]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/netdevice.h
rhashtable: Mark internal/private inline functions as such
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
36
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
42
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
45 #include <net/dsa.h>
46 #ifdef CONFIG_DCB
47 #include <net/dcbnl.h>
48 #endif
49 #include <net/netprio_cgroup.h>
50
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
55
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_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 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
772 /*
773 * This structure defines the management hooks for network devices.
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
776 *
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
782 *
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
786 *
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
789 * state.
790 *
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
793 * state.
794 *
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
797 * Called when a packet needs to be transmitted.
798 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
799 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
800 * Required can not be NULL.
801 *
802 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
803 * void *accel_priv, select_queue_fallback_t fallback);
804 * Called to decide which queue to when device supports multiple
805 * transmit queues.
806 *
807 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
808 * This function is called to allow device receiver to make
809 * changes to configuration when multicast or promiscious is enabled.
810 *
811 * void (*ndo_set_rx_mode)(struct net_device *dev);
812 * This function is called device changes address list filtering.
813 * If driver handles unicast address filtering, it should set
814 * IFF_UNICAST_FLT to its priv_flags.
815 *
816 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
817 * This function is called when the Media Access Control address
818 * needs to be changed. If this interface is not defined, the
819 * mac address can not be changed.
820 *
821 * int (*ndo_validate_addr)(struct net_device *dev);
822 * Test if Media Access Control address is valid for the device.
823 *
824 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
825 * Called when a user request an ioctl which can't be handled by
826 * the generic interface code. If not defined ioctl's return
827 * not supported error code.
828 *
829 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
830 * Used to set network devices bus interface parameters. This interface
831 * is retained for legacy reason, new devices should use the bus
832 * interface (PCI) for low level management.
833 *
834 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
835 * Called when a user wants to change the Maximum Transfer Unit
836 * of a device. If not defined, any request to change MTU will
837 * will return an error.
838 *
839 * void (*ndo_tx_timeout)(struct net_device *dev);
840 * Callback uses when the transmitter has not made any progress
841 * for dev->watchdog ticks.
842 *
843 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
844 * struct rtnl_link_stats64 *storage);
845 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
846 * Called when a user wants to get the network device usage
847 * statistics. Drivers must do one of the following:
848 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
849 * rtnl_link_stats64 structure passed by the caller.
850 * 2. Define @ndo_get_stats to update a net_device_stats structure
851 * (which should normally be dev->stats) and return a pointer to
852 * it. The structure may be changed asynchronously only if each
853 * field is written atomically.
854 * 3. Update dev->stats asynchronously and atomically, and define
855 * neither operation.
856 *
857 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
858 * If device support VLAN filtering this function is called when a
859 * VLAN id is registered.
860 *
861 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
862 * If device support VLAN filtering this function is called when a
863 * VLAN id is unregistered.
864 *
865 * void (*ndo_poll_controller)(struct net_device *dev);
866 *
867 * SR-IOV management functions.
868 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
869 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
870 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
871 * int max_tx_rate);
872 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
873 * int (*ndo_get_vf_config)(struct net_device *dev,
874 * int vf, struct ifla_vf_info *ivf);
875 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
876 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
877 * struct nlattr *port[]);
878 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
879 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
880 * Called to setup 'tc' number of traffic classes in the net device. This
881 * is always called from the stack with the rtnl lock held and netif tx
882 * queues stopped. This allows the netdevice to perform queue management
883 * safely.
884 *
885 * Fiber Channel over Ethernet (FCoE) offload functions.
886 * int (*ndo_fcoe_enable)(struct net_device *dev);
887 * Called when the FCoE protocol stack wants to start using LLD for FCoE
888 * so the underlying device can perform whatever needed configuration or
889 * initialization to support acceleration of FCoE traffic.
890 *
891 * int (*ndo_fcoe_disable)(struct net_device *dev);
892 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
893 * so the underlying device can perform whatever needed clean-ups to
894 * stop supporting acceleration of FCoE traffic.
895 *
896 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
897 * struct scatterlist *sgl, unsigned int sgc);
898 * Called when the FCoE Initiator wants to initialize an I/O that
899 * is a possible candidate for Direct Data Placement (DDP). The LLD can
900 * perform necessary setup and returns 1 to indicate the device is set up
901 * successfully to perform DDP on this I/O, otherwise this returns 0.
902 *
903 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
904 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
905 * indicated by the FC exchange id 'xid', so the underlying device can
906 * clean up and reuse resources for later DDP requests.
907 *
908 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
909 * struct scatterlist *sgl, unsigned int sgc);
910 * Called when the FCoE Target wants to initialize an I/O that
911 * is a possible candidate for Direct Data Placement (DDP). The LLD can
912 * perform necessary setup and returns 1 to indicate the device is set up
913 * successfully to perform DDP on this I/O, otherwise this returns 0.
914 *
915 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
916 * struct netdev_fcoe_hbainfo *hbainfo);
917 * Called when the FCoE Protocol stack wants information on the underlying
918 * device. This information is utilized by the FCoE protocol stack to
919 * register attributes with Fiber Channel management service as per the
920 * FC-GS Fabric Device Management Information(FDMI) specification.
921 *
922 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
923 * Called when the underlying device wants to override default World Wide
924 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
925 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
926 * protocol stack to use.
927 *
928 * RFS acceleration.
929 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
930 * u16 rxq_index, u32 flow_id);
931 * Set hardware filter for RFS. rxq_index is the target queue index;
932 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
933 * Return the filter ID on success, or a negative error code.
934 *
935 * Slave management functions (for bridge, bonding, etc).
936 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
937 * Called to make another netdev an underling.
938 *
939 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
940 * Called to release previously enslaved netdev.
941 *
942 * Feature/offload setting functions.
943 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
944 * netdev_features_t features);
945 * Adjusts the requested feature flags according to device-specific
946 * constraints, and returns the resulting flags. Must not modify
947 * the device state.
948 *
949 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
950 * Called to update device configuration to new features. Passed
951 * feature set might be less than what was returned by ndo_fix_features()).
952 * Must return >0 or -errno if it changed dev->features itself.
953 *
954 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
955 * struct net_device *dev,
956 * const unsigned char *addr, u16 vid, u16 flags)
957 * Adds an FDB entry to dev for addr.
958 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
959 * struct net_device *dev,
960 * const unsigned char *addr, u16 vid)
961 * Deletes the FDB entry from dev coresponding to addr.
962 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
963 * struct net_device *dev, struct net_device *filter_dev,
964 * int idx)
965 * Used to add FDB entries to dump requests. Implementers should add
966 * entries to skb and update idx with the number of entries.
967 *
968 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
969 * u16 flags)
970 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
971 * struct net_device *dev, u32 filter_mask)
972 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
973 * u16 flags);
974 *
975 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
976 * Called to change device carrier. Soft-devices (like dummy, team, etc)
977 * which do not represent real hardware may define this to allow their
978 * userspace components to manage their virtual carrier state. Devices
979 * that determine carrier state from physical hardware properties (eg
980 * network cables) or protocol-dependent mechanisms (eg
981 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
982 *
983 * int (*ndo_get_phys_port_id)(struct net_device *dev,
984 * struct netdev_phys_item_id *ppid);
985 * Called to get ID of physical port of this device. If driver does
986 * not implement this, it is assumed that the hw is not able to have
987 * multiple net devices on single physical port.
988 *
989 * void (*ndo_add_vxlan_port)(struct net_device *dev,
990 * sa_family_t sa_family, __be16 port);
991 * Called by vxlan to notiy a driver about the UDP port and socket
992 * address family that vxlan is listnening to. It is called only when
993 * a new port starts listening. The operation is protected by the
994 * vxlan_net->sock_lock.
995 *
996 * void (*ndo_del_vxlan_port)(struct net_device *dev,
997 * sa_family_t sa_family, __be16 port);
998 * Called by vxlan to notify the driver about a UDP port and socket
999 * address family that vxlan is not listening to anymore. The operation
1000 * is protected by the vxlan_net->sock_lock.
1001 *
1002 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1003 * struct net_device *dev)
1004 * Called by upper layer devices to accelerate switching or other
1005 * station functionality into hardware. 'pdev is the lowerdev
1006 * to use for the offload and 'dev' is the net device that will
1007 * back the offload. Returns a pointer to the private structure
1008 * the upper layer will maintain.
1009 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1010 * Called by upper layer device to delete the station created
1011 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1012 * the station and priv is the structure returned by the add
1013 * operation.
1014 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1015 * struct net_device *dev,
1016 * void *priv);
1017 * Callback to use for xmit over the accelerated station. This
1018 * is used in place of ndo_start_xmit on accelerated net
1019 * devices.
1020 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1021 * struct net_device *dev
1022 * netdev_features_t features);
1023 * Called by core transmit path to determine if device is capable of
1024 * performing offload operations on a given packet. This is to give
1025 * the device an opportunity to implement any restrictions that cannot
1026 * be otherwise expressed by feature flags. The check is called with
1027 * the set of features that the stack has calculated and it returns
1028 * those the driver believes to be appropriate.
1029 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1030 * int queue_index, u32 maxrate);
1031 * Called when a user wants to set a max-rate limitation of specific
1032 * TX queue.
1033 */
1034 struct net_device_ops {
1035 int (*ndo_init)(struct net_device *dev);
1036 void (*ndo_uninit)(struct net_device *dev);
1037 int (*ndo_open)(struct net_device *dev);
1038 int (*ndo_stop)(struct net_device *dev);
1039 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1040 struct net_device *dev);
1041 u16 (*ndo_select_queue)(struct net_device *dev,
1042 struct sk_buff *skb,
1043 void *accel_priv,
1044 select_queue_fallback_t fallback);
1045 void (*ndo_change_rx_flags)(struct net_device *dev,
1046 int flags);
1047 void (*ndo_set_rx_mode)(struct net_device *dev);
1048 int (*ndo_set_mac_address)(struct net_device *dev,
1049 void *addr);
1050 int (*ndo_validate_addr)(struct net_device *dev);
1051 int (*ndo_do_ioctl)(struct net_device *dev,
1052 struct ifreq *ifr, int cmd);
1053 int (*ndo_set_config)(struct net_device *dev,
1054 struct ifmap *map);
1055 int (*ndo_change_mtu)(struct net_device *dev,
1056 int new_mtu);
1057 int (*ndo_neigh_setup)(struct net_device *dev,
1058 struct neigh_parms *);
1059 void (*ndo_tx_timeout) (struct net_device *dev);
1060
1061 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1062 struct rtnl_link_stats64 *storage);
1063 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1064
1065 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1066 __be16 proto, u16 vid);
1067 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1068 __be16 proto, u16 vid);
1069 #ifdef CONFIG_NET_POLL_CONTROLLER
1070 void (*ndo_poll_controller)(struct net_device *dev);
1071 int (*ndo_netpoll_setup)(struct net_device *dev,
1072 struct netpoll_info *info);
1073 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1074 #endif
1075 #ifdef CONFIG_NET_RX_BUSY_POLL
1076 int (*ndo_busy_poll)(struct napi_struct *dev);
1077 #endif
1078 int (*ndo_set_vf_mac)(struct net_device *dev,
1079 int queue, u8 *mac);
1080 int (*ndo_set_vf_vlan)(struct net_device *dev,
1081 int queue, u16 vlan, u8 qos);
1082 int (*ndo_set_vf_rate)(struct net_device *dev,
1083 int vf, int min_tx_rate,
1084 int max_tx_rate);
1085 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1086 int vf, bool setting);
1087 int (*ndo_get_vf_config)(struct net_device *dev,
1088 int vf,
1089 struct ifla_vf_info *ivf);
1090 int (*ndo_set_vf_link_state)(struct net_device *dev,
1091 int vf, int link_state);
1092 int (*ndo_set_vf_port)(struct net_device *dev,
1093 int vf,
1094 struct nlattr *port[]);
1095 int (*ndo_get_vf_port)(struct net_device *dev,
1096 int vf, struct sk_buff *skb);
1097 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1098 #if IS_ENABLED(CONFIG_FCOE)
1099 int (*ndo_fcoe_enable)(struct net_device *dev);
1100 int (*ndo_fcoe_disable)(struct net_device *dev);
1101 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1102 u16 xid,
1103 struct scatterlist *sgl,
1104 unsigned int sgc);
1105 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1106 u16 xid);
1107 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1108 u16 xid,
1109 struct scatterlist *sgl,
1110 unsigned int sgc);
1111 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1112 struct netdev_fcoe_hbainfo *hbainfo);
1113 #endif
1114
1115 #if IS_ENABLED(CONFIG_LIBFCOE)
1116 #define NETDEV_FCOE_WWNN 0
1117 #define NETDEV_FCOE_WWPN 1
1118 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1119 u64 *wwn, int type);
1120 #endif
1121
1122 #ifdef CONFIG_RFS_ACCEL
1123 int (*ndo_rx_flow_steer)(struct net_device *dev,
1124 const struct sk_buff *skb,
1125 u16 rxq_index,
1126 u32 flow_id);
1127 #endif
1128 int (*ndo_add_slave)(struct net_device *dev,
1129 struct net_device *slave_dev);
1130 int (*ndo_del_slave)(struct net_device *dev,
1131 struct net_device *slave_dev);
1132 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1133 netdev_features_t features);
1134 int (*ndo_set_features)(struct net_device *dev,
1135 netdev_features_t features);
1136 int (*ndo_neigh_construct)(struct neighbour *n);
1137 void (*ndo_neigh_destroy)(struct neighbour *n);
1138
1139 int (*ndo_fdb_add)(struct ndmsg *ndm,
1140 struct nlattr *tb[],
1141 struct net_device *dev,
1142 const unsigned char *addr,
1143 u16 vid,
1144 u16 flags);
1145 int (*ndo_fdb_del)(struct ndmsg *ndm,
1146 struct nlattr *tb[],
1147 struct net_device *dev,
1148 const unsigned char *addr,
1149 u16 vid);
1150 int (*ndo_fdb_dump)(struct sk_buff *skb,
1151 struct netlink_callback *cb,
1152 struct net_device *dev,
1153 struct net_device *filter_dev,
1154 int idx);
1155
1156 int (*ndo_bridge_setlink)(struct net_device *dev,
1157 struct nlmsghdr *nlh,
1158 u16 flags);
1159 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1160 u32 pid, u32 seq,
1161 struct net_device *dev,
1162 u32 filter_mask);
1163 int (*ndo_bridge_dellink)(struct net_device *dev,
1164 struct nlmsghdr *nlh,
1165 u16 flags);
1166 int (*ndo_change_carrier)(struct net_device *dev,
1167 bool new_carrier);
1168 int (*ndo_get_phys_port_id)(struct net_device *dev,
1169 struct netdev_phys_item_id *ppid);
1170 int (*ndo_get_phys_port_name)(struct net_device *dev,
1171 char *name, size_t len);
1172 void (*ndo_add_vxlan_port)(struct net_device *dev,
1173 sa_family_t sa_family,
1174 __be16 port);
1175 void (*ndo_del_vxlan_port)(struct net_device *dev,
1176 sa_family_t sa_family,
1177 __be16 port);
1178
1179 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1180 struct net_device *dev);
1181 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1182 void *priv);
1183
1184 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1185 struct net_device *dev,
1186 void *priv);
1187 int (*ndo_get_lock_subclass)(struct net_device *dev);
1188 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1189 struct net_device *dev,
1190 netdev_features_t features);
1191 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1192 int queue_index,
1193 u32 maxrate);
1194 };
1195
1196 /**
1197 * enum net_device_priv_flags - &struct net_device priv_flags
1198 *
1199 * These are the &struct net_device, they are only set internally
1200 * by drivers and used in the kernel. These flags are invisible to
1201 * userspace, this means that the order of these flags can change
1202 * during any kernel release.
1203 *
1204 * You should have a pretty good reason to be extending these flags.
1205 *
1206 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1207 * @IFF_EBRIDGE: Ethernet bridging device
1208 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1209 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1210 * @IFF_MASTER_ALB: bonding master, balance-alb
1211 * @IFF_BONDING: bonding master or slave
1212 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1213 * @IFF_ISATAP: ISATAP interface (RFC4214)
1214 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1215 * @IFF_WAN_HDLC: WAN HDLC device
1216 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1217 * release skb->dst
1218 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1219 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1220 * @IFF_MACVLAN_PORT: device used as macvlan port
1221 * @IFF_BRIDGE_PORT: device used as bridge port
1222 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1223 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1224 * @IFF_UNICAST_FLT: Supports unicast filtering
1225 * @IFF_TEAM_PORT: device used as team port
1226 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1227 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1228 * change when it's running
1229 * @IFF_MACVLAN: Macvlan device
1230 */
1231 enum netdev_priv_flags {
1232 IFF_802_1Q_VLAN = 1<<0,
1233 IFF_EBRIDGE = 1<<1,
1234 IFF_SLAVE_INACTIVE = 1<<2,
1235 IFF_MASTER_8023AD = 1<<3,
1236 IFF_MASTER_ALB = 1<<4,
1237 IFF_BONDING = 1<<5,
1238 IFF_SLAVE_NEEDARP = 1<<6,
1239 IFF_ISATAP = 1<<7,
1240 IFF_MASTER_ARPMON = 1<<8,
1241 IFF_WAN_HDLC = 1<<9,
1242 IFF_XMIT_DST_RELEASE = 1<<10,
1243 IFF_DONT_BRIDGE = 1<<11,
1244 IFF_DISABLE_NETPOLL = 1<<12,
1245 IFF_MACVLAN_PORT = 1<<13,
1246 IFF_BRIDGE_PORT = 1<<14,
1247 IFF_OVS_DATAPATH = 1<<15,
1248 IFF_TX_SKB_SHARING = 1<<16,
1249 IFF_UNICAST_FLT = 1<<17,
1250 IFF_TEAM_PORT = 1<<18,
1251 IFF_SUPP_NOFCS = 1<<19,
1252 IFF_LIVE_ADDR_CHANGE = 1<<20,
1253 IFF_MACVLAN = 1<<21,
1254 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1255 IFF_IPVLAN_MASTER = 1<<23,
1256 IFF_IPVLAN_SLAVE = 1<<24,
1257 };
1258
1259 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1260 #define IFF_EBRIDGE IFF_EBRIDGE
1261 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1262 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1263 #define IFF_MASTER_ALB IFF_MASTER_ALB
1264 #define IFF_BONDING IFF_BONDING
1265 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1266 #define IFF_ISATAP IFF_ISATAP
1267 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1268 #define IFF_WAN_HDLC IFF_WAN_HDLC
1269 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1270 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1271 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1272 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1273 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1274 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1275 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1276 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1277 #define IFF_TEAM_PORT IFF_TEAM_PORT
1278 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1279 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1280 #define IFF_MACVLAN IFF_MACVLAN
1281 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1282 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1283 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1284
1285 /**
1286 * struct net_device - The DEVICE structure.
1287 * Actually, this whole structure is a big mistake. It mixes I/O
1288 * data with strictly "high-level" data, and it has to know about
1289 * almost every data structure used in the INET module.
1290 *
1291 * @name: This is the first field of the "visible" part of this structure
1292 * (i.e. as seen by users in the "Space.c" file). It is the name
1293 * of the interface.
1294 *
1295 * @name_hlist: Device name hash chain, please keep it close to name[]
1296 * @ifalias: SNMP alias
1297 * @mem_end: Shared memory end
1298 * @mem_start: Shared memory start
1299 * @base_addr: Device I/O address
1300 * @irq: Device IRQ number
1301 *
1302 * @state: Generic network queuing layer state, see netdev_state_t
1303 * @dev_list: The global list of network devices
1304 * @napi_list: List entry, that is used for polling napi devices
1305 * @unreg_list: List entry, that is used, when we are unregistering the
1306 * device, see the function unregister_netdev
1307 * @close_list: List entry, that is used, when we are closing the device
1308 *
1309 * @adj_list: Directly linked devices, like slaves for bonding
1310 * @all_adj_list: All linked devices, *including* neighbours
1311 * @features: Currently active device features
1312 * @hw_features: User-changeable features
1313 *
1314 * @wanted_features: User-requested features
1315 * @vlan_features: Mask of features inheritable by VLAN devices
1316 *
1317 * @hw_enc_features: Mask of features inherited by encapsulating devices
1318 * This field indicates what encapsulation
1319 * offloads the hardware is capable of doing,
1320 * and drivers will need to set them appropriately.
1321 *
1322 * @mpls_features: Mask of features inheritable by MPLS
1323 *
1324 * @ifindex: interface index
1325 * @iflink: unique device identifier
1326 *
1327 * @stats: Statistics struct, which was left as a legacy, use
1328 * rtnl_link_stats64 instead
1329 *
1330 * @rx_dropped: Dropped packets by core network,
1331 * do not use this in drivers
1332 * @tx_dropped: Dropped packets by core network,
1333 * do not use this in drivers
1334 *
1335 * @carrier_changes: Stats to monitor carrier on<->off transitions
1336 *
1337 * @wireless_handlers: List of functions to handle Wireless Extensions,
1338 * instead of ioctl,
1339 * see <net/iw_handler.h> for details.
1340 * @wireless_data: Instance data managed by the core of wireless extensions
1341 *
1342 * @netdev_ops: Includes several pointers to callbacks,
1343 * if one wants to override the ndo_*() functions
1344 * @ethtool_ops: Management operations
1345 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1346 * of Layer 2 headers.
1347 *
1348 * @flags: Interface flags (a la BSD)
1349 * @priv_flags: Like 'flags' but invisible to userspace,
1350 * see if.h for the definitions
1351 * @gflags: Global flags ( kept as legacy )
1352 * @padded: How much padding added by alloc_netdev()
1353 * @operstate: RFC2863 operstate
1354 * @link_mode: Mapping policy to operstate
1355 * @if_port: Selectable AUI, TP, ...
1356 * @dma: DMA channel
1357 * @mtu: Interface MTU value
1358 * @type: Interface hardware type
1359 * @hard_header_len: Hardware header length
1360 *
1361 * @needed_headroom: Extra headroom the hardware may need, but not in all
1362 * cases can this be guaranteed
1363 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1364 * cases can this be guaranteed. Some cases also use
1365 * LL_MAX_HEADER instead to allocate the skb
1366 *
1367 * interface address info:
1368 *
1369 * @perm_addr: Permanent hw address
1370 * @addr_assign_type: Hw address assignment type
1371 * @addr_len: Hardware address length
1372 * @neigh_priv_len; Used in neigh_alloc(),
1373 * initialized only in atm/clip.c
1374 * @dev_id: Used to differentiate devices that share
1375 * the same link layer address
1376 * @dev_port: Used to differentiate devices that share
1377 * the same function
1378 * @addr_list_lock: XXX: need comments on this one
1379 * @uc: unicast mac addresses
1380 * @mc: multicast mac addresses
1381 * @dev_addrs: list of device hw addresses
1382 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1383 * @uc_promisc: Counter, that indicates, that promiscuous mode
1384 * has been enabled due to the need to listen to
1385 * additional unicast addresses in a device that
1386 * does not implement ndo_set_rx_mode()
1387 * @promiscuity: Number of times, the NIC is told to work in
1388 * Promiscuous mode, if it becomes 0 the NIC will
1389 * exit from working in Promiscuous mode
1390 * @allmulti: Counter, enables or disables allmulticast mode
1391 *
1392 * @vlan_info: VLAN info
1393 * @dsa_ptr: dsa specific data
1394 * @tipc_ptr: TIPC specific data
1395 * @atalk_ptr: AppleTalk link
1396 * @ip_ptr: IPv4 specific data
1397 * @dn_ptr: DECnet specific data
1398 * @ip6_ptr: IPv6 specific data
1399 * @ax25_ptr: AX.25 specific data
1400 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1401 *
1402 * @last_rx: Time of last Rx
1403 * @dev_addr: Hw address (before bcast,
1404 * because most packets are unicast)
1405 *
1406 * @_rx: Array of RX queues
1407 * @num_rx_queues: Number of RX queues
1408 * allocated at register_netdev() time
1409 * @real_num_rx_queues: Number of RX queues currently active in device
1410 *
1411 * @rx_handler: handler for received packets
1412 * @rx_handler_data: XXX: need comments on this one
1413 * @ingress_queue: XXX: need comments on this one
1414 * @broadcast: hw bcast address
1415 *
1416 * @_tx: Array of TX queues
1417 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1418 * @real_num_tx_queues: Number of TX queues currently active in device
1419 * @qdisc: Root qdisc from userspace point of view
1420 * @tx_queue_len: Max frames per queue allowed
1421 * @tx_global_lock: XXX: need comments on this one
1422 *
1423 * @xps_maps: XXX: need comments on this one
1424 *
1425 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1426 * indexed by RX queue number. Assigned by driver.
1427 * This must only be set if the ndo_rx_flow_steer
1428 * operation is defined
1429 *
1430 * @trans_start: Time (in jiffies) of last Tx
1431 * @watchdog_timeo: Represents the timeout that is used by
1432 * the watchdog ( see dev_watchdog() )
1433 * @watchdog_timer: List of timers
1434 *
1435 * @pcpu_refcnt: Number of references to this device
1436 * @todo_list: Delayed register/unregister
1437 * @index_hlist: Device index hash chain
1438 * @link_watch_list: XXX: need comments on this one
1439 *
1440 * @reg_state: Register/unregister state machine
1441 * @dismantle: Device is going to be freed
1442 * @rtnl_link_state: This enum represents the phases of creating
1443 * a new link
1444 *
1445 * @destructor: Called from unregister,
1446 * can be used to call free_netdev
1447 * @npinfo: XXX: need comments on this one
1448 * @nd_net: Network namespace this network device is inside
1449 *
1450 * @ml_priv: Mid-layer private
1451 * @lstats: Loopback statistics
1452 * @tstats: Tunnel statistics
1453 * @dstats: Dummy statistics
1454 * @vstats: Virtual ethernet statistics
1455 *
1456 * @garp_port: GARP
1457 * @mrp_port: MRP
1458 *
1459 * @dev: Class/net/name entry
1460 * @sysfs_groups: Space for optional device, statistics and wireless
1461 * sysfs groups
1462 *
1463 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1464 * @rtnl_link_ops: Rtnl_link_ops
1465 *
1466 * @gso_max_size: Maximum size of generic segmentation offload
1467 * @gso_max_segs: Maximum number of segments that can be passed to the
1468 * NIC for GSO
1469 * @gso_min_segs: Minimum number of segments that can be passed to the
1470 * NIC for GSO
1471 *
1472 * @dcbnl_ops: Data Center Bridging netlink ops
1473 * @num_tc: Number of traffic classes in the net device
1474 * @tc_to_txq: XXX: need comments on this one
1475 * @prio_tc_map XXX: need comments on this one
1476 *
1477 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1478 *
1479 * @priomap: XXX: need comments on this one
1480 * @phydev: Physical device may attach itself
1481 * for hardware timestamping
1482 *
1483 * @qdisc_tx_busylock: XXX: need comments on this one
1484 *
1485 * @group: The group, that the device belongs to
1486 * @pm_qos_req: Power Management QoS object
1487 *
1488 * FIXME: cleanup struct net_device such that network protocol info
1489 * moves out.
1490 */
1491
1492 struct net_device {
1493 char name[IFNAMSIZ];
1494 struct hlist_node name_hlist;
1495 char *ifalias;
1496 /*
1497 * I/O specific fields
1498 * FIXME: Merge these and struct ifmap into one
1499 */
1500 unsigned long mem_end;
1501 unsigned long mem_start;
1502 unsigned long base_addr;
1503 int irq;
1504
1505 /*
1506 * Some hardware also needs these fields (state,dev_list,
1507 * napi_list,unreg_list,close_list) but they are not
1508 * part of the usual set specified in Space.c.
1509 */
1510
1511 unsigned long state;
1512
1513 struct list_head dev_list;
1514 struct list_head napi_list;
1515 struct list_head unreg_list;
1516 struct list_head close_list;
1517 struct list_head ptype_all;
1518 struct list_head ptype_specific;
1519
1520 struct {
1521 struct list_head upper;
1522 struct list_head lower;
1523 } adj_list;
1524
1525 struct {
1526 struct list_head upper;
1527 struct list_head lower;
1528 } all_adj_list;
1529
1530 netdev_features_t features;
1531 netdev_features_t hw_features;
1532 netdev_features_t wanted_features;
1533 netdev_features_t vlan_features;
1534 netdev_features_t hw_enc_features;
1535 netdev_features_t mpls_features;
1536
1537 int ifindex;
1538 int iflink;
1539
1540 struct net_device_stats stats;
1541
1542 atomic_long_t rx_dropped;
1543 atomic_long_t tx_dropped;
1544
1545 atomic_t carrier_changes;
1546
1547 #ifdef CONFIG_WIRELESS_EXT
1548 const struct iw_handler_def * wireless_handlers;
1549 struct iw_public_data * wireless_data;
1550 #endif
1551 const struct net_device_ops *netdev_ops;
1552 const struct ethtool_ops *ethtool_ops;
1553 #ifdef CONFIG_NET_SWITCHDEV
1554 const struct swdev_ops *swdev_ops;
1555 #endif
1556
1557 const struct header_ops *header_ops;
1558
1559 unsigned int flags;
1560 unsigned int priv_flags;
1561
1562 unsigned short gflags;
1563 unsigned short padded;
1564
1565 unsigned char operstate;
1566 unsigned char link_mode;
1567
1568 unsigned char if_port;
1569 unsigned char dma;
1570
1571 unsigned int mtu;
1572 unsigned short type;
1573 unsigned short hard_header_len;
1574
1575 unsigned short needed_headroom;
1576 unsigned short needed_tailroom;
1577
1578 /* Interface address info. */
1579 unsigned char perm_addr[MAX_ADDR_LEN];
1580 unsigned char addr_assign_type;
1581 unsigned char addr_len;
1582 unsigned short neigh_priv_len;
1583 unsigned short dev_id;
1584 unsigned short dev_port;
1585 spinlock_t addr_list_lock;
1586 struct netdev_hw_addr_list uc;
1587 struct netdev_hw_addr_list mc;
1588 struct netdev_hw_addr_list dev_addrs;
1589
1590 #ifdef CONFIG_SYSFS
1591 struct kset *queues_kset;
1592 #endif
1593
1594 unsigned char name_assign_type;
1595
1596 bool uc_promisc;
1597 unsigned int promiscuity;
1598 unsigned int allmulti;
1599
1600
1601 /* Protocol specific pointers */
1602
1603 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1604 struct vlan_info __rcu *vlan_info;
1605 #endif
1606 #if IS_ENABLED(CONFIG_NET_DSA)
1607 struct dsa_switch_tree *dsa_ptr;
1608 #endif
1609 #if IS_ENABLED(CONFIG_TIPC)
1610 struct tipc_bearer __rcu *tipc_ptr;
1611 #endif
1612 void *atalk_ptr;
1613 struct in_device __rcu *ip_ptr;
1614 struct dn_dev __rcu *dn_ptr;
1615 struct inet6_dev __rcu *ip6_ptr;
1616 void *ax25_ptr;
1617 struct wireless_dev *ieee80211_ptr;
1618 struct wpan_dev *ieee802154_ptr;
1619
1620 /*
1621 * Cache lines mostly used on receive path (including eth_type_trans())
1622 */
1623 unsigned long last_rx;
1624
1625 /* Interface address info used in eth_type_trans() */
1626 unsigned char *dev_addr;
1627
1628
1629 #ifdef CONFIG_SYSFS
1630 struct netdev_rx_queue *_rx;
1631
1632 unsigned int num_rx_queues;
1633 unsigned int real_num_rx_queues;
1634
1635 #endif
1636
1637 unsigned long gro_flush_timeout;
1638 rx_handler_func_t __rcu *rx_handler;
1639 void __rcu *rx_handler_data;
1640
1641 struct netdev_queue __rcu *ingress_queue;
1642 unsigned char broadcast[MAX_ADDR_LEN];
1643
1644
1645 /*
1646 * Cache lines mostly used on transmit path
1647 */
1648 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1649 unsigned int num_tx_queues;
1650 unsigned int real_num_tx_queues;
1651 struct Qdisc *qdisc;
1652 unsigned long tx_queue_len;
1653 spinlock_t tx_global_lock;
1654
1655 #ifdef CONFIG_XPS
1656 struct xps_dev_maps __rcu *xps_maps;
1657 #endif
1658 #ifdef CONFIG_RFS_ACCEL
1659 struct cpu_rmap *rx_cpu_rmap;
1660 #endif
1661
1662 /* These may be needed for future network-power-down code. */
1663
1664 /*
1665 * trans_start here is expensive for high speed devices on SMP,
1666 * please use netdev_queue->trans_start instead.
1667 */
1668 unsigned long trans_start;
1669
1670 int watchdog_timeo;
1671 struct timer_list watchdog_timer;
1672
1673 int __percpu *pcpu_refcnt;
1674 struct list_head todo_list;
1675
1676 struct hlist_node index_hlist;
1677 struct list_head link_watch_list;
1678
1679 enum { NETREG_UNINITIALIZED=0,
1680 NETREG_REGISTERED, /* completed register_netdevice */
1681 NETREG_UNREGISTERING, /* called unregister_netdevice */
1682 NETREG_UNREGISTERED, /* completed unregister todo */
1683 NETREG_RELEASED, /* called free_netdev */
1684 NETREG_DUMMY, /* dummy device for NAPI poll */
1685 } reg_state:8;
1686
1687 bool dismantle;
1688
1689 enum {
1690 RTNL_LINK_INITIALIZED,
1691 RTNL_LINK_INITIALIZING,
1692 } rtnl_link_state:16;
1693
1694 void (*destructor)(struct net_device *dev);
1695
1696 #ifdef CONFIG_NETPOLL
1697 struct netpoll_info __rcu *npinfo;
1698 #endif
1699
1700 possible_net_t nd_net;
1701
1702 /* mid-layer private */
1703 union {
1704 void *ml_priv;
1705 struct pcpu_lstats __percpu *lstats;
1706 struct pcpu_sw_netstats __percpu *tstats;
1707 struct pcpu_dstats __percpu *dstats;
1708 struct pcpu_vstats __percpu *vstats;
1709 };
1710
1711 struct garp_port __rcu *garp_port;
1712 struct mrp_port __rcu *mrp_port;
1713
1714 struct device dev;
1715 const struct attribute_group *sysfs_groups[4];
1716 const struct attribute_group *sysfs_rx_queue_group;
1717
1718 const struct rtnl_link_ops *rtnl_link_ops;
1719
1720 /* for setting kernel sock attribute on TCP connection setup */
1721 #define GSO_MAX_SIZE 65536
1722 unsigned int gso_max_size;
1723 #define GSO_MAX_SEGS 65535
1724 u16 gso_max_segs;
1725 u16 gso_min_segs;
1726 #ifdef CONFIG_DCB
1727 const struct dcbnl_rtnl_ops *dcbnl_ops;
1728 #endif
1729 u8 num_tc;
1730 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1731 u8 prio_tc_map[TC_BITMASK + 1];
1732
1733 #if IS_ENABLED(CONFIG_FCOE)
1734 unsigned int fcoe_ddp_xid;
1735 #endif
1736 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1737 struct netprio_map __rcu *priomap;
1738 #endif
1739 struct phy_device *phydev;
1740 struct lock_class_key *qdisc_tx_busylock;
1741 int group;
1742 struct pm_qos_request pm_qos_req;
1743 };
1744 #define to_net_dev(d) container_of(d, struct net_device, dev)
1745
1746 #define NETDEV_ALIGN 32
1747
1748 static inline
1749 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1750 {
1751 return dev->prio_tc_map[prio & TC_BITMASK];
1752 }
1753
1754 static inline
1755 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1756 {
1757 if (tc >= dev->num_tc)
1758 return -EINVAL;
1759
1760 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1761 return 0;
1762 }
1763
1764 static inline
1765 void netdev_reset_tc(struct net_device *dev)
1766 {
1767 dev->num_tc = 0;
1768 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1769 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1770 }
1771
1772 static inline
1773 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1774 {
1775 if (tc >= dev->num_tc)
1776 return -EINVAL;
1777
1778 dev->tc_to_txq[tc].count = count;
1779 dev->tc_to_txq[tc].offset = offset;
1780 return 0;
1781 }
1782
1783 static inline
1784 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1785 {
1786 if (num_tc > TC_MAX_QUEUE)
1787 return -EINVAL;
1788
1789 dev->num_tc = num_tc;
1790 return 0;
1791 }
1792
1793 static inline
1794 int netdev_get_num_tc(struct net_device *dev)
1795 {
1796 return dev->num_tc;
1797 }
1798
1799 static inline
1800 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1801 unsigned int index)
1802 {
1803 return &dev->_tx[index];
1804 }
1805
1806 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1807 const struct sk_buff *skb)
1808 {
1809 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1810 }
1811
1812 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1813 void (*f)(struct net_device *,
1814 struct netdev_queue *,
1815 void *),
1816 void *arg)
1817 {
1818 unsigned int i;
1819
1820 for (i = 0; i < dev->num_tx_queues; i++)
1821 f(dev, &dev->_tx[i], arg);
1822 }
1823
1824 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1825 struct sk_buff *skb,
1826 void *accel_priv);
1827
1828 /*
1829 * Net namespace inlines
1830 */
1831 static inline
1832 struct net *dev_net(const struct net_device *dev)
1833 {
1834 return read_pnet(&dev->nd_net);
1835 }
1836
1837 static inline
1838 void dev_net_set(struct net_device *dev, struct net *net)
1839 {
1840 write_pnet(&dev->nd_net, net);
1841 }
1842
1843 static inline bool netdev_uses_dsa(struct net_device *dev)
1844 {
1845 #if IS_ENABLED(CONFIG_NET_DSA)
1846 if (dev->dsa_ptr != NULL)
1847 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1848 #endif
1849 return false;
1850 }
1851
1852 /**
1853 * netdev_priv - access network device private data
1854 * @dev: network device
1855 *
1856 * Get network device private data
1857 */
1858 static inline void *netdev_priv(const struct net_device *dev)
1859 {
1860 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1861 }
1862
1863 /* Set the sysfs physical device reference for the network logical device
1864 * if set prior to registration will cause a symlink during initialization.
1865 */
1866 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1867
1868 /* Set the sysfs device type for the network logical device to allow
1869 * fine-grained identification of different network device types. For
1870 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1871 */
1872 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1873
1874 /* Default NAPI poll() weight
1875 * Device drivers are strongly advised to not use bigger value
1876 */
1877 #define NAPI_POLL_WEIGHT 64
1878
1879 /**
1880 * netif_napi_add - initialize a napi context
1881 * @dev: network device
1882 * @napi: napi context
1883 * @poll: polling function
1884 * @weight: default weight
1885 *
1886 * netif_napi_add() must be used to initialize a napi context prior to calling
1887 * *any* of the other napi related functions.
1888 */
1889 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1890 int (*poll)(struct napi_struct *, int), int weight);
1891
1892 /**
1893 * netif_napi_del - remove a napi context
1894 * @napi: napi context
1895 *
1896 * netif_napi_del() removes a napi context from the network device napi list
1897 */
1898 void netif_napi_del(struct napi_struct *napi);
1899
1900 struct napi_gro_cb {
1901 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1902 void *frag0;
1903
1904 /* Length of frag0. */
1905 unsigned int frag0_len;
1906
1907 /* This indicates where we are processing relative to skb->data. */
1908 int data_offset;
1909
1910 /* This is non-zero if the packet cannot be merged with the new skb. */
1911 u16 flush;
1912
1913 /* Save the IP ID here and check when we get to the transport layer */
1914 u16 flush_id;
1915
1916 /* Number of segments aggregated. */
1917 u16 count;
1918
1919 /* Start offset for remote checksum offload */
1920 u16 gro_remcsum_start;
1921
1922 /* jiffies when first packet was created/queued */
1923 unsigned long age;
1924
1925 /* Used in ipv6_gro_receive() and foo-over-udp */
1926 u16 proto;
1927
1928 /* This is non-zero if the packet may be of the same flow. */
1929 u8 same_flow:1;
1930
1931 /* Used in udp_gro_receive */
1932 u8 udp_mark:1;
1933
1934 /* GRO checksum is valid */
1935 u8 csum_valid:1;
1936
1937 /* Number of checksums via CHECKSUM_UNNECESSARY */
1938 u8 csum_cnt:3;
1939
1940 /* Free the skb? */
1941 u8 free:2;
1942 #define NAPI_GRO_FREE 1
1943 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1944
1945 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1946 u8 is_ipv6:1;
1947
1948 /* 7 bit hole */
1949
1950 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1951 __wsum csum;
1952
1953 /* used in skb_gro_receive() slow path */
1954 struct sk_buff *last;
1955 };
1956
1957 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1958
1959 struct packet_type {
1960 __be16 type; /* This is really htons(ether_type). */
1961 struct net_device *dev; /* NULL is wildcarded here */
1962 int (*func) (struct sk_buff *,
1963 struct net_device *,
1964 struct packet_type *,
1965 struct net_device *);
1966 bool (*id_match)(struct packet_type *ptype,
1967 struct sock *sk);
1968 void *af_packet_priv;
1969 struct list_head list;
1970 };
1971
1972 struct offload_callbacks {
1973 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1974 netdev_features_t features);
1975 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1976 struct sk_buff *skb);
1977 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1978 };
1979
1980 struct packet_offload {
1981 __be16 type; /* This is really htons(ether_type). */
1982 struct offload_callbacks callbacks;
1983 struct list_head list;
1984 };
1985
1986 struct udp_offload;
1987
1988 struct udp_offload_callbacks {
1989 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1990 struct sk_buff *skb,
1991 struct udp_offload *uoff);
1992 int (*gro_complete)(struct sk_buff *skb,
1993 int nhoff,
1994 struct udp_offload *uoff);
1995 };
1996
1997 struct udp_offload {
1998 __be16 port;
1999 u8 ipproto;
2000 struct udp_offload_callbacks callbacks;
2001 };
2002
2003 /* often modified stats are per cpu, other are shared (netdev->stats) */
2004 struct pcpu_sw_netstats {
2005 u64 rx_packets;
2006 u64 rx_bytes;
2007 u64 tx_packets;
2008 u64 tx_bytes;
2009 struct u64_stats_sync syncp;
2010 };
2011
2012 #define netdev_alloc_pcpu_stats(type) \
2013 ({ \
2014 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2015 if (pcpu_stats) { \
2016 int i; \
2017 for_each_possible_cpu(i) { \
2018 typeof(type) *stat; \
2019 stat = per_cpu_ptr(pcpu_stats, i); \
2020 u64_stats_init(&stat->syncp); \
2021 } \
2022 } \
2023 pcpu_stats; \
2024 })
2025
2026 #include <linux/notifier.h>
2027
2028 /* netdevice notifier chain. Please remember to update the rtnetlink
2029 * notification exclusion list in rtnetlink_event() when adding new
2030 * types.
2031 */
2032 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2033 #define NETDEV_DOWN 0x0002
2034 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2035 detected a hardware crash and restarted
2036 - we can use this eg to kick tcp sessions
2037 once done */
2038 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2039 #define NETDEV_REGISTER 0x0005
2040 #define NETDEV_UNREGISTER 0x0006
2041 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2042 #define NETDEV_CHANGEADDR 0x0008
2043 #define NETDEV_GOING_DOWN 0x0009
2044 #define NETDEV_CHANGENAME 0x000A
2045 #define NETDEV_FEAT_CHANGE 0x000B
2046 #define NETDEV_BONDING_FAILOVER 0x000C
2047 #define NETDEV_PRE_UP 0x000D
2048 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2049 #define NETDEV_POST_TYPE_CHANGE 0x000F
2050 #define NETDEV_POST_INIT 0x0010
2051 #define NETDEV_UNREGISTER_FINAL 0x0011
2052 #define NETDEV_RELEASE 0x0012
2053 #define NETDEV_NOTIFY_PEERS 0x0013
2054 #define NETDEV_JOIN 0x0014
2055 #define NETDEV_CHANGEUPPER 0x0015
2056 #define NETDEV_RESEND_IGMP 0x0016
2057 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2058 #define NETDEV_CHANGEINFODATA 0x0018
2059 #define NETDEV_BONDING_INFO 0x0019
2060
2061 int register_netdevice_notifier(struct notifier_block *nb);
2062 int unregister_netdevice_notifier(struct notifier_block *nb);
2063
2064 struct netdev_notifier_info {
2065 struct net_device *dev;
2066 };
2067
2068 struct netdev_notifier_change_info {
2069 struct netdev_notifier_info info; /* must be first */
2070 unsigned int flags_changed;
2071 };
2072
2073 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2074 struct net_device *dev)
2075 {
2076 info->dev = dev;
2077 }
2078
2079 static inline struct net_device *
2080 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2081 {
2082 return info->dev;
2083 }
2084
2085 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2086
2087
2088 extern rwlock_t dev_base_lock; /* Device list lock */
2089
2090 #define for_each_netdev(net, d) \
2091 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2092 #define for_each_netdev_reverse(net, d) \
2093 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2094 #define for_each_netdev_rcu(net, d) \
2095 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2096 #define for_each_netdev_safe(net, d, n) \
2097 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2098 #define for_each_netdev_continue(net, d) \
2099 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2100 #define for_each_netdev_continue_rcu(net, d) \
2101 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2102 #define for_each_netdev_in_bond_rcu(bond, slave) \
2103 for_each_netdev_rcu(&init_net, slave) \
2104 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2105 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2106
2107 static inline struct net_device *next_net_device(struct net_device *dev)
2108 {
2109 struct list_head *lh;
2110 struct net *net;
2111
2112 net = dev_net(dev);
2113 lh = dev->dev_list.next;
2114 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2115 }
2116
2117 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2118 {
2119 struct list_head *lh;
2120 struct net *net;
2121
2122 net = dev_net(dev);
2123 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2124 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2125 }
2126
2127 static inline struct net_device *first_net_device(struct net *net)
2128 {
2129 return list_empty(&net->dev_base_head) ? NULL :
2130 net_device_entry(net->dev_base_head.next);
2131 }
2132
2133 static inline struct net_device *first_net_device_rcu(struct net *net)
2134 {
2135 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2136
2137 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2138 }
2139
2140 int netdev_boot_setup_check(struct net_device *dev);
2141 unsigned long netdev_boot_base(const char *prefix, int unit);
2142 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2143 const char *hwaddr);
2144 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2145 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2146 void dev_add_pack(struct packet_type *pt);
2147 void dev_remove_pack(struct packet_type *pt);
2148 void __dev_remove_pack(struct packet_type *pt);
2149 void dev_add_offload(struct packet_offload *po);
2150 void dev_remove_offload(struct packet_offload *po);
2151
2152 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2153 unsigned short mask);
2154 struct net_device *dev_get_by_name(struct net *net, const char *name);
2155 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2156 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2157 int dev_alloc_name(struct net_device *dev, const char *name);
2158 int dev_open(struct net_device *dev);
2159 int dev_close(struct net_device *dev);
2160 int dev_close_many(struct list_head *head, bool unlink);
2161 void dev_disable_lro(struct net_device *dev);
2162 int dev_loopback_xmit(struct sk_buff *newskb);
2163 int dev_queue_xmit(struct sk_buff *skb);
2164 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2165 int register_netdevice(struct net_device *dev);
2166 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2167 void unregister_netdevice_many(struct list_head *head);
2168 static inline void unregister_netdevice(struct net_device *dev)
2169 {
2170 unregister_netdevice_queue(dev, NULL);
2171 }
2172
2173 int netdev_refcnt_read(const struct net_device *dev);
2174 void free_netdev(struct net_device *dev);
2175 void netdev_freemem(struct net_device *dev);
2176 void synchronize_net(void);
2177 int init_dummy_netdev(struct net_device *dev);
2178
2179 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2180 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2181 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2182 int netdev_get_name(struct net *net, char *name, int ifindex);
2183 int dev_restart(struct net_device *dev);
2184 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2185
2186 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2187 {
2188 return NAPI_GRO_CB(skb)->data_offset;
2189 }
2190
2191 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2192 {
2193 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2194 }
2195
2196 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2197 {
2198 NAPI_GRO_CB(skb)->data_offset += len;
2199 }
2200
2201 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2202 unsigned int offset)
2203 {
2204 return NAPI_GRO_CB(skb)->frag0 + offset;
2205 }
2206
2207 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2208 {
2209 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2210 }
2211
2212 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2213 unsigned int offset)
2214 {
2215 if (!pskb_may_pull(skb, hlen))
2216 return NULL;
2217
2218 NAPI_GRO_CB(skb)->frag0 = NULL;
2219 NAPI_GRO_CB(skb)->frag0_len = 0;
2220 return skb->data + offset;
2221 }
2222
2223 static inline void *skb_gro_network_header(struct sk_buff *skb)
2224 {
2225 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2226 skb_network_offset(skb);
2227 }
2228
2229 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2230 const void *start, unsigned int len)
2231 {
2232 if (NAPI_GRO_CB(skb)->csum_valid)
2233 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2234 csum_partial(start, len, 0));
2235 }
2236
2237 /* GRO checksum functions. These are logical equivalents of the normal
2238 * checksum functions (in skbuff.h) except that they operate on the GRO
2239 * offsets and fields in sk_buff.
2240 */
2241
2242 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2243
2244 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2245 {
2246 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2247 skb_gro_offset(skb));
2248 }
2249
2250 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2251 bool zero_okay,
2252 __sum16 check)
2253 {
2254 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2255 skb_checksum_start_offset(skb) <
2256 skb_gro_offset(skb)) &&
2257 !skb_at_gro_remcsum_start(skb) &&
2258 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2259 (!zero_okay || check));
2260 }
2261
2262 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2263 __wsum psum)
2264 {
2265 if (NAPI_GRO_CB(skb)->csum_valid &&
2266 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2267 return 0;
2268
2269 NAPI_GRO_CB(skb)->csum = psum;
2270
2271 return __skb_gro_checksum_complete(skb);
2272 }
2273
2274 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2275 {
2276 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2277 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2278 NAPI_GRO_CB(skb)->csum_cnt--;
2279 } else {
2280 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2281 * verified a new top level checksum or an encapsulated one
2282 * during GRO. This saves work if we fallback to normal path.
2283 */
2284 __skb_incr_checksum_unnecessary(skb);
2285 }
2286 }
2287
2288 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2289 compute_pseudo) \
2290 ({ \
2291 __sum16 __ret = 0; \
2292 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2293 __ret = __skb_gro_checksum_validate_complete(skb, \
2294 compute_pseudo(skb, proto)); \
2295 if (__ret) \
2296 __skb_mark_checksum_bad(skb); \
2297 else \
2298 skb_gro_incr_csum_unnecessary(skb); \
2299 __ret; \
2300 })
2301
2302 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2303 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2304
2305 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2306 compute_pseudo) \
2307 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2308
2309 #define skb_gro_checksum_simple_validate(skb) \
2310 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2311
2312 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2313 {
2314 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2315 !NAPI_GRO_CB(skb)->csum_valid);
2316 }
2317
2318 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2319 __sum16 check, __wsum pseudo)
2320 {
2321 NAPI_GRO_CB(skb)->csum = ~pseudo;
2322 NAPI_GRO_CB(skb)->csum_valid = 1;
2323 }
2324
2325 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2326 do { \
2327 if (__skb_gro_checksum_convert_check(skb)) \
2328 __skb_gro_checksum_convert(skb, check, \
2329 compute_pseudo(skb, proto)); \
2330 } while (0)
2331
2332 struct gro_remcsum {
2333 int offset;
2334 __wsum delta;
2335 };
2336
2337 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2338 {
2339 grc->offset = 0;
2340 grc->delta = 0;
2341 }
2342
2343 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2344 int start, int offset,
2345 struct gro_remcsum *grc,
2346 bool nopartial)
2347 {
2348 __wsum delta;
2349
2350 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2351
2352 if (!nopartial) {
2353 NAPI_GRO_CB(skb)->gro_remcsum_start =
2354 ((unsigned char *)ptr + start) - skb->head;
2355 return;
2356 }
2357
2358 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2359
2360 /* Adjust skb->csum since we changed the packet */
2361 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2362
2363 grc->offset = (ptr + offset) - (void *)skb->head;
2364 grc->delta = delta;
2365 }
2366
2367 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2368 struct gro_remcsum *grc)
2369 {
2370 if (!grc->delta)
2371 return;
2372
2373 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2374 }
2375
2376 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2377 unsigned short type,
2378 const void *daddr, const void *saddr,
2379 unsigned int len)
2380 {
2381 if (!dev->header_ops || !dev->header_ops->create)
2382 return 0;
2383
2384 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2385 }
2386
2387 static inline int dev_parse_header(const struct sk_buff *skb,
2388 unsigned char *haddr)
2389 {
2390 const struct net_device *dev = skb->dev;
2391
2392 if (!dev->header_ops || !dev->header_ops->parse)
2393 return 0;
2394 return dev->header_ops->parse(skb, haddr);
2395 }
2396
2397 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2398 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2399 static inline int unregister_gifconf(unsigned int family)
2400 {
2401 return register_gifconf(family, NULL);
2402 }
2403
2404 #ifdef CONFIG_NET_FLOW_LIMIT
2405 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2406 struct sd_flow_limit {
2407 u64 count;
2408 unsigned int num_buckets;
2409 unsigned int history_head;
2410 u16 history[FLOW_LIMIT_HISTORY];
2411 u8 buckets[];
2412 };
2413
2414 extern int netdev_flow_limit_table_len;
2415 #endif /* CONFIG_NET_FLOW_LIMIT */
2416
2417 /*
2418 * Incoming packets are placed on per-cpu queues
2419 */
2420 struct softnet_data {
2421 struct list_head poll_list;
2422 struct sk_buff_head process_queue;
2423
2424 /* stats */
2425 unsigned int processed;
2426 unsigned int time_squeeze;
2427 unsigned int cpu_collision;
2428 unsigned int received_rps;
2429 #ifdef CONFIG_RPS
2430 struct softnet_data *rps_ipi_list;
2431 #endif
2432 #ifdef CONFIG_NET_FLOW_LIMIT
2433 struct sd_flow_limit __rcu *flow_limit;
2434 #endif
2435 struct Qdisc *output_queue;
2436 struct Qdisc **output_queue_tailp;
2437 struct sk_buff *completion_queue;
2438
2439 #ifdef CONFIG_RPS
2440 /* Elements below can be accessed between CPUs for RPS */
2441 struct call_single_data csd ____cacheline_aligned_in_smp;
2442 struct softnet_data *rps_ipi_next;
2443 unsigned int cpu;
2444 unsigned int input_queue_head;
2445 unsigned int input_queue_tail;
2446 #endif
2447 unsigned int dropped;
2448 struct sk_buff_head input_pkt_queue;
2449 struct napi_struct backlog;
2450
2451 };
2452
2453 static inline void input_queue_head_incr(struct softnet_data *sd)
2454 {
2455 #ifdef CONFIG_RPS
2456 sd->input_queue_head++;
2457 #endif
2458 }
2459
2460 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2461 unsigned int *qtail)
2462 {
2463 #ifdef CONFIG_RPS
2464 *qtail = ++sd->input_queue_tail;
2465 #endif
2466 }
2467
2468 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2469
2470 void __netif_schedule(struct Qdisc *q);
2471 void netif_schedule_queue(struct netdev_queue *txq);
2472
2473 static inline void netif_tx_schedule_all(struct net_device *dev)
2474 {
2475 unsigned int i;
2476
2477 for (i = 0; i < dev->num_tx_queues; i++)
2478 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2479 }
2480
2481 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2482 {
2483 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2484 }
2485
2486 /**
2487 * netif_start_queue - allow transmit
2488 * @dev: network device
2489 *
2490 * Allow upper layers to call the device hard_start_xmit routine.
2491 */
2492 static inline void netif_start_queue(struct net_device *dev)
2493 {
2494 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2495 }
2496
2497 static inline void netif_tx_start_all_queues(struct net_device *dev)
2498 {
2499 unsigned int i;
2500
2501 for (i = 0; i < dev->num_tx_queues; i++) {
2502 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2503 netif_tx_start_queue(txq);
2504 }
2505 }
2506
2507 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2508
2509 /**
2510 * netif_wake_queue - restart transmit
2511 * @dev: network device
2512 *
2513 * Allow upper layers to call the device hard_start_xmit routine.
2514 * Used for flow control when transmit resources are available.
2515 */
2516 static inline void netif_wake_queue(struct net_device *dev)
2517 {
2518 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2519 }
2520
2521 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2522 {
2523 unsigned int i;
2524
2525 for (i = 0; i < dev->num_tx_queues; i++) {
2526 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2527 netif_tx_wake_queue(txq);
2528 }
2529 }
2530
2531 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2532 {
2533 if (WARN_ON(!dev_queue)) {
2534 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2535 return;
2536 }
2537 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2538 }
2539
2540 /**
2541 * netif_stop_queue - stop transmitted packets
2542 * @dev: network device
2543 *
2544 * Stop upper layers calling the device hard_start_xmit routine.
2545 * Used for flow control when transmit resources are unavailable.
2546 */
2547 static inline void netif_stop_queue(struct net_device *dev)
2548 {
2549 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2550 }
2551
2552 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2553 {
2554 unsigned int i;
2555
2556 for (i = 0; i < dev->num_tx_queues; i++) {
2557 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2558 netif_tx_stop_queue(txq);
2559 }
2560 }
2561
2562 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2563 {
2564 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2565 }
2566
2567 /**
2568 * netif_queue_stopped - test if transmit queue is flowblocked
2569 * @dev: network device
2570 *
2571 * Test if transmit queue on device is currently unable to send.
2572 */
2573 static inline bool netif_queue_stopped(const struct net_device *dev)
2574 {
2575 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2576 }
2577
2578 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2579 {
2580 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2581 }
2582
2583 static inline bool
2584 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2585 {
2586 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2587 }
2588
2589 static inline bool
2590 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2591 {
2592 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2593 }
2594
2595 /**
2596 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2597 * @dev_queue: pointer to transmit queue
2598 *
2599 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2600 * to give appropriate hint to the cpu.
2601 */
2602 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2603 {
2604 #ifdef CONFIG_BQL
2605 prefetchw(&dev_queue->dql.num_queued);
2606 #endif
2607 }
2608
2609 /**
2610 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2611 * @dev_queue: pointer to transmit queue
2612 *
2613 * BQL enabled drivers might use this helper in their TX completion path,
2614 * to give appropriate hint to the cpu.
2615 */
2616 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2617 {
2618 #ifdef CONFIG_BQL
2619 prefetchw(&dev_queue->dql.limit);
2620 #endif
2621 }
2622
2623 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2624 unsigned int bytes)
2625 {
2626 #ifdef CONFIG_BQL
2627 dql_queued(&dev_queue->dql, bytes);
2628
2629 if (likely(dql_avail(&dev_queue->dql) >= 0))
2630 return;
2631
2632 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2633
2634 /*
2635 * The XOFF flag must be set before checking the dql_avail below,
2636 * because in netdev_tx_completed_queue we update the dql_completed
2637 * before checking the XOFF flag.
2638 */
2639 smp_mb();
2640
2641 /* check again in case another CPU has just made room avail */
2642 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2643 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2644 #endif
2645 }
2646
2647 /**
2648 * netdev_sent_queue - report the number of bytes queued to hardware
2649 * @dev: network device
2650 * @bytes: number of bytes queued to the hardware device queue
2651 *
2652 * Report the number of bytes queued for sending/completion to the network
2653 * device hardware queue. @bytes should be a good approximation and should
2654 * exactly match netdev_completed_queue() @bytes
2655 */
2656 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2657 {
2658 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2659 }
2660
2661 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2662 unsigned int pkts, unsigned int bytes)
2663 {
2664 #ifdef CONFIG_BQL
2665 if (unlikely(!bytes))
2666 return;
2667
2668 dql_completed(&dev_queue->dql, bytes);
2669
2670 /*
2671 * Without the memory barrier there is a small possiblity that
2672 * netdev_tx_sent_queue will miss the update and cause the queue to
2673 * be stopped forever
2674 */
2675 smp_mb();
2676
2677 if (dql_avail(&dev_queue->dql) < 0)
2678 return;
2679
2680 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2681 netif_schedule_queue(dev_queue);
2682 #endif
2683 }
2684
2685 /**
2686 * netdev_completed_queue - report bytes and packets completed by device
2687 * @dev: network device
2688 * @pkts: actual number of packets sent over the medium
2689 * @bytes: actual number of bytes sent over the medium
2690 *
2691 * Report the number of bytes and packets transmitted by the network device
2692 * hardware queue over the physical medium, @bytes must exactly match the
2693 * @bytes amount passed to netdev_sent_queue()
2694 */
2695 static inline void netdev_completed_queue(struct net_device *dev,
2696 unsigned int pkts, unsigned int bytes)
2697 {
2698 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2699 }
2700
2701 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2702 {
2703 #ifdef CONFIG_BQL
2704 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2705 dql_reset(&q->dql);
2706 #endif
2707 }
2708
2709 /**
2710 * netdev_reset_queue - reset the packets and bytes count of a network device
2711 * @dev_queue: network device
2712 *
2713 * Reset the bytes and packet count of a network device and clear the
2714 * software flow control OFF bit for this network device
2715 */
2716 static inline void netdev_reset_queue(struct net_device *dev_queue)
2717 {
2718 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2719 }
2720
2721 /**
2722 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2723 * @dev: network device
2724 * @queue_index: given tx queue index
2725 *
2726 * Returns 0 if given tx queue index >= number of device tx queues,
2727 * otherwise returns the originally passed tx queue index.
2728 */
2729 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2730 {
2731 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2732 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2733 dev->name, queue_index,
2734 dev->real_num_tx_queues);
2735 return 0;
2736 }
2737
2738 return queue_index;
2739 }
2740
2741 /**
2742 * netif_running - test if up
2743 * @dev: network device
2744 *
2745 * Test if the device has been brought up.
2746 */
2747 static inline bool netif_running(const struct net_device *dev)
2748 {
2749 return test_bit(__LINK_STATE_START, &dev->state);
2750 }
2751
2752 /*
2753 * Routines to manage the subqueues on a device. We only need start
2754 * stop, and a check if it's stopped. All other device management is
2755 * done at the overall netdevice level.
2756 * Also test the device if we're multiqueue.
2757 */
2758
2759 /**
2760 * netif_start_subqueue - allow sending packets on subqueue
2761 * @dev: network device
2762 * @queue_index: sub queue index
2763 *
2764 * Start individual transmit queue of a device with multiple transmit queues.
2765 */
2766 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2767 {
2768 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2769
2770 netif_tx_start_queue(txq);
2771 }
2772
2773 /**
2774 * netif_stop_subqueue - stop sending packets on subqueue
2775 * @dev: network device
2776 * @queue_index: sub queue index
2777 *
2778 * Stop individual transmit queue of a device with multiple transmit queues.
2779 */
2780 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2781 {
2782 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2783 netif_tx_stop_queue(txq);
2784 }
2785
2786 /**
2787 * netif_subqueue_stopped - test status of subqueue
2788 * @dev: network device
2789 * @queue_index: sub queue index
2790 *
2791 * Check individual transmit queue of a device with multiple transmit queues.
2792 */
2793 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2794 u16 queue_index)
2795 {
2796 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2797
2798 return netif_tx_queue_stopped(txq);
2799 }
2800
2801 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2802 struct sk_buff *skb)
2803 {
2804 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2805 }
2806
2807 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2808
2809 #ifdef CONFIG_XPS
2810 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2811 u16 index);
2812 #else
2813 static inline int netif_set_xps_queue(struct net_device *dev,
2814 const struct cpumask *mask,
2815 u16 index)
2816 {
2817 return 0;
2818 }
2819 #endif
2820
2821 /*
2822 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2823 * as a distribution range limit for the returned value.
2824 */
2825 static inline u16 skb_tx_hash(const struct net_device *dev,
2826 struct sk_buff *skb)
2827 {
2828 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2829 }
2830
2831 /**
2832 * netif_is_multiqueue - test if device has multiple transmit queues
2833 * @dev: network device
2834 *
2835 * Check if device has multiple transmit queues
2836 */
2837 static inline bool netif_is_multiqueue(const struct net_device *dev)
2838 {
2839 return dev->num_tx_queues > 1;
2840 }
2841
2842 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2843
2844 #ifdef CONFIG_SYSFS
2845 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2846 #else
2847 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2848 unsigned int rxq)
2849 {
2850 return 0;
2851 }
2852 #endif
2853
2854 #ifdef CONFIG_SYSFS
2855 static inline unsigned int get_netdev_rx_queue_index(
2856 struct netdev_rx_queue *queue)
2857 {
2858 struct net_device *dev = queue->dev;
2859 int index = queue - dev->_rx;
2860
2861 BUG_ON(index >= dev->num_rx_queues);
2862 return index;
2863 }
2864 #endif
2865
2866 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2867 int netif_get_num_default_rss_queues(void);
2868
2869 enum skb_free_reason {
2870 SKB_REASON_CONSUMED,
2871 SKB_REASON_DROPPED,
2872 };
2873
2874 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2875 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2876
2877 /*
2878 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2879 * interrupt context or with hardware interrupts being disabled.
2880 * (in_irq() || irqs_disabled())
2881 *
2882 * We provide four helpers that can be used in following contexts :
2883 *
2884 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2885 * replacing kfree_skb(skb)
2886 *
2887 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2888 * Typically used in place of consume_skb(skb) in TX completion path
2889 *
2890 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2891 * replacing kfree_skb(skb)
2892 *
2893 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2894 * and consumed a packet. Used in place of consume_skb(skb)
2895 */
2896 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2897 {
2898 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2899 }
2900
2901 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2902 {
2903 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2904 }
2905
2906 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2907 {
2908 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2909 }
2910
2911 static inline void dev_consume_skb_any(struct sk_buff *skb)
2912 {
2913 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2914 }
2915
2916 int netif_rx(struct sk_buff *skb);
2917 int netif_rx_ni(struct sk_buff *skb);
2918 int netif_receive_skb(struct sk_buff *skb);
2919 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2920 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2921 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2922 gro_result_t napi_gro_frags(struct napi_struct *napi);
2923 struct packet_offload *gro_find_receive_by_type(__be16 type);
2924 struct packet_offload *gro_find_complete_by_type(__be16 type);
2925
2926 static inline void napi_free_frags(struct napi_struct *napi)
2927 {
2928 kfree_skb(napi->skb);
2929 napi->skb = NULL;
2930 }
2931
2932 int netdev_rx_handler_register(struct net_device *dev,
2933 rx_handler_func_t *rx_handler,
2934 void *rx_handler_data);
2935 void netdev_rx_handler_unregister(struct net_device *dev);
2936
2937 bool dev_valid_name(const char *name);
2938 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2939 int dev_ethtool(struct net *net, struct ifreq *);
2940 unsigned int dev_get_flags(const struct net_device *);
2941 int __dev_change_flags(struct net_device *, unsigned int flags);
2942 int dev_change_flags(struct net_device *, unsigned int);
2943 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2944 unsigned int gchanges);
2945 int dev_change_name(struct net_device *, const char *);
2946 int dev_set_alias(struct net_device *, const char *, size_t);
2947 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2948 int dev_set_mtu(struct net_device *, int);
2949 void dev_set_group(struct net_device *, int);
2950 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2951 int dev_change_carrier(struct net_device *, bool new_carrier);
2952 int dev_get_phys_port_id(struct net_device *dev,
2953 struct netdev_phys_item_id *ppid);
2954 int dev_get_phys_port_name(struct net_device *dev,
2955 char *name, size_t len);
2956 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2957 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2958 struct netdev_queue *txq, int *ret);
2959 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2960 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2961 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2962
2963 extern int netdev_budget;
2964
2965 /* Called by rtnetlink.c:rtnl_unlock() */
2966 void netdev_run_todo(void);
2967
2968 /**
2969 * dev_put - release reference to device
2970 * @dev: network device
2971 *
2972 * Release reference to device to allow it to be freed.
2973 */
2974 static inline void dev_put(struct net_device *dev)
2975 {
2976 this_cpu_dec(*dev->pcpu_refcnt);
2977 }
2978
2979 /**
2980 * dev_hold - get reference to device
2981 * @dev: network device
2982 *
2983 * Hold reference to device to keep it from being freed.
2984 */
2985 static inline void dev_hold(struct net_device *dev)
2986 {
2987 this_cpu_inc(*dev->pcpu_refcnt);
2988 }
2989
2990 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2991 * and _off may be called from IRQ context, but it is caller
2992 * who is responsible for serialization of these calls.
2993 *
2994 * The name carrier is inappropriate, these functions should really be
2995 * called netif_lowerlayer_*() because they represent the state of any
2996 * kind of lower layer not just hardware media.
2997 */
2998
2999 void linkwatch_init_dev(struct net_device *dev);
3000 void linkwatch_fire_event(struct net_device *dev);
3001 void linkwatch_forget_dev(struct net_device *dev);
3002
3003 /**
3004 * netif_carrier_ok - test if carrier present
3005 * @dev: network device
3006 *
3007 * Check if carrier is present on device
3008 */
3009 static inline bool netif_carrier_ok(const struct net_device *dev)
3010 {
3011 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3012 }
3013
3014 unsigned long dev_trans_start(struct net_device *dev);
3015
3016 void __netdev_watchdog_up(struct net_device *dev);
3017
3018 void netif_carrier_on(struct net_device *dev);
3019
3020 void netif_carrier_off(struct net_device *dev);
3021
3022 /**
3023 * netif_dormant_on - mark device as dormant.
3024 * @dev: network device
3025 *
3026 * Mark device as dormant (as per RFC2863).
3027 *
3028 * The dormant state indicates that the relevant interface is not
3029 * actually in a condition to pass packets (i.e., it is not 'up') but is
3030 * in a "pending" state, waiting for some external event. For "on-
3031 * demand" interfaces, this new state identifies the situation where the
3032 * interface is waiting for events to place it in the up state.
3033 *
3034 */
3035 static inline void netif_dormant_on(struct net_device *dev)
3036 {
3037 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3038 linkwatch_fire_event(dev);
3039 }
3040
3041 /**
3042 * netif_dormant_off - set device as not dormant.
3043 * @dev: network device
3044 *
3045 * Device is not in dormant state.
3046 */
3047 static inline void netif_dormant_off(struct net_device *dev)
3048 {
3049 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3050 linkwatch_fire_event(dev);
3051 }
3052
3053 /**
3054 * netif_dormant - test if carrier present
3055 * @dev: network device
3056 *
3057 * Check if carrier is present on device
3058 */
3059 static inline bool netif_dormant(const struct net_device *dev)
3060 {
3061 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3062 }
3063
3064
3065 /**
3066 * netif_oper_up - test if device is operational
3067 * @dev: network device
3068 *
3069 * Check if carrier is operational
3070 */
3071 static inline bool netif_oper_up(const struct net_device *dev)
3072 {
3073 return (dev->operstate == IF_OPER_UP ||
3074 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3075 }
3076
3077 /**
3078 * netif_device_present - is device available or removed
3079 * @dev: network device
3080 *
3081 * Check if device has not been removed from system.
3082 */
3083 static inline bool netif_device_present(struct net_device *dev)
3084 {
3085 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3086 }
3087
3088 void netif_device_detach(struct net_device *dev);
3089
3090 void netif_device_attach(struct net_device *dev);
3091
3092 /*
3093 * Network interface message level settings
3094 */
3095
3096 enum {
3097 NETIF_MSG_DRV = 0x0001,
3098 NETIF_MSG_PROBE = 0x0002,
3099 NETIF_MSG_LINK = 0x0004,
3100 NETIF_MSG_TIMER = 0x0008,
3101 NETIF_MSG_IFDOWN = 0x0010,
3102 NETIF_MSG_IFUP = 0x0020,
3103 NETIF_MSG_RX_ERR = 0x0040,
3104 NETIF_MSG_TX_ERR = 0x0080,
3105 NETIF_MSG_TX_QUEUED = 0x0100,
3106 NETIF_MSG_INTR = 0x0200,
3107 NETIF_MSG_TX_DONE = 0x0400,
3108 NETIF_MSG_RX_STATUS = 0x0800,
3109 NETIF_MSG_PKTDATA = 0x1000,
3110 NETIF_MSG_HW = 0x2000,
3111 NETIF_MSG_WOL = 0x4000,
3112 };
3113
3114 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3115 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3116 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3117 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3118 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3119 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3120 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3121 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3122 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3123 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3124 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3125 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3126 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3127 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3128 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3129
3130 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3131 {
3132 /* use default */
3133 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3134 return default_msg_enable_bits;
3135 if (debug_value == 0) /* no output */
3136 return 0;
3137 /* set low N bits */
3138 return (1 << debug_value) - 1;
3139 }
3140
3141 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3142 {
3143 spin_lock(&txq->_xmit_lock);
3144 txq->xmit_lock_owner = cpu;
3145 }
3146
3147 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3148 {
3149 spin_lock_bh(&txq->_xmit_lock);
3150 txq->xmit_lock_owner = smp_processor_id();
3151 }
3152
3153 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3154 {
3155 bool ok = spin_trylock(&txq->_xmit_lock);
3156 if (likely(ok))
3157 txq->xmit_lock_owner = smp_processor_id();
3158 return ok;
3159 }
3160
3161 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3162 {
3163 txq->xmit_lock_owner = -1;
3164 spin_unlock(&txq->_xmit_lock);
3165 }
3166
3167 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3168 {
3169 txq->xmit_lock_owner = -1;
3170 spin_unlock_bh(&txq->_xmit_lock);
3171 }
3172
3173 static inline void txq_trans_update(struct netdev_queue *txq)
3174 {
3175 if (txq->xmit_lock_owner != -1)
3176 txq->trans_start = jiffies;
3177 }
3178
3179 /**
3180 * netif_tx_lock - grab network device transmit lock
3181 * @dev: network device
3182 *
3183 * Get network device transmit lock
3184 */
3185 static inline void netif_tx_lock(struct net_device *dev)
3186 {
3187 unsigned int i;
3188 int cpu;
3189
3190 spin_lock(&dev->tx_global_lock);
3191 cpu = smp_processor_id();
3192 for (i = 0; i < dev->num_tx_queues; i++) {
3193 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3194
3195 /* We are the only thread of execution doing a
3196 * freeze, but we have to grab the _xmit_lock in
3197 * order to synchronize with threads which are in
3198 * the ->hard_start_xmit() handler and already
3199 * checked the frozen bit.
3200 */
3201 __netif_tx_lock(txq, cpu);
3202 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3203 __netif_tx_unlock(txq);
3204 }
3205 }
3206
3207 static inline void netif_tx_lock_bh(struct net_device *dev)
3208 {
3209 local_bh_disable();
3210 netif_tx_lock(dev);
3211 }
3212
3213 static inline void netif_tx_unlock(struct net_device *dev)
3214 {
3215 unsigned int i;
3216
3217 for (i = 0; i < dev->num_tx_queues; i++) {
3218 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3219
3220 /* No need to grab the _xmit_lock here. If the
3221 * queue is not stopped for another reason, we
3222 * force a schedule.
3223 */
3224 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3225 netif_schedule_queue(txq);
3226 }
3227 spin_unlock(&dev->tx_global_lock);
3228 }
3229
3230 static inline void netif_tx_unlock_bh(struct net_device *dev)
3231 {
3232 netif_tx_unlock(dev);
3233 local_bh_enable();
3234 }
3235
3236 #define HARD_TX_LOCK(dev, txq, cpu) { \
3237 if ((dev->features & NETIF_F_LLTX) == 0) { \
3238 __netif_tx_lock(txq, cpu); \
3239 } \
3240 }
3241
3242 #define HARD_TX_TRYLOCK(dev, txq) \
3243 (((dev->features & NETIF_F_LLTX) == 0) ? \
3244 __netif_tx_trylock(txq) : \
3245 true )
3246
3247 #define HARD_TX_UNLOCK(dev, txq) { \
3248 if ((dev->features & NETIF_F_LLTX) == 0) { \
3249 __netif_tx_unlock(txq); \
3250 } \
3251 }
3252
3253 static inline void netif_tx_disable(struct net_device *dev)
3254 {
3255 unsigned int i;
3256 int cpu;
3257
3258 local_bh_disable();
3259 cpu = smp_processor_id();
3260 for (i = 0; i < dev->num_tx_queues; i++) {
3261 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3262
3263 __netif_tx_lock(txq, cpu);
3264 netif_tx_stop_queue(txq);
3265 __netif_tx_unlock(txq);
3266 }
3267 local_bh_enable();
3268 }
3269
3270 static inline void netif_addr_lock(struct net_device *dev)
3271 {
3272 spin_lock(&dev->addr_list_lock);
3273 }
3274
3275 static inline void netif_addr_lock_nested(struct net_device *dev)
3276 {
3277 int subclass = SINGLE_DEPTH_NESTING;
3278
3279 if (dev->netdev_ops->ndo_get_lock_subclass)
3280 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3281
3282 spin_lock_nested(&dev->addr_list_lock, subclass);
3283 }
3284
3285 static inline void netif_addr_lock_bh(struct net_device *dev)
3286 {
3287 spin_lock_bh(&dev->addr_list_lock);
3288 }
3289
3290 static inline void netif_addr_unlock(struct net_device *dev)
3291 {
3292 spin_unlock(&dev->addr_list_lock);
3293 }
3294
3295 static inline void netif_addr_unlock_bh(struct net_device *dev)
3296 {
3297 spin_unlock_bh(&dev->addr_list_lock);
3298 }
3299
3300 /*
3301 * dev_addrs walker. Should be used only for read access. Call with
3302 * rcu_read_lock held.
3303 */
3304 #define for_each_dev_addr(dev, ha) \
3305 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3306
3307 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3308
3309 void ether_setup(struct net_device *dev);
3310
3311 /* Support for loadable net-drivers */
3312 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3313 unsigned char name_assign_type,
3314 void (*setup)(struct net_device *),
3315 unsigned int txqs, unsigned int rxqs);
3316 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3317 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3318
3319 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3320 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3321 count)
3322
3323 int register_netdev(struct net_device *dev);
3324 void unregister_netdev(struct net_device *dev);
3325
3326 /* General hardware address lists handling functions */
3327 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3328 struct netdev_hw_addr_list *from_list, int addr_len);
3329 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3330 struct netdev_hw_addr_list *from_list, int addr_len);
3331 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3332 struct net_device *dev,
3333 int (*sync)(struct net_device *, const unsigned char *),
3334 int (*unsync)(struct net_device *,
3335 const unsigned char *));
3336 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3337 struct net_device *dev,
3338 int (*unsync)(struct net_device *,
3339 const unsigned char *));
3340 void __hw_addr_init(struct netdev_hw_addr_list *list);
3341
3342 /* Functions used for device addresses handling */
3343 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3344 unsigned char addr_type);
3345 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3346 unsigned char addr_type);
3347 void dev_addr_flush(struct net_device *dev);
3348 int dev_addr_init(struct net_device *dev);
3349
3350 /* Functions used for unicast addresses handling */
3351 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3352 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3353 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3354 int dev_uc_sync(struct net_device *to, struct net_device *from);
3355 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3356 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3357 void dev_uc_flush(struct net_device *dev);
3358 void dev_uc_init(struct net_device *dev);
3359
3360 /**
3361 * __dev_uc_sync - Synchonize device's unicast list
3362 * @dev: device to sync
3363 * @sync: function to call if address should be added
3364 * @unsync: function to call if address should be removed
3365 *
3366 * Add newly added addresses to the interface, and release
3367 * addresses that have been deleted.
3368 **/
3369 static inline int __dev_uc_sync(struct net_device *dev,
3370 int (*sync)(struct net_device *,
3371 const unsigned char *),
3372 int (*unsync)(struct net_device *,
3373 const unsigned char *))
3374 {
3375 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3376 }
3377
3378 /**
3379 * __dev_uc_unsync - Remove synchronized addresses from device
3380 * @dev: device to sync
3381 * @unsync: function to call if address should be removed
3382 *
3383 * Remove all addresses that were added to the device by dev_uc_sync().
3384 **/
3385 static inline void __dev_uc_unsync(struct net_device *dev,
3386 int (*unsync)(struct net_device *,
3387 const unsigned char *))
3388 {
3389 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3390 }
3391
3392 /* Functions used for multicast addresses handling */
3393 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3394 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3395 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3396 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3397 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3398 int dev_mc_sync(struct net_device *to, struct net_device *from);
3399 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3400 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3401 void dev_mc_flush(struct net_device *dev);
3402 void dev_mc_init(struct net_device *dev);
3403
3404 /**
3405 * __dev_mc_sync - Synchonize device's multicast list
3406 * @dev: device to sync
3407 * @sync: function to call if address should be added
3408 * @unsync: function to call if address should be removed
3409 *
3410 * Add newly added addresses to the interface, and release
3411 * addresses that have been deleted.
3412 **/
3413 static inline int __dev_mc_sync(struct net_device *dev,
3414 int (*sync)(struct net_device *,
3415 const unsigned char *),
3416 int (*unsync)(struct net_device *,
3417 const unsigned char *))
3418 {
3419 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3420 }
3421
3422 /**
3423 * __dev_mc_unsync - Remove synchronized addresses from device
3424 * @dev: device to sync
3425 * @unsync: function to call if address should be removed
3426 *
3427 * Remove all addresses that were added to the device by dev_mc_sync().
3428 **/
3429 static inline void __dev_mc_unsync(struct net_device *dev,
3430 int (*unsync)(struct net_device *,
3431 const unsigned char *))
3432 {
3433 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3434 }
3435
3436 /* Functions used for secondary unicast and multicast support */
3437 void dev_set_rx_mode(struct net_device *dev);
3438 void __dev_set_rx_mode(struct net_device *dev);
3439 int dev_set_promiscuity(struct net_device *dev, int inc);
3440 int dev_set_allmulti(struct net_device *dev, int inc);
3441 void netdev_state_change(struct net_device *dev);
3442 void netdev_notify_peers(struct net_device *dev);
3443 void netdev_features_change(struct net_device *dev);
3444 /* Load a device via the kmod */
3445 void dev_load(struct net *net, const char *name);
3446 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3447 struct rtnl_link_stats64 *storage);
3448 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3449 const struct net_device_stats *netdev_stats);
3450
3451 extern int netdev_max_backlog;
3452 extern int netdev_tstamp_prequeue;
3453 extern int weight_p;
3454 extern int bpf_jit_enable;
3455
3456 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3457 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3458 struct list_head **iter);
3459 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3460 struct list_head **iter);
3461
3462 /* iterate through upper list, must be called under RCU read lock */
3463 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3464 for (iter = &(dev)->adj_list.upper, \
3465 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3466 updev; \
3467 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3468
3469 /* iterate through upper list, must be called under RCU read lock */
3470 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3471 for (iter = &(dev)->all_adj_list.upper, \
3472 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3473 updev; \
3474 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3475
3476 void *netdev_lower_get_next_private(struct net_device *dev,
3477 struct list_head **iter);
3478 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3479 struct list_head **iter);
3480
3481 #define netdev_for_each_lower_private(dev, priv, iter) \
3482 for (iter = (dev)->adj_list.lower.next, \
3483 priv = netdev_lower_get_next_private(dev, &(iter)); \
3484 priv; \
3485 priv = netdev_lower_get_next_private(dev, &(iter)))
3486
3487 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3488 for (iter = &(dev)->adj_list.lower, \
3489 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3490 priv; \
3491 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3492
3493 void *netdev_lower_get_next(struct net_device *dev,
3494 struct list_head **iter);
3495 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3496 for (iter = &(dev)->adj_list.lower, \
3497 ldev = netdev_lower_get_next(dev, &(iter)); \
3498 ldev; \
3499 ldev = netdev_lower_get_next(dev, &(iter)))
3500
3501 void *netdev_adjacent_get_private(struct list_head *adj_list);
3502 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3503 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3504 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3505 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3506 int netdev_master_upper_dev_link(struct net_device *dev,
3507 struct net_device *upper_dev);
3508 int netdev_master_upper_dev_link_private(struct net_device *dev,
3509 struct net_device *upper_dev,
3510 void *private);
3511 void netdev_upper_dev_unlink(struct net_device *dev,
3512 struct net_device *upper_dev);
3513 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3514 void *netdev_lower_dev_get_private(struct net_device *dev,
3515 struct net_device *lower_dev);
3516
3517 /* RSS keys are 40 or 52 bytes long */
3518 #define NETDEV_RSS_KEY_LEN 52
3519 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3520 void netdev_rss_key_fill(void *buffer, size_t len);
3521
3522 int dev_get_nest_level(struct net_device *dev,
3523 bool (*type_check)(struct net_device *dev));
3524 int skb_checksum_help(struct sk_buff *skb);
3525 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3526 netdev_features_t features, bool tx_path);
3527 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3528 netdev_features_t features);
3529
3530 struct netdev_bonding_info {
3531 ifslave slave;
3532 ifbond master;
3533 };
3534
3535 struct netdev_notifier_bonding_info {
3536 struct netdev_notifier_info info; /* must be first */
3537 struct netdev_bonding_info bonding_info;
3538 };
3539
3540 void netdev_bonding_info_change(struct net_device *dev,
3541 struct netdev_bonding_info *bonding_info);
3542
3543 static inline
3544 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3545 {
3546 return __skb_gso_segment(skb, features, true);
3547 }
3548 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3549
3550 static inline bool can_checksum_protocol(netdev_features_t features,
3551 __be16 protocol)
3552 {
3553 return ((features & NETIF_F_GEN_CSUM) ||
3554 ((features & NETIF_F_V4_CSUM) &&
3555 protocol == htons(ETH_P_IP)) ||
3556 ((features & NETIF_F_V6_CSUM) &&
3557 protocol == htons(ETH_P_IPV6)) ||
3558 ((features & NETIF_F_FCOE_CRC) &&
3559 protocol == htons(ETH_P_FCOE)));
3560 }
3561
3562 #ifdef CONFIG_BUG
3563 void netdev_rx_csum_fault(struct net_device *dev);
3564 #else
3565 static inline void netdev_rx_csum_fault(struct net_device *dev)
3566 {
3567 }
3568 #endif
3569 /* rx skb timestamps */
3570 void net_enable_timestamp(void);
3571 void net_disable_timestamp(void);
3572
3573 #ifdef CONFIG_PROC_FS
3574 int __init dev_proc_init(void);
3575 #else
3576 #define dev_proc_init() 0
3577 #endif
3578
3579 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3580 struct sk_buff *skb, struct net_device *dev,
3581 bool more)
3582 {
3583 skb->xmit_more = more ? 1 : 0;
3584 return ops->ndo_start_xmit(skb, dev);
3585 }
3586
3587 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3588 struct netdev_queue *txq, bool more)
3589 {
3590 const struct net_device_ops *ops = dev->netdev_ops;
3591 int rc;
3592
3593 rc = __netdev_start_xmit(ops, skb, dev, more);
3594 if (rc == NETDEV_TX_OK)
3595 txq_trans_update(txq);
3596
3597 return rc;
3598 }
3599
3600 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3601 const void *ns);
3602 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3603 const void *ns);
3604
3605 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3606 {
3607 return netdev_class_create_file_ns(class_attr, NULL);
3608 }
3609
3610 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3611 {
3612 netdev_class_remove_file_ns(class_attr, NULL);
3613 }
3614
3615 extern struct kobj_ns_type_operations net_ns_type_operations;
3616
3617 const char *netdev_drivername(const struct net_device *dev);
3618
3619 void linkwatch_run_queue(void);
3620
3621 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3622 netdev_features_t f2)
3623 {
3624 if (f1 & NETIF_F_GEN_CSUM)
3625 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3626 if (f2 & NETIF_F_GEN_CSUM)
3627 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3628 f1 &= f2;
3629 if (f1 & NETIF_F_GEN_CSUM)
3630 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3631
3632 return f1;
3633 }
3634
3635 static inline netdev_features_t netdev_get_wanted_features(
3636 struct net_device *dev)
3637 {
3638 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3639 }
3640 netdev_features_t netdev_increment_features(netdev_features_t all,
3641 netdev_features_t one, netdev_features_t mask);
3642
3643 /* Allow TSO being used on stacked device :
3644 * Performing the GSO segmentation before last device
3645 * is a performance improvement.
3646 */
3647 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3648 netdev_features_t mask)
3649 {
3650 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3651 }
3652
3653 int __netdev_update_features(struct net_device *dev);
3654 void netdev_update_features(struct net_device *dev);
3655 void netdev_change_features(struct net_device *dev);
3656
3657 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3658 struct net_device *dev);
3659
3660 netdev_features_t netif_skb_features(struct sk_buff *skb);
3661
3662 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3663 {
3664 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3665
3666 /* check flags correspondence */
3667 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3668 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3669 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3670 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3671 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3672 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3673 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3674 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3675 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3676 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3677 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3678 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3679 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3680
3681 return (features & feature) == feature;
3682 }
3683
3684 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3685 {
3686 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3687 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3688 }
3689
3690 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3691 netdev_features_t features)
3692 {
3693 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3694 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3695 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3696 }
3697
3698 static inline void netif_set_gso_max_size(struct net_device *dev,
3699 unsigned int size)
3700 {
3701 dev->gso_max_size = size;
3702 }
3703
3704 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3705 int pulled_hlen, u16 mac_offset,
3706 int mac_len)
3707 {
3708 skb->protocol = protocol;
3709 skb->encapsulation = 1;
3710 skb_push(skb, pulled_hlen);
3711 skb_reset_transport_header(skb);
3712 skb->mac_header = mac_offset;
3713 skb->network_header = skb->mac_header + mac_len;
3714 skb->mac_len = mac_len;
3715 }
3716
3717 static inline bool netif_is_macvlan(struct net_device *dev)
3718 {
3719 return dev->priv_flags & IFF_MACVLAN;
3720 }
3721
3722 static inline bool netif_is_macvlan_port(struct net_device *dev)
3723 {
3724 return dev->priv_flags & IFF_MACVLAN_PORT;
3725 }
3726
3727 static inline bool netif_is_ipvlan(struct net_device *dev)
3728 {
3729 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3730 }
3731
3732 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3733 {
3734 return dev->priv_flags & IFF_IPVLAN_MASTER;
3735 }
3736
3737 static inline bool netif_is_bond_master(struct net_device *dev)
3738 {
3739 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3740 }
3741
3742 static inline bool netif_is_bond_slave(struct net_device *dev)
3743 {
3744 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3745 }
3746
3747 static inline bool netif_supports_nofcs(struct net_device *dev)
3748 {
3749 return dev->priv_flags & IFF_SUPP_NOFCS;
3750 }
3751
3752 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3753 static inline void netif_keep_dst(struct net_device *dev)
3754 {
3755 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3756 }
3757
3758 extern struct pernet_operations __net_initdata loopback_net_ops;
3759
3760 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3761
3762 /* netdev_printk helpers, similar to dev_printk */
3763
3764 static inline const char *netdev_name(const struct net_device *dev)
3765 {
3766 if (!dev->name[0] || strchr(dev->name, '%'))
3767 return "(unnamed net_device)";
3768 return dev->name;
3769 }
3770
3771 static inline const char *netdev_reg_state(const struct net_device *dev)
3772 {
3773 switch (dev->reg_state) {
3774 case NETREG_UNINITIALIZED: return " (uninitialized)";
3775 case NETREG_REGISTERED: return "";
3776 case NETREG_UNREGISTERING: return " (unregistering)";
3777 case NETREG_UNREGISTERED: return " (unregistered)";
3778 case NETREG_RELEASED: return " (released)";
3779 case NETREG_DUMMY: return " (dummy)";
3780 }
3781
3782 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3783 return " (unknown)";
3784 }
3785
3786 __printf(3, 4)
3787 void netdev_printk(const char *level, const struct net_device *dev,
3788 const char *format, ...);
3789 __printf(2, 3)
3790 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3791 __printf(2, 3)
3792 void netdev_alert(const struct net_device *dev, const char *format, ...);
3793 __printf(2, 3)
3794 void netdev_crit(const struct net_device *dev, const char *format, ...);
3795 __printf(2, 3)
3796 void netdev_err(const struct net_device *dev, const char *format, ...);
3797 __printf(2, 3)
3798 void netdev_warn(const struct net_device *dev, const char *format, ...);
3799 __printf(2, 3)
3800 void netdev_notice(const struct net_device *dev, const char *format, ...);
3801 __printf(2, 3)
3802 void netdev_info(const struct net_device *dev, const char *format, ...);
3803
3804 #define MODULE_ALIAS_NETDEV(device) \
3805 MODULE_ALIAS("netdev-" device)
3806
3807 #if defined(CONFIG_DYNAMIC_DEBUG)
3808 #define netdev_dbg(__dev, format, args...) \
3809 do { \
3810 dynamic_netdev_dbg(__dev, format, ##args); \
3811 } while (0)
3812 #elif defined(DEBUG)
3813 #define netdev_dbg(__dev, format, args...) \
3814 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3815 #else
3816 #define netdev_dbg(__dev, format, args...) \
3817 ({ \
3818 if (0) \
3819 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3820 })
3821 #endif
3822
3823 #if defined(VERBOSE_DEBUG)
3824 #define netdev_vdbg netdev_dbg
3825 #else
3826
3827 #define netdev_vdbg(dev, format, args...) \
3828 ({ \
3829 if (0) \
3830 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3831 0; \
3832 })
3833 #endif
3834
3835 /*
3836 * netdev_WARN() acts like dev_printk(), but with the key difference
3837 * of using a WARN/WARN_ON to get the message out, including the
3838 * file/line information and a backtrace.
3839 */
3840 #define netdev_WARN(dev, format, args...) \
3841 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3842 netdev_reg_state(dev), ##args)
3843
3844 /* netif printk helpers, similar to netdev_printk */
3845
3846 #define netif_printk(priv, type, level, dev, fmt, args...) \
3847 do { \
3848 if (netif_msg_##type(priv)) \
3849 netdev_printk(level, (dev), fmt, ##args); \
3850 } while (0)
3851
3852 #define netif_level(level, priv, type, dev, fmt, args...) \
3853 do { \
3854 if (netif_msg_##type(priv)) \
3855 netdev_##level(dev, fmt, ##args); \
3856 } while (0)
3857
3858 #define netif_emerg(priv, type, dev, fmt, args...) \
3859 netif_level(emerg, priv, type, dev, fmt, ##args)
3860 #define netif_alert(priv, type, dev, fmt, args...) \
3861 netif_level(alert, priv, type, dev, fmt, ##args)
3862 #define netif_crit(priv, type, dev, fmt, args...) \
3863 netif_level(crit, priv, type, dev, fmt, ##args)
3864 #define netif_err(priv, type, dev, fmt, args...) \
3865 netif_level(err, priv, type, dev, fmt, ##args)
3866 #define netif_warn(priv, type, dev, fmt, args...) \
3867 netif_level(warn, priv, type, dev, fmt, ##args)
3868 #define netif_notice(priv, type, dev, fmt, args...) \
3869 netif_level(notice, priv, type, dev, fmt, ##args)
3870 #define netif_info(priv, type, dev, fmt, args...) \
3871 netif_level(info, priv, type, dev, fmt, ##args)
3872
3873 #if defined(CONFIG_DYNAMIC_DEBUG)
3874 #define netif_dbg(priv, type, netdev, format, args...) \
3875 do { \
3876 if (netif_msg_##type(priv)) \
3877 dynamic_netdev_dbg(netdev, format, ##args); \
3878 } while (0)
3879 #elif defined(DEBUG)
3880 #define netif_dbg(priv, type, dev, format, args...) \
3881 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3882 #else
3883 #define netif_dbg(priv, type, dev, format, args...) \
3884 ({ \
3885 if (0) \
3886 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3887 0; \
3888 })
3889 #endif
3890
3891 #if defined(VERBOSE_DEBUG)
3892 #define netif_vdbg netif_dbg
3893 #else
3894 #define netif_vdbg(priv, type, dev, format, args...) \
3895 ({ \
3896 if (0) \
3897 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3898 0; \
3899 })
3900 #endif
3901
3902 /*
3903 * The list of packet types we will receive (as opposed to discard)
3904 * and the routines to invoke.
3905 *
3906 * Why 16. Because with 16 the only overlap we get on a hash of the
3907 * low nibble of the protocol value is RARP/SNAP/X.25.
3908 *
3909 * NOTE: That is no longer true with the addition of VLAN tags. Not
3910 * sure which should go first, but I bet it won't make much
3911 * difference if we are running VLANs. The good news is that
3912 * this protocol won't be in the list unless compiled in, so
3913 * the average user (w/out VLANs) will not be adversely affected.
3914 * --BLG
3915 *
3916 * 0800 IP
3917 * 8100 802.1Q VLAN
3918 * 0001 802.3
3919 * 0002 AX.25
3920 * 0004 802.2
3921 * 8035 RARP
3922 * 0005 SNAP
3923 * 0805 X.25
3924 * 0806 ARP
3925 * 8137 IPX
3926 * 0009 Localtalk
3927 * 86DD IPv6
3928 */
3929 #define PTYPE_HASH_SIZE (16)
3930 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3931
3932 #endif /* _LINUX_NETDEVICE_H */