_PORT: device used as team port
1257 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1258 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1259 * change when it's running
1260 * @IFF_MACVLAN: Macvlan device
1261 * @IFF_L3MDEV_MASTER: device is an L3 master device
1262 * @IFF_NO_QUEUE: device can run without qdisc attached
1263 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1264 */
1265 enum netdev_priv_flags {
1266 IFF_802_1Q_VLAN = 1<<0,
1267 IFF_EBRIDGE = 1<<1,
1268 IFF_BONDING = 1<<2,
1269 IFF_ISATAP = 1<<3,
1270 IFF_WAN_HDLC = 1<<4,
1271 IFF_XMIT_DST_RELEASE = 1<<5,
1272 IFF_DONT_BRIDGE = 1<<6,
1273 IFF_DISABLE_NETPOLL = 1<<7,
1274 IFF_MACVLAN_PORT = 1<<8,
1275 IFF_BRIDGE_PORT = 1<<9,
1276 IFF_OVS_DATAPATH = 1<<10,
1277 IFF_TX_SKB_SHARING = 1<<11,
1278 IFF_UNICAST_FLT = 1<<12,
1279 IFF_TEAM_PORT = 1<<13,
1280 IFF_SUPP_NOFCS = 1<<14,
1281 IFF_LIVE_ADDR_CHANGE = 1<<15,
1282 IFF_MACVLAN = 1<<16,
1283 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1284 IFF_IPVLAN_MASTER = 1<<18,
1285 IFF_IPVLAN_SLAVE = 1<<19,
1286 IFF_L3MDEV_MASTER = 1<<20,
1287 IFF_NO_QUEUE = 1<<21,
1288 IFF_OPENVSWITCH = 1<<22,
1289 };
1290
1291 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1292 #define IFF_EBRIDGE IFF_EBRIDGE
1293 #define IFF_BONDING IFF_BONDING
1294 #define IFF_ISATAP IFF_ISATAP
1295 #define IFF_WAN_HDLC IFF_WAN_HDLC
1296 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1297 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1298 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1299 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1300 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1301 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1302 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1303 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1304 #define IFF_TEAM_PORT IFF_TEAM_PORT
1305 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1306 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1307 #define IFF_MACVLAN IFF_MACVLAN
1308 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1309 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1310 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1311 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1312 #define IFF_NO_QUEUE IFF_NO_QUEUE
1313 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1314
1315 /**
1316 * struct net_device - The DEVICE structure.
1317 * Actually, this whole structure is a big mistake. It mixes I/O
1318 * data with strictly "high-level" data, and it has to know about
1319 * almost every data structure used in the INET module.
1320 *
1321 * @name: This is the first field of the "visible" part of this structure
1322 * (i.e. as seen by users in the "Space.c" file). It is the name
1323 * of the interface.
1324 *
1325 * @name_hlist: Device name hash chain, please keep it close to name[]
1326 * @ifalias: SNMP alias
1327 * @mem_end: Shared memory end
1328 * @mem_start: Shared memory start
1329 * @base_addr: Device I/O address
1330 * @irq: Device IRQ number
1331 *
1332 * @carrier_changes: Stats to monitor carrier on<->off transitions
1333 *
1334 * @state: Generic network queuing layer state, see netdev_state_t
1335 * @dev_list: The global list of network devices
1336 * @napi_list: List entry, that is used for polling napi devices
1337 * @unreg_list: List entry, that is used, when we are unregistering the
1338 * device, see the function unregister_netdev
1339 * @close_list: List entry, that is used, when we are closing the device
1340 *
1341 * @adj_list: Directly linked devices, like slaves for bonding
1342 * @all_adj_list: All linked devices, *including* neighbours
1343 * @features: Currently active device features
1344 * @hw_features: User-changeable features
1345 *
1346 * @wanted_features: User-requested features
1347 * @vlan_features: Mask of features inheritable by VLAN devices
1348 *
1349 * @hw_enc_features: Mask of features inherited by encapsulating devices
1350 * This field indicates what encapsulation
1351 * offloads the hardware is capable of doing,
1352 * and drivers will need to set them appropriately.
1353 *
1354 * @mpls_features: Mask of features inheritable by MPLS
1355 *
1356 * @ifindex: interface index
1357 * @group: The group, that the device belongs to
1358 *
1359 * @stats: Statistics struct, which was left as a legacy, use
1360 * rtnl_link_stats64 instead
1361 *
1362 * @rx_dropped: Dropped packets by core network,
1363 * do not use this in drivers
1364 * @tx_dropped: Dropped packets by core network,
1365 * do not use this in drivers
1366 *
1367 * @wireless_handlers: List of functions to handle Wireless Extensions,
1368 * instead of ioctl,
1369 * see <net/iw_handler.h> for details.
1370 * @wireless_data: Instance data managed by the core of wireless extensions
1371 *
1372 * @netdev_ops: Includes several pointers to callbacks,
1373 * if one wants to override the ndo_*() functions
1374 * @ethtool_ops: Management operations
1375 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1376 * of Layer 2 headers.
1377 *
1378 * @flags: Interface flags (a la BSD)
1379 * @priv_flags: Like 'flags' but invisible to userspace,
1380 * see if.h for the definitions
1381 * @gflags: Global flags ( kept as legacy )
1382 * @padded: How much padding added by alloc_netdev()
1383 * @operstate: RFC2863 operstate
1384 * @link_mode: Mapping policy to operstate
1385 * @if_port: Selectable AUI, TP, ...
1386 * @dma: DMA channel
1387 * @mtu: Interface MTU value
1388 * @type: Interface hardware type
1389 * @hard_header_len: Hardware header length
1390 *
1391 * @needed_headroom: Extra headroom the hardware may need, but not in all
1392 * cases can this be guaranteed
1393 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1394 * cases can this be guaranteed. Some cases also use
1395 * LL_MAX_HEADER instead to allocate the skb
1396 *
1397 * interface address info:
1398 *
1399 * @perm_addr: Permanent hw address
1400 * @addr_assign_type: Hw address assignment type
1401 * @addr_len: Hardware address length
1402 * @neigh_priv_len; Used in neigh_alloc(),
1403 * initialized only in atm/clip.c
1404 * @dev_id: Used to differentiate devices that share
1405 * the same link layer address
1406 * @dev_port: Used to differentiate devices that share
1407 * the same function
1408 * @addr_list_lock: XXX: need comments on this one
1409 * @uc_promisc: Counter, that indicates, that promiscuous mode
1410 * has been enabled due to the need to listen to
1411 * additional unicast addresses in a device that
1412 * does not implement ndo_set_rx_mode()
1413 * @uc: unicast mac addresses
1414 * @mc: multicast mac addresses
1415 * @dev_addrs: list of device hw addresses
1416 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1417 * @promiscuity: Number of times, the NIC is told to work in
1418 * Promiscuous mode, if it becomes 0 the NIC will
1419 * exit from working in Promiscuous mode
1420 * @allmulti: Counter, enables or disables allmulticast mode
1421 *
1422 * @vlan_info: VLAN info
1423 * @dsa_ptr: dsa specific data
1424 * @tipc_ptr: TIPC specific data
1425 * @atalk_ptr: AppleTalk link
1426 * @ip_ptr: IPv4 specific data
1427 * @dn_ptr: DECnet specific data
1428 * @ip6_ptr: IPv6 specific data
1429 * @ax25_ptr: AX.25 specific data
1430 * @vrf_ptr: VRF specific data
1431 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1432 *
1433 * @last_rx: Time of last Rx
1434 * @dev_addr: Hw address (before bcast,
1435 * because most packets are unicast)
1436 *
1437 * @_rx: Array of RX queues
1438 * @num_rx_queues: Number of RX queues
1439 * allocated at register_netdev() time
1440 * @real_num_rx_queues: Number of RX queues currently active in device
1441 *
1442 * @rx_handler: handler for received packets
1443 * @rx_handler_data: XXX: need comments on this one
1444 * @ingress_queue: XXX: need comments on this one
1445 * @broadcast: hw bcast address
1446 *
1447 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1448 * indexed by RX queue number. Assigned by driver.
1449 * This must only be set if the ndo_rx_flow_steer
1450 * operation is defined
1451 * @index_hlist: Device index hash chain
1452 *
1453 * @_tx: Array of TX queues
1454 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1455 * @real_num_tx_queues: Number of TX queues currently active in device
1456 * @qdisc: Root qdisc from userspace point of view
1457 * @tx_queue_len: Max frames per queue allowed
1458 * @tx_global_lock: XXX: need comments on this one
1459 *
1460 * @xps_maps: XXX: need comments on this one
1461 *
1462 * @offload_fwd_mark: Offload device fwding mark
1463 *
1464 * @trans_start: Time (in jiffies) of last Tx
1465 * @watchdog_timeo: Represents the timeout that is used by
1466 * the watchdog ( see dev_watchdog() )
1467 * @watchdog_timer: List of timers
1468 *
1469 * @pcpu_refcnt: Number of references to this device
1470 * @todo_list: Delayed register/unregister
1471 * @link_watch_list: XXX: need comments on this one
1472 *
1473 * @reg_state: Register/unregister state machine
1474 * @dismantle: Device is going to be freed
1475 * @rtnl_link_state: This enum represents the phases of creating
1476 * a new link
1477 *
1478 * @destructor: Called from unregister,
1479 * can be used to call free_netdev
1480 * @npinfo: XXX: need comments on this one
1481 * @nd_net: Network namespace this network device is inside
1482 *
1483 * @ml_priv: Mid-layer private
1484 * @lstats: Loopback statistics
1485 * @tstats: Tunnel statistics
1486 * @dstats: Dummy statistics
1487 * @vstats: Virtual ethernet statistics
1488 *
1489 * @garp_port: GARP
1490 * @mrp_port: MRP
1491 *
1492 * @dev: Class/net/name entry
1493 * @sysfs_groups: Space for optional device, statistics and wireless
1494 * sysfs groups
1495 *
1496 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1497 * @rtnl_link_ops: Rtnl_link_ops
1498 *
1499 * @gso_max_size: Maximum size of generic segmentation offload
1500 * @gso_max_segs: Maximum number of segments that can be passed to the
1501 * NIC for GSO
1502 * @gso_min_segs: Minimum number of segments that can be passed to the
1503 * NIC for GSO
1504 *
1505 * @dcbnl_ops: Data Center Bridging netlink ops
1506 * @num_tc: Number of traffic classes in the net device
1507 * @tc_to_txq: XXX: need comments on this one
1508 * @prio_tc_map XXX: need comments on this one
1509 *
1510 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1511 *
1512 * @priomap: XXX: need comments on this one
1513 * @phydev: Physical device may attach itself
1514 * for hardware timestamping
1515 *
1516 * @qdisc_tx_busylock: XXX: need comments on this one
1517 *
1518 * @proto_down: protocol port state information can be sent to the
1519 * switch driver and used to set the phys state of the
1520 * switch port.
1521 *
1522 * FIXME: cleanup struct net_device such that network protocol info
1523 * moves out.
1524 */
1525
1526 struct net_device {
1527 char name[IFNAMSIZ];
1528 struct hlist_node name_hlist;
1529 char *ifalias;
1530 /*
1531 * I/O specific fields
1532 * FIXME: Merge these and struct ifmap into one
1533 */
1534 unsigned long mem_end;
1535 unsigned long mem_start;
1536 unsigned long base_addr;
1537 int irq;
1538
1539 atomic_t carrier_changes;
1540
1541 /*
1542 * Some hardware also needs these fields (state,dev_list,
1543 * napi_list,unreg_list,close_list) but they are not
1544 * part of the usual set specified in Space.c.
1545 */
1546
1547 unsigned long state;
1548
1549 struct list_head dev_list;
1550 struct list_head napi_list;
1551 struct list_head unreg_list;
1552 struct list_head close_list;
1553 struct list_head ptype_all;
1554 struct list_head ptype_specific;
1555
1556 struct {
1557 struct list_head upper;
1558 struct list_head lower;
1559 } adj_list;
1560
1561 struct {
1562 struct list_head upper;
1563 struct list_head lower;
1564 } all_adj_list;
1565
1566 netdev_features_t features;
1567 netdev_features_t hw_features;
1568 netdev_features_t wanted_features;
1569 netdev_features_t vlan_features;
1570 netdev_features_t hw_enc_features;
1571 netdev_features_t mpls_features;
1572
1573 int ifindex;
1574 int group;
1575
1576 struct net_device_stats stats;
1577
1578 atomic_long_t rx_dropped;
1579 atomic_long_t tx_dropped;
1580
1581 #ifdef CONFIG_WIRELESS_EXT
1582 const struct iw_handler_def * wireless_handlers;
1583 struct iw_public_data * wireless_data;
1584 #endif
1585 const struct net_device_ops *netdev_ops;
1586 const struct ethtool_ops *ethtool_ops;
1587 #ifdef CONFIG_NET_SWITCHDEV
1588 const struct switchdev_ops *switchdev_ops;
1589 #endif
1590
1591 const struct header_ops *header_ops;
1592
1593 unsigned int flags;
1594 unsigned int priv_flags;
1595
1596 unsigned short gflags;
1597 unsigned short padded;
1598
1599 unsigned char operstate;
1600 unsigned char link_mode;
1601
1602 unsigned char if_port;
1603 unsigned char dma;
1604
1605 unsigned int mtu;
1606 unsigned short type;
1607 unsigned short hard_header_len;
1608
1609 unsigned short needed_headroom;
1610 unsigned short needed_tailroom;
1611
1612 /* Interface address info. */
1613 unsigned char perm_addr[MAX_ADDR_LEN];
1614 unsigned char addr_assign_type;
1615 unsigned char addr_len;
1616 unsigned short neigh_priv_len;
1617 unsigned short dev_id;
1618 unsigned short dev_port;
1619 spinlock_t addr_list_lock;
1620 unsigned char name_assign_type;
1621 bool uc_promisc;
1622 struct netdev_hw_addr_list uc;
1623 struct netdev_hw_addr_list mc;
1624 struct netdev_hw_addr_list dev_addrs;
1625
1626 #ifdef CONFIG_SYSFS
1627 struct kset *queues_kset;
1628 #endif
1629 unsigned int promiscuity;
1630 unsigned int allmulti;
1631
1632
1633 /* Protocol specific pointers */
1634
1635 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1636 struct vlan_info __rcu *vlan_info;
1637 #endif
1638 #if IS_ENABLED(CONFIG_NET_DSA)
1639 struct dsa_switch_tree *dsa_ptr;
1640 #endif
1641 #if IS_ENABLED(CONFIG_TIPC)
1642 struct tipc_bearer __rcu *tipc_ptr;
1643 #endif
1644 void *atalk_ptr;
1645 struct in_device __rcu *ip_ptr;
1646 struct dn_dev __rcu *dn_ptr;
1647 struct inet6_dev __rcu *ip6_ptr;
1648 void *ax25_ptr;
1649 struct net_vrf_dev __rcu *vrf_ptr;
1650 struct wireless_dev *ieee80211_ptr;
1651 struct wpan_dev *ieee802154_ptr;
1652 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1653 struct mpls_dev __rcu *mpls_ptr;
1654 #endif
1655
1656 /*
1657 * Cache lines mostly used on receive path (including eth_type_trans())
1658 */
1659 unsigned long last_rx;
1660
1661 /* Interface address info used in eth_type_trans() */
1662 unsigned char *dev_addr;
1663
1664
1665 #ifdef CONFIG_SYSFS
1666 struct netdev_rx_queue *_rx;
1667
1668 unsigned int num_rx_queues;
1669 unsigned int real_num_rx_queues;
1670
1671 #endif
1672
1673 unsigned long gro_flush_timeout;
1674 rx_handler_func_t __rcu *rx_handler;
1675 void __rcu *rx_handler_data;
1676
1677 #ifdef CONFIG_NET_CLS_ACT
1678 struct tcf_proto __rcu *ingress_cl_list;
1679 #endif
1680 struct netdev_queue __rcu *ingress_queue;
1681 #ifdef CONFIG_NETFILTER_INGRESS
1682 struct list_head nf_hooks_ingress;
1683 #endif
1684
1685 unsigned char broadcast[MAX_ADDR_LEN];
1686 #ifdef CONFIG_RFS_ACCEL
1687 struct cpu_rmap *rx_cpu_rmap;
1688 #endif
1689 struct hlist_node index_hlist;
1690
1691 /*
1692 * Cache lines mostly used on transmit path
1693 */
1694 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1695 unsigned int num_tx_queues;
1696 unsigned int real_num_tx_queues;
1697 struct Qdisc *qdisc;
1698 unsigned long tx_queue_len;
1699 spinlock_t tx_global_lock;
1700 int watchdog_timeo;
1701
1702 #ifdef CONFIG_XPS
1703 struct xps_dev_maps __rcu *xps_maps;
1704 #endif
1705
1706 #ifdef CONFIG_NET_SWITCHDEV
1707 u32 offload_fwd_mark;
1708 #endif
1709
1710 /* These may be needed for future network-power-down code. */
1711
1712 /*
1713 * trans_start here is expensive for high speed devices on SMP,
1714 * please use netdev_queue->trans_start instead.
1715 */
1716 unsigned long trans_start;
1717
1718 struct timer_list watchdog_timer;
1719
1720 int __percpu *pcpu_refcnt;
1721 struct list_head todo_list;
1722
1723 struct list_head link_watch_list;
1724
1725 enum { NETREG_UNINITIALIZED=0,
1726 NETREG_REGISTERED, /* completed register_netdevice */
1727 NETREG_UNREGISTERING, /* called unregister_netdevice */
1728 NETREG_UNREGISTERED, /* completed unregister todo */
1729 NETREG_RELEASED, /* called free_netdev */
1730 NETREG_DUMMY, /* dummy device for NAPI poll */
1731 } reg_state:8;
1732
1733 bool dismantle;
1734
1735 enum {
1736 RTNL_LINK_INITIALIZED,
1737 RTNL_LINK_INITIALIZING,
1738 } rtnl_link_state:16;
1739
1740 void (*destructor)(struct net_device *dev);
1741
1742 #ifdef CONFIG_NETPOLL
1743 struct netpoll_info __rcu *npinfo;
1744 #endif
1745
1746 possible_net_t nd_net;
1747
1748 /* mid-layer private */
1749 union {
1750 void *ml_priv;
1751 struct pcpu_lstats __percpu *lstats;
1752 struct pcpu_sw_netstats __percpu *tstats;
1753 struct pcpu_dstats __percpu *dstats;
1754 struct pcpu_vstats __percpu *vstats;
1755 };
1756
1757 struct garp_port __rcu *garp_port;
1758 struct mrp_port __rcu *mrp_port;
1759
1760 struct device dev;
1761 const struct attribute_group *sysfs_groups[4];
1762 const struct attribute_group *sysfs_rx_queue_group;
1763
1764 const struct rtnl_link_ops *rtnl_link_ops;
1765
1766 /* for setting kernel sock attribute on TCP connection setup */
1767 #define GSO_MAX_SIZE 65536
1768 unsigned int gso_max_size;
1769 #define GSO_MAX_SEGS 65535
1770 u16 gso_max_segs;
1771 u16 gso_min_segs;
1772 #ifdef CONFIG_DCB
1773 const struct dcbnl_rtnl_ops *dcbnl_ops;
1774 #endif
1775 u8 num_tc;
1776 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1777 u8 prio_tc_map[TC_BITMASK + 1];
1778
1779 #if IS_ENABLED(CONFIG_FCOE)
1780 unsigned int fcoe_ddp_xid;
1781 #endif
1782 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1783 struct netprio_map __rcu *priomap;
1784 #endif
1785 struct phy_device *phydev;
1786 struct lock_class_key *qdisc_tx_busylock;
1787 bool proto_down;
1788 };
1789 #define to_net_dev(d) container_of(d, struct net_device, dev)
1790
1791 #define NETDEV_ALIGN 32
1792
1793 static inline
1794 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1795 {
1796 return dev->prio_tc_map[prio & TC_BITMASK];
1797 }
1798
1799 static inline
1800 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1801 {
1802 if (tc >= dev->num_tc)
1803 return -EINVAL;
1804
1805 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1806 return 0;
1807 }
1808
1809 static inline
1810 void netdev_reset_tc(struct net_device *dev)
1811 {
1812 dev->num_tc = 0;
1813 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1814 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1815 }
1816
1817 static inline
1818 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1819 {
1820 if (tc >= dev->num_tc)
1821 return -EINVAL;
1822
1823 dev->tc_to_txq[tc].count = count;
1824 dev->tc_to_txq[tc].offset = offset;
1825 return 0;
1826 }
1827
1828 static inline
1829 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1830 {
1831 if (num_tc > TC_MAX_QUEUE)
1832 return -EINVAL;
1833
1834 dev->num_tc = num_tc;
1835 return 0;
1836 }
1837
1838 static inline
1839 int netdev_get_num_tc(struct net_device *dev)
1840 {
1841 return dev->num_tc;
1842 }
1843
1844 static inline
1845 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1846 unsigned int index)
1847 {
1848 return &dev->_tx[index];
1849 }
1850
1851 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1852 const struct sk_buff *skb)
1853 {
1854 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1855 }
1856
1857 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1858 void (*f)(struct net_device *,
1859 struct netdev_queue *,
1860 void *),
1861 void *arg)
1862 {
1863 unsigned int i;
1864
1865 for (i = 0; i < dev->num_tx_queues; i++)
1866 f(dev, &dev->_tx[i], arg);
1867 }
1868
1869 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1870 struct sk_buff *skb,
1871 void *accel_priv);
1872
1873 /*
1874 * Net namespace inlines
1875 */
1876 static inline
1877 struct net *dev_net(const struct net_device *dev)
1878 {
1879 return read_pnet(&dev->nd_net);
1880 }
1881
1882 static inline
1883 void dev_net_set(struct net_device *dev, struct net *net)
1884 {
1885 write_pnet(&dev->nd_net, net);
1886 }
1887
1888 static inline bool netdev_uses_dsa(struct net_device *dev)
1889 {
1890 #if IS_ENABLED(CONFIG_NET_DSA)
1891 if (dev->dsa_ptr != NULL)
1892 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1893 #endif
1894 return false;
1895 }
1896
1897 /**
1898 * netdev_priv - access network device private data
1899 * @dev: network device
1900 *
1901 * Get network device private data
1902 */
1903 static inline void *netdev_priv(const struct net_device *dev)
1904 {
1905 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1906 }
1907
1908 /* Set the sysfs physical device reference for the network logical device
1909 * if set prior to registration will cause a symlink during initialization.
1910 */
1911 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1912
1913 /* Set the sysfs device type for the network logical device to allow
1914 * fine-grained identification of different network device types. For
1915 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1916 */
1917 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1918
1919 /* Default NAPI poll() weight
1920 * Device drivers are strongly advised to not use bigger value
1921 */
1922 #define NAPI_POLL_WEIGHT 64
1923
1924 /**
1925 * netif_napi_add - initialize a napi context
1926 * @dev: network device
1927 * @napi: napi context
1928 * @poll: polling function
1929 * @weight: default weight
1930 *
1931 * netif_napi_add() must be used to initialize a napi context prior to calling
1932 * *any* of the other napi related functions.
1933 */
1934 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1935 int (*poll)(struct napi_struct *, int), int weight);
1936
1937 /**
1938 * netif_napi_del - remove a napi context
1939 * @napi: napi context
1940 *
1941 * netif_napi_del() removes a napi context from the network device napi list
1942 */
1943 void netif_napi_del(struct napi_struct *napi);
1944
1945 struct napi_gro_cb {
1946 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1947 void *frag0;
1948
1949 /* Length of frag0. */
1950 unsigned int frag0_len;
1951
1952 /* This indicates where we are processing relative to skb->data. */
1953 int data_offset;
1954
1955 /* This is non-zero if the packet cannot be merged with the new skb. */
1956 u16 flush;
1957
1958 /* Save the IP ID here and check when we get to the transport layer */
1959 u16 flush_id;
1960
1961 /* Number of segments aggregated. */
1962 u16 count;
1963
1964 /* Start offset for remote checksum offload */
1965 u16 gro_remcsum_start;
1966
1967 /* jiffies when first packet was created/queued */
1968 unsigned long age;
1969
1970 /* Used in ipv6_gro_receive() and foo-over-udp */
1971 u16 proto;
1972
1973 /* This is non-zero if the packet may be of the same flow. */
1974 u8 same_flow:1;
1975
1976 /* Used in udp_gro_receive */
1977 u8 udp_mark:1;
1978
1979 /* GRO checksum is valid */
1980 u8 csum_valid:1;
1981
1982 /* Number of checksums via CHECKSUM_UNNECESSARY */
1983 u8 csum_cnt:3;
1984
1985 /* Free the skb? */
1986 u8 free:2;
1987 #define NAPI_GRO_FREE 1
1988 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1989
1990 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1991 u8 is_ipv6:1;
1992
1993 /* 7 bit hole */
1994
1995 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1996 __wsum csum;
1997
1998 /* used in skb_gro_receive() slow path */
1999 struct sk_buff *last;
2000 };
2001
2002 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2003
2004 struct packet_type {
2005 __be16 type; /* This is really htons(ether_type). */
2006 struct net_device *dev; /* NULL is wildcarded here */
2007 int (*func) (struct sk_buff *,
2008 struct net_device *,
2009 struct packet_type *,
2010 struct net_device *);
2011 bool (*id_match)(struct packet_type *ptype,
2012 struct sock *sk);
2013 void *af_packet_priv;
2014 struct list_head list;
2015 };
2016
2017 struct offload_callbacks {
2018 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2019 netdev_features_t features);
2020 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2021 struct sk_buff *skb);
2022 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2023 };
2024
2025 struct packet_offload {
2026 __be16 type; /* This is really htons(ether_type). */
2027 u16 priority;
2028 struct offload_callbacks callbacks;
2029 struct list_head list;
2030 };
2031
2032 struct udp_offload;
2033
2034 struct udp_offload_callbacks {
2035 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2036 struct sk_buff *skb,
2037 struct udp_offload *uoff);
2038 int (*gro_complete)(struct sk_buff *skb,
2039 int nhoff,
2040 struct udp_offload *uoff);
2041 };
2042
2043 struct udp_offload {
2044 __be16 port;
2045 u8 ipproto;
2046 struct udp_offload_callbacks callbacks;
2047 };
2048
2049 /* often modified stats are per cpu, other are shared (netdev->stats) */
2050 struct pcpu_sw_netstats {
2051 u64 rx_packets;
2052 u64 rx_bytes;
2053 u64 tx_packets;
2054 u64 tx_bytes;
2055 struct u64_stats_sync syncp;
2056 };
2057
2058 #define netdev_alloc_pcpu_stats(type) \
2059 ({ \
2060 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2061 if (pcpu_stats) { \
2062 int __cpu; \
2063 for_each_possible_cpu(__cpu) { \
2064 typeof(type) *stat; \
2065 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2066 u64_stats_init(&stat->syncp); \
2067 } \
2068 } \
2069 pcpu_stats; \
2070 })
2071
2072 #include <linux/notifier.h>
2073
2074 /* netdevice notifier chain. Please remember to update the rtnetlink
2075 * notification exclusion list in rtnetlink_event() when adding new
2076 * types.
2077 */
2078 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2079 #define NETDEV_DOWN 0x0002
2080 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2081 detected a hardware crash and restarted
2082 - we can use this eg to kick tcp sessions
2083 once done */
2084 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2085 #define NETDEV_REGISTER 0x0005
2086 #define NETDEV_UNREGISTER 0x0006
2087 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2088 #define NETDEV_CHANGEADDR 0x0008
2089 #define NETDEV_GOING_DOWN 0x0009
2090 #define NETDEV_CHANGENAME 0x000A
2091 #define NETDEV_FEAT_CHANGE 0x000B
2092 #define NETDEV_BONDING_FAILOVER 0x000C
2093 #define NETDEV_PRE_UP 0x000D
2094 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2095 #define NETDEV_POST_TYPE_CHANGE 0x000F
2096 #define NETDEV_POST_INIT 0x0010
2097 #define NETDEV_UNREGISTER_FINAL 0x0011
2098 #define NETDEV_RELEASE 0x0012
2099 #define NETDEV_NOTIFY_PEERS 0x0013
2100 #define NETDEV_JOIN 0x0014
2101 #define NETDEV_CHANGEUPPER 0x0015
2102 #define NETDEV_RESEND_IGMP 0x0016
2103 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2104 #define NETDEV_CHANGEINFODATA 0x0018
2105 #define NETDEV_BONDING_INFO 0x0019
2106
2107 int register_netdevice_notifier(struct notifier_block *nb);
2108 int unregister_netdevice_notifier(struct notifier_block *nb);
2109
2110 struct netdev_notifier_info {
2111 struct net_device *dev;
2112 };
2113
2114 struct netdev_notifier_change_info {
2115 struct netdev_notifier_info info; /* must be first */
2116 unsigned int flags_changed;
2117 };
2118
2119 struct netdev_notifier_changeupper_info {
2120 struct netdev_notifier_info info; /* must be first */
2121 struct net_device *upper_dev; /* new upper dev */
2122 bool master; /* is upper dev master */
2123 bool linking; /* is the nofication for link or unlink */
2124 };
2125
2126 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2127 struct net_device *dev)
2128 {
2129 info->dev = dev;
2130 }
2131
2132 static inline struct net_device *
2133 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2134 {
2135 return info->dev;
2136 }
2137
2138 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2139
2140
2141 extern rwlock_t dev_base_lock; /* Device list lock */
2142
2143 #define for_each_netdev(net, d) \
2144 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2145 #define for_each_netdev_reverse(net, d) \
2146 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2147 #define for_each_netdev_rcu(net, d) \
2148 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2149 #define for_each_netdev_safe(net, d, n) \
2150 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2151 #define for_each_netdev_continue(net, d) \
2152 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2153 #define for_each_netdev_continue_rcu(net, d) \
2154 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2155 #define for_each_netdev_in_bond_rcu(bond, slave) \
2156 for_each_netdev_rcu(&init_net, slave) \
2157 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2158 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2159
2160 static inline struct net_device *next_net_device(struct net_device *dev)
2161 {
2162 struct list_head *lh;
2163 struct net *net;
2164
2165 net = dev_net(dev);
2166 lh = dev->dev_list.next;
2167 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2168 }
2169
2170 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2171 {
2172 struct list_head *lh;
2173 struct net *net;
2174
2175 net = dev_net(dev);
2176 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2177 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2178 }
2179
2180 static inline struct net_device *first_net_device(struct net *net)
2181 {
2182 return list_empty(&net->dev_base_head) ? NULL :
2183 net_device_entry(net->dev_base_head.next);
2184 }
2185
2186 static inline struct net_device *first_net_device_rcu(struct net *net)
2187 {
2188 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2189
2190 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2191 }
2192
2193 int netdev_boot_setup_check(struct net_device *dev);
2194 unsigned long netdev_boot_base(const char *prefix, int unit);
2195 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2196 const char *hwaddr);
2197 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2198 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2199 void dev_add_pack(struct packet_type *pt);
2200 void dev_remove_pack(struct packet_type *pt);
2201 void __dev_remove_pack(struct packet_type *pt);
2202 void dev_add_offload(struct packet_offload *po);
2203 void dev_remove_offload(struct packet_offload *po);
2204
2205 int dev_get_iflink(const struct net_device *dev);
2206 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2207 unsigned short mask);
2208 struct net_device *dev_get_by_name(struct net *net, const char *name);
2209 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2210 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2211 int dev_alloc_name(struct net_device *dev, const char *name);
2212 int dev_open(struct net_device *dev);
2213 int dev_close(struct net_device *dev);
2214 int dev_close_many(struct list_head *head, bool unlink);
2215 void dev_disable_lro(struct net_device *dev);
2216 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2217 int dev_queue_xmit(struct sk_buff *skb);
2218 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2219 int register_netdevice(struct net_device *dev);
2220 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2221 void unregister_netdevice_many(struct list_head *head);
2222 static inline void unregister_netdevice(struct net_device *dev)
2223 {
2224 unregister_netdevice_queue(dev, NULL);
2225 }
2226
2227 int netdev_refcnt_read(const struct net_device *dev);
2228 void free_netdev(struct net_device *dev);
2229 void netdev_freemem(struct net_device *dev);
2230 void synchronize_net(void);
2231 int init_dummy_netdev(struct net_device *dev);
2232
2233 DECLARE_PER_CPU(int, xmit_recursion);
2234 static inline int dev_recursion_level(void)
2235 {
2236 return this_cpu_read(xmit_recursion);
2237 }
2238
2239 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2240 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2241 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2242 int netdev_get_name(struct net *net, char *name, int ifindex);
2243 int dev_restart(struct net_device *dev);
2244 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2245
2246 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2247 {
2248 return NAPI_GRO_CB(skb)->data_offset;
2249 }
2250
2251 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2252 {
2253 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2254 }
2255
2256 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2257 {
2258 NAPI_GRO_CB(skb)->data_offset += len;
2259 }
2260
2261 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2262 unsigned int offset)
2263 {
2264 return NAPI_GRO_CB(skb)->frag0 + offset;
2265 }
2266
2267 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2268 {
2269 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2270 }
2271
2272 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2273 unsigned int offset)
2274 {
2275 if (!pskb_may_pull(skb, hlen))
2276 return NULL;
2277
2278 NAPI_GRO_CB(skb)->frag0 = NULL;
2279 NAPI_GRO_CB(skb)->frag0_len = 0;
2280 return skb->data + offset;
2281 }
2282
2283 static inline void *skb_gro_network_header(struct sk_buff *skb)
2284 {
2285 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2286 skb_network_offset(skb);
2287 }
2288
2289 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2290 const void *start, unsigned int len)
2291 {
2292 if (NAPI_GRO_CB(skb)->csum_valid)
2293 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2294 csum_partial(start, len, 0));
2295 }
2296
2297 /* GRO checksum functions. These are logical equivalents of the normal
2298 * checksum functions (in skbuff.h) except that they operate on the GRO
2299 * offsets and fields in sk_buff.
2300 */
2301
2302 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2303
2304 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2305 {
2306 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2307 }
2308
2309 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2310 bool zero_okay,
2311 __sum16 check)
2312 {
2313 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2314 skb_checksum_start_offset(skb) <
2315 skb_gro_offset(skb)) &&
2316 !skb_at_gro_remcsum_start(skb) &&
2317 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2318 (!zero_okay || check));
2319 }
2320
2321 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2322 __wsum psum)
2323 {
2324 if (NAPI_GRO_CB(skb)->csum_valid &&
2325 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2326 return 0;
2327
2328 NAPI_GRO_CB(skb)->csum = psum;
2329
2330 return __skb_gro_checksum_complete(skb);
2331 }
2332
2333 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2334 {
2335 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2336 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2337 NAPI_GRO_CB(skb)->csum_cnt--;
2338 } else {
2339 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2340 * verified a new top level checksum or an encapsulated one
2341 * during GRO. This saves work if we fallback to normal path.
2342 */
2343 __skb_incr_checksum_unnecessary(skb);
2344 }
2345 }
2346
2347 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2348 compute_pseudo) \
2349 ({ \
2350 __sum16 __ret = 0; \
2351 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2352 __ret = __skb_gro_checksum_validate_complete(skb, \
2353 compute_pseudo(skb, proto)); \
2354 if (__ret) \
2355 __skb_mark_checksum_bad(skb); \
2356 else \
2357 skb_gro_incr_csum_unnecessary(skb); \
2358 __ret; \
2359 })
2360
2361 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2362 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2363
2364 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2365 compute_pseudo) \
2366 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2367
2368 #define skb_gro_checksum_simple_validate(skb) \
2369 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2370
2371 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2372 {
2373 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2374 !NAPI_GRO_CB(skb)->csum_valid);
2375 }
2376
2377 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2378 __sum16 check, __wsum pseudo)
2379 {
2380 NAPI_GRO_CB(skb)->csum = ~pseudo;
2381 NAPI_GRO_CB(skb)->csum_valid = 1;
2382 }
2383
2384 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2385 do { \
2386 if (__skb_gro_checksum_convert_check(skb)) \
2387 __skb_gro_checksum_convert(skb, check, \
2388 compute_pseudo(skb, proto)); \
2389 } while (0)
2390
2391 struct gro_remcsum {
2392 int offset;
2393 __wsum delta;
2394 };
2395
2396 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2397 {
2398 grc->offset = 0;
2399 grc->delta = 0;
2400 }
2401
2402 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2403 unsigned int off, size_t hdrlen,
2404 int start, int offset,
2405 struct gro_remcsum *grc,
2406 bool nopartial)
2407 {
2408 __wsum delta;
2409 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2410
2411 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2412
2413 if (!nopartial) {
2414 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2415 return ptr;
2416 }
2417
2418 ptr = skb_gro_header_fast(skb, off);
2419 if (skb_gro_header_hard(skb, off + plen)) {
2420 ptr = skb_gro_header_slow(skb, off + plen, off);
2421 if (!ptr)
2422 return NULL;
2423 }
2424
2425 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2426 start, offset);
2427
2428 /* Adjust skb->csum since we changed the packet */
2429 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2430
2431 grc->offset = off + hdrlen + offset;
2432 grc->delta = delta;
2433
2434 return ptr;
2435 }
2436
2437 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2438 struct gro_remcsum *grc)
2439 {
2440 void *ptr;
2441 size_t plen = grc->offset + sizeof(u16);
2442
2443 if (!grc->delta)
2444 return;
2445
2446 ptr = skb_gro_header_fast(skb, grc->offset);
2447 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2448 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2449 if (!ptr)
2450 return;
2451 }
2452
2453 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2454 }
2455
2456 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2457 unsigned short type,
2458 const void *daddr, const void *saddr,
2459 unsigned int len)
2460 {
2461 if (!dev->header_ops || !dev->header_ops->create)
2462 return 0;
2463
2464 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2465 }
2466
2467 static inline int dev_parse_header(const struct sk_buff *skb,
2468 unsigned char *haddr)
2469 {
2470 const struct net_device *dev = skb->dev;
2471
2472 if (!dev->header_ops || !dev->header_ops->parse)
2473 return 0;
2474 return dev->header_ops->parse(skb, haddr);
2475 }
2476
2477 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2478 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2479 static inline int unregister_gifconf(unsigned int family)
2480 {
2481 return register_gifconf(family, NULL);
2482 }
2483
2484 #ifdef CONFIG_NET_FLOW_LIMIT
2485 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2486 struct sd_flow_limit {
2487 u64 count;
2488 unsigned int num_buckets;
2489 unsigned int history_head;
2490 u16 history[FLOW_LIMIT_HISTORY];
2491 u8 buckets[];
2492 };
2493
2494 extern int netdev_flow_limit_table_len;
2495 #endif /* CONFIG_NET_FLOW_LIMIT */
2496
2497 /*
2498 * Incoming packets are placed on per-cpu queues
2499 */
2500 struct softnet_data {
2501 struct list_head poll_list;
2502 struct sk_buff_head process_queue;
2503
2504 /* stats */
2505 unsigned int processed;
2506 unsigned int time_squeeze;
2507 unsigned int cpu_collision;
2508 unsigned int received_rps;
2509 #ifdef CONFIG_RPS
2510 struct softnet_data *rps_ipi_list;
2511 #endif
2512 #ifdef CONFIG_NET_FLOW_LIMIT
2513 struct sd_flow_limit __rcu *flow_limit;
2514 #endif
2515 struct Qdisc *output_queue;
2516 struct Qdisc **output_queue_tailp;
2517 struct sk_buff *completion_queue;
2518
2519 #ifdef CONFIG_RPS
2520 /* Elements below can be accessed between CPUs for RPS */
2521 struct call_single_data csd ____cacheline_aligned_in_smp;
2522 struct softnet_data *rps_ipi_next;
2523 unsigned int cpu;
2524 unsigned int input_queue_head;
2525 unsigned int input_queue_tail;
2526 #endif
2527 unsigned int dropped;
2528 struct sk_buff_head input_pkt_queue;
2529 struct napi_struct backlog;
2530
2531 };
2532
2533 static inline void input_queue_head_incr(struct softnet_data *sd)
2534 {
2535 #ifdef CONFIG_RPS
2536 sd->input_queue_head++;
2537 #endif
2538 }
2539
2540 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2541 unsigned int *qtail)
2542 {
2543 #ifdef CONFIG_RPS
2544 *qtail = ++sd->input_queue_tail;
2545 #endif
2546 }
2547
2548 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2549
2550 void __netif_schedule(struct Qdisc *q);
2551 void netif_schedule_queue(struct netdev_queue *txq);
2552
2553 static inline void netif_tx_schedule_all(struct net_device *dev)
2554 {
2555 unsigned int i;
2556
2557 for (i = 0; i < dev->num_tx_queues; i++)
2558 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2559 }
2560
2561 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2562 {
2563 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2564 }
2565
2566 /**
2567 * netif_start_queue - allow transmit
2568 * @dev: network device
2569 *
2570 * Allow upper layers to call the device hard_start_xmit routine.
2571 */
2572 static inline void netif_start_queue(struct net_device *dev)
2573 {
2574 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2575 }
2576
2577 static inline void netif_tx_start_all_queues(struct net_device *dev)
2578 {
2579 unsigned int i;
2580
2581 for (i = 0; i < dev->num_tx_queues; i++) {
2582 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2583 netif_tx_start_queue(txq);
2584 }
2585 }
2586
2587 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2588
2589 /**
2590 * netif_wake_queue - restart transmit
2591 * @dev: network device
2592 *
2593 * Allow upper layers to call the device hard_start_xmit routine.
2594 * Used for flow control when transmit resources are available.
2595 */
2596 static inline void netif_wake_queue(struct net_device *dev)
2597 {
2598 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2599 }
2600
2601 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2602 {
2603 unsigned int i;
2604
2605 for (i = 0; i < dev->num_tx_queues; i++) {
2606 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2607 netif_tx_wake_queue(txq);
2608 }
2609 }
2610
2611 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2612 {
2613 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2614 }
2615
2616 /**
2617 * netif_stop_queue - stop transmitted packets
2618 * @dev: network device
2619 *
2620 * Stop upper layers calling the device hard_start_xmit routine.
2621 * Used for flow control when transmit resources are unavailable.
2622 */
2623 static inline void netif_stop_queue(struct net_device *dev)
2624 {
2625 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2626 }
2627
2628 void netif_tx_stop_all_queues(struct net_device *dev);
2629
2630 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2631 {
2632 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2633 }
2634
2635 /**
2636 * netif_queue_stopped - test if transmit queue is flowblocked
2637 * @dev: network device
2638 *
2639 * Test if transmit queue on device is currently unable to send.
2640 */
2641 static inline bool netif_queue_stopped(const struct net_device *dev)
2642 {
2643 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2644 }
2645
2646 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2647 {
2648 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2649 }
2650
2651 static inline bool
2652 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2653 {
2654 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2655 }
2656
2657 static inline bool
2658 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2659 {
2660 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2661 }
2662
2663 /**
2664 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2665 * @dev_queue: pointer to transmit queue
2666 *
2667 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2668 * to give appropriate hint to the cpu.
2669 */
2670 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2671 {
2672 #ifdef CONFIG_BQL
2673 prefetchw(&dev_queue->dql.num_queued);
2674 #endif
2675 }
2676
2677 /**
2678 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2679 * @dev_queue: pointer to transmit queue
2680 *
2681 * BQL enabled drivers might use this helper in their TX completion path,
2682 * to give appropriate hint to the cpu.
2683 */
2684 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2685 {
2686 #ifdef CONFIG_BQL
2687 prefetchw(&dev_queue->dql.limit);
2688 #endif
2689 }
2690
2691 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2692 unsigned int bytes)
2693 {
2694 #ifdef CONFIG_BQL
2695 dql_queued(&dev_queue->dql, bytes);
2696
2697 if (likely(dql_avail(&dev_queue->dql) >= 0))
2698 return;
2699
2700 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2701
2702 /*
2703 * The XOFF flag must be set before checking the dql_avail below,
2704 * because in netdev_tx_completed_queue we update the dql_completed
2705 * before checking the XOFF flag.
2706 */
2707 smp_mb();
2708
2709 /* check again in case another CPU has just made room avail */
2710 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2711 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2712 #endif
2713 }
2714
2715 /**
2716 * netdev_sent_queue - report the number of bytes queued to hardware
2717 * @dev: network device
2718 * @bytes: number of bytes queued to the hardware device queue
2719 *
2720 * Report the number of bytes queued for sending/completion to the network
2721 * device hardware queue. @bytes should be a good approximation and should
2722 * exactly match netdev_completed_queue() @bytes
2723 */
2724 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2725 {
2726 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2727 }
2728
2729 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2730 unsigned int pkts, unsigned int bytes)
2731 {
2732 #ifdef CONFIG_BQL
2733 if (unlikely(!bytes))
2734 return;
2735
2736 dql_completed(&dev_queue->dql, bytes);
2737
2738 /*
2739 * Without the memory barrier there is a small possiblity that
2740 * netdev_tx_sent_queue will miss the update and cause the queue to
2741 * be stopped forever
2742 */
2743 smp_mb();
2744
2745 if (dql_avail(&dev_queue->dql) < 0)
2746 return;
2747
2748 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2749 netif_schedule_queue(dev_queue);
2750 #endif
2751 }
2752
2753 /**
2754 * netdev_completed_queue - report bytes and packets completed by device
2755 * @dev: network device
2756 * @pkts: actual number of packets sent over the medium
2757 * @bytes: actual number of bytes sent over the medium
2758 *
2759 * Report the number of bytes and packets transmitted by the network device
2760 * hardware queue over the physical medium, @bytes must exactly match the
2761 * @bytes amount passed to netdev_sent_queue()
2762 */
2763 static inline void netdev_completed_queue(struct net_device *dev,
2764 unsigned int pkts, unsigned int bytes)
2765 {
2766 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2767 }
2768
2769 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2770 {
2771 #ifdef CONFIG_BQL
2772 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2773 dql_reset(&q->dql);
2774 #endif
2775 }
2776
2777 /**
2778 * netdev_reset_queue - reset the packets and bytes count of a network device
2779 * @dev_queue: network device
2780 *
2781 * Reset the bytes and packet count of a network device and clear the
2782 * software flow control OFF bit for this network device
2783 */
2784 static inline void netdev_reset_queue(struct net_device *dev_queue)
2785 {
2786 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2787 }
2788
2789 /**
2790 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2791 * @dev: network device
2792 * @queue_index: given tx queue index
2793 *
2794 * Returns 0 if given tx queue index >= number of device tx queues,
2795 * otherwise returns the originally passed tx queue index.
2796 */
2797 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2798 {
2799 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2800 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2801 dev->name, queue_index,
2802 dev->real_num_tx_queues);
2803 return 0;
2804 }
2805
2806 return queue_index;
2807 }
2808
2809 /**
2810 * netif_running - test if up
2811 * @dev: network device
2812 *
2813 * Test if the device has been brought up.
2814 */
2815 static inline bool netif_running(const struct net_device *dev)
2816 {
2817 return test_bit(__LINK_STATE_START, &dev->state);
2818 }
2819
2820 /*
2821 * Routines to manage the subqueues on a device. We only need start
2822 * stop, and a check if it's stopped. All other device management is
2823 * done at the overall netdevice level.
2824 * Also test the device if we're multiqueue.
2825 */
2826
2827 /**
2828 * netif_start_subqueue - allow sending packets on subqueue
2829 * @dev: network device
2830 * @queue_index: sub queue index
2831 *
2832 * Start individual transmit queue of a device with multiple transmit queues.
2833 */
2834 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2835 {
2836 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2837
2838 netif_tx_start_queue(txq);
2839 }
2840
2841 /**
2842 * netif_stop_subqueue - stop sending packets on subqueue
2843 * @dev: network device
2844 * @queue_index: sub queue index
2845 *
2846 * Stop individual transmit queue of a device with multiple transmit queues.
2847 */
2848 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2849 {
2850 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2851 netif_tx_stop_queue(txq);
2852 }
2853
2854 /**
2855 * netif_subqueue_stopped - test status of subqueue
2856 * @dev: network device
2857 * @queue_index: sub queue index
2858 *
2859 * Check individual transmit queue of a device with multiple transmit queues.
2860 */
2861 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2862 u16 queue_index)
2863 {
2864 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2865
2866 return netif_tx_queue_stopped(txq);
2867 }
2868
2869 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2870 struct sk_buff *skb)
2871 {
2872 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2873 }
2874
2875 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2876
2877 #ifdef CONFIG_XPS
2878 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2879 u16 index);
2880 #else
2881 static inline int netif_set_xps_queue(struct net_device *dev,
2882 const struct cpumask *mask,
2883 u16 index)
2884 {
2885 return 0;
2886 }
2887 #endif
2888
2889 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2890 unsigned int num_tx_queues);
2891
2892 /*
2893 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2894 * as a distribution range limit for the returned value.
2895 */
2896 static inline u16 skb_tx_hash(const struct net_device *dev,
2897 struct sk_buff *skb)
2898 {
2899 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2900 }
2901
2902 /**
2903 * netif_is_multiqueue - test if device has multiple transmit queues
2904 * @dev: network device
2905 *
2906 * Check if device has multiple transmit queues
2907 */
2908 static inline bool netif_is_multiqueue(const struct net_device *dev)
2909 {
2910 return dev->num_tx_queues > 1;
2911 }
2912
2913 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2914
2915 #ifdef CONFIG_SYSFS
2916 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2917 #else
2918 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2919 unsigned int rxq)
2920 {
2921 return 0;
2922 }
2923 #endif
2924
2925 #ifdef CONFIG_SYSFS
2926 static inline unsigned int get_netdev_rx_queue_index(
2927 struct netdev_rx_queue *queue)
2928 {
2929 struct net_device *dev = queue->dev;
2930 int index = queue - dev->_rx;
2931
2932 BUG_ON(index >= dev->num_rx_queues);
2933 return index;
2934 }
2935 #endif
2936
2937 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2938 int netif_get_num_default_rss_queues(void);
2939
2940 enum skb_free_reason {
2941 SKB_REASON_CONSUMED,
2942 SKB_REASON_DROPPED,
2943 };
2944
2945 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2946 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2947
2948 /*
2949 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2950 * interrupt context or with hardware interrupts being disabled.
2951 * (in_irq() || irqs_disabled())
2952 *
2953 * We provide four helpers that can be used in following contexts :
2954 *
2955 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2956 * replacing kfree_skb(skb)
2957 *
2958 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2959 * Typically used in place of consume_skb(skb) in TX completion path
2960 *
2961 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2962 * replacing kfree_skb(skb)
2963 *
2964 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2965 * and consumed a packet. Used in place of consume_skb(skb)
2966 */
2967 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2968 {
2969 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2970 }
2971
2972 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2973 {
2974 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2975 }
2976
2977 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2978 {
2979 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2980 }
2981
2982 static inline void dev_consume_skb_any(struct sk_buff *skb)
2983 {
2984 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2985 }
2986
2987 int netif_rx(struct sk_buff *skb);
2988 int netif_rx_ni(struct sk_buff *skb);
2989 int netif_receive_skb(struct sk_buff *skb);
2990 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2991 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2992 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2993 gro_result_t napi_gro_frags(struct napi_struct *napi);
2994 struct packet_offload *gro_find_receive_by_type(__be16 type);
2995 struct packet_offload *gro_find_complete_by_type(__be16 type);
2996
2997 static inline void napi_free_frags(struct napi_struct *napi)
2998 {
2999 kfree_skb(napi->skb);
3000 napi->skb = NULL;
3001 }
3002
3003 int netdev_rx_handler_register(struct net_device *dev,
3004 rx_handler_func_t *rx_handler,
3005 void *rx_handler_data);
3006 void netdev_rx_handler_unregister(struct net_device *dev);
3007
3008 bool dev_valid_name(const char *name);
3009 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3010 int dev_ethtool(struct net *net, struct ifreq *);
3011 unsigned int dev_get_flags(const struct net_device *);
3012 int __dev_change_flags(struct net_device *, unsigned int flags);
3013 int dev_change_flags(struct net_device *, unsigned int);
3014 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3015 unsigned int gchanges);
3016 int dev_change_name(struct net_device *, const char *);
3017 int dev_set_alias(struct net_device *, const char *, size_t);
3018 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3019 int dev_set_mtu(struct net_device *, int);
3020 void dev_set_group(struct net_device *, int);
3021 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3022 int dev_change_carrier(struct net_device *, bool new_carrier);
3023 int dev_get_phys_port_id(struct net_device *dev,
3024 struct netdev_phys_item_id *ppid);
3025 int dev_get_phys_port_name(struct net_device *dev,
3026 char *name, size_t len);
3027 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3028 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3029 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3030 struct netdev_queue *txq, int *ret);
3031 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3032 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3033 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3034
3035 extern int netdev_budget;
3036
3037 /* Called by rtnetlink.c:rtnl_unlock() */
3038 void netdev_run_todo(void);
3039
3040 /**
3041 * dev_put - release reference to device
3042 * @dev: network device
3043 *
3044 * Release reference to device to allow it to be freed.
3045 */
3046 static inline void dev_put(struct net_device *dev)
3047 {
3048 this_cpu_dec(*dev->pcpu_refcnt);
3049 }
3050
3051 /**
3052 * dev_hold - get reference to device
3053 * @dev: network device
3054 *
3055 * Hold reference to device to keep it from being freed.
3056 */
3057 static inline void dev_hold(struct net_device *dev)
3058 {
3059 this_cpu_inc(*dev->pcpu_refcnt);
3060 }
3061
3062 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3063 * and _off may be called from IRQ context, but it is caller
3064 * who is responsible for serialization of these calls.
3065 *
3066 * The name carrier is inappropriate, these functions should really be
3067 * called netif_lowerlayer_*() because they represent the state of any
3068 * kind of lower layer not just hardware media.
3069 */
3070
3071 void linkwatch_init_dev(struct net_device *dev);
3072 void linkwatch_fire_event(struct net_device *dev);
3073 void linkwatch_forget_dev(struct net_device *dev);
3074
3075 /**
3076 * netif_carrier_ok - test if carrier present
3077 * @dev: network device
3078 *
3079 * Check if carrier is present on device
3080 */
3081 static inline bool netif_carrier_ok(const struct net_device *dev)
3082 {
3083 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3084 }
3085
3086 unsigned long dev_trans_start(struct net_device *dev);
3087
3088 void __netdev_watchdog_up(struct net_device *dev);
3089
3090 void netif_carrier_on(struct net_device *dev);
3091
3092 void netif_carrier_off(struct net_device *dev);
3093
3094 /**
3095 * netif_dormant_on - mark device as dormant.
3096 * @dev: network device
3097 *
3098 * Mark device as dormant (as per RFC2863).
3099 *
3100 * The dormant state indicates that the relevant interface is not
3101 * actually in a condition to pass packets (i.e., it is not 'up') but is
3102 * in a "pending" state, waiting for some external event. For "on-
3103 * demand" interfaces, this new state identifies the situation where the
3104 * interface is waiting for events to place it in the up state.
3105 *
3106 */
3107 static inline void netif_dormant_on(struct net_device *dev)
3108 {
3109 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3110 linkwatch_fire_event(dev);
3111 }
3112
3113 /**
3114 * netif_dormant_off - set device as not dormant.
3115 * @dev: network device
3116 *
3117 * Device is not in dormant state.
3118 */
3119 static inline void netif_dormant_off(struct net_device *dev)
3120 {
3121 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3122 linkwatch_fire_event(dev);
3123 }
3124
3125 /**
3126 * netif_dormant - test if carrier present
3127 * @dev: network device
3128 *
3129 * Check if carrier is present on device
3130 */
3131 static inline bool netif_dormant(const struct net_device *dev)
3132 {
3133 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3134 }
3135
3136
3137 /**
3138 * netif_oper_up - test if device is operational
3139 * @dev: network device
3140 *
3141 * Check if carrier is operational
3142 */
3143 static inline bool netif_oper_up(const struct net_device *dev)
3144 {
3145 return (dev->operstate == IF_OPER_UP ||
3146 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3147 }
3148
3149 /**
3150 * netif_device_present - is device available or removed
3151 * @dev: network device
3152 *
3153 * Check if device has not been removed from system.
3154 */
3155 static inline bool netif_device_present(struct net_device *dev)
3156 {
3157 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3158 }
3159
3160 void netif_device_detach(struct net_device *dev);
3161
3162 void netif_device_attach(struct net_device *dev);
3163
3164 /*
3165 * Network interface message level settings
3166 */
3167
3168 enum {
3169 NETIF_MSG_DRV = 0x0001,
3170 NETIF_MSG_PROBE = 0x0002,
3171 NETIF_MSG_LINK = 0x0004,
3172 NETIF_MSG_TIMER = 0x0008,
3173 NETIF_MSG_IFDOWN = 0x0010,
3174 NETIF_MSG_IFUP = 0x0020,
3175 NETIF_MSG_RX_ERR = 0x0040,
3176 NETIF_MSG_TX_ERR = 0x0080,
3177 NETIF_MSG_TX_QUEUED = 0x0100,
3178 NETIF_MSG_INTR = 0x0200,
3179 NETIF_MSG_TX_DONE = 0x0400,
3180 NETIF_MSG_RX_STATUS = 0x0800,
3181 NETIF_MSG_PKTDATA = 0x1000,
3182 NETIF_MSG_HW = 0x2000,
3183 NETIF_MSG_WOL = 0x4000,
3184 };
3185
3186 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3187 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3188 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3189 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3190 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3191 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3192 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3193 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3194 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3195 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3196 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3197 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3198 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3199 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3200 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3201
3202 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3203 {
3204 /* use default */
3205 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3206 return default_msg_enable_bits;
3207 if (debug_value == 0) /* no output */
3208 return 0;
3209 /* set low N bits */
3210 return (1 << debug_value) - 1;
3211 }
3212
3213 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3214 {
3215 spin_lock(&txq->_xmit_lock);
3216 txq->xmit_lock_owner = cpu;
3217 }
3218
3219 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3220 {
3221 spin_lock_bh(&txq->_xmit_lock);
3222 txq->xmit_lock_owner = smp_processor_id();
3223 }
3224
3225 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3226 {
3227 bool ok = spin_trylock(&txq->_xmit_lock);
3228 if (likely(ok))
3229 txq->xmit_lock_owner = smp_processor_id();
3230 return ok;
3231 }
3232
3233 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3234 {
3235 txq->xmit_lock_owner = -1;
3236 spin_unlock(&txq->_xmit_lock);
3237 }
3238
3239 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3240 {
3241 txq->xmit_lock_owner = -1;
3242 spin_unlock_bh(&txq->_xmit_lock);
3243 }
3244
3245 static inline void txq_trans_update(struct netdev_queue *txq)
3246 {
3247 if (txq->xmit_lock_owner != -1)
3248 txq->trans_start = jiffies;
3249 }
3250
3251 /**
3252 * netif_tx_lock - grab network device transmit lock
3253 * @dev: network device
3254 *
3255 * Get network device transmit lock
3256 */
3257 static inline void netif_tx_lock(struct net_device *dev)
3258 {
3259 unsigned int i;
3260 int cpu;
3261
3262 spin_lock(&dev->tx_global_lock);
3263 cpu = smp_processor_id();
3264 for (i = 0; i < dev->num_tx_queues; i++) {
3265 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3266
3267 /* We are the only thread of execution doing a
3268 * freeze, but we have to grab the _xmit_lock in
3269 * order to synchronize with threads which are in
3270 * the ->hard_start_xmit() handler and already
3271 * checked the frozen bit.
3272 */
3273 __netif_tx_lock(txq, cpu);
3274 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3275 __netif_tx_unlock(txq);
3276 }
3277 }
3278
3279 static inline void netif_tx_lock_bh(struct net_device *dev)
3280 {
3281 local_bh_disable();
3282 netif_tx_lock(dev);
3283 }
3284
3285 static inline void netif_tx_unlock(struct net_device *dev)
3286 {
3287 unsigned int i;
3288
3289 for (i = 0; i < dev->num_tx_queues; i++) {
3290 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3291
3292 /* No need to grab the _xmit_lock here. If the
3293 * queue is not stopped for another reason, we
3294 * force a schedule.
3295 */
3296 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3297 netif_schedule_queue(txq);
3298 }
3299 spin_unlock(&dev->tx_global_lock);
3300 }
3301
3302 static inline void netif_tx_unlock_bh(struct net_device *dev)
3303 {
3304 netif_tx_unlock(dev);
3305 local_bh_enable();
3306 }
3307
3308 #define HARD_TX_LOCK(dev, txq, cpu) { \
3309 if ((dev->features & NETIF_F_LLTX) == 0) { \
3310 __netif_tx_lock(txq, cpu); \
3311 } \
3312 }
3313
3314 #define HARD_TX_TRYLOCK(dev, txq) \
3315 (((dev->features & NETIF_F_LLTX) == 0) ? \
3316 __netif_tx_trylock(txq) : \
3317 true )
3318
3319 #define HARD_TX_UNLOCK(dev, txq) { \
3320 if ((dev->features & NETIF_F_LLTX) == 0) { \
3321 __netif_tx_unlock(txq); \
3322 } \
3323 }
3324
3325 static inline void netif_tx_disable(struct net_device *dev)
3326 {
3327 unsigned int i;
3328 int cpu;
3329
3330 local_bh_disable();
3331 cpu = smp_processor_id();
3332 for (i = 0; i < dev->num_tx_queues; i++) {
3333 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3334
3335 __netif_tx_lock(txq, cpu);
3336 netif_tx_stop_queue(txq);
3337 __netif_tx_unlock(txq);
3338 }
3339 local_bh_enable();
3340 }
3341
3342 static inline void netif_addr_lock(struct net_device *dev)
3343 {
3344 spin_lock(&dev->addr_list_lock);
3345 }
3346
3347 static inline void netif_addr_lock_nested(struct net_device *dev)
3348 {
3349 int subclass = SINGLE_DEPTH_NESTING;
3350
3351 if (dev->netdev_ops->ndo_get_lock_subclass)
3352 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3353
3354 spin_lock_nested(&dev->addr_list_lock, subclass);
3355 }
3356
3357 static inline void netif_addr_lock_bh(struct net_device *dev)
3358 {
3359 spin_lock_bh(&dev->addr_list_lock);
3360 }
3361
3362 static inline void netif_addr_unlock(struct net_device *dev)
3363 {
3364 spin_unlock(&dev->addr_list_lock);
3365 }
3366
3367 static inline void netif_addr_unlock_bh(struct net_device *dev)
3368 {
3369 spin_unlock_bh(&dev->addr_list_lock);
3370 }
3371
3372 /*
3373 * dev_addrs walker. Should be used only for read access. Call with
3374 * rcu_read_lock held.
3375 */
3376 #define for_each_dev_addr(dev, ha) \
3377 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3378
3379 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3380
3381 void ether_setup(struct net_device *dev);
3382
3383 /* Support for loadable net-drivers */
3384 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3385 unsigned char name_assign_type,
3386 void (*setup)(struct net_device *),
3387 unsigned int txqs, unsigned int rxqs);
3388 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3389 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3390
3391 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3392 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3393 count)
3394
3395 int register_netdev(struct net_device *dev);
3396 void unregister_netdev(struct net_device *dev);
3397
3398 /* General hardware address lists handling functions */
3399 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3400 struct netdev_hw_addr_list *from_list, int addr_len);
3401 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3402 struct netdev_hw_addr_list *from_list, int addr_len);
3403 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3404 struct net_device *dev,
3405 int (*sync)(struct net_device *, const unsigned char *),
3406 int (*unsync)(struct net_device *,
3407 const unsigned char *));
3408 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3409 struct net_device *dev,
3410 int (*unsync)(struct net_device *,
3411 const unsigned char *));
3412 void __hw_addr_init(struct netdev_hw_addr_list *list);
3413
3414 /* Functions used for device addresses handling */
3415 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3416 unsigned char addr_type);
3417 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3418 unsigned char addr_type);
3419 void dev_addr_flush(struct net_device *dev);
3420 int dev_addr_init(struct net_device *dev);
3421
3422 /* Functions used for unicast addresses handling */
3423 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3424 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3425 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3426 int dev_uc_sync(struct net_device *to, struct net_device *from);
3427 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3428 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3429 void dev_uc_flush(struct net_device *dev);
3430 void dev_uc_init(struct net_device *dev);
3431
3432 /**
3433 * __dev_uc_sync - Synchonize device's unicast list
3434 * @dev: device to sync
3435 * @sync: function to call if address should be added
3436 * @unsync: function to call if address should be removed
3437 *
3438 * Add newly added addresses to the interface, and release
3439 * addresses that have been deleted.
3440 **/
3441 static inline int __dev_uc_sync(struct net_device *dev,
3442 int (*sync)(struct net_device *,
3443 const unsigned char *),
3444 int (*unsync)(struct net_device *,
3445 const unsigned char *))
3446 {
3447 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3448 }
3449
3450 /**
3451 * __dev_uc_unsync - Remove synchronized addresses from device
3452 * @dev: device to sync
3453 * @unsync: function to call if address should be removed
3454 *
3455 * Remove all addresses that were added to the device by dev_uc_sync().
3456 **/
3457 static inline void __dev_uc_unsync(struct net_device *dev,
3458 int (*unsync)(struct net_device *,
3459 const unsigned char *))
3460 {
3461 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3462 }
3463
3464 /* Functions used for multicast addresses handling */
3465 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3466 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3467 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3468 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3469 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3470 int dev_mc_sync(struct net_device *to, struct net_device *from);
3471 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3472 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3473 void dev_mc_flush(struct net_device *dev);
3474 void dev_mc_init(struct net_device *dev);
3475
3476 /**
3477 * __dev_mc_sync - Synchonize device's multicast list
3478 * @dev: device to sync
3479 * @sync: function to call if address should be added
3480 * @unsync: function to call if address should be removed
3481 *
3482 * Add newly added addresses to the interface, and release
3483 * addresses that have been deleted.
3484 **/
3485 static inline int __dev_mc_sync(struct net_device *dev,
3486 int (*sync)(struct net_device *,
3487 const unsigned char *),
3488 int (*unsync)(struct net_device *,
3489 const unsigned char *))
3490 {
3491 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3492 }
3493
3494 /**
3495 * __dev_mc_unsync - Remove synchronized addresses from device
3496 * @dev: device to sync
3497 * @unsync: function to call if address should be removed
3498 *
3499 * Remove all addresses that were added to the device by dev_mc_sync().
3500 **/
3501 static inline void __dev_mc_unsync(struct net_device *dev,
3502 int (*unsync)(struct net_device *,
3503 const unsigned char *))
3504 {
3505 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3506 }
3507
3508 /* Functions used for secondary unicast and multicast support */
3509 void dev_set_rx_mode(struct net_device *dev);
3510 void __dev_set_rx_mode(struct net_device *dev);
3511 int dev_set_promiscuity(struct net_device *dev, int inc);
3512 int dev_set_allmulti(struct net_device *dev, int inc);
3513 void netdev_state_change(struct net_device *dev);
3514 void netdev_notify_peers(struct net_device *dev);
3515 void netdev_features_change(struct net_device *dev);
3516 /* Load a device via the kmod */
3517 void dev_load(struct net *net, const char *name);
3518 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3519 struct rtnl_link_stats64 *storage);
3520 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3521 const struct net_device_stats *netdev_stats);
3522
3523 extern int netdev_max_backlog;
3524 extern int netdev_tstamp_prequeue;
3525 extern int weight_p;
3526 extern int bpf_jit_enable;
3527
3528 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3529 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3530 struct list_head **iter);
3531 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3532 struct list_head **iter);
3533
3534 /* iterate through upper list, must be called under RCU read lock */
3535 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3536 for (iter = &(dev)->adj_list.upper, \
3537 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3538 updev; \
3539 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3540
3541 /* iterate through upper list, must be called under RCU read lock */
3542 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3543 for (iter = &(dev)->all_adj_list.upper, \
3544 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3545 updev; \
3546 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3547
3548 void *netdev_lower_get_next_private(struct net_device *dev,
3549 struct list_head **iter);
3550 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3551 struct list_head **iter);
3552
3553 #define netdev_for_each_lower_private(dev, priv, iter) \
3554 for (iter = (dev)->adj_list.lower.next, \
3555 priv = netdev_lower_get_next_private(dev, &(iter)); \
3556 priv; \
3557 priv = netdev_lower_get_next_private(dev, &(iter)))
3558
3559 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3560 for (iter = &(dev)->adj_list.lower, \
3561 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3562 priv; \
3563 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3564
3565 void *netdev_lower_get_next(struct net_device *dev,
3566 struct list_head **iter);
3567 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3568 for (iter = &(dev)->adj_list.lower, \
3569 ldev = netdev_lower_get_next(dev, &(iter)); \
3570 ldev; \
3571 ldev = netdev_lower_get_next(dev, &(iter)))
3572
3573 void *netdev_adjacent_get_private(struct list_head *adj_list);
3574 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3575 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3576 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3577 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3578 int netdev_master_upper_dev_link(struct net_device *dev,
3579 struct net_device *upper_dev);
3580 int netdev_master_upper_dev_link_private(struct net_device *dev,
3581 struct net_device *upper_dev,
3582 void *private);
3583 void netdev_upper_dev_unlink(struct net_device *dev,
3584 struct net_device *upper_dev);
3585 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3586 void *netdev_lower_dev_get_private(struct net_device *dev,
3587 struct net_device *lower_dev);
3588
3589 /* RSS keys are 40 or 52 bytes long */
3590 #define NETDEV_RSS_KEY_LEN 52
3591 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3592 void netdev_rss_key_fill(void *buffer, size_t len);
3593
3594 int dev_get_nest_level(struct net_device *dev,
3595 bool (*type_check)(struct net_device *dev));
3596 int skb_checksum_help(struct sk_buff *skb);
3597 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3598 netdev_features_t features, bool tx_path);
3599 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3600 netdev_features_t features);
3601
3602 struct netdev_bonding_info {
3603 ifslave slave;
3604 ifbond master;
3605 };
3606
3607 struct netdev_notifier_bonding_info {
3608 struct netdev_notifier_info info; /* must be first */
3609 struct netdev_bonding_info bonding_info;
3610 };
3611
3612 void netdev_bonding_info_change(struct net_device *dev,
3613 struct netdev_bonding_info *bonding_info);
3614
3615 static inline
3616 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3617 {
3618 return __skb_gso_segment(skb, features, true);
3619 }
3620 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3621
3622 static inline bool can_checksum_protocol(netdev_features_t features,
3623 __be16 protocol)
3624 {
3625 return ((features & NETIF_F_GEN_CSUM) ||
3626 ((features & NETIF_F_V4_CSUM) &&
3627 protocol == htons(ETH_P_IP)) ||
3628 ((features & NETIF_F_V6_CSUM) &&
3629 protocol == htons(ETH_P_IPV6)) ||
3630 ((features & NETIF_F_FCOE_CRC) &&
3631 protocol == htons(ETH_P_FCOE)));
3632 }
3633
3634 #ifdef CONFIG_BUG
3635 void netdev_rx_csum_fault(struct net_device *dev);
3636 #else
3637 static inline void netdev_rx_csum_fault(struct net_device *dev)
3638 {
3639 }
3640 #endif
3641 /* rx skb timestamps */
3642 void net_enable_timestamp(void);
3643 void net_disable_timestamp(void);
3644
3645 #ifdef CONFIG_PROC_FS
3646 int __init dev_proc_init(void);
3647 #else
3648 #define dev_proc_init() 0
3649 #endif
3650
3651 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3652 struct sk_buff *skb, struct net_device *dev,
3653 bool more)
3654 {
3655 skb->xmit_more = more ? 1 : 0;
3656 return ops->ndo_start_xmit(skb, dev);
3657 }
3658
3659 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3660 struct netdev_queue *txq, bool more)
3661 {
3662 const struct net_device_ops *ops = dev->netdev_ops;
3663 int rc;
3664
3665 rc = __netdev_start_xmit(ops, skb, dev, more);
3666 if (rc == NETDEV_TX_OK)
3667 txq_trans_update(txq);
3668
3669 return rc;
3670 }
3671
3672 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3673 const void *ns);
3674 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3675 const void *ns);
3676
3677 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3678 {
3679 return netdev_class_create_file_ns(class_attr, NULL);
3680 }
3681
3682 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3683 {
3684 netdev_class_remove_file_ns(class_attr, NULL);
3685 }
3686
3687 extern struct kobj_ns_type_operations net_ns_type_operations;
3688
3689 const char *netdev_drivername(const struct net_device *dev);
3690
3691 void linkwatch_run_queue(void);
3692
3693 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3694 netdev_features_t f2)
3695 {
3696 if (f1 & NETIF_F_GEN_CSUM)
3697 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3698 if (f2 & NETIF_F_GEN_CSUM)
3699 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3700 f1 &= f2;
3701 if (f1 & NETIF_F_GEN_CSUM)
3702 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3703
3704 return f1;
3705 }
3706
3707 static inline netdev_features_t netdev_get_wanted_features(
3708 struct net_device *dev)
3709 {
3710 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3711 }
3712 netdev_features_t netdev_increment_features(netdev_features_t all,
3713 netdev_features_t one, netdev_features_t mask);
3714
3715 /* Allow TSO being used on stacked device :
3716 * Performing the GSO segmentation before last device
3717 * is a performance improvement.
3718 */
3719 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3720 netdev_features_t mask)
3721 {
3722 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3723 }
3724
3725 int __netdev_update_features(struct net_device *dev);
3726 void netdev_update_features(struct net_device *dev);
3727 void netdev_change_features(struct net_device *dev);
3728
3729 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3730 struct net_device *dev);
3731
3732 netdev_features_t passthru_features_check(struct sk_buff *skb,
3733 struct net_device *dev,
3734 netdev_features_t features);
3735 netdev_features_t netif_skb_features(struct sk_buff *skb);
3736
3737 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3738 {
3739 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3740
3741 /* check flags correspondence */
3742 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3743 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3744 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3745 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3746 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3747 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3748 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3749 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3750 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3751 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3752 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3753 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3754 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3755
3756 return (features & feature) == feature;
3757 }
3758
3759 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3760 {
3761 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3762 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3763 }
3764
3765 static inline bool netif_needs_gso(struct sk_buff *skb,
3766 netdev_features_t features)
3767 {
3768 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3769 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3770 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3771 }
3772
3773 static inline void netif_set_gso_max_size(struct net_device *dev,
3774 unsigned int size)
3775 {
3776 dev->gso_max_size = size;
3777 }
3778
3779 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3780 int pulled_hlen, u16 mac_offset,
3781 int mac_len)
3782 {
3783 skb->protocol = protocol;
3784 skb->encapsulation = 1;
3785 skb_push(skb, pulled_hlen);
3786 skb_reset_transport_header(skb);
3787 skb->mac_header = mac_offset;
3788 skb->network_header = skb->mac_header + mac_len;
3789 skb->mac_len = mac_len;
3790 }
3791
3792 static inline bool netif_is_macvlan(struct net_device *dev)
3793 {
3794 return dev->priv_flags & IFF_MACVLAN;
3795 }
3796
3797 static inline bool netif_is_macvlan_port(struct net_device *dev)
3798 {
3799 return dev->priv_flags & IFF_MACVLAN_PORT;
3800 }
3801
3802 static inline bool netif_is_ipvlan(struct net_device *dev)
3803 {
3804 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3805 }
3806
3807 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3808 {
3809 return dev->priv_flags & IFF_IPVLAN_MASTER;
3810 }
3811
3812 static inline bool netif_is_bond_master(struct net_device *dev)
3813 {
3814 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3815 }
3816
3817 static inline bool netif_is_bond_slave(struct net_device *dev)
3818 {
3819 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3820 }
3821
3822 static inline bool netif_supports_nofcs(struct net_device *dev)
3823 {
3824 return dev->priv_flags & IFF_SUPP_NOFCS;
3825 }
3826
3827 static inline bool netif_is_l3_master(const struct net_device *dev)
3828 {
3829 return dev->priv_flags & IFF_L3MDEV_MASTER;
3830 }
3831
3832 static inline bool netif_is_bridge_master(const struct net_device *dev)
3833 {
3834 return dev->priv_flags & IFF_EBRIDGE;
3835 }
3836
3837 static inline bool netif_is_ovs_master(const struct net_device *dev)
3838 {
3839 return dev->priv_flags & IFF_OPENVSWITCH;
3840 }
3841
3842 static inline bool netif_index_is_l3_master(struct net *net, int ifindex)
3843 {
3844 bool rc = false;
3845
3846 #if IS_ENABLED(CONFIG_NET_VRF)
3847 struct net_device *dev;
3848
3849 if (ifindex == 0)
3850 return false;
3851
3852 rcu_read_lock();
3853
3854 dev = dev_get_by_index_rcu(net, ifindex);
3855 if (dev)
3856 rc = netif_is_l3_master(dev);
3857
3858 rcu_read_unlock();
3859 #endif
3860 return rc;
3861 }
3862
3863 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3864 static inline void netif_keep_dst(struct net_device *dev)
3865 {
3866 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3867 }
3868
3869 extern struct pernet_operations __net_initdata loopback_net_ops;
3870
3871 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3872
3873 /* netdev_printk helpers, similar to dev_printk */
3874
3875 static inline const char *netdev_name(const struct net_device *dev)
3876 {
3877 if (!dev->name[0] || strchr(dev->name, '%'))
3878 return "(unnamed net_device)";
3879 return dev->name;
3880 }
3881
3882 static inline const char *netdev_reg_state(const struct net_device *dev)
3883 {
3884 switch (dev->reg_state) {
3885 case NETREG_UNINITIALIZED: return " (uninitialized)";
3886 case NETREG_REGISTERED: return "";
3887 case NETREG_UNREGISTERING: return " (unregistering)";
3888 case NETREG_UNREGISTERED: return " (unregistered)";
3889 case NETREG_RELEASED: return " (released)";
3890 case NETREG_DUMMY: return " (dummy)";
3891 }
3892
3893 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3894 return " (unknown)";
3895 }
3896
3897 __printf(3, 4)
3898 void netdev_printk(const char *level, const struct net_device *dev,
3899 const char *format, ...);
3900 __printf(2, 3)
3901 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3902 __printf(2, 3)
3903 void netdev_alert(const struct net_device *dev, const char *format, ...);
3904 __printf(2, 3)
3905 void netdev_crit(const struct net_device *dev, const char *format, ...);
3906 __printf(2, 3)
3907 void netdev_err(const struct net_device *dev, const char *format, ...);
3908 __printf(2, 3)
3909 void netdev_warn(const struct net_device *dev, const char *format, ...);
3910 __printf(2, 3)
3911 void netdev_notice(const struct net_device *dev, const char *format, ...);
3912 __printf(2, 3)
3913 void netdev_info(const struct net_device *dev, const char *format, ...);
3914
3915 #define MODULE_ALIAS_NETDEV(device) \
3916 MODULE_ALIAS("netdev-" device)
3917
3918 #if defined(CONFIG_DYNAMIC_DEBUG)
3919 #define netdev_dbg(__dev, format, args...) \
3920 do { \
3921 dynamic_netdev_dbg(__dev, format, ##args); \
3922 } while (0)
3923 #elif defined(DEBUG)
3924 #define netdev_dbg(__dev, format, args...) \
3925 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3926 #else
3927 #define netdev_dbg(__dev, format, args...) \
3928 ({ \
3929 if (0) \
3930 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3931 })
3932 #endif
3933
3934 #if defined(VERBOSE_DEBUG)
3935 #define netdev_vdbg netdev_dbg
3936 #else
3937
3938 #define netdev_vdbg(dev, format, args...) \
3939 ({ \
3940 if (0) \
3941 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3942 0; \
3943 })
3944 #endif
3945
3946 /*
3947 * netdev_WARN() acts like dev_printk(), but with the key difference
3948 * of using a WARN/WARN_ON to get the message out, including the
3949 * file/line information and a backtrace.
3950 */
3951 #define netdev_WARN(dev, format, args...) \
3952 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3953 netdev_reg_state(dev), ##args)
3954
3955 /* netif printk helpers, similar to netdev_printk */
3956
3957 #define netif_printk(priv, type, level, dev, fmt, args...) \
3958 do { \
3959 if (netif_msg_##type(priv)) \
3960 netdev_printk(level, (dev), fmt, ##args); \
3961 } while (0)
3962
3963 #define netif_level(level, priv, type, dev, fmt, args...) \
3964 do { \
3965 if (netif_msg_##type(priv)) \
3966 netdev_##level(dev, fmt, ##args); \
3967 } while (0)
3968
3969 #define netif_emerg(priv, type, dev, fmt, args...) \
3970 netif_level(emerg, priv, type, dev, fmt, ##args)
3971 #define netif_alert(priv, type, dev, fmt, args...) \
3972 netif_level(alert, priv, type, dev, fmt, ##args)
3973 #define netif_crit(priv, type, dev, fmt, args...) \
3974 netif_level(crit, priv, type, dev, fmt, ##args)
3975 #define netif_err(priv, type, dev, fmt, args...) \
3976 netif_level(err, priv, type, dev, fmt, ##args)
3977 #define netif_warn(priv, type, dev, fmt, args...) \
3978 netif_level(warn, priv, type, dev, fmt, ##args)
3979 #define netif_notice(priv, type, dev, fmt, args...) \
3980 netif_level(notice, priv, type, dev, fmt, ##args)
3981 #define netif_info(priv, type, dev, fmt, args...) \
3982 netif_level(info, priv, type, dev, fmt, ##args)
3983
3984 #if defined(CONFIG_DYNAMIC_DEBUG)
3985 #define netif_dbg(priv, type, netdev, format, args...) \
3986 do { \
3987 if (netif_msg_##type(priv)) \
3988 dynamic_netdev_dbg(netdev, format, ##args); \
3989 } while (0)
3990 #elif defined(DEBUG)
3991 #define netif_dbg(priv, type, dev, format, args...) \
3992 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3993 #else
3994 #define netif_dbg(priv, type, dev, format, args...) \
3995 ({ \
3996 if (0) \
3997 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3998 0; \
3999 })
4000 #endif
4001
4002 #if defined(VERBOSE_DEBUG)
4003 #define netif_vdbg netif_dbg
4004 #else
4005 #define netif_vdbg(priv, type, dev, format, args...) \
4006 ({ \
4007 if (0) \
4008 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4009 0; \
4010 })
4011 #endif
4012
4013 /*
4014 * The list of packet types we will receive (as opposed to discard)
4015 * and the routines to invoke.
4016 *
4017 * Why 16. Because with 16 the only overlap we get on a hash of the
4018 * low nibble of the protocol value is RARP/SNAP/X.25.
4019 *
4020 * NOTE: That is no longer true with the addition of VLAN tags. Not
4021 * sure which should go first, but I bet it won't make much
4022 * difference if we are running VLANs. The good news is that
4023 * this protocol won't be in the list unless compiled in, so
4024 * the average user (w/out VLANs) will not be adversely affected.
4025 * --BLG
4026 *
4027 * 0800 IP
4028 * 8100 802.1Q VLAN
4029 * 0001 802.3
4030 * 0002 AX.25
4031 * 0004 802.2
4032 * 8035 RARP
4033 * 0005 SNAP
4034 * 0805 X.25
4035 * 0806 ARP
4036 * 8137 IPX
4037 * 0009 Localtalk
4038 * 86DD IPv6
4039 */
4040 #define PTYPE_HASH_SIZE (16)
4041 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4042
4043 #endif /* _LINUX_NETDEVICE_H */
This page took 0.113859 seconds and 6 git commands to generate.