]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/netdevice.h
netfilter: Use nf_hook_state.net
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/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 }
511
512 #ifdef CONFIG_SMP
513 /**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521 static inline void napi_synchronize(const struct napi_struct *n)
522 {
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525 }
526 #else
527 # define napi_synchronize(n) barrier()
528 #endif
529
530 enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
534 };
535
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
546 /*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
555
556 struct netdev_queue {
557 /*
558 * read mostly part
559 */
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
563 #ifdef CONFIG_SYSFS
564 struct kobject kobj;
565 #endif
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568 #endif
569 /*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
584
585 unsigned long state;
586
587 #ifdef CONFIG_BQL
588 struct dql dql;
589 #endif
590 unsigned long tx_maxrate;
591 } ____cacheline_aligned_in_smp;
592
593 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594 {
595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597 #else
598 return NUMA_NO_NODE;
599 #endif
600 }
601
602 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603 {
604 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606 #endif
607 }
608
609 #ifdef CONFIG_RPS
610 /*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614 struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618 };
619 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
620
621 /*
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
625 */
626 struct rps_dev_flow {
627 u16 cpu;
628 u16 filter;
629 unsigned int last_qtail;
630 };
631 #define RPS_NO_FILTER 0xffff
632
633 /*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636 struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
639 struct rps_dev_flow flows[0];
640 };
641 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
642 ((_num) * sizeof(struct rps_dev_flow)))
643
644 /*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
653 */
654 struct rps_sock_flow_table {
655 u32 mask;
656
657 u32 ents[0] ____cacheline_aligned_in_smp;
658 };
659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
660
661 #define RPS_NO_CPU 0xffff
662
663 extern u32 rps_cpu_mask;
664 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
666 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668 {
669 if (table && hash) {
670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
672
673 /* We only give a hint, preemption can change cpu under us */
674 val |= raw_smp_processor_id();
675
676 if (table->ents[index] != val)
677 table->ents[index] = val;
678 }
679 }
680
681 #ifdef CONFIG_RFS_ACCEL
682 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
684 #endif
685 #endif /* CONFIG_RPS */
686
687 /* This structure contains an instance of an RX queue. */
688 struct netdev_rx_queue {
689 #ifdef CONFIG_RPS
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
692 #endif
693 struct kobject kobj;
694 struct net_device *dev;
695 } ____cacheline_aligned_in_smp;
696
697 /*
698 * RX queue sysfs structures and functions.
699 */
700 struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706 };
707
708 #ifdef CONFIG_XPS
709 /*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713 struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718 };
719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723 /*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726 struct xps_dev_maps {
727 struct rcu_head rcu;
728 struct xps_map __rcu *cpu_map[0];
729 };
730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732 #endif /* CONFIG_XPS */
733
734 #define TC_MAX_QUEUE 16
735 #define TC_BITMASK 15
736 /* HW offloaded queuing disciplines txq count and offset maps */
737 struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740 };
741
742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 /*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747 struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756 };
757 #endif
758
759 #define MAX_PHYS_ITEM_ID_LEN 32
760
761 /* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
763 */
764 struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
766 unsigned char id_len;
767 };
768
769 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
770 struct netdev_phys_item_id *b)
771 {
772 return a->id_len == b->id_len &&
773 memcmp(a->id, b->id, a->id_len) == 0;
774 }
775
776 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
777 struct sk_buff *skb);
778
779 /*
780 * This structure defines the management hooks for network devices.
781 * The following hooks can be defined; unless noted otherwise, they are
782 * optional and can be filled with a null pointer.
783 *
784 * int (*ndo_init)(struct net_device *dev);
785 * This function is called once when network device is registered.
786 * The network device can use this to any late stage initializaton
787 * or semantic validattion. It can fail with an error code which will
788 * be propogated back to register_netdev
789 *
790 * void (*ndo_uninit)(struct net_device *dev);
791 * This function is called when device is unregistered or when registration
792 * fails. It is not called if init fails.
793 *
794 * int (*ndo_open)(struct net_device *dev);
795 * This function is called when network device transistions to the up
796 * state.
797 *
798 * int (*ndo_stop)(struct net_device *dev);
799 * This function is called when network device transistions to the down
800 * state.
801 *
802 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
803 * struct net_device *dev);
804 * Called when a packet needs to be transmitted.
805 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
806 * the queue before that can happen; it's for obsolete devices and weird
807 * corner cases, but the stack really does a non-trivial amount
808 * of useless work if you return NETDEV_TX_BUSY.
809 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
810 * Required can not be NULL.
811 *
812 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
813 * void *accel_priv, select_queue_fallback_t fallback);
814 * Called to decide which queue to when device supports multiple
815 * transmit queues.
816 *
817 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
818 * This function is called to allow device receiver to make
819 * changes to configuration when multicast or promiscious is enabled.
820 *
821 * void (*ndo_set_rx_mode)(struct net_device *dev);
822 * This function is called device changes address list filtering.
823 * If driver handles unicast address filtering, it should set
824 * IFF_UNICAST_FLT to its priv_flags.
825 *
826 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
827 * This function is called when the Media Access Control address
828 * needs to be changed. If this interface is not defined, the
829 * mac address can not be changed.
830 *
831 * int (*ndo_validate_addr)(struct net_device *dev);
832 * Test if Media Access Control address is valid for the device.
833 *
834 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
835 * Called when a user request an ioctl which can't be handled by
836 * the generic interface code. If not defined ioctl's return
837 * not supported error code.
838 *
839 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
840 * Used to set network devices bus interface parameters. This interface
841 * is retained for legacy reason, new devices should use the bus
842 * interface (PCI) for low level management.
843 *
844 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
845 * Called when a user wants to change the Maximum Transfer Unit
846 * of a device. If not defined, any request to change MTU will
847 * will return an error.
848 *
849 * void (*ndo_tx_timeout)(struct net_device *dev);
850 * Callback uses when the transmitter has not made any progress
851 * for dev->watchdog ticks.
852 *
853 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
854 * struct rtnl_link_stats64 *storage);
855 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
856 * Called when a user wants to get the network device usage
857 * statistics. Drivers must do one of the following:
858 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
859 * rtnl_link_stats64 structure passed by the caller.
860 * 2. Define @ndo_get_stats to update a net_device_stats structure
861 * (which should normally be dev->stats) and return a pointer to
862 * it. The structure may be changed asynchronously only if each
863 * field is written atomically.
864 * 3. Update dev->stats asynchronously and atomically, and define
865 * neither operation.
866 *
867 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
868 * If device support VLAN filtering this function is called when a
869 * VLAN id is registered.
870 *
871 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
872 * If device support VLAN filtering this function is called when a
873 * VLAN id is unregistered.
874 *
875 * void (*ndo_poll_controller)(struct net_device *dev);
876 *
877 * SR-IOV management functions.
878 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
879 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
880 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
881 * int max_tx_rate);
882 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
883 * int (*ndo_get_vf_config)(struct net_device *dev,
884 * int vf, struct ifla_vf_info *ivf);
885 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
886 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
887 * struct nlattr *port[]);
888 *
889 * Enable or disable the VF ability to query its RSS Redirection Table and
890 * Hash Key. This is needed since on some devices VF share this information
891 * with PF and querying it may adduce a theoretical security risk.
892 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
893 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
894 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
895 * Called to setup 'tc' number of traffic classes in the net device. This
896 * is always called from the stack with the rtnl lock held and netif tx
897 * queues stopped. This allows the netdevice to perform queue management
898 * safely.
899 *
900 * Fiber Channel over Ethernet (FCoE) offload functions.
901 * int (*ndo_fcoe_enable)(struct net_device *dev);
902 * Called when the FCoE protocol stack wants to start using LLD for FCoE
903 * so the underlying device can perform whatever needed configuration or
904 * initialization to support acceleration of FCoE traffic.
905 *
906 * int (*ndo_fcoe_disable)(struct net_device *dev);
907 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
908 * so the underlying device can perform whatever needed clean-ups to
909 * stop supporting acceleration of FCoE traffic.
910 *
911 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
912 * struct scatterlist *sgl, unsigned int sgc);
913 * Called when the FCoE Initiator wants to initialize an I/O that
914 * is a possible candidate for Direct Data Placement (DDP). The LLD can
915 * perform necessary setup and returns 1 to indicate the device is set up
916 * successfully to perform DDP on this I/O, otherwise this returns 0.
917 *
918 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
919 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
920 * indicated by the FC exchange id 'xid', so the underlying device can
921 * clean up and reuse resources for later DDP requests.
922 *
923 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
924 * struct scatterlist *sgl, unsigned int sgc);
925 * Called when the FCoE Target wants to initialize an I/O that
926 * is a possible candidate for Direct Data Placement (DDP). The LLD can
927 * perform necessary setup and returns 1 to indicate the device is set up
928 * successfully to perform DDP on this I/O, otherwise this returns 0.
929 *
930 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
931 * struct netdev_fcoe_hbainfo *hbainfo);
932 * Called when the FCoE Protocol stack wants information on the underlying
933 * device. This information is utilized by the FCoE protocol stack to
934 * register attributes with Fiber Channel management service as per the
935 * FC-GS Fabric Device Management Information(FDMI) specification.
936 *
937 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
938 * Called when the underlying device wants to override default World Wide
939 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
940 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
941 * protocol stack to use.
942 *
943 * RFS acceleration.
944 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
945 * u16 rxq_index, u32 flow_id);
946 * Set hardware filter for RFS. rxq_index is the target queue index;
947 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
948 * Return the filter ID on success, or a negative error code.
949 *
950 * Slave management functions (for bridge, bonding, etc).
951 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
952 * Called to make another netdev an underling.
953 *
954 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
955 * Called to release previously enslaved netdev.
956 *
957 * Feature/offload setting functions.
958 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
959 * netdev_features_t features);
960 * Adjusts the requested feature flags according to device-specific
961 * constraints, and returns the resulting flags. Must not modify
962 * the device state.
963 *
964 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
965 * Called to update device configuration to new features. Passed
966 * feature set might be less than what was returned by ndo_fix_features()).
967 * Must return >0 or -errno if it changed dev->features itself.
968 *
969 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
970 * struct net_device *dev,
971 * const unsigned char *addr, u16 vid, u16 flags)
972 * Adds an FDB entry to dev for addr.
973 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
974 * struct net_device *dev,
975 * const unsigned char *addr, u16 vid)
976 * Deletes the FDB entry from dev coresponding to addr.
977 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
978 * struct net_device *dev, struct net_device *filter_dev,
979 * int idx)
980 * Used to add FDB entries to dump requests. Implementers should add
981 * entries to skb and update idx with the number of entries.
982 *
983 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
984 * u16 flags)
985 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
986 * struct net_device *dev, u32 filter_mask,
987 * int nlflags)
988 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
989 * u16 flags);
990 *
991 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
992 * Called to change device carrier. Soft-devices (like dummy, team, etc)
993 * which do not represent real hardware may define this to allow their
994 * userspace components to manage their virtual carrier state. Devices
995 * that determine carrier state from physical hardware properties (eg
996 * network cables) or protocol-dependent mechanisms (eg
997 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
998 *
999 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1000 * struct netdev_phys_item_id *ppid);
1001 * Called to get ID of physical port of this device. If driver does
1002 * not implement this, it is assumed that the hw is not able to have
1003 * multiple net devices on single physical port.
1004 *
1005 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1006 * sa_family_t sa_family, __be16 port);
1007 * Called by vxlan to notiy a driver about the UDP port and socket
1008 * address family that vxlan is listnening to. It is called only when
1009 * a new port starts listening. The operation is protected by the
1010 * vxlan_net->sock_lock.
1011 *
1012 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1013 * sa_family_t sa_family, __be16 port);
1014 * Called by vxlan to notify the driver about a UDP port and socket
1015 * address family that vxlan is not listening to anymore. The operation
1016 * is protected by the vxlan_net->sock_lock.
1017 *
1018 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1019 * struct net_device *dev)
1020 * Called by upper layer devices to accelerate switching or other
1021 * station functionality into hardware. 'pdev is the lowerdev
1022 * to use for the offload and 'dev' is the net device that will
1023 * back the offload. Returns a pointer to the private structure
1024 * the upper layer will maintain.
1025 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1026 * Called by upper layer device to delete the station created
1027 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1028 * the station and priv is the structure returned by the add
1029 * operation.
1030 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1031 * struct net_device *dev,
1032 * void *priv);
1033 * Callback to use for xmit over the accelerated station. This
1034 * is used in place of ndo_start_xmit on accelerated net
1035 * devices.
1036 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1037 * struct net_device *dev
1038 * netdev_features_t features);
1039 * Called by core transmit path to determine if device is capable of
1040 * performing offload operations on a given packet. This is to give
1041 * the device an opportunity to implement any restrictions that cannot
1042 * be otherwise expressed by feature flags. The check is called with
1043 * the set of features that the stack has calculated and it returns
1044 * those the driver believes to be appropriate.
1045 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1046 * int queue_index, u32 maxrate);
1047 * Called when a user wants to set a max-rate limitation of specific
1048 * TX queue.
1049 * int (*ndo_get_iflink)(const struct net_device *dev);
1050 * Called to get the iflink value of this device.
1051 * void (*ndo_change_proto_down)(struct net_device *dev,
1052 * bool proto_down);
1053 * This function is used to pass protocol port error state information
1054 * to the switch driver. The switch driver can react to the proto_down
1055 * by doing a phys down on the associated switch port.
1056 *
1057 */
1058 struct net_device_ops {
1059 int (*ndo_init)(struct net_device *dev);
1060 void (*ndo_uninit)(struct net_device *dev);
1061 int (*ndo_open)(struct net_device *dev);
1062 int (*ndo_stop)(struct net_device *dev);
1063 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1064 struct net_device *dev);
1065 u16 (*ndo_select_queue)(struct net_device *dev,
1066 struct sk_buff *skb,
1067 void *accel_priv,
1068 select_queue_fallback_t fallback);
1069 void (*ndo_change_rx_flags)(struct net_device *dev,
1070 int flags);
1071 void (*ndo_set_rx_mode)(struct net_device *dev);
1072 int (*ndo_set_mac_address)(struct net_device *dev,
1073 void *addr);
1074 int (*ndo_validate_addr)(struct net_device *dev);
1075 int (*ndo_do_ioctl)(struct net_device *dev,
1076 struct ifreq *ifr, int cmd);
1077 int (*ndo_set_config)(struct net_device *dev,
1078 struct ifmap *map);
1079 int (*ndo_change_mtu)(struct net_device *dev,
1080 int new_mtu);
1081 int (*ndo_neigh_setup)(struct net_device *dev,
1082 struct neigh_parms *);
1083 void (*ndo_tx_timeout) (struct net_device *dev);
1084
1085 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1086 struct rtnl_link_stats64 *storage);
1087 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1088
1089 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1090 __be16 proto, u16 vid);
1091 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1092 __be16 proto, u16 vid);
1093 #ifdef CONFIG_NET_POLL_CONTROLLER
1094 void (*ndo_poll_controller)(struct net_device *dev);
1095 int (*ndo_netpoll_setup)(struct net_device *dev,
1096 struct netpoll_info *info);
1097 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1098 #endif
1099 #ifdef CONFIG_NET_RX_BUSY_POLL
1100 int (*ndo_busy_poll)(struct napi_struct *dev);
1101 #endif
1102 int (*ndo_set_vf_mac)(struct net_device *dev,
1103 int queue, u8 *mac);
1104 int (*ndo_set_vf_vlan)(struct net_device *dev,
1105 int queue, u16 vlan, u8 qos);
1106 int (*ndo_set_vf_rate)(struct net_device *dev,
1107 int vf, int min_tx_rate,
1108 int max_tx_rate);
1109 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1110 int vf, bool setting);
1111 int (*ndo_get_vf_config)(struct net_device *dev,
1112 int vf,
1113 struct ifla_vf_info *ivf);
1114 int (*ndo_set_vf_link_state)(struct net_device *dev,
1115 int vf, int link_state);
1116 int (*ndo_get_vf_stats)(struct net_device *dev,
1117 int vf,
1118 struct ifla_vf_stats
1119 *vf_stats);
1120 int (*ndo_set_vf_port)(struct net_device *dev,
1121 int vf,
1122 struct nlattr *port[]);
1123 int (*ndo_get_vf_port)(struct net_device *dev,
1124 int vf, struct sk_buff *skb);
1125 int (*ndo_set_vf_rss_query_en)(
1126 struct net_device *dev,
1127 int vf, bool setting);
1128 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1129 #if IS_ENABLED(CONFIG_FCOE)
1130 int (*ndo_fcoe_enable)(struct net_device *dev);
1131 int (*ndo_fcoe_disable)(struct net_device *dev);
1132 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1133 u16 xid,
1134 struct scatterlist *sgl,
1135 unsigned int sgc);
1136 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1137 u16 xid);
1138 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1139 u16 xid,
1140 struct scatterlist *sgl,
1141 unsigned int sgc);
1142 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1143 struct netdev_fcoe_hbainfo *hbainfo);
1144 #endif
1145
1146 #if IS_ENABLED(CONFIG_LIBFCOE)
1147 #define NETDEV_FCOE_WWNN 0
1148 #define NETDEV_FCOE_WWPN 1
1149 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1150 u64 *wwn, int type);
1151 #endif
1152
1153 #ifdef CONFIG_RFS_ACCEL
1154 int (*ndo_rx_flow_steer)(struct net_device *dev,
1155 const struct sk_buff *skb,
1156 u16 rxq_index,
1157 u32 flow_id);
1158 #endif
1159 int (*ndo_add_slave)(struct net_device *dev,
1160 struct net_device *slave_dev);
1161 int (*ndo_del_slave)(struct net_device *dev,
1162 struct net_device *slave_dev);
1163 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1164 netdev_features_t features);
1165 int (*ndo_set_features)(struct net_device *dev,
1166 netdev_features_t features);
1167 int (*ndo_neigh_construct)(struct neighbour *n);
1168 void (*ndo_neigh_destroy)(struct neighbour *n);
1169
1170 int (*ndo_fdb_add)(struct ndmsg *ndm,
1171 struct nlattr *tb[],
1172 struct net_device *dev,
1173 const unsigned char *addr,
1174 u16 vid,
1175 u16 flags);
1176 int (*ndo_fdb_del)(struct ndmsg *ndm,
1177 struct nlattr *tb[],
1178 struct net_device *dev,
1179 const unsigned char *addr,
1180 u16 vid);
1181 int (*ndo_fdb_dump)(struct sk_buff *skb,
1182 struct netlink_callback *cb,
1183 struct net_device *dev,
1184 struct net_device *filter_dev,
1185 int idx);
1186
1187 int (*ndo_bridge_setlink)(struct net_device *dev,
1188 struct nlmsghdr *nlh,
1189 u16 flags);
1190 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1191 u32 pid, u32 seq,
1192 struct net_device *dev,
1193 u32 filter_mask,
1194 int nlflags);
1195 int (*ndo_bridge_dellink)(struct net_device *dev,
1196 struct nlmsghdr *nlh,
1197 u16 flags);
1198 int (*ndo_change_carrier)(struct net_device *dev,
1199 bool new_carrier);
1200 int (*ndo_get_phys_port_id)(struct net_device *dev,
1201 struct netdev_phys_item_id *ppid);
1202 int (*ndo_get_phys_port_name)(struct net_device *dev,
1203 char *name, size_t len);
1204 void (*ndo_add_vxlan_port)(struct net_device *dev,
1205 sa_family_t sa_family,
1206 __be16 port);
1207 void (*ndo_del_vxlan_port)(struct net_device *dev,
1208 sa_family_t sa_family,
1209 __be16 port);
1210
1211 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1212 struct net_device *dev);
1213 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1214 void *priv);
1215
1216 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1217 struct net_device *dev,
1218 void *priv);
1219 int (*ndo_get_lock_subclass)(struct net_device *dev);
1220 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1221 struct net_device *dev,
1222 netdev_features_t features);
1223 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1224 int queue_index,
1225 u32 maxrate);
1226 int (*ndo_get_iflink)(const struct net_device *dev);
1227 int (*ndo_change_proto_down)(struct net_device *dev,
1228 bool proto_down);
1229 };
1230
1231 /**
1232 * enum net_device_priv_flags - &struct net_device priv_flags
1233 *
1234 * These are the &struct net_device, they are only set internally
1235 * by drivers and used in the kernel. These flags are invisible to
1236 * userspace, this means that the order of these flags can change
1237 * during any kernel release.
1238 *
1239 * You should have a pretty good reason to be extending these flags.
1240 *
1241 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1242 * @IFF_EBRIDGE: Ethernet bridging device
1243 * @IFF_BONDING: bonding master or slave
1244 * @IFF_ISATAP: ISATAP interface (RFC4214)
1245 * @IFF_WAN_HDLC: WAN HDLC device
1246 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1247 * release skb->dst
1248 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1249 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1250 * @IFF_MACVLAN_PORT: device used as macvlan port
1251 * @IFF_BRIDGE_PORT: device used as bridge port
1252 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1253 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1254 * @IFF_UNICAST_FLT: Supports unicast filtering
1255 * @IFF_TEAM_PORT: device used as team port
1256 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1257 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1258 * change when it's running
1259 * @IFF_MACVLAN: Macvlan device
1260 * @IFF_VRF_MASTER: device is a VRF master
1261 * @IFF_NO_QUEUE: device can run without qdisc attached
1262 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1263 */
1264 enum netdev_priv_flags {
1265 IFF_802_1Q_VLAN = 1<<0,
1266 IFF_EBRIDGE = 1<<1,
1267 IFF_BONDING = 1<<2,
1268 IFF_ISATAP = 1<<3,
1269 IFF_WAN_HDLC = 1<<4,
1270 IFF_XMIT_DST_RELEASE = 1<<5,
1271 IFF_DONT_BRIDGE = 1<<6,
1272 IFF_DISABLE_NETPOLL = 1<<7,
1273 IFF_MACVLAN_PORT = 1<<8,
1274 IFF_BRIDGE_PORT = 1<<9,
1275 IFF_OVS_DATAPATH = 1<<10,
1276 IFF_TX_SKB_SHARING = 1<<11,
1277 IFF_UNICAST_FLT = 1<<12,
1278 IFF_TEAM_PORT = 1<<13,
1279 IFF_SUPP_NOFCS = 1<<14,
1280 IFF_LIVE_ADDR_CHANGE = 1<<15,
1281 IFF_MACVLAN = 1<<16,
1282 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1283 IFF_IPVLAN_MASTER = 1<<18,
1284 IFF_IPVLAN_SLAVE = 1<<19,
1285 IFF_VRF_MASTER = 1<<20,
1286 IFF_NO_QUEUE = 1<<21,
1287 IFF_OPENVSWITCH = 1<<22,
1288 };
1289
1290 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1291 #define IFF_EBRIDGE IFF_EBRIDGE
1292 #define IFF_BONDING IFF_BONDING
1293 #define IFF_ISATAP IFF_ISATAP
1294 #define IFF_WAN_HDLC IFF_WAN_HDLC
1295 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1296 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1297 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1298 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1299 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1300 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1301 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1302 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1303 #define IFF_TEAM_PORT IFF_TEAM_PORT
1304 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1305 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1306 #define IFF_MACVLAN IFF_MACVLAN
1307 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1308 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1309 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1310 #define IFF_VRF_MASTER IFF_VRF_MASTER
1311 #define IFF_NO_QUEUE IFF_NO_QUEUE
1312 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1313
1314 /**
1315 * struct net_device - The DEVICE structure.
1316 * Actually, this whole structure is a big mistake. It mixes I/O
1317 * data with strictly "high-level" data, and it has to know about
1318 * almost every data structure used in the INET module.
1319 *
1320 * @name: This is the first field of the "visible" part of this structure
1321 * (i.e. as seen by users in the "Space.c" file). It is the name
1322 * of the interface.
1323 *
1324 * @name_hlist: Device name hash chain, please keep it close to name[]
1325 * @ifalias: SNMP alias
1326 * @mem_end: Shared memory end
1327 * @mem_start: Shared memory start
1328 * @base_addr: Device I/O address
1329 * @irq: Device IRQ number
1330 *
1331 * @carrier_changes: Stats to monitor carrier on<->off transitions
1332 *
1333 * @state: Generic network queuing layer state, see netdev_state_t
1334 * @dev_list: The global list of network devices
1335 * @napi_list: List entry, that is used for polling napi devices
1336 * @unreg_list: List entry, that is used, when we are unregistering the
1337 * device, see the function unregister_netdev
1338 * @close_list: List entry, that is used, when we are closing the device
1339 *
1340 * @adj_list: Directly linked devices, like slaves for bonding
1341 * @all_adj_list: All linked devices, *including* neighbours
1342 * @features: Currently active device features
1343 * @hw_features: User-changeable features
1344 *
1345 * @wanted_features: User-requested features
1346 * @vlan_features: Mask of features inheritable by VLAN devices
1347 *
1348 * @hw_enc_features: Mask of features inherited by encapsulating devices
1349 * This field indicates what encapsulation
1350 * offloads the hardware is capable of doing,
1351 * and drivers will need to set them appropriately.
1352 *
1353 * @mpls_features: Mask of features inheritable by MPLS
1354 *
1355 * @ifindex: interface index
1356 * @group: The group, that the device belongs to
1357 *
1358 * @stats: Statistics struct, which was left as a legacy, use
1359 * rtnl_link_stats64 instead
1360 *
1361 * @rx_dropped: Dropped packets by core network,
1362 * do not use this in drivers
1363 * @tx_dropped: Dropped packets by core network,
1364 * do not use this in drivers
1365 *
1366 * @wireless_handlers: List of functions to handle Wireless Extensions,
1367 * instead of ioctl,
1368 * see <net/iw_handler.h> for details.
1369 * @wireless_data: Instance data managed by the core of wireless extensions
1370 *
1371 * @netdev_ops: Includes several pointers to callbacks,
1372 * if one wants to override the ndo_*() functions
1373 * @ethtool_ops: Management operations
1374 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1375 * of Layer 2 headers.
1376 *
1377 * @flags: Interface flags (a la BSD)
1378 * @priv_flags: Like 'flags' but invisible to userspace,
1379 * see if.h for the definitions
1380 * @gflags: Global flags ( kept as legacy )
1381 * @padded: How much padding added by alloc_netdev()
1382 * @operstate: RFC2863 operstate
1383 * @link_mode: Mapping policy to operstate
1384 * @if_port: Selectable AUI, TP, ...
1385 * @dma: DMA channel
1386 * @mtu: Interface MTU value
1387 * @type: Interface hardware type
1388 * @hard_header_len: Hardware header length
1389 *
1390 * @needed_headroom: Extra headroom the hardware may need, but not in all
1391 * cases can this be guaranteed
1392 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1393 * cases can this be guaranteed. Some cases also use
1394 * LL_MAX_HEADER instead to allocate the skb
1395 *
1396 * interface address info:
1397 *
1398 * @perm_addr: Permanent hw address
1399 * @addr_assign_type: Hw address assignment type
1400 * @addr_len: Hardware address length
1401 * @neigh_priv_len; Used in neigh_alloc(),
1402 * initialized only in atm/clip.c
1403 * @dev_id: Used to differentiate devices that share
1404 * the same link layer address
1405 * @dev_port: Used to differentiate devices that share
1406 * the same function
1407 * @addr_list_lock: XXX: need comments on this one
1408 * @uc_promisc: Counter, that indicates, that promiscuous mode
1409 * has been enabled due to the need to listen to
1410 * additional unicast addresses in a device that
1411 * does not implement ndo_set_rx_mode()
1412 * @uc: unicast mac addresses
1413 * @mc: multicast mac addresses
1414 * @dev_addrs: list of device hw addresses
1415 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1416 * @promiscuity: Number of times, the NIC is told to work in
1417 * Promiscuous mode, if it becomes 0 the NIC will
1418 * exit from working in Promiscuous mode
1419 * @allmulti: Counter, enables or disables allmulticast mode
1420 *
1421 * @vlan_info: VLAN info
1422 * @dsa_ptr: dsa specific data
1423 * @tipc_ptr: TIPC specific data
1424 * @atalk_ptr: AppleTalk link
1425 * @ip_ptr: IPv4 specific data
1426 * @dn_ptr: DECnet specific data
1427 * @ip6_ptr: IPv6 specific data
1428 * @ax25_ptr: AX.25 specific data
1429 * @vrf_ptr: VRF specific data
1430 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1431 *
1432 * @last_rx: Time of last Rx
1433 * @dev_addr: Hw address (before bcast,
1434 * because most packets are unicast)
1435 *
1436 * @_rx: Array of RX queues
1437 * @num_rx_queues: Number of RX queues
1438 * allocated at register_netdev() time
1439 * @real_num_rx_queues: Number of RX queues currently active in device
1440 *
1441 * @rx_handler: handler for received packets
1442 * @rx_handler_data: XXX: need comments on this one
1443 * @ingress_queue: XXX: need comments on this one
1444 * @broadcast: hw bcast address
1445 *
1446 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1447 * indexed by RX queue number. Assigned by driver.
1448 * This must only be set if the ndo_rx_flow_steer
1449 * operation is defined
1450 * @index_hlist: Device index hash chain
1451 *
1452 * @_tx: Array of TX queues
1453 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1454 * @real_num_tx_queues: Number of TX queues currently active in device
1455 * @qdisc: Root qdisc from userspace point of view
1456 * @tx_queue_len: Max frames per queue allowed
1457 * @tx_global_lock: XXX: need comments on this one
1458 *
1459 * @xps_maps: XXX: need comments on this one
1460 *
1461 * @offload_fwd_mark: Offload device fwding mark
1462 *
1463 * @trans_start: Time (in jiffies) of last Tx
1464 * @watchdog_timeo: Represents the timeout that is used by
1465 * the watchdog ( see dev_watchdog() )
1466 * @watchdog_timer: List of timers
1467 *
1468 * @pcpu_refcnt: Number of references to this device
1469 * @todo_list: Delayed register/unregister
1470 * @link_watch_list: XXX: need comments on this one
1471 *
1472 * @reg_state: Register/unregister state machine
1473 * @dismantle: Device is going to be freed
1474 * @rtnl_link_state: This enum represents the phases of creating
1475 * a new link
1476 *
1477 * @destructor: Called from unregister,
1478 * can be used to call free_netdev
1479 * @npinfo: XXX: need comments on this one
1480 * @nd_net: Network namespace this network device is inside
1481 *
1482 * @ml_priv: Mid-layer private
1483 * @lstats: Loopback statistics
1484 * @tstats: Tunnel statistics
1485 * @dstats: Dummy statistics
1486 * @vstats: Virtual ethernet statistics
1487 *
1488 * @garp_port: GARP
1489 * @mrp_port: MRP
1490 *
1491 * @dev: Class/net/name entry
1492 * @sysfs_groups: Space for optional device, statistics and wireless
1493 * sysfs groups
1494 *
1495 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1496 * @rtnl_link_ops: Rtnl_link_ops
1497 *
1498 * @gso_max_size: Maximum size of generic segmentation offload
1499 * @gso_max_segs: Maximum number of segments that can be passed to the
1500 * NIC for GSO
1501 * @gso_min_segs: Minimum number of segments that can be passed to the
1502 * NIC for GSO
1503 *
1504 * @dcbnl_ops: Data Center Bridging netlink ops
1505 * @num_tc: Number of traffic classes in the net device
1506 * @tc_to_txq: XXX: need comments on this one
1507 * @prio_tc_map XXX: need comments on this one
1508 *
1509 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1510 *
1511 * @priomap: XXX: need comments on this one
1512 * @phydev: Physical device may attach itself
1513 * for hardware timestamping
1514 *
1515 * @qdisc_tx_busylock: XXX: need comments on this one
1516 *
1517 * @proto_down: protocol port state information can be sent to the
1518 * switch driver and used to set the phys state of the
1519 * switch port.
1520 *
1521 * FIXME: cleanup struct net_device such that network protocol info
1522 * moves out.
1523 */
1524
1525 struct net_device {
1526 char name[IFNAMSIZ];
1527 struct hlist_node name_hlist;
1528 char *ifalias;
1529 /*
1530 * I/O specific fields
1531 * FIXME: Merge these and struct ifmap into one
1532 */
1533 unsigned long mem_end;
1534 unsigned long mem_start;
1535 unsigned long base_addr;
1536 int irq;
1537
1538 atomic_t carrier_changes;
1539
1540 /*
1541 * Some hardware also needs these fields (state,dev_list,
1542 * napi_list,unreg_list,close_list) but they are not
1543 * part of the usual set specified in Space.c.
1544 */
1545
1546 unsigned long state;
1547
1548 struct list_head dev_list;
1549 struct list_head napi_list;
1550 struct list_head unreg_list;
1551 struct list_head close_list;
1552 struct list_head ptype_all;
1553 struct list_head ptype_specific;
1554
1555 struct {
1556 struct list_head upper;
1557 struct list_head lower;
1558 } adj_list;
1559
1560 struct {
1561 struct list_head upper;
1562 struct list_head lower;
1563 } all_adj_list;
1564
1565 netdev_features_t features;
1566 netdev_features_t hw_features;
1567 netdev_features_t wanted_features;
1568 netdev_features_t vlan_features;
1569 netdev_features_t hw_enc_features;
1570 netdev_features_t mpls_features;
1571
1572 int ifindex;
1573 int group;
1574
1575 struct net_device_stats stats;
1576
1577 atomic_long_t rx_dropped;
1578 atomic_long_t tx_dropped;
1579
1580 #ifdef CONFIG_WIRELESS_EXT
1581 const struct iw_handler_def * wireless_handlers;
1582 struct iw_public_data * wireless_data;
1583 #endif
1584 const struct net_device_ops *netdev_ops;
1585 const struct ethtool_ops *ethtool_ops;
1586 #ifdef CONFIG_NET_SWITCHDEV
1587 const struct switchdev_ops *switchdev_ops;
1588 #endif
1589
1590 const struct header_ops *header_ops;
1591
1592 unsigned int flags;
1593 unsigned int priv_flags;
1594
1595 unsigned short gflags;
1596 unsigned short padded;
1597
1598 unsigned char operstate;
1599 unsigned char link_mode;
1600
1601 unsigned char if_port;
1602 unsigned char dma;
1603
1604 unsigned int mtu;
1605 unsigned short type;
1606 unsigned short hard_header_len;
1607
1608 unsigned short needed_headroom;
1609 unsigned short needed_tailroom;
1610
1611 /* Interface address info. */
1612 unsigned char perm_addr[MAX_ADDR_LEN];
1613 unsigned char addr_assign_type;
1614 unsigned char addr_len;
1615 unsigned short neigh_priv_len;
1616 unsigned short dev_id;
1617 unsigned short dev_port;
1618 spinlock_t addr_list_lock;
1619 unsigned char name_assign_type;
1620 bool uc_promisc;
1621 struct netdev_hw_addr_list uc;
1622 struct netdev_hw_addr_list mc;
1623 struct netdev_hw_addr_list dev_addrs;
1624
1625 #ifdef CONFIG_SYSFS
1626 struct kset *queues_kset;
1627 #endif
1628 unsigned int promiscuity;
1629 unsigned int allmulti;
1630
1631
1632 /* Protocol specific pointers */
1633
1634 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1635 struct vlan_info __rcu *vlan_info;
1636 #endif
1637 #if IS_ENABLED(CONFIG_NET_DSA)
1638 struct dsa_switch_tree *dsa_ptr;
1639 #endif
1640 #if IS_ENABLED(CONFIG_TIPC)
1641 struct tipc_bearer __rcu *tipc_ptr;
1642 #endif
1643 void *atalk_ptr;
1644 struct in_device __rcu *ip_ptr;
1645 struct dn_dev __rcu *dn_ptr;
1646 struct inet6_dev __rcu *ip6_ptr;
1647 void *ax25_ptr;
1648 struct net_vrf_dev __rcu *vrf_ptr;
1649 struct wireless_dev *ieee80211_ptr;
1650 struct wpan_dev *ieee802154_ptr;
1651 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1652 struct mpls_dev __rcu *mpls_ptr;
1653 #endif
1654
1655 /*
1656 * Cache lines mostly used on receive path (including eth_type_trans())
1657 */
1658 unsigned long last_rx;
1659
1660 /* Interface address info used in eth_type_trans() */
1661 unsigned char *dev_addr;
1662
1663
1664 #ifdef CONFIG_SYSFS
1665 struct netdev_rx_queue *_rx;
1666
1667 unsigned int num_rx_queues;
1668 unsigned int real_num_rx_queues;
1669
1670 #endif
1671
1672 unsigned long gro_flush_timeout;
1673 rx_handler_func_t __rcu *rx_handler;
1674 void __rcu *rx_handler_data;
1675
1676 #ifdef CONFIG_NET_CLS_ACT
1677 struct tcf_proto __rcu *ingress_cl_list;
1678 #endif
1679 struct netdev_queue __rcu *ingress_queue;
1680 #ifdef CONFIG_NETFILTER_INGRESS
1681 struct list_head nf_hooks_ingress;
1682 #endif
1683
1684 unsigned char broadcast[MAX_ADDR_LEN];
1685 #ifdef CONFIG_RFS_ACCEL
1686 struct cpu_rmap *rx_cpu_rmap;
1687 #endif
1688 struct hlist_node index_hlist;
1689
1690 /*
1691 * Cache lines mostly used on transmit path
1692 */
1693 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1694 unsigned int num_tx_queues;
1695 unsigned int real_num_tx_queues;
1696 struct Qdisc *qdisc;
1697 unsigned long tx_queue_len;
1698 spinlock_t tx_global_lock;
1699 int watchdog_timeo;
1700
1701 #ifdef CONFIG_XPS
1702 struct xps_dev_maps __rcu *xps_maps;
1703 #endif
1704
1705 #ifdef CONFIG_NET_SWITCHDEV
1706 u32 offload_fwd_mark;
1707 #endif
1708
1709 /* These may be needed for future network-power-down code. */
1710
1711 /*
1712 * trans_start here is expensive for high speed devices on SMP,
1713 * please use netdev_queue->trans_start instead.
1714 */
1715 unsigned long trans_start;
1716
1717 struct timer_list watchdog_timer;
1718
1719 int __percpu *pcpu_refcnt;
1720 struct list_head todo_list;
1721
1722 struct list_head link_watch_list;
1723
1724 enum { NETREG_UNINITIALIZED=0,
1725 NETREG_REGISTERED, /* completed register_netdevice */
1726 NETREG_UNREGISTERING, /* called unregister_netdevice */
1727 NETREG_UNREGISTERED, /* completed unregister todo */
1728 NETREG_RELEASED, /* called free_netdev */
1729 NETREG_DUMMY, /* dummy device for NAPI poll */
1730 } reg_state:8;
1731
1732 bool dismantle;
1733
1734 enum {
1735 RTNL_LINK_INITIALIZED,
1736 RTNL_LINK_INITIALIZING,
1737 } rtnl_link_state:16;
1738
1739 void (*destructor)(struct net_device *dev);
1740
1741 #ifdef CONFIG_NETPOLL
1742 struct netpoll_info __rcu *npinfo;
1743 #endif
1744
1745 possible_net_t nd_net;
1746
1747 /* mid-layer private */
1748 union {
1749 void *ml_priv;
1750 struct pcpu_lstats __percpu *lstats;
1751 struct pcpu_sw_netstats __percpu *tstats;
1752 struct pcpu_dstats __percpu *dstats;
1753 struct pcpu_vstats __percpu *vstats;
1754 };
1755
1756 struct garp_port __rcu *garp_port;
1757 struct mrp_port __rcu *mrp_port;
1758
1759 struct device dev;
1760 const struct attribute_group *sysfs_groups[4];
1761 const struct attribute_group *sysfs_rx_queue_group;
1762
1763 const struct rtnl_link_ops *rtnl_link_ops;
1764
1765 /* for setting kernel sock attribute on TCP connection setup */
1766 #define GSO_MAX_SIZE 65536
1767 unsigned int gso_max_size;
1768 #define GSO_MAX_SEGS 65535
1769 u16 gso_max_segs;
1770 u16 gso_min_segs;
1771 #ifdef CONFIG_DCB
1772 const struct dcbnl_rtnl_ops *dcbnl_ops;
1773 #endif
1774 u8 num_tc;
1775 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1776 u8 prio_tc_map[TC_BITMASK + 1];
1777
1778 #if IS_ENABLED(CONFIG_FCOE)
1779 unsigned int fcoe_ddp_xid;
1780 #endif
1781 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1782 struct netprio_map __rcu *priomap;
1783 #endif
1784 struct phy_device *phydev;
1785 struct lock_class_key *qdisc_tx_busylock;
1786 bool proto_down;
1787 };
1788 #define to_net_dev(d) container_of(d, struct net_device, dev)
1789
1790 #define NETDEV_ALIGN 32
1791
1792 static inline
1793 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1794 {
1795 return dev->prio_tc_map[prio & TC_BITMASK];
1796 }
1797
1798 static inline
1799 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1800 {
1801 if (tc >= dev->num_tc)
1802 return -EINVAL;
1803
1804 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1805 return 0;
1806 }
1807
1808 static inline
1809 void netdev_reset_tc(struct net_device *dev)
1810 {
1811 dev->num_tc = 0;
1812 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1813 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1814 }
1815
1816 static inline
1817 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1818 {
1819 if (tc >= dev->num_tc)
1820 return -EINVAL;
1821
1822 dev->tc_to_txq[tc].count = count;
1823 dev->tc_to_txq[tc].offset = offset;
1824 return 0;
1825 }
1826
1827 static inline
1828 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1829 {
1830 if (num_tc > TC_MAX_QUEUE)
1831 return -EINVAL;
1832
1833 dev->num_tc = num_tc;
1834 return 0;
1835 }
1836
1837 static inline
1838 int netdev_get_num_tc(struct net_device *dev)
1839 {
1840 return dev->num_tc;
1841 }
1842
1843 static inline
1844 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1845 unsigned int index)
1846 {
1847 return &dev->_tx[index];
1848 }
1849
1850 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1851 const struct sk_buff *skb)
1852 {
1853 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1854 }
1855
1856 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1857 void (*f)(struct net_device *,
1858 struct netdev_queue *,
1859 void *),
1860 void *arg)
1861 {
1862 unsigned int i;
1863
1864 for (i = 0; i < dev->num_tx_queues; i++)
1865 f(dev, &dev->_tx[i], arg);
1866 }
1867
1868 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1869 struct sk_buff *skb,
1870 void *accel_priv);
1871
1872 /*
1873 * Net namespace inlines
1874 */
1875 static inline
1876 struct net *dev_net(const struct net_device *dev)
1877 {
1878 return read_pnet(&dev->nd_net);
1879 }
1880
1881 static inline
1882 void dev_net_set(struct net_device *dev, struct net *net)
1883 {
1884 write_pnet(&dev->nd_net, net);
1885 }
1886
1887 static inline bool netdev_uses_dsa(struct net_device *dev)
1888 {
1889 #if IS_ENABLED(CONFIG_NET_DSA)
1890 if (dev->dsa_ptr != NULL)
1891 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1892 #endif
1893 return false;
1894 }
1895
1896 /**
1897 * netdev_priv - access network device private data
1898 * @dev: network device
1899 *
1900 * Get network device private data
1901 */
1902 static inline void *netdev_priv(const struct net_device *dev)
1903 {
1904 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1905 }
1906
1907 /* Set the sysfs physical device reference for the network logical device
1908 * if set prior to registration will cause a symlink during initialization.
1909 */
1910 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1911
1912 /* Set the sysfs device type for the network logical device to allow
1913 * fine-grained identification of different network device types. For
1914 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1915 */
1916 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1917
1918 /* Default NAPI poll() weight
1919 * Device drivers are strongly advised to not use bigger value
1920 */
1921 #define NAPI_POLL_WEIGHT 64
1922
1923 /**
1924 * netif_napi_add - initialize a napi context
1925 * @dev: network device
1926 * @napi: napi context
1927 * @poll: polling function
1928 * @weight: default weight
1929 *
1930 * netif_napi_add() must be used to initialize a napi context prior to calling
1931 * *any* of the other napi related functions.
1932 */
1933 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1934 int (*poll)(struct napi_struct *, int), int weight);
1935
1936 /**
1937 * netif_napi_del - remove a napi context
1938 * @napi: napi context
1939 *
1940 * netif_napi_del() removes a napi context from the network device napi list
1941 */
1942 void netif_napi_del(struct napi_struct *napi);
1943
1944 struct napi_gro_cb {
1945 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1946 void *frag0;
1947
1948 /* Length of frag0. */
1949 unsigned int frag0_len;
1950
1951 /* This indicates where we are processing relative to skb->data. */
1952 int data_offset;
1953
1954 /* This is non-zero if the packet cannot be merged with the new skb. */
1955 u16 flush;
1956
1957 /* Save the IP ID here and check when we get to the transport layer */
1958 u16 flush_id;
1959
1960 /* Number of segments aggregated. */
1961 u16 count;
1962
1963 /* Start offset for remote checksum offload */
1964 u16 gro_remcsum_start;
1965
1966 /* jiffies when first packet was created/queued */
1967 unsigned long age;
1968
1969 /* Used in ipv6_gro_receive() and foo-over-udp */
1970 u16 proto;
1971
1972 /* This is non-zero if the packet may be of the same flow. */
1973 u8 same_flow:1;
1974
1975 /* Used in udp_gro_receive */
1976 u8 udp_mark:1;
1977
1978 /* GRO checksum is valid */
1979 u8 csum_valid:1;
1980
1981 /* Number of checksums via CHECKSUM_UNNECESSARY */
1982 u8 csum_cnt:3;
1983
1984 /* Free the skb? */
1985 u8 free:2;
1986 #define NAPI_GRO_FREE 1
1987 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1988
1989 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1990 u8 is_ipv6:1;
1991
1992 /* 7 bit hole */
1993
1994 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1995 __wsum csum;
1996
1997 /* used in skb_gro_receive() slow path */
1998 struct sk_buff *last;
1999 };
2000
2001 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2002
2003 struct packet_type {
2004 __be16 type; /* This is really htons(ether_type). */
2005 struct net_device *dev; /* NULL is wildcarded here */
2006 int (*func) (struct sk_buff *,
2007 struct net_device *,
2008 struct packet_type *,
2009 struct net_device *);
2010 bool (*id_match)(struct packet_type *ptype,
2011 struct sock *sk);
2012 void *af_packet_priv;
2013 struct list_head list;
2014 };
2015
2016 struct offload_callbacks {
2017 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2018 netdev_features_t features);
2019 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2020 struct sk_buff *skb);
2021 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2022 };
2023
2024 struct packet_offload {
2025 __be16 type; /* This is really htons(ether_type). */
2026 u16 priority;
2027 struct offload_callbacks callbacks;
2028 struct list_head list;
2029 };
2030
2031 struct udp_offload;
2032
2033 struct udp_offload_callbacks {
2034 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2035 struct sk_buff *skb,
2036 struct udp_offload *uoff);
2037 int (*gro_complete)(struct sk_buff *skb,
2038 int nhoff,
2039 struct udp_offload *uoff);
2040 };
2041
2042 struct udp_offload {
2043 __be16 port;
2044 u8 ipproto;
2045 struct udp_offload_callbacks callbacks;
2046 };
2047
2048 /* often modified stats are per cpu, other are shared (netdev->stats) */
2049 struct pcpu_sw_netstats {
2050 u64 rx_packets;
2051 u64 rx_bytes;
2052 u64 tx_packets;
2053 u64 tx_bytes;
2054 struct u64_stats_sync syncp;
2055 };
2056
2057 #define netdev_alloc_pcpu_stats(type) \
2058 ({ \
2059 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2060 if (pcpu_stats) { \
2061 int __cpu; \
2062 for_each_possible_cpu(__cpu) { \
2063 typeof(type) *stat; \
2064 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2065 u64_stats_init(&stat->syncp); \
2066 } \
2067 } \
2068 pcpu_stats; \
2069 })
2070
2071 #include <linux/notifier.h>
2072
2073 /* netdevice notifier chain. Please remember to update the rtnetlink
2074 * notification exclusion list in rtnetlink_event() when adding new
2075 * types.
2076 */
2077 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2078 #define NETDEV_DOWN 0x0002
2079 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2080 detected a hardware crash and restarted
2081 - we can use this eg to kick tcp sessions
2082 once done */
2083 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2084 #define NETDEV_REGISTER 0x0005
2085 #define NETDEV_UNREGISTER 0x0006
2086 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2087 #define NETDEV_CHANGEADDR 0x0008
2088 #define NETDEV_GOING_DOWN 0x0009
2089 #define NETDEV_CHANGENAME 0x000A
2090 #define NETDEV_FEAT_CHANGE 0x000B
2091 #define NETDEV_BONDING_FAILOVER 0x000C
2092 #define NETDEV_PRE_UP 0x000D
2093 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2094 #define NETDEV_POST_TYPE_CHANGE 0x000F
2095 #define NETDEV_POST_INIT 0x0010
2096 #define NETDEV_UNREGISTER_FINAL 0x0011
2097 #define NETDEV_RELEASE 0x0012
2098 #define NETDEV_NOTIFY_PEERS 0x0013
2099 #define NETDEV_JOIN 0x0014
2100 #define NETDEV_CHANGEUPPER 0x0015
2101 #define NETDEV_RESEND_IGMP 0x0016
2102 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2103 #define NETDEV_CHANGEINFODATA 0x0018
2104 #define NETDEV_BONDING_INFO 0x0019
2105
2106 int register_netdevice_notifier(struct notifier_block *nb);
2107 int unregister_netdevice_notifier(struct notifier_block *nb);
2108
2109 struct netdev_notifier_info {
2110 struct net_device *dev;
2111 };
2112
2113 struct netdev_notifier_change_info {
2114 struct netdev_notifier_info info; /* must be first */
2115 unsigned int flags_changed;
2116 };
2117
2118 struct netdev_notifier_changeupper_info {
2119 struct netdev_notifier_info info; /* must be first */
2120 struct net_device *upper_dev; /* new upper dev */
2121 bool master; /* is upper dev master */
2122 bool linking; /* is the nofication for link or unlink */
2123 };
2124
2125 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2126 struct net_device *dev)
2127 {
2128 info->dev = dev;
2129 }
2130
2131 static inline struct net_device *
2132 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2133 {
2134 return info->dev;
2135 }
2136
2137 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2138
2139
2140 extern rwlock_t dev_base_lock; /* Device list lock */
2141
2142 #define for_each_netdev(net, d) \
2143 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2144 #define for_each_netdev_reverse(net, d) \
2145 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2146 #define for_each_netdev_rcu(net, d) \
2147 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2148 #define for_each_netdev_safe(net, d, n) \
2149 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2150 #define for_each_netdev_continue(net, d) \
2151 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2152 #define for_each_netdev_continue_rcu(net, d) \
2153 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2154 #define for_each_netdev_in_bond_rcu(bond, slave) \
2155 for_each_netdev_rcu(&init_net, slave) \
2156 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2157 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2158
2159 static inline struct net_device *next_net_device(struct net_device *dev)
2160 {
2161 struct list_head *lh;
2162 struct net *net;
2163
2164 net = dev_net(dev);
2165 lh = dev->dev_list.next;
2166 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2167 }
2168
2169 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2170 {
2171 struct list_head *lh;
2172 struct net *net;
2173
2174 net = dev_net(dev);
2175 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2176 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2177 }
2178
2179 static inline struct net_device *first_net_device(struct net *net)
2180 {
2181 return list_empty(&net->dev_base_head) ? NULL :
2182 net_device_entry(net->dev_base_head.next);
2183 }
2184
2185 static inline struct net_device *first_net_device_rcu(struct net *net)
2186 {
2187 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2188
2189 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2190 }
2191
2192 int netdev_boot_setup_check(struct net_device *dev);
2193 unsigned long netdev_boot_base(const char *prefix, int unit);
2194 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2195 const char *hwaddr);
2196 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2197 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2198 void dev_add_pack(struct packet_type *pt);
2199 void dev_remove_pack(struct packet_type *pt);
2200 void __dev_remove_pack(struct packet_type *pt);
2201 void dev_add_offload(struct packet_offload *po);
2202 void dev_remove_offload(struct packet_offload *po);
2203
2204 int dev_get_iflink(const struct net_device *dev);
2205 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2206 unsigned short mask);
2207 struct net_device *dev_get_by_name(struct net *net, const char *name);
2208 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2209 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2210 int dev_alloc_name(struct net_device *dev, const char *name);
2211 int dev_open(struct net_device *dev);
2212 int dev_close(struct net_device *dev);
2213 int dev_close_many(struct list_head *head, bool unlink);
2214 void dev_disable_lro(struct net_device *dev);
2215 int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2216 int dev_queue_xmit(struct sk_buff *skb);
2217 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2218 int register_netdevice(struct net_device *dev);
2219 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2220 void unregister_netdevice_many(struct list_head *head);
2221 static inline void unregister_netdevice(struct net_device *dev)
2222 {
2223 unregister_netdevice_queue(dev, NULL);
2224 }
2225
2226 int netdev_refcnt_read(const struct net_device *dev);
2227 void free_netdev(struct net_device *dev);
2228 void netdev_freemem(struct net_device *dev);
2229 void synchronize_net(void);
2230 int init_dummy_netdev(struct net_device *dev);
2231
2232 DECLARE_PER_CPU(int, xmit_recursion);
2233 static inline int dev_recursion_level(void)
2234 {
2235 return this_cpu_read(xmit_recursion);
2236 }
2237
2238 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2239 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2240 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2241 int netdev_get_name(struct net *net, char *name, int ifindex);
2242 int dev_restart(struct net_device *dev);
2243 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2244
2245 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2246 {
2247 return NAPI_GRO_CB(skb)->data_offset;
2248 }
2249
2250 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2251 {
2252 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2253 }
2254
2255 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2256 {
2257 NAPI_GRO_CB(skb)->data_offset += len;
2258 }
2259
2260 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2261 unsigned int offset)
2262 {
2263 return NAPI_GRO_CB(skb)->frag0 + offset;
2264 }
2265
2266 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2267 {
2268 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2269 }
2270
2271 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2272 unsigned int offset)
2273 {
2274 if (!pskb_may_pull(skb, hlen))
2275 return NULL;
2276
2277 NAPI_GRO_CB(skb)->frag0 = NULL;
2278 NAPI_GRO_CB(skb)->frag0_len = 0;
2279 return skb->data + offset;
2280 }
2281
2282 static inline void *skb_gro_network_header(struct sk_buff *skb)
2283 {
2284 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2285 skb_network_offset(skb);
2286 }
2287
2288 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2289 const void *start, unsigned int len)
2290 {
2291 if (NAPI_GRO_CB(skb)->csum_valid)
2292 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2293 csum_partial(start, len, 0));
2294 }
2295
2296 /* GRO checksum functions. These are logical equivalents of the normal
2297 * checksum functions (in skbuff.h) except that they operate on the GRO
2298 * offsets and fields in sk_buff.
2299 */
2300
2301 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2302
2303 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2304 {
2305 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2306 }
2307
2308 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2309 bool zero_okay,
2310 __sum16 check)
2311 {
2312 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2313 skb_checksum_start_offset(skb) <
2314 skb_gro_offset(skb)) &&
2315 !skb_at_gro_remcsum_start(skb) &&
2316 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2317 (!zero_okay || check));
2318 }
2319
2320 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2321 __wsum psum)
2322 {
2323 if (NAPI_GRO_CB(skb)->csum_valid &&
2324 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2325 return 0;
2326
2327 NAPI_GRO_CB(skb)->csum = psum;
2328
2329 return __skb_gro_checksum_complete(skb);
2330 }
2331
2332 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2333 {
2334 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2335 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2336 NAPI_GRO_CB(skb)->csum_cnt--;
2337 } else {
2338 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2339 * verified a new top level checksum or an encapsulated one
2340 * during GRO. This saves work if we fallback to normal path.
2341 */
2342 __skb_incr_checksum_unnecessary(skb);
2343 }
2344 }
2345
2346 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2347 compute_pseudo) \
2348 ({ \
2349 __sum16 __ret = 0; \
2350 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2351 __ret = __skb_gro_checksum_validate_complete(skb, \
2352 compute_pseudo(skb, proto)); \
2353 if (__ret) \
2354 __skb_mark_checksum_bad(skb); \
2355 else \
2356 skb_gro_incr_csum_unnecessary(skb); \
2357 __ret; \
2358 })
2359
2360 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2361 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2362
2363 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2364 compute_pseudo) \
2365 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2366
2367 #define skb_gro_checksum_simple_validate(skb) \
2368 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2369
2370 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2371 {
2372 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2373 !NAPI_GRO_CB(skb)->csum_valid);
2374 }
2375
2376 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2377 __sum16 check, __wsum pseudo)
2378 {
2379 NAPI_GRO_CB(skb)->csum = ~pseudo;
2380 NAPI_GRO_CB(skb)->csum_valid = 1;
2381 }
2382
2383 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2384 do { \
2385 if (__skb_gro_checksum_convert_check(skb)) \
2386 __skb_gro_checksum_convert(skb, check, \
2387 compute_pseudo(skb, proto)); \
2388 } while (0)
2389
2390 struct gro_remcsum {
2391 int offset;
2392 __wsum delta;
2393 };
2394
2395 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2396 {
2397 grc->offset = 0;
2398 grc->delta = 0;
2399 }
2400
2401 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2402 unsigned int off, size_t hdrlen,
2403 int start, int offset,
2404 struct gro_remcsum *grc,
2405 bool nopartial)
2406 {
2407 __wsum delta;
2408 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2409
2410 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2411
2412 if (!nopartial) {
2413 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2414 return ptr;
2415 }
2416
2417 ptr = skb_gro_header_fast(skb, off);
2418 if (skb_gro_header_hard(skb, off + plen)) {
2419 ptr = skb_gro_header_slow(skb, off + plen, off);
2420 if (!ptr)
2421 return NULL;
2422 }
2423
2424 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2425 start, offset);
2426
2427 /* Adjust skb->csum since we changed the packet */
2428 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2429
2430 grc->offset = off + hdrlen + offset;
2431 grc->delta = delta;
2432
2433 return ptr;
2434 }
2435
2436 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2437 struct gro_remcsum *grc)
2438 {
2439 void *ptr;
2440 size_t plen = grc->offset + sizeof(u16);
2441
2442 if (!grc->delta)
2443 return;
2444
2445 ptr = skb_gro_header_fast(skb, grc->offset);
2446 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2447 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2448 if (!ptr)
2449 return;
2450 }
2451
2452 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2453 }
2454
2455 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2456 unsigned short type,
2457 const void *daddr, const void *saddr,
2458 unsigned int len)
2459 {
2460 if (!dev->header_ops || !dev->header_ops->create)
2461 return 0;
2462
2463 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2464 }
2465
2466 static inline int dev_parse_header(const struct sk_buff *skb,
2467 unsigned char *haddr)
2468 {
2469 const struct net_device *dev = skb->dev;
2470
2471 if (!dev->header_ops || !dev->header_ops->parse)
2472 return 0;
2473 return dev->header_ops->parse(skb, haddr);
2474 }
2475
2476 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2477 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2478 static inline int unregister_gifconf(unsigned int family)
2479 {
2480 return register_gifconf(family, NULL);
2481 }
2482
2483 #ifdef CONFIG_NET_FLOW_LIMIT
2484 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2485 struct sd_flow_limit {
2486 u64 count;
2487 unsigned int num_buckets;
2488 unsigned int history_head;
2489 u16 history[FLOW_LIMIT_HISTORY];
2490 u8 buckets[];
2491 };
2492
2493 extern int netdev_flow_limit_table_len;
2494 #endif /* CONFIG_NET_FLOW_LIMIT */
2495
2496 /*
2497 * Incoming packets are placed on per-cpu queues
2498 */
2499 struct softnet_data {
2500 struct list_head poll_list;
2501 struct sk_buff_head process_queue;
2502
2503 /* stats */
2504 unsigned int processed;
2505 unsigned int time_squeeze;
2506 unsigned int cpu_collision;
2507 unsigned int received_rps;
2508 #ifdef CONFIG_RPS
2509 struct softnet_data *rps_ipi_list;
2510 #endif
2511 #ifdef CONFIG_NET_FLOW_LIMIT
2512 struct sd_flow_limit __rcu *flow_limit;
2513 #endif
2514 struct Qdisc *output_queue;
2515 struct Qdisc **output_queue_tailp;
2516 struct sk_buff *completion_queue;
2517
2518 #ifdef CONFIG_RPS
2519 /* Elements below can be accessed between CPUs for RPS */
2520 struct call_single_data csd ____cacheline_aligned_in_smp;
2521 struct softnet_data *rps_ipi_next;
2522 unsigned int cpu;
2523 unsigned int input_queue_head;
2524 unsigned int input_queue_tail;
2525 #endif
2526 unsigned int dropped;
2527 struct sk_buff_head input_pkt_queue;
2528 struct napi_struct backlog;
2529
2530 };
2531
2532 static inline void input_queue_head_incr(struct softnet_data *sd)
2533 {
2534 #ifdef CONFIG_RPS
2535 sd->input_queue_head++;
2536 #endif
2537 }
2538
2539 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2540 unsigned int *qtail)
2541 {
2542 #ifdef CONFIG_RPS
2543 *qtail = ++sd->input_queue_tail;
2544 #endif
2545 }
2546
2547 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2548
2549 void __netif_schedule(struct Qdisc *q);
2550 void netif_schedule_queue(struct netdev_queue *txq);
2551
2552 static inline void netif_tx_schedule_all(struct net_device *dev)
2553 {
2554 unsigned int i;
2555
2556 for (i = 0; i < dev->num_tx_queues; i++)
2557 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2558 }
2559
2560 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2561 {
2562 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2563 }
2564
2565 /**
2566 * netif_start_queue - allow transmit
2567 * @dev: network device
2568 *
2569 * Allow upper layers to call the device hard_start_xmit routine.
2570 */
2571 static inline void netif_start_queue(struct net_device *dev)
2572 {
2573 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2574 }
2575
2576 static inline void netif_tx_start_all_queues(struct net_device *dev)
2577 {
2578 unsigned int i;
2579
2580 for (i = 0; i < dev->num_tx_queues; i++) {
2581 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2582 netif_tx_start_queue(txq);
2583 }
2584 }
2585
2586 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2587
2588 /**
2589 * netif_wake_queue - restart transmit
2590 * @dev: network device
2591 *
2592 * Allow upper layers to call the device hard_start_xmit routine.
2593 * Used for flow control when transmit resources are available.
2594 */
2595 static inline void netif_wake_queue(struct net_device *dev)
2596 {
2597 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2598 }
2599
2600 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2601 {
2602 unsigned int i;
2603
2604 for (i = 0; i < dev->num_tx_queues; i++) {
2605 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2606 netif_tx_wake_queue(txq);
2607 }
2608 }
2609
2610 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2611 {
2612 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2613 }
2614
2615 /**
2616 * netif_stop_queue - stop transmitted packets
2617 * @dev: network device
2618 *
2619 * Stop upper layers calling the device hard_start_xmit routine.
2620 * Used for flow control when transmit resources are unavailable.
2621 */
2622 static inline void netif_stop_queue(struct net_device *dev)
2623 {
2624 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2625 }
2626
2627 void netif_tx_stop_all_queues(struct net_device *dev);
2628
2629 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2630 {
2631 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2632 }
2633
2634 /**
2635 * netif_queue_stopped - test if transmit queue is flowblocked
2636 * @dev: network device
2637 *
2638 * Test if transmit queue on device is currently unable to send.
2639 */
2640 static inline bool netif_queue_stopped(const struct net_device *dev)
2641 {
2642 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2643 }
2644
2645 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2646 {
2647 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2648 }
2649
2650 static inline bool
2651 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2652 {
2653 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2654 }
2655
2656 static inline bool
2657 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2658 {
2659 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2660 }
2661
2662 /**
2663 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2664 * @dev_queue: pointer to transmit queue
2665 *
2666 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2667 * to give appropriate hint to the cpu.
2668 */
2669 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2670 {
2671 #ifdef CONFIG_BQL
2672 prefetchw(&dev_queue->dql.num_queued);
2673 #endif
2674 }
2675
2676 /**
2677 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2678 * @dev_queue: pointer to transmit queue
2679 *
2680 * BQL enabled drivers might use this helper in their TX completion path,
2681 * to give appropriate hint to the cpu.
2682 */
2683 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2684 {
2685 #ifdef CONFIG_BQL
2686 prefetchw(&dev_queue->dql.limit);
2687 #endif
2688 }
2689
2690 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2691 unsigned int bytes)
2692 {
2693 #ifdef CONFIG_BQL
2694 dql_queued(&dev_queue->dql, bytes);
2695
2696 if (likely(dql_avail(&dev_queue->dql) >= 0))
2697 return;
2698
2699 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2700
2701 /*
2702 * The XOFF flag must be set before checking the dql_avail below,
2703 * because in netdev_tx_completed_queue we update the dql_completed
2704 * before checking the XOFF flag.
2705 */
2706 smp_mb();
2707
2708 /* check again in case another CPU has just made room avail */
2709 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2710 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2711 #endif
2712 }
2713
2714 /**
2715 * netdev_sent_queue - report the number of bytes queued to hardware
2716 * @dev: network device
2717 * @bytes: number of bytes queued to the hardware device queue
2718 *
2719 * Report the number of bytes queued for sending/completion to the network
2720 * device hardware queue. @bytes should be a good approximation and should
2721 * exactly match netdev_completed_queue() @bytes
2722 */
2723 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2724 {
2725 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2726 }
2727
2728 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2729 unsigned int pkts, unsigned int bytes)
2730 {
2731 #ifdef CONFIG_BQL
2732 if (unlikely(!bytes))
2733 return;
2734
2735 dql_completed(&dev_queue->dql, bytes);
2736
2737 /*
2738 * Without the memory barrier there is a small possiblity that
2739 * netdev_tx_sent_queue will miss the update and cause the queue to
2740 * be stopped forever
2741 */
2742 smp_mb();
2743
2744 if (dql_avail(&dev_queue->dql) < 0)
2745 return;
2746
2747 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2748 netif_schedule_queue(dev_queue);
2749 #endif
2750 }
2751
2752 /**
2753 * netdev_completed_queue - report bytes and packets completed by device
2754 * @dev: network device
2755 * @pkts: actual number of packets sent over the medium
2756 * @bytes: actual number of bytes sent over the medium
2757 *
2758 * Report the number of bytes and packets transmitted by the network device
2759 * hardware queue over the physical medium, @bytes must exactly match the
2760 * @bytes amount passed to netdev_sent_queue()
2761 */
2762 static inline void netdev_completed_queue(struct net_device *dev,
2763 unsigned int pkts, unsigned int bytes)
2764 {
2765 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2766 }
2767
2768 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2769 {
2770 #ifdef CONFIG_BQL
2771 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2772 dql_reset(&q->dql);
2773 #endif
2774 }
2775
2776 /**
2777 * netdev_reset_queue - reset the packets and bytes count of a network device
2778 * @dev_queue: network device
2779 *
2780 * Reset the bytes and packet count of a network device and clear the
2781 * software flow control OFF bit for this network device
2782 */
2783 static inline void netdev_reset_queue(struct net_device *dev_queue)
2784 {
2785 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2786 }
2787
2788 /**
2789 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2790 * @dev: network device
2791 * @queue_index: given tx queue index
2792 *
2793 * Returns 0 if given tx queue index >= number of device tx queues,
2794 * otherwise returns the originally passed tx queue index.
2795 */
2796 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2797 {
2798 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2799 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2800 dev->name, queue_index,
2801 dev->real_num_tx_queues);
2802 return 0;
2803 }
2804
2805 return queue_index;
2806 }
2807
2808 /**
2809 * netif_running - test if up
2810 * @dev: network device
2811 *
2812 * Test if the device has been brought up.
2813 */
2814 static inline bool netif_running(const struct net_device *dev)
2815 {
2816 return test_bit(__LINK_STATE_START, &dev->state);
2817 }
2818
2819 /*
2820 * Routines to manage the subqueues on a device. We only need start
2821 * stop, and a check if it's stopped. All other device management is
2822 * done at the overall netdevice level.
2823 * Also test the device if we're multiqueue.
2824 */
2825
2826 /**
2827 * netif_start_subqueue - allow sending packets on subqueue
2828 * @dev: network device
2829 * @queue_index: sub queue index
2830 *
2831 * Start individual transmit queue of a device with multiple transmit queues.
2832 */
2833 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2834 {
2835 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2836
2837 netif_tx_start_queue(txq);
2838 }
2839
2840 /**
2841 * netif_stop_subqueue - stop sending packets on subqueue
2842 * @dev: network device
2843 * @queue_index: sub queue index
2844 *
2845 * Stop individual transmit queue of a device with multiple transmit queues.
2846 */
2847 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2848 {
2849 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2850 netif_tx_stop_queue(txq);
2851 }
2852
2853 /**
2854 * netif_subqueue_stopped - test status of subqueue
2855 * @dev: network device
2856 * @queue_index: sub queue index
2857 *
2858 * Check individual transmit queue of a device with multiple transmit queues.
2859 */
2860 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2861 u16 queue_index)
2862 {
2863 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2864
2865 return netif_tx_queue_stopped(txq);
2866 }
2867
2868 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2869 struct sk_buff *skb)
2870 {
2871 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2872 }
2873
2874 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2875
2876 #ifdef CONFIG_XPS
2877 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2878 u16 index);
2879 #else
2880 static inline int netif_set_xps_queue(struct net_device *dev,
2881 const struct cpumask *mask,
2882 u16 index)
2883 {
2884 return 0;
2885 }
2886 #endif
2887
2888 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2889 unsigned int num_tx_queues);
2890
2891 /*
2892 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2893 * as a distribution range limit for the returned value.
2894 */
2895 static inline u16 skb_tx_hash(const struct net_device *dev,
2896 struct sk_buff *skb)
2897 {
2898 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2899 }
2900
2901 /**
2902 * netif_is_multiqueue - test if device has multiple transmit queues
2903 * @dev: network device
2904 *
2905 * Check if device has multiple transmit queues
2906 */
2907 static inline bool netif_is_multiqueue(const struct net_device *dev)
2908 {
2909 return dev->num_tx_queues > 1;
2910 }
2911
2912 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2913
2914 #ifdef CONFIG_SYSFS
2915 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2916 #else
2917 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2918 unsigned int rxq)
2919 {
2920 return 0;
2921 }
2922 #endif
2923
2924 #ifdef CONFIG_SYSFS
2925 static inline unsigned int get_netdev_rx_queue_index(
2926 struct netdev_rx_queue *queue)
2927 {
2928 struct net_device *dev = queue->dev;
2929 int index = queue - dev->_rx;
2930
2931 BUG_ON(index >= dev->num_rx_queues);
2932 return index;
2933 }
2934 #endif
2935
2936 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2937 int netif_get_num_default_rss_queues(void);
2938
2939 enum skb_free_reason {
2940 SKB_REASON_CONSUMED,
2941 SKB_REASON_DROPPED,
2942 };
2943
2944 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2945 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2946
2947 /*
2948 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2949 * interrupt context or with hardware interrupts being disabled.
2950 * (in_irq() || irqs_disabled())
2951 *
2952 * We provide four helpers that can be used in following contexts :
2953 *
2954 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2955 * replacing kfree_skb(skb)
2956 *
2957 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2958 * Typically used in place of consume_skb(skb) in TX completion path
2959 *
2960 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2961 * replacing kfree_skb(skb)
2962 *
2963 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2964 * and consumed a packet. Used in place of consume_skb(skb)
2965 */
2966 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2967 {
2968 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2969 }
2970
2971 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2972 {
2973 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2974 }
2975
2976 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2977 {
2978 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2979 }
2980
2981 static inline void dev_consume_skb_any(struct sk_buff *skb)
2982 {
2983 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2984 }
2985
2986 int netif_rx(struct sk_buff *skb);
2987 int netif_rx_ni(struct sk_buff *skb);
2988 int netif_receive_skb(struct sk_buff *skb);
2989 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2990 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2991 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2992 gro_result_t napi_gro_frags(struct napi_struct *napi);
2993 struct packet_offload *gro_find_receive_by_type(__be16 type);
2994 struct packet_offload *gro_find_complete_by_type(__be16 type);
2995
2996 static inline void napi_free_frags(struct napi_struct *napi)
2997 {
2998 kfree_skb(napi->skb);
2999 napi->skb = NULL;
3000 }
3001
3002 int netdev_rx_handler_register(struct net_device *dev,
3003 rx_handler_func_t *rx_handler,
3004 void *rx_handler_data);
3005 void netdev_rx_handler_unregister(struct net_device *dev);
3006
3007 bool dev_valid_name(const char *name);
3008 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3009 int dev_ethtool(struct net *net, struct ifreq *);
3010 unsigned int dev_get_flags(const struct net_device *);
3011 int __dev_change_flags(struct net_device *, unsigned int flags);
3012 int dev_change_flags(struct net_device *, unsigned int);
3013 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3014 unsigned int gchanges);
3015 int dev_change_name(struct net_device *, const char *);
3016 int dev_set_alias(struct net_device *, const char *, size_t);
3017 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3018 int dev_set_mtu(struct net_device *, int);
3019 void dev_set_group(struct net_device *, int);
3020 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3021 int dev_change_carrier(struct net_device *, bool new_carrier);
3022 int dev_get_phys_port_id(struct net_device *dev,
3023 struct netdev_phys_item_id *ppid);
3024 int dev_get_phys_port_name(struct net_device *dev,
3025 char *name, size_t len);
3026 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3027 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3028 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3029 struct netdev_queue *txq, int *ret);
3030 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3031 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3032 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3033
3034 extern int netdev_budget;
3035
3036 /* Called by rtnetlink.c:rtnl_unlock() */
3037 void netdev_run_todo(void);
3038
3039 /**
3040 * dev_put - release reference to device
3041 * @dev: network device
3042 *
3043 * Release reference to device to allow it to be freed.
3044 */
3045 static inline void dev_put(struct net_device *dev)
3046 {
3047 this_cpu_dec(*dev->pcpu_refcnt);
3048 }
3049
3050 /**
3051 * dev_hold - get reference to device
3052 * @dev: network device
3053 *
3054 * Hold reference to device to keep it from being freed.
3055 */
3056 static inline void dev_hold(struct net_device *dev)
3057 {
3058 this_cpu_inc(*dev->pcpu_refcnt);
3059 }
3060
3061 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3062 * and _off may be called from IRQ context, but it is caller
3063 * who is responsible for serialization of these calls.
3064 *
3065 * The name carrier is inappropriate, these functions should really be
3066 * called netif_lowerlayer_*() because they represent the state of any
3067 * kind of lower layer not just hardware media.
3068 */
3069
3070 void linkwatch_init_dev(struct net_device *dev);
3071 void linkwatch_fire_event(struct net_device *dev);
3072 void linkwatch_forget_dev(struct net_device *dev);
3073
3074 /**
3075 * netif_carrier_ok - test if carrier present
3076 * @dev: network device
3077 *
3078 * Check if carrier is present on device
3079 */
3080 static inline bool netif_carrier_ok(const struct net_device *dev)
3081 {
3082 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3083 }
3084
3085 unsigned long dev_trans_start(struct net_device *dev);
3086
3087 void __netdev_watchdog_up(struct net_device *dev);
3088
3089 void netif_carrier_on(struct net_device *dev);
3090
3091 void netif_carrier_off(struct net_device *dev);
3092
3093 /**
3094 * netif_dormant_on - mark device as dormant.
3095 * @dev: network device
3096 *
3097 * Mark device as dormant (as per RFC2863).
3098 *
3099 * The dormant state indicates that the relevant interface is not
3100 * actually in a condition to pass packets (i.e., it is not 'up') but is
3101 * in a "pending" state, waiting for some external event. For "on-
3102 * demand" interfaces, this new state identifies the situation where the
3103 * interface is waiting for events to place it in the up state.
3104 *
3105 */
3106 static inline void netif_dormant_on(struct net_device *dev)
3107 {
3108 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3109 linkwatch_fire_event(dev);
3110 }
3111
3112 /**
3113 * netif_dormant_off - set device as not dormant.
3114 * @dev: network device
3115 *
3116 * Device is not in dormant state.
3117 */
3118 static inline void netif_dormant_off(struct net_device *dev)
3119 {
3120 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3121 linkwatch_fire_event(dev);
3122 }
3123
3124 /**
3125 * netif_dormant - test if carrier present
3126 * @dev: network device
3127 *
3128 * Check if carrier is present on device
3129 */
3130 static inline bool netif_dormant(const struct net_device *dev)
3131 {
3132 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3133 }
3134
3135
3136 /**
3137 * netif_oper_up - test if device is operational
3138 * @dev: network device
3139 *
3140 * Check if carrier is operational
3141 */
3142 static inline bool netif_oper_up(const struct net_device *dev)
3143 {
3144 return (dev->operstate == IF_OPER_UP ||
3145 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3146 }
3147
3148 /**
3149 * netif_device_present - is device available or removed
3150 * @dev: network device
3151 *
3152 * Check if device has not been removed from system.
3153 */
3154 static inline bool netif_device_present(struct net_device *dev)
3155 {
3156 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3157 }
3158
3159 void netif_device_detach(struct net_device *dev);
3160
3161 void netif_device_attach(struct net_device *dev);
3162
3163 /*
3164 * Network interface message level settings
3165 */
3166
3167 enum {
3168 NETIF_MSG_DRV = 0x0001,
3169 NETIF_MSG_PROBE = 0x0002,
3170 NETIF_MSG_LINK = 0x0004,
3171 NETIF_MSG_TIMER = 0x0008,
3172 NETIF_MSG_IFDOWN = 0x0010,
3173 NETIF_MSG_IFUP = 0x0020,
3174 NETIF_MSG_RX_ERR = 0x0040,
3175 NETIF_MSG_TX_ERR = 0x0080,
3176 NETIF_MSG_TX_QUEUED = 0x0100,
3177 NETIF_MSG_INTR = 0x0200,
3178 NETIF_MSG_TX_DONE = 0x0400,
3179 NETIF_MSG_RX_STATUS = 0x0800,
3180 NETIF_MSG_PKTDATA = 0x1000,
3181 NETIF_MSG_HW = 0x2000,
3182 NETIF_MSG_WOL = 0x4000,
3183 };
3184
3185 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3186 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3187 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3188 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3189 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3190 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3191 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3192 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3193 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3194 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3195 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3196 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3197 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3198 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3199 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3200
3201 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3202 {
3203 /* use default */
3204 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3205 return default_msg_enable_bits;
3206 if (debug_value == 0) /* no output */
3207 return 0;
3208 /* set low N bits */
3209 return (1 << debug_value) - 1;
3210 }
3211
3212 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3213 {
3214 spin_lock(&txq->_xmit_lock);
3215 txq->xmit_lock_owner = cpu;
3216 }
3217
3218 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3219 {
3220 spin_lock_bh(&txq->_xmit_lock);
3221 txq->xmit_lock_owner = smp_processor_id();
3222 }
3223
3224 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3225 {
3226 bool ok = spin_trylock(&txq->_xmit_lock);
3227 if (likely(ok))
3228 txq->xmit_lock_owner = smp_processor_id();
3229 return ok;
3230 }
3231
3232 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3233 {
3234 txq->xmit_lock_owner = -1;
3235 spin_unlock(&txq->_xmit_lock);
3236 }
3237
3238 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3239 {
3240 txq->xmit_lock_owner = -1;
3241 spin_unlock_bh(&txq->_xmit_lock);
3242 }
3243
3244 static inline void txq_trans_update(struct netdev_queue *txq)
3245 {
3246 if (txq->xmit_lock_owner != -1)
3247 txq->trans_start = jiffies;
3248 }
3249
3250 /**
3251 * netif_tx_lock - grab network device transmit lock
3252 * @dev: network device
3253 *
3254 * Get network device transmit lock
3255 */
3256 static inline void netif_tx_lock(struct net_device *dev)
3257 {
3258 unsigned int i;
3259 int cpu;
3260
3261 spin_lock(&dev->tx_global_lock);
3262 cpu = smp_processor_id();
3263 for (i = 0; i < dev->num_tx_queues; i++) {
3264 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3265
3266 /* We are the only thread of execution doing a
3267 * freeze, but we have to grab the _xmit_lock in
3268 * order to synchronize with threads which are in
3269 * the ->hard_start_xmit() handler and already
3270 * checked the frozen bit.
3271 */
3272 __netif_tx_lock(txq, cpu);
3273 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3274 __netif_tx_unlock(txq);
3275 }
3276 }
3277
3278 static inline void netif_tx_lock_bh(struct net_device *dev)
3279 {
3280 local_bh_disable();
3281 netif_tx_lock(dev);
3282 }
3283
3284 static inline void netif_tx_unlock(struct net_device *dev)
3285 {
3286 unsigned int i;
3287
3288 for (i = 0; i < dev->num_tx_queues; i++) {
3289 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3290
3291 /* No need to grab the _xmit_lock here. If the
3292 * queue is not stopped for another reason, we
3293 * force a schedule.
3294 */
3295 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3296 netif_schedule_queue(txq);
3297 }
3298 spin_unlock(&dev->tx_global_lock);
3299 }
3300
3301 static inline void netif_tx_unlock_bh(struct net_device *dev)
3302 {
3303 netif_tx_unlock(dev);
3304 local_bh_enable();
3305 }
3306
3307 #define HARD_TX_LOCK(dev, txq, cpu) { \
3308 if ((dev->features & NETIF_F_LLTX) == 0) { \
3309 __netif_tx_lock(txq, cpu); \
3310 } \
3311 }
3312
3313 #define HARD_TX_TRYLOCK(dev, txq) \
3314 (((dev->features & NETIF_F_LLTX) == 0) ? \
3315 __netif_tx_trylock(txq) : \
3316 true )
3317
3318 #define HARD_TX_UNLOCK(dev, txq) { \
3319 if ((dev->features & NETIF_F_LLTX) == 0) { \
3320 __netif_tx_unlock(txq); \
3321 } \
3322 }
3323
3324 static inline void netif_tx_disable(struct net_device *dev)
3325 {
3326 unsigned int i;
3327 int cpu;
3328
3329 local_bh_disable();
3330 cpu = smp_processor_id();
3331 for (i = 0; i < dev->num_tx_queues; i++) {
3332 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3333
3334 __netif_tx_lock(txq, cpu);
3335 netif_tx_stop_queue(txq);
3336 __netif_tx_unlock(txq);
3337 }
3338 local_bh_enable();
3339 }
3340
3341 static inline void netif_addr_lock(struct net_device *dev)
3342 {
3343 spin_lock(&dev->addr_list_lock);
3344 }
3345
3346 static inline void netif_addr_lock_nested(struct net_device *dev)
3347 {
3348 int subclass = SINGLE_DEPTH_NESTING;
3349
3350 if (dev->netdev_ops->ndo_get_lock_subclass)
3351 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3352
3353 spin_lock_nested(&dev->addr_list_lock, subclass);
3354 }
3355
3356 static inline void netif_addr_lock_bh(struct net_device *dev)
3357 {
3358 spin_lock_bh(&dev->addr_list_lock);
3359 }
3360
3361 static inline void netif_addr_unlock(struct net_device *dev)
3362 {
3363 spin_unlock(&dev->addr_list_lock);
3364 }
3365
3366 static inline void netif_addr_unlock_bh(struct net_device *dev)
3367 {
3368 spin_unlock_bh(&dev->addr_list_lock);
3369 }
3370
3371 /*
3372 * dev_addrs walker. Should be used only for read access. Call with
3373 * rcu_read_lock held.
3374 */
3375 #define for_each_dev_addr(dev, ha) \
3376 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3377
3378 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3379
3380 void ether_setup(struct net_device *dev);
3381
3382 /* Support for loadable net-drivers */
3383 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3384 unsigned char name_assign_type,
3385 void (*setup)(struct net_device *),
3386 unsigned int txqs, unsigned int rxqs);
3387 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3388 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3389
3390 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3391 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3392 count)
3393
3394 int register_netdev(struct net_device *dev);
3395 void unregister_netdev(struct net_device *dev);
3396
3397 /* General hardware address lists handling functions */
3398 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3399 struct netdev_hw_addr_list *from_list, int addr_len);
3400 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3401 struct netdev_hw_addr_list *from_list, int addr_len);
3402 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3403 struct net_device *dev,
3404 int (*sync)(struct net_device *, const unsigned char *),
3405 int (*unsync)(struct net_device *,
3406 const unsigned char *));
3407 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3408 struct net_device *dev,
3409 int (*unsync)(struct net_device *,
3410 const unsigned char *));
3411 void __hw_addr_init(struct netdev_hw_addr_list *list);
3412
3413 /* Functions used for device addresses handling */
3414 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3415 unsigned char addr_type);
3416 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3417 unsigned char addr_type);
3418 void dev_addr_flush(struct net_device *dev);
3419 int dev_addr_init(struct net_device *dev);
3420
3421 /* Functions used for unicast addresses handling */
3422 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3423 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3424 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3425 int dev_uc_sync(struct net_device *to, struct net_device *from);
3426 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3427 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3428 void dev_uc_flush(struct net_device *dev);
3429 void dev_uc_init(struct net_device *dev);
3430
3431 /**
3432 * __dev_uc_sync - Synchonize device's unicast list
3433 * @dev: device to sync
3434 * @sync: function to call if address should be added
3435 * @unsync: function to call if address should be removed
3436 *
3437 * Add newly added addresses to the interface, and release
3438 * addresses that have been deleted.
3439 **/
3440 static inline int __dev_uc_sync(struct net_device *dev,
3441 int (*sync)(struct net_device *,
3442 const unsigned char *),
3443 int (*unsync)(struct net_device *,
3444 const unsigned char *))
3445 {
3446 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3447 }
3448
3449 /**
3450 * __dev_uc_unsync - Remove synchronized addresses from device
3451 * @dev: device to sync
3452 * @unsync: function to call if address should be removed
3453 *
3454 * Remove all addresses that were added to the device by dev_uc_sync().
3455 **/
3456 static inline void __dev_uc_unsync(struct net_device *dev,
3457 int (*unsync)(struct net_device *,
3458 const unsigned char *))
3459 {
3460 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3461 }
3462
3463 /* Functions used for multicast addresses handling */
3464 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3465 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3466 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3467 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3468 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3469 int dev_mc_sync(struct net_device *to, struct net_device *from);
3470 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3471 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3472 void dev_mc_flush(struct net_device *dev);
3473 void dev_mc_init(struct net_device *dev);
3474
3475 /**
3476 * __dev_mc_sync - Synchonize device's multicast list
3477 * @dev: device to sync
3478 * @sync: function to call if address should be added
3479 * @unsync: function to call if address should be removed
3480 *
3481 * Add newly added addresses to the interface, and release
3482 * addresses that have been deleted.
3483 **/
3484 static inline int __dev_mc_sync(struct net_device *dev,
3485 int (*sync)(struct net_device *,
3486 const unsigned char *),
3487 int (*unsync)(struct net_device *,
3488 const unsigned char *))
3489 {
3490 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3491 }
3492
3493 /**
3494 * __dev_mc_unsync - Remove synchronized addresses from device
3495 * @dev: device to sync
3496 * @unsync: function to call if address should be removed
3497 *
3498 * Remove all addresses that were added to the device by dev_mc_sync().
3499 **/
3500 static inline void __dev_mc_unsync(struct net_device *dev,
3501 int (*unsync)(struct net_device *,
3502 const unsigned char *))
3503 {
3504 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3505 }
3506
3507 /* Functions used for secondary unicast and multicast support */
3508 void dev_set_rx_mode(struct net_device *dev);
3509 void __dev_set_rx_mode(struct net_device *dev);
3510 int dev_set_promiscuity(struct net_device *dev, int inc);
3511 int dev_set_allmulti(struct net_device *dev, int inc);
3512 void netdev_state_change(struct net_device *dev);
3513 void netdev_notify_peers(struct net_device *dev);
3514 void netdev_features_change(struct net_device *dev);
3515 /* Load a device via the kmod */
3516 void dev_load(struct net *net, const char *name);
3517 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3518 struct rtnl_link_stats64 *storage);
3519 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3520 const struct net_device_stats *netdev_stats);
3521
3522 extern int netdev_max_backlog;
3523 extern int netdev_tstamp_prequeue;
3524 extern int weight_p;
3525 extern int bpf_jit_enable;
3526
3527 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3528 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3529 struct list_head **iter);
3530 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3531 struct list_head **iter);
3532
3533 /* iterate through upper list, must be called under RCU read lock */
3534 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3535 for (iter = &(dev)->adj_list.upper, \
3536 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3537 updev; \
3538 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3539
3540 /* iterate through upper list, must be called under RCU read lock */
3541 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3542 for (iter = &(dev)->all_adj_list.upper, \
3543 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3544 updev; \
3545 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3546
3547 void *netdev_lower_get_next_private(struct net_device *dev,
3548 struct list_head **iter);
3549 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3550 struct list_head **iter);
3551
3552 #define netdev_for_each_lower_private(dev, priv, iter) \
3553 for (iter = (dev)->adj_list.lower.next, \
3554 priv = netdev_lower_get_next_private(dev, &(iter)); \
3555 priv; \
3556 priv = netdev_lower_get_next_private(dev, &(iter)))
3557
3558 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3559 for (iter = &(dev)->adj_list.lower, \
3560 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3561 priv; \
3562 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3563
3564 void *netdev_lower_get_next(struct net_device *dev,
3565 struct list_head **iter);
3566 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3567 for (iter = &(dev)->adj_list.lower, \
3568 ldev = netdev_lower_get_next(dev, &(iter)); \
3569 ldev; \
3570 ldev = netdev_lower_get_next(dev, &(iter)))
3571
3572 void *netdev_adjacent_get_private(struct list_head *adj_list);
3573 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3574 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3575 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3576 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3577 int netdev_master_upper_dev_link(struct net_device *dev,
3578 struct net_device *upper_dev);
3579 int netdev_master_upper_dev_link_private(struct net_device *dev,
3580 struct net_device *upper_dev,
3581 void *private);
3582 void netdev_upper_dev_unlink(struct net_device *dev,
3583 struct net_device *upper_dev);
3584 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3585 void *netdev_lower_dev_get_private(struct net_device *dev,
3586 struct net_device *lower_dev);
3587
3588 /* RSS keys are 40 or 52 bytes long */
3589 #define NETDEV_RSS_KEY_LEN 52
3590 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3591 void netdev_rss_key_fill(void *buffer, size_t len);
3592
3593 int dev_get_nest_level(struct net_device *dev,
3594 bool (*type_check)(struct net_device *dev));
3595 int skb_checksum_help(struct sk_buff *skb);
3596 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3597 netdev_features_t features, bool tx_path);
3598 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3599 netdev_features_t features);
3600
3601 struct netdev_bonding_info {
3602 ifslave slave;
3603 ifbond master;
3604 };
3605
3606 struct netdev_notifier_bonding_info {
3607 struct netdev_notifier_info info; /* must be first */
3608 struct netdev_bonding_info bonding_info;
3609 };
3610
3611 void netdev_bonding_info_change(struct net_device *dev,
3612 struct netdev_bonding_info *bonding_info);
3613
3614 static inline
3615 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3616 {
3617 return __skb_gso_segment(skb, features, true);
3618 }
3619 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3620
3621 static inline bool can_checksum_protocol(netdev_features_t features,
3622 __be16 protocol)
3623 {
3624 return ((features & NETIF_F_GEN_CSUM) ||
3625 ((features & NETIF_F_V4_CSUM) &&
3626 protocol == htons(ETH_P_IP)) ||
3627 ((features & NETIF_F_V6_CSUM) &&
3628 protocol == htons(ETH_P_IPV6)) ||
3629 ((features & NETIF_F_FCOE_CRC) &&
3630 protocol == htons(ETH_P_FCOE)));
3631 }
3632
3633 #ifdef CONFIG_BUG
3634 void netdev_rx_csum_fault(struct net_device *dev);
3635 #else
3636 static inline void netdev_rx_csum_fault(struct net_device *dev)
3637 {
3638 }
3639 #endif
3640 /* rx skb timestamps */
3641 void net_enable_timestamp(void);
3642 void net_disable_timestamp(void);
3643
3644 #ifdef CONFIG_PROC_FS
3645 int __init dev_proc_init(void);
3646 #else
3647 #define dev_proc_init() 0
3648 #endif
3649
3650 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3651 struct sk_buff *skb, struct net_device *dev,
3652 bool more)
3653 {
3654 skb->xmit_more = more ? 1 : 0;
3655 return ops->ndo_start_xmit(skb, dev);
3656 }
3657
3658 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3659 struct netdev_queue *txq, bool more)
3660 {
3661 const struct net_device_ops *ops = dev->netdev_ops;
3662 int rc;
3663
3664 rc = __netdev_start_xmit(ops, skb, dev, more);
3665 if (rc == NETDEV_TX_OK)
3666 txq_trans_update(txq);
3667
3668 return rc;
3669 }
3670
3671 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3672 const void *ns);
3673 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3674 const void *ns);
3675
3676 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3677 {
3678 return netdev_class_create_file_ns(class_attr, NULL);
3679 }
3680
3681 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3682 {
3683 netdev_class_remove_file_ns(class_attr, NULL);
3684 }
3685
3686 extern struct kobj_ns_type_operations net_ns_type_operations;
3687
3688 const char *netdev_drivername(const struct net_device *dev);
3689
3690 void linkwatch_run_queue(void);
3691
3692 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3693 netdev_features_t f2)
3694 {
3695 if (f1 & NETIF_F_GEN_CSUM)
3696 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3697 if (f2 & NETIF_F_GEN_CSUM)
3698 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3699 f1 &= f2;
3700 if (f1 & NETIF_F_GEN_CSUM)
3701 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3702
3703 return f1;
3704 }
3705
3706 static inline netdev_features_t netdev_get_wanted_features(
3707 struct net_device *dev)
3708 {
3709 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3710 }
3711 netdev_features_t netdev_increment_features(netdev_features_t all,
3712 netdev_features_t one, netdev_features_t mask);
3713
3714 /* Allow TSO being used on stacked device :
3715 * Performing the GSO segmentation before last device
3716 * is a performance improvement.
3717 */
3718 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3719 netdev_features_t mask)
3720 {
3721 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3722 }
3723
3724 int __netdev_update_features(struct net_device *dev);
3725 void netdev_update_features(struct net_device *dev);
3726 void netdev_change_features(struct net_device *dev);
3727
3728 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3729 struct net_device *dev);
3730
3731 netdev_features_t passthru_features_check(struct sk_buff *skb,
3732 struct net_device *dev,
3733 netdev_features_t features);
3734 netdev_features_t netif_skb_features(struct sk_buff *skb);
3735
3736 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3737 {
3738 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3739
3740 /* check flags correspondence */
3741 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3742 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3743 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3744 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3745 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3746 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3747 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3748 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3749 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3750 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3751 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3752 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3753 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3754
3755 return (features & feature) == feature;
3756 }
3757
3758 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3759 {
3760 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3761 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3762 }
3763
3764 static inline bool netif_needs_gso(struct sk_buff *skb,
3765 netdev_features_t features)
3766 {
3767 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3768 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3769 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3770 }
3771
3772 static inline void netif_set_gso_max_size(struct net_device *dev,
3773 unsigned int size)
3774 {
3775 dev->gso_max_size = size;
3776 }
3777
3778 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3779 int pulled_hlen, u16 mac_offset,
3780 int mac_len)
3781 {
3782 skb->protocol = protocol;
3783 skb->encapsulation = 1;
3784 skb_push(skb, pulled_hlen);
3785 skb_reset_transport_header(skb);
3786 skb->mac_header = mac_offset;
3787 skb->network_header = skb->mac_header + mac_len;
3788 skb->mac_len = mac_len;
3789 }
3790
3791 static inline bool netif_is_macvlan(struct net_device *dev)
3792 {
3793 return dev->priv_flags & IFF_MACVLAN;
3794 }
3795
3796 static inline bool netif_is_macvlan_port(struct net_device *dev)
3797 {
3798 return dev->priv_flags & IFF_MACVLAN_PORT;
3799 }
3800
3801 static inline bool netif_is_ipvlan(struct net_device *dev)
3802 {
3803 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3804 }
3805
3806 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3807 {
3808 return dev->priv_flags & IFF_IPVLAN_MASTER;
3809 }
3810
3811 static inline bool netif_is_bond_master(struct net_device *dev)
3812 {
3813 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3814 }
3815
3816 static inline bool netif_is_bond_slave(struct net_device *dev)
3817 {
3818 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3819 }
3820
3821 static inline bool netif_supports_nofcs(struct net_device *dev)
3822 {
3823 return dev->priv_flags & IFF_SUPP_NOFCS;
3824 }
3825
3826 static inline bool netif_is_vrf(const struct net_device *dev)
3827 {
3828 return dev->priv_flags & IFF_VRF_MASTER;
3829 }
3830
3831 static inline bool netif_is_bridge_master(const struct net_device *dev)
3832 {
3833 return dev->priv_flags & IFF_EBRIDGE;
3834 }
3835
3836 static inline bool netif_is_ovs_master(const struct net_device *dev)
3837 {
3838 return dev->priv_flags & IFF_OPENVSWITCH;
3839 }
3840
3841 static inline bool netif_index_is_vrf(struct net *net, int ifindex)
3842 {
3843 bool rc = false;
3844
3845 #if IS_ENABLED(CONFIG_NET_VRF)
3846 struct net_device *dev;
3847
3848 if (ifindex == 0)
3849 return false;
3850
3851 rcu_read_lock();
3852
3853 dev = dev_get_by_index_rcu(net, ifindex);
3854 if (dev)
3855 rc = netif_is_vrf(dev);
3856
3857 rcu_read_unlock();
3858 #endif
3859 return rc;
3860 }
3861
3862 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3863 static inline void netif_keep_dst(struct net_device *dev)
3864 {
3865 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3866 }
3867
3868 extern struct pernet_operations __net_initdata loopback_net_ops;
3869
3870 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3871
3872 /* netdev_printk helpers, similar to dev_printk */
3873
3874 static inline const char *netdev_name(const struct net_device *dev)
3875 {
3876 if (!dev->name[0] || strchr(dev->name, '%'))
3877 return "(unnamed net_device)";
3878 return dev->name;
3879 }
3880
3881 static inline const char *netdev_reg_state(const struct net_device *dev)
3882 {
3883 switch (dev->reg_state) {
3884 case NETREG_UNINITIALIZED: return " (uninitialized)";
3885 case NETREG_REGISTERED: return "";
3886 case NETREG_UNREGISTERING: return " (unregistering)";
3887 case NETREG_UNREGISTERED: return " (unregistered)";
3888 case NETREG_RELEASED: return " (released)";
3889 case NETREG_DUMMY: return " (dummy)";
3890 }
3891
3892 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3893 return " (unknown)";
3894 }
3895
3896 __printf(3, 4)
3897 void netdev_printk(const char *level, const struct net_device *dev,
3898 const char *format, ...);
3899 __printf(2, 3)
3900 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3901 __printf(2, 3)
3902 void netdev_alert(const struct net_device *dev, const char *format, ...);
3903 __printf(2, 3)
3904 void netdev_crit(const struct net_device *dev, const char *format, ...);
3905 __printf(2, 3)
3906 void netdev_err(const struct net_device *dev, const char *format, ...);
3907 __printf(2, 3)
3908 void netdev_warn(const struct net_device *dev, const char *format, ...);
3909 __printf(2, 3)
3910 void netdev_notice(const struct net_device *dev, const char *format, ...);
3911 __printf(2, 3)
3912 void netdev_info(const struct net_device *dev, const char *format, ...);
3913
3914 #define MODULE_ALIAS_NETDEV(device) \
3915 MODULE_ALIAS("netdev-" device)
3916
3917 #if defined(CONFIG_DYNAMIC_DEBUG)
3918 #define netdev_dbg(__dev, format, args...) \
3919 do { \
3920 dynamic_netdev_dbg(__dev, format, ##args); \
3921 } while (0)
3922 #elif defined(DEBUG)
3923 #define netdev_dbg(__dev, format, args...) \
3924 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3925 #else
3926 #define netdev_dbg(__dev, format, args...) \
3927 ({ \
3928 if (0) \
3929 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3930 })
3931 #endif
3932
3933 #if defined(VERBOSE_DEBUG)
3934 #define netdev_vdbg netdev_dbg
3935 #else
3936
3937 #define netdev_vdbg(dev, format, args...) \
3938 ({ \
3939 if (0) \
3940 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3941 0; \
3942 })
3943 #endif
3944
3945 /*
3946 * netdev_WARN() acts like dev_printk(), but with the key difference
3947 * of using a WARN/WARN_ON to get the message out, including the
3948 * file/line information and a backtrace.
3949 */
3950 #define netdev_WARN(dev, format, args...) \
3951 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3952 netdev_reg_state(dev), ##args)
3953
3954 /* netif printk helpers, similar to netdev_printk */
3955
3956 #define netif_printk(priv, type, level, dev, fmt, args...) \
3957 do { \
3958 if (netif_msg_##type(priv)) \
3959 netdev_printk(level, (dev), fmt, ##args); \
3960 } while (0)
3961
3962 #define netif_level(level, priv, type, dev, fmt, args...) \
3963 do { \
3964 if (netif_msg_##type(priv)) \
3965 netdev_##level(dev, fmt, ##args); \
3966 } while (0)
3967
3968 #define netif_emerg(priv, type, dev, fmt, args...) \
3969 netif_level(emerg, priv, type, dev, fmt, ##args)
3970 #define netif_alert(priv, type, dev, fmt, args...) \
3971 netif_level(alert, priv, type, dev, fmt, ##args)
3972 #define netif_crit(priv, type, dev, fmt, args...) \
3973 netif_level(crit, priv, type, dev, fmt, ##args)
3974 #define netif_err(priv, type, dev, fmt, args...) \
3975 netif_level(err, priv, type, dev, fmt, ##args)
3976 #define netif_warn(priv, type, dev, fmt, args...) \
3977 netif_level(warn, priv, type, dev, fmt, ##args)
3978 #define netif_notice(priv, type, dev, fmt, args...) \
3979 netif_level(notice, priv, type, dev, fmt, ##args)
3980 #define netif_info(priv, type, dev, fmt, args...) \
3981 netif_level(info, priv, type, dev, fmt, ##args)
3982
3983 #if defined(CONFIG_DYNAMIC_DEBUG)
3984 #define netif_dbg(priv, type, netdev, format, args...) \
3985 do { \
3986 if (netif_msg_##type(priv)) \
3987 dynamic_netdev_dbg(netdev, format, ##args); \
3988 } while (0)
3989 #elif defined(DEBUG)
3990 #define netif_dbg(priv, type, dev, format, args...) \
3991 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3992 #else
3993 #define netif_dbg(priv, type, dev, format, args...) \
3994 ({ \
3995 if (0) \
3996 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3997 0; \
3998 })
3999 #endif
4000
4001 #if defined(VERBOSE_DEBUG)
4002 #define netif_vdbg netif_dbg
4003 #else
4004 #define netif_vdbg(priv, type, dev, format, args...) \
4005 ({ \
4006 if (0) \
4007 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4008 0; \
4009 })
4010 #endif
4011
4012 /*
4013 * The list of packet types we will receive (as opposed to discard)
4014 * and the routines to invoke.
4015 *
4016 * Why 16. Because with 16 the only overlap we get on a hash of the
4017 * low nibble of the protocol value is RARP/SNAP/X.25.
4018 *
4019 * NOTE: That is no longer true with the addition of VLAN tags. Not
4020 * sure which should go first, but I bet it won't make much
4021 * difference if we are running VLANs. The good news is that
4022 * this protocol won't be in the list unless compiled in, so
4023 * the average user (w/out VLANs) will not be adversely affected.
4024 * --BLG
4025 *
4026 * 0800 IP
4027 * 8100 802.1Q VLAN
4028 * 0001 802.3
4029 * 0002 AX.25
4030 * 0004 802.2
4031 * 8035 RARP
4032 * 0005 SNAP
4033 * 0805 X.25
4034 * 0806 ARP
4035 * 8137 IPX
4036 * 0009 Localtalk
4037 * 86DD IPv6
4038 */
4039 #define PTYPE_HASH_SIZE (16)
4040 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4041
4042 #endif /* _LINUX_NETDEVICE_H */