net: Fix high overhead of vlan sub-device teardown.
[deliverable/linux.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/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 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
970 * struct net_device *dev, u32 filter_mask)
971 *
972 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
973 * Called to change device carrier. Soft-devices (like dummy, team, etc)
974 * which do not represent real hardware may define this to allow their
975 * userspace components to manage their virtual carrier state. Devices
976 * that determine carrier state from physical hardware properties (eg
977 * network cables) or protocol-dependent mechanisms (eg
978 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
979 *
980 * int (*ndo_get_phys_port_id)(struct net_device *dev,
981 * struct netdev_phys_item_id *ppid);
982 * Called to get ID of physical port of this device. If driver does
983 * not implement this, it is assumed that the hw is not able to have
984 * multiple net devices on single physical port.
985 *
986 * void (*ndo_add_vxlan_port)(struct net_device *dev,
987 * sa_family_t sa_family, __be16 port);
988 * Called by vxlan to notiy a driver about the UDP port and socket
989 * address family that vxlan is listnening to. It is called only when
990 * a new port starts listening. The operation is protected by the
991 * vxlan_net->sock_lock.
992 *
993 * void (*ndo_del_vxlan_port)(struct net_device *dev,
994 * sa_family_t sa_family, __be16 port);
995 * Called by vxlan to notify the driver about a UDP port and socket
996 * address family that vxlan is not listening to anymore. The operation
997 * is protected by the vxlan_net->sock_lock.
998 *
999 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1000 * struct net_device *dev)
1001 * Called by upper layer devices to accelerate switching or other
1002 * station functionality into hardware. 'pdev is the lowerdev
1003 * to use for the offload and 'dev' is the net device that will
1004 * back the offload. Returns a pointer to the private structure
1005 * the upper layer will maintain.
1006 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1007 * Called by upper layer device to delete the station created
1008 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1009 * the station and priv is the structure returned by the add
1010 * operation.
1011 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1012 * struct net_device *dev,
1013 * void *priv);
1014 * Callback to use for xmit over the accelerated station. This
1015 * is used in place of ndo_start_xmit on accelerated net
1016 * devices.
1017 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1018 * struct net_device *dev
1019 * netdev_features_t features);
1020 * Called by core transmit path to determine if device is capable of
1021 * performing offload operations on a given packet. This is to give
1022 * the device an opportunity to implement any restrictions that cannot
1023 * be otherwise expressed by feature flags. The check is called with
1024 * the set of features that the stack has calculated and it returns
1025 * those the driver believes to be appropriate.
1026 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1027 * int queue_index, u32 maxrate);
1028 * Called when a user wants to set a max-rate limitation of specific
1029 * TX queue.
1030 */
1031 struct net_device_ops {
1032 int (*ndo_init)(struct net_device *dev);
1033 void (*ndo_uninit)(struct net_device *dev);
1034 int (*ndo_open)(struct net_device *dev);
1035 int (*ndo_stop)(struct net_device *dev);
1036 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1037 struct net_device *dev);
1038 u16 (*ndo_select_queue)(struct net_device *dev,
1039 struct sk_buff *skb,
1040 void *accel_priv,
1041 select_queue_fallback_t fallback);
1042 void (*ndo_change_rx_flags)(struct net_device *dev,
1043 int flags);
1044 void (*ndo_set_rx_mode)(struct net_device *dev);
1045 int (*ndo_set_mac_address)(struct net_device *dev,
1046 void *addr);
1047 int (*ndo_validate_addr)(struct net_device *dev);
1048 int (*ndo_do_ioctl)(struct net_device *dev,
1049 struct ifreq *ifr, int cmd);
1050 int (*ndo_set_config)(struct net_device *dev,
1051 struct ifmap *map);
1052 int (*ndo_change_mtu)(struct net_device *dev,
1053 int new_mtu);
1054 int (*ndo_neigh_setup)(struct net_device *dev,
1055 struct neigh_parms *);
1056 void (*ndo_tx_timeout) (struct net_device *dev);
1057
1058 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1059 struct rtnl_link_stats64 *storage);
1060 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1061
1062 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1063 __be16 proto, u16 vid);
1064 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1065 __be16 proto, u16 vid);
1066 #ifdef CONFIG_NET_POLL_CONTROLLER
1067 void (*ndo_poll_controller)(struct net_device *dev);
1068 int (*ndo_netpoll_setup)(struct net_device *dev,
1069 struct netpoll_info *info);
1070 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1071 #endif
1072 #ifdef CONFIG_NET_RX_BUSY_POLL
1073 int (*ndo_busy_poll)(struct napi_struct *dev);
1074 #endif
1075 int (*ndo_set_vf_mac)(struct net_device *dev,
1076 int queue, u8 *mac);
1077 int (*ndo_set_vf_vlan)(struct net_device *dev,
1078 int queue, u16 vlan, u8 qos);
1079 int (*ndo_set_vf_rate)(struct net_device *dev,
1080 int vf, int min_tx_rate,
1081 int max_tx_rate);
1082 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1083 int vf, bool setting);
1084 int (*ndo_get_vf_config)(struct net_device *dev,
1085 int vf,
1086 struct ifla_vf_info *ivf);
1087 int (*ndo_set_vf_link_state)(struct net_device *dev,
1088 int vf, int link_state);
1089 int (*ndo_set_vf_port)(struct net_device *dev,
1090 int vf,
1091 struct nlattr *port[]);
1092 int (*ndo_get_vf_port)(struct net_device *dev,
1093 int vf, struct sk_buff *skb);
1094 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1095 #if IS_ENABLED(CONFIG_FCOE)
1096 int (*ndo_fcoe_enable)(struct net_device *dev);
1097 int (*ndo_fcoe_disable)(struct net_device *dev);
1098 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1099 u16 xid,
1100 struct scatterlist *sgl,
1101 unsigned int sgc);
1102 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1103 u16 xid);
1104 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1105 u16 xid,
1106 struct scatterlist *sgl,
1107 unsigned int sgc);
1108 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1109 struct netdev_fcoe_hbainfo *hbainfo);
1110 #endif
1111
1112 #if IS_ENABLED(CONFIG_LIBFCOE)
1113 #define NETDEV_FCOE_WWNN 0
1114 #define NETDEV_FCOE_WWPN 1
1115 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1116 u64 *wwn, int type);
1117 #endif
1118
1119 #ifdef CONFIG_RFS_ACCEL
1120 int (*ndo_rx_flow_steer)(struct net_device *dev,
1121 const struct sk_buff *skb,
1122 u16 rxq_index,
1123 u32 flow_id);
1124 #endif
1125 int (*ndo_add_slave)(struct net_device *dev,
1126 struct net_device *slave_dev);
1127 int (*ndo_del_slave)(struct net_device *dev,
1128 struct net_device *slave_dev);
1129 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1130 netdev_features_t features);
1131 int (*ndo_set_features)(struct net_device *dev,
1132 netdev_features_t features);
1133 int (*ndo_neigh_construct)(struct neighbour *n);
1134 void (*ndo_neigh_destroy)(struct neighbour *n);
1135
1136 int (*ndo_fdb_add)(struct ndmsg *ndm,
1137 struct nlattr *tb[],
1138 struct net_device *dev,
1139 const unsigned char *addr,
1140 u16 vid,
1141 u16 flags);
1142 int (*ndo_fdb_del)(struct ndmsg *ndm,
1143 struct nlattr *tb[],
1144 struct net_device *dev,
1145 const unsigned char *addr,
1146 u16 vid);
1147 int (*ndo_fdb_dump)(struct sk_buff *skb,
1148 struct netlink_callback *cb,
1149 struct net_device *dev,
1150 struct net_device *filter_dev,
1151 int idx);
1152
1153 int (*ndo_bridge_setlink)(struct net_device *dev,
1154 struct nlmsghdr *nlh,
1155 u16 flags);
1156 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1157 u32 pid, u32 seq,
1158 struct net_device *dev,
1159 u32 filter_mask);
1160 int (*ndo_bridge_dellink)(struct net_device *dev,
1161 struct nlmsghdr *nlh,
1162 u16 flags);
1163 int (*ndo_change_carrier)(struct net_device *dev,
1164 bool new_carrier);
1165 int (*ndo_get_phys_port_id)(struct net_device *dev,
1166 struct netdev_phys_item_id *ppid);
1167 int (*ndo_get_phys_port_name)(struct net_device *dev,
1168 char *name, size_t len);
1169 void (*ndo_add_vxlan_port)(struct net_device *dev,
1170 sa_family_t sa_family,
1171 __be16 port);
1172 void (*ndo_del_vxlan_port)(struct net_device *dev,
1173 sa_family_t sa_family,
1174 __be16 port);
1175
1176 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1177 struct net_device *dev);
1178 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1179 void *priv);
1180
1181 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1182 struct net_device *dev,
1183 void *priv);
1184 int (*ndo_get_lock_subclass)(struct net_device *dev);
1185 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1186 struct net_device *dev,
1187 netdev_features_t features);
1188 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1189 int queue_index,
1190 u32 maxrate);
1191 };
1192
1193 /**
1194 * enum net_device_priv_flags - &struct net_device priv_flags
1195 *
1196 * These are the &struct net_device, they are only set internally
1197 * by drivers and used in the kernel. These flags are invisible to
1198 * userspace, this means that the order of these flags can change
1199 * during any kernel release.
1200 *
1201 * You should have a pretty good reason to be extending these flags.
1202 *
1203 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1204 * @IFF_EBRIDGE: Ethernet bridging device
1205 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1206 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1207 * @IFF_MASTER_ALB: bonding master, balance-alb
1208 * @IFF_BONDING: bonding master or slave
1209 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1210 * @IFF_ISATAP: ISATAP interface (RFC4214)
1211 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1212 * @IFF_WAN_HDLC: WAN HDLC device
1213 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1214 * release skb->dst
1215 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1216 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1217 * @IFF_MACVLAN_PORT: device used as macvlan port
1218 * @IFF_BRIDGE_PORT: device used as bridge port
1219 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1220 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1221 * @IFF_UNICAST_FLT: Supports unicast filtering
1222 * @IFF_TEAM_PORT: device used as team port
1223 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1224 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1225 * change when it's running
1226 * @IFF_MACVLAN: Macvlan device
1227 */
1228 enum netdev_priv_flags {
1229 IFF_802_1Q_VLAN = 1<<0,
1230 IFF_EBRIDGE = 1<<1,
1231 IFF_SLAVE_INACTIVE = 1<<2,
1232 IFF_MASTER_8023AD = 1<<3,
1233 IFF_MASTER_ALB = 1<<4,
1234 IFF_BONDING = 1<<5,
1235 IFF_SLAVE_NEEDARP = 1<<6,
1236 IFF_ISATAP = 1<<7,
1237 IFF_MASTER_ARPMON = 1<<8,
1238 IFF_WAN_HDLC = 1<<9,
1239 IFF_XMIT_DST_RELEASE = 1<<10,
1240 IFF_DONT_BRIDGE = 1<<11,
1241 IFF_DISABLE_NETPOLL = 1<<12,
1242 IFF_MACVLAN_PORT = 1<<13,
1243 IFF_BRIDGE_PORT = 1<<14,
1244 IFF_OVS_DATAPATH = 1<<15,
1245 IFF_TX_SKB_SHARING = 1<<16,
1246 IFF_UNICAST_FLT = 1<<17,
1247 IFF_TEAM_PORT = 1<<18,
1248 IFF_SUPP_NOFCS = 1<<19,
1249 IFF_LIVE_ADDR_CHANGE = 1<<20,
1250 IFF_MACVLAN = 1<<21,
1251 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1252 IFF_IPVLAN_MASTER = 1<<23,
1253 IFF_IPVLAN_SLAVE = 1<<24,
1254 };
1255
1256 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1257 #define IFF_EBRIDGE IFF_EBRIDGE
1258 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1259 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1260 #define IFF_MASTER_ALB IFF_MASTER_ALB
1261 #define IFF_BONDING IFF_BONDING
1262 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1263 #define IFF_ISATAP IFF_ISATAP
1264 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1265 #define IFF_WAN_HDLC IFF_WAN_HDLC
1266 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1267 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1268 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1269 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1270 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1271 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1272 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1273 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1274 #define IFF_TEAM_PORT IFF_TEAM_PORT
1275 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1276 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1277 #define IFF_MACVLAN IFF_MACVLAN
1278 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1279 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1280 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1281
1282 /**
1283 * struct net_device - The DEVICE structure.
1284 * Actually, this whole structure is a big mistake. It mixes I/O
1285 * data with strictly "high-level" data, and it has to know about
1286 * almost every data structure used in the INET module.
1287 *
1288 * @name: This is the first field of the "visible" part of this structure
1289 * (i.e. as seen by users in the "Space.c" file). It is the name
1290 * of the interface.
1291 *
1292 * @name_hlist: Device name hash chain, please keep it close to name[]
1293 * @ifalias: SNMP alias
1294 * @mem_end: Shared memory end
1295 * @mem_start: Shared memory start
1296 * @base_addr: Device I/O address
1297 * @irq: Device IRQ number
1298 *
1299 * @state: Generic network queuing layer state, see netdev_state_t
1300 * @dev_list: The global list of network devices
1301 * @napi_list: List entry, that is used for polling napi devices
1302 * @unreg_list: List entry, that is used, when we are unregistering the
1303 * device, see the function unregister_netdev
1304 * @close_list: List entry, that is used, when we are closing the device
1305 *
1306 * @adj_list: Directly linked devices, like slaves for bonding
1307 * @all_adj_list: All linked devices, *including* neighbours
1308 * @features: Currently active device features
1309 * @hw_features: User-changeable features
1310 *
1311 * @wanted_features: User-requested features
1312 * @vlan_features: Mask of features inheritable by VLAN devices
1313 *
1314 * @hw_enc_features: Mask of features inherited by encapsulating devices
1315 * This field indicates what encapsulation
1316 * offloads the hardware is capable of doing,
1317 * and drivers will need to set them appropriately.
1318 *
1319 * @mpls_features: Mask of features inheritable by MPLS
1320 *
1321 * @ifindex: interface index
1322 * @iflink: unique device identifier
1323 *
1324 * @stats: Statistics struct, which was left as a legacy, use
1325 * rtnl_link_stats64 instead
1326 *
1327 * @rx_dropped: Dropped packets by core network,
1328 * do not use this in drivers
1329 * @tx_dropped: Dropped packets by core network,
1330 * do not use this in drivers
1331 *
1332 * @carrier_changes: Stats to monitor carrier on<->off transitions
1333 *
1334 * @wireless_handlers: List of functions to handle Wireless Extensions,
1335 * instead of ioctl,
1336 * see <net/iw_handler.h> for details.
1337 * @wireless_data: Instance data managed by the core of wireless extensions
1338 *
1339 * @netdev_ops: Includes several pointers to callbacks,
1340 * if one wants to override the ndo_*() functions
1341 * @ethtool_ops: Management operations
1342 * @fwd_ops: Management operations
1343 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1344 * of Layer 2 headers.
1345 *
1346 * @flags: Interface flags (a la BSD)
1347 * @priv_flags: Like 'flags' but invisible to userspace,
1348 * see if.h for the definitions
1349 * @gflags: Global flags ( kept as legacy )
1350 * @padded: How much padding added by alloc_netdev()
1351 * @operstate: RFC2863 operstate
1352 * @link_mode: Mapping policy to operstate
1353 * @if_port: Selectable AUI, TP, ...
1354 * @dma: DMA channel
1355 * @mtu: Interface MTU value
1356 * @type: Interface hardware type
1357 * @hard_header_len: Hardware header length
1358 *
1359 * @needed_headroom: Extra headroom the hardware may need, but not in all
1360 * cases can this be guaranteed
1361 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1362 * cases can this be guaranteed. Some cases also use
1363 * LL_MAX_HEADER instead to allocate the skb
1364 *
1365 * interface address info:
1366 *
1367 * @perm_addr: Permanent hw address
1368 * @addr_assign_type: Hw address assignment type
1369 * @addr_len: Hardware address length
1370 * @neigh_priv_len; Used in neigh_alloc(),
1371 * initialized only in atm/clip.c
1372 * @dev_id: Used to differentiate devices that share
1373 * the same link layer address
1374 * @dev_port: Used to differentiate devices that share
1375 * the same function
1376 * @addr_list_lock: XXX: need comments on this one
1377 * @uc: unicast mac addresses
1378 * @mc: multicast mac addresses
1379 * @dev_addrs: list of device hw addresses
1380 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1381 * @uc_promisc: Counter, that indicates, that promiscuous mode
1382 * has been enabled due to the need to listen to
1383 * additional unicast addresses in a device that
1384 * does not implement ndo_set_rx_mode()
1385 * @promiscuity: Number of times, the NIC is told to work in
1386 * Promiscuous mode, if it becomes 0 the NIC will
1387 * exit from working in Promiscuous mode
1388 * @allmulti: Counter, enables or disables allmulticast mode
1389 *
1390 * @vlan_info: VLAN info
1391 * @dsa_ptr: dsa specific data
1392 * @tipc_ptr: TIPC specific data
1393 * @atalk_ptr: AppleTalk link
1394 * @ip_ptr: IPv4 specific data
1395 * @dn_ptr: DECnet specific data
1396 * @ip6_ptr: IPv6 specific data
1397 * @ax25_ptr: AX.25 specific data
1398 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1399 *
1400 * @last_rx: Time of last Rx
1401 * @dev_addr: Hw address (before bcast,
1402 * because most packets are unicast)
1403 *
1404 * @_rx: Array of RX queues
1405 * @num_rx_queues: Number of RX queues
1406 * allocated at register_netdev() time
1407 * @real_num_rx_queues: Number of RX queues currently active in device
1408 *
1409 * @rx_handler: handler for received packets
1410 * @rx_handler_data: XXX: need comments on this one
1411 * @ingress_queue: XXX: need comments on this one
1412 * @broadcast: hw bcast address
1413 *
1414 * @_tx: Array of TX queues
1415 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1416 * @real_num_tx_queues: Number of TX queues currently active in device
1417 * @qdisc: Root qdisc from userspace point of view
1418 * @tx_queue_len: Max frames per queue allowed
1419 * @tx_global_lock: XXX: need comments on this one
1420 *
1421 * @xps_maps: XXX: need comments on this one
1422 *
1423 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1424 * indexed by RX queue number. Assigned by driver.
1425 * This must only be set if the ndo_rx_flow_steer
1426 * operation is defined
1427 *
1428 * @trans_start: Time (in jiffies) of last Tx
1429 * @watchdog_timeo: Represents the timeout that is used by
1430 * the watchdog ( see dev_watchdog() )
1431 * @watchdog_timer: List of timers
1432 *
1433 * @pcpu_refcnt: Number of references to this device
1434 * @todo_list: Delayed register/unregister
1435 * @index_hlist: Device index hash chain
1436 * @link_watch_list: XXX: need comments on this one
1437 *
1438 * @reg_state: Register/unregister state machine
1439 * @dismantle: Device is going to be freed
1440 * @rtnl_link_state: This enum represents the phases of creating
1441 * a new link
1442 *
1443 * @destructor: Called from unregister,
1444 * can be used to call free_netdev
1445 * @npinfo: XXX: need comments on this one
1446 * @nd_net: Network namespace this network device is inside
1447 *
1448 * @ml_priv: Mid-layer private
1449 * @lstats: Loopback statistics
1450 * @tstats: Tunnel statistics
1451 * @dstats: Dummy statistics
1452 * @vstats: Virtual ethernet statistics
1453 *
1454 * @garp_port: GARP
1455 * @mrp_port: MRP
1456 *
1457 * @dev: Class/net/name entry
1458 * @sysfs_groups: Space for optional device, statistics and wireless
1459 * sysfs groups
1460 *
1461 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1462 * @rtnl_link_ops: Rtnl_link_ops
1463 *
1464 * @gso_max_size: Maximum size of generic segmentation offload
1465 * @gso_max_segs: Maximum number of segments that can be passed to the
1466 * NIC for GSO
1467 * @gso_min_segs: Minimum number of segments that can be passed to the
1468 * NIC for GSO
1469 *
1470 * @dcbnl_ops: Data Center Bridging netlink ops
1471 * @num_tc: Number of traffic classes in the net device
1472 * @tc_to_txq: XXX: need comments on this one
1473 * @prio_tc_map XXX: need comments on this one
1474 *
1475 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1476 *
1477 * @priomap: XXX: need comments on this one
1478 * @phydev: Physical device may attach itself
1479 * for hardware timestamping
1480 *
1481 * @qdisc_tx_busylock: XXX: need comments on this one
1482 *
1483 * @group: The group, that the device belongs to
1484 * @pm_qos_req: Power Management QoS object
1485 *
1486 * FIXME: cleanup struct net_device such that network protocol info
1487 * moves out.
1488 */
1489
1490 struct net_device {
1491 char name[IFNAMSIZ];
1492 struct hlist_node name_hlist;
1493 char *ifalias;
1494 /*
1495 * I/O specific fields
1496 * FIXME: Merge these and struct ifmap into one
1497 */
1498 unsigned long mem_end;
1499 unsigned long mem_start;
1500 unsigned long base_addr;
1501 int irq;
1502
1503 /*
1504 * Some hardware also needs these fields (state,dev_list,
1505 * napi_list,unreg_list,close_list) but they are not
1506 * part of the usual set specified in Space.c.
1507 */
1508
1509 unsigned long state;
1510
1511 struct list_head dev_list;
1512 struct list_head napi_list;
1513 struct list_head unreg_list;
1514 struct list_head close_list;
1515 struct list_head ptype_all;
1516 struct list_head ptype_specific;
1517
1518 struct {
1519 struct list_head upper;
1520 struct list_head lower;
1521 } adj_list;
1522
1523 struct {
1524 struct list_head upper;
1525 struct list_head lower;
1526 } all_adj_list;
1527
1528 netdev_features_t features;
1529 netdev_features_t hw_features;
1530 netdev_features_t wanted_features;
1531 netdev_features_t vlan_features;
1532 netdev_features_t hw_enc_features;
1533 netdev_features_t mpls_features;
1534
1535 int ifindex;
1536 int iflink;
1537
1538 struct net_device_stats stats;
1539
1540 atomic_long_t rx_dropped;
1541 atomic_long_t tx_dropped;
1542
1543 atomic_t carrier_changes;
1544
1545 #ifdef CONFIG_WIRELESS_EXT
1546 const struct iw_handler_def * wireless_handlers;
1547 struct iw_public_data * wireless_data;
1548 #endif
1549 const struct net_device_ops *netdev_ops;
1550 const struct ethtool_ops *ethtool_ops;
1551 const struct forwarding_accel_ops *fwd_ops;
1552 #ifdef CONFIG_NET_SWITCHDEV
1553 const struct swdev_ops *swdev_ops;
1554 #endif
1555
1556 const struct header_ops *header_ops;
1557
1558 unsigned int flags;
1559 unsigned int priv_flags;
1560
1561 unsigned short gflags;
1562 unsigned short padded;
1563
1564 unsigned char operstate;
1565 unsigned char link_mode;
1566
1567 unsigned char if_port;
1568 unsigned char dma;
1569
1570 unsigned int mtu;
1571 unsigned short type;
1572 unsigned short hard_header_len;
1573
1574 unsigned short needed_headroom;
1575 unsigned short needed_tailroom;
1576
1577 /* Interface address info. */
1578 unsigned char perm_addr[MAX_ADDR_LEN];
1579 unsigned char addr_assign_type;
1580 unsigned char addr_len;
1581 unsigned short neigh_priv_len;
1582 unsigned short dev_id;
1583 unsigned short dev_port;
1584 spinlock_t addr_list_lock;
1585 struct netdev_hw_addr_list uc;
1586 struct netdev_hw_addr_list mc;
1587 struct netdev_hw_addr_list dev_addrs;
1588
1589 #ifdef CONFIG_SYSFS
1590 struct kset *queues_kset;
1591 #endif
1592
1593 unsigned char name_assign_type;
1594
1595 bool uc_promisc;
1596 unsigned int promiscuity;
1597 unsigned int allmulti;
1598
1599
1600 /* Protocol specific pointers */
1601
1602 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1603 struct vlan_info __rcu *vlan_info;
1604 #endif
1605 #if IS_ENABLED(CONFIG_NET_DSA)
1606 struct dsa_switch_tree *dsa_ptr;
1607 #endif
1608 #if IS_ENABLED(CONFIG_TIPC)
1609 struct tipc_bearer __rcu *tipc_ptr;
1610 #endif
1611 void *atalk_ptr;
1612 struct in_device __rcu *ip_ptr;
1613 struct dn_dev __rcu *dn_ptr;
1614 struct inet6_dev __rcu *ip6_ptr;
1615 void *ax25_ptr;
1616 struct wireless_dev *ieee80211_ptr;
1617 struct wpan_dev *ieee802154_ptr;
1618
1619 /*
1620 * Cache lines mostly used on receive path (including eth_type_trans())
1621 */
1622 unsigned long last_rx;
1623
1624 /* Interface address info used in eth_type_trans() */
1625 unsigned char *dev_addr;
1626
1627
1628 #ifdef CONFIG_SYSFS
1629 struct netdev_rx_queue *_rx;
1630
1631 unsigned int num_rx_queues;
1632 unsigned int real_num_rx_queues;
1633
1634 #endif
1635
1636 unsigned long gro_flush_timeout;
1637 rx_handler_func_t __rcu *rx_handler;
1638 void __rcu *rx_handler_data;
1639
1640 struct netdev_queue __rcu *ingress_queue;
1641 unsigned char broadcast[MAX_ADDR_LEN];
1642
1643
1644 /*
1645 * Cache lines mostly used on transmit path
1646 */
1647 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1648 unsigned int num_tx_queues;
1649 unsigned int real_num_tx_queues;
1650 struct Qdisc *qdisc;
1651 unsigned long tx_queue_len;
1652 spinlock_t tx_global_lock;
1653
1654 #ifdef CONFIG_XPS
1655 struct xps_dev_maps __rcu *xps_maps;
1656 #endif
1657 #ifdef CONFIG_RFS_ACCEL
1658 struct cpu_rmap *rx_cpu_rmap;
1659 #endif
1660
1661 /* These may be needed for future network-power-down code. */
1662
1663 /*
1664 * trans_start here is expensive for high speed devices on SMP,
1665 * please use netdev_queue->trans_start instead.
1666 */
1667 unsigned long trans_start;
1668
1669 int watchdog_timeo;
1670 struct timer_list watchdog_timer;
1671
1672 int __percpu *pcpu_refcnt;
1673 struct list_head todo_list;
1674
1675 struct hlist_node index_hlist;
1676 struct list_head link_watch_list;
1677
1678 enum { NETREG_UNINITIALIZED=0,
1679 NETREG_REGISTERED, /* completed register_netdevice */
1680 NETREG_UNREGISTERING, /* called unregister_netdevice */
1681 NETREG_UNREGISTERED, /* completed unregister todo */
1682 NETREG_RELEASED, /* called free_netdev */
1683 NETREG_DUMMY, /* dummy device for NAPI poll */
1684 } reg_state:8;
1685
1686 bool dismantle;
1687
1688 enum {
1689 RTNL_LINK_INITIALIZED,
1690 RTNL_LINK_INITIALIZING,
1691 } rtnl_link_state:16;
1692
1693 void (*destructor)(struct net_device *dev);
1694
1695 #ifdef CONFIG_NETPOLL
1696 struct netpoll_info __rcu *npinfo;
1697 #endif
1698
1699 possible_net_t nd_net;
1700
1701 /* mid-layer private */
1702 union {
1703 void *ml_priv;
1704 struct pcpu_lstats __percpu *lstats;
1705 struct pcpu_sw_netstats __percpu *tstats;
1706 struct pcpu_dstats __percpu *dstats;
1707 struct pcpu_vstats __percpu *vstats;
1708 };
1709
1710 struct garp_port __rcu *garp_port;
1711 struct mrp_port __rcu *mrp_port;
1712
1713 struct device dev;
1714 const struct attribute_group *sysfs_groups[4];
1715 const struct attribute_group *sysfs_rx_queue_group;
1716
1717 const struct rtnl_link_ops *rtnl_link_ops;
1718
1719 /* for setting kernel sock attribute on TCP connection setup */
1720 #define GSO_MAX_SIZE 65536
1721 unsigned int gso_max_size;
1722 #define GSO_MAX_SEGS 65535
1723 u16 gso_max_segs;
1724 u16 gso_min_segs;
1725 #ifdef CONFIG_DCB
1726 const struct dcbnl_rtnl_ops *dcbnl_ops;
1727 #endif
1728 u8 num_tc;
1729 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1730 u8 prio_tc_map[TC_BITMASK + 1];
1731
1732 #if IS_ENABLED(CONFIG_FCOE)
1733 unsigned int fcoe_ddp_xid;
1734 #endif
1735 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1736 struct netprio_map __rcu *priomap;
1737 #endif
1738 struct phy_device *phydev;
1739 struct lock_class_key *qdisc_tx_busylock;
1740 int group;
1741 struct pm_qos_request pm_qos_req;
1742 };
1743 #define to_net_dev(d) container_of(d, struct net_device, dev)
1744
1745 #define NETDEV_ALIGN 32
1746
1747 static inline
1748 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1749 {
1750 return dev->prio_tc_map[prio & TC_BITMASK];
1751 }
1752
1753 static inline
1754 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1755 {
1756 if (tc >= dev->num_tc)
1757 return -EINVAL;
1758
1759 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1760 return 0;
1761 }
1762
1763 static inline
1764 void netdev_reset_tc(struct net_device *dev)
1765 {
1766 dev->num_tc = 0;
1767 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1768 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1769 }
1770
1771 static inline
1772 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1773 {
1774 if (tc >= dev->num_tc)
1775 return -EINVAL;
1776
1777 dev->tc_to_txq[tc].count = count;
1778 dev->tc_to_txq[tc].offset = offset;
1779 return 0;
1780 }
1781
1782 static inline
1783 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1784 {
1785 if (num_tc > TC_MAX_QUEUE)
1786 return -EINVAL;
1787
1788 dev->num_tc = num_tc;
1789 return 0;
1790 }
1791
1792 static inline
1793 int netdev_get_num_tc(struct net_device *dev)
1794 {
1795 return dev->num_tc;
1796 }
1797
1798 static inline
1799 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1800 unsigned int index)
1801 {
1802 return &dev->_tx[index];
1803 }
1804
1805 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1806 const struct sk_buff *skb)
1807 {
1808 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1809 }
1810
1811 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1812 void (*f)(struct net_device *,
1813 struct netdev_queue *,
1814 void *),
1815 void *arg)
1816 {
1817 unsigned int i;
1818
1819 for (i = 0; i < dev->num_tx_queues; i++)
1820 f(dev, &dev->_tx[i], arg);
1821 }
1822
1823 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1824 struct sk_buff *skb,
1825 void *accel_priv);
1826
1827 /*
1828 * Net namespace inlines
1829 */
1830 static inline
1831 struct net *dev_net(const struct net_device *dev)
1832 {
1833 return read_pnet(&dev->nd_net);
1834 }
1835
1836 static inline
1837 void dev_net_set(struct net_device *dev, struct net *net)
1838 {
1839 write_pnet(&dev->nd_net, net);
1840 }
1841
1842 static inline bool netdev_uses_dsa(struct net_device *dev)
1843 {
1844 #if IS_ENABLED(CONFIG_NET_DSA)
1845 if (dev->dsa_ptr != NULL)
1846 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1847 #endif
1848 return false;
1849 }
1850
1851 /**
1852 * netdev_priv - access network device private data
1853 * @dev: network device
1854 *
1855 * Get network device private data
1856 */
1857 static inline void *netdev_priv(const struct net_device *dev)
1858 {
1859 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1860 }
1861
1862 /* Set the sysfs physical device reference for the network logical device
1863 * if set prior to registration will cause a symlink during initialization.
1864 */
1865 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1866
1867 /* Set the sysfs device type for the network logical device to allow
1868 * fine-grained identification of different network device types. For
1869 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1870 */
1871 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1872
1873 /* Default NAPI poll() weight
1874 * Device drivers are strongly advised to not use bigger value
1875 */
1876 #define NAPI_POLL_WEIGHT 64
1877
1878 /**
1879 * netif_napi_add - initialize a napi context
1880 * @dev: network device
1881 * @napi: napi context
1882 * @poll: polling function
1883 * @weight: default weight
1884 *
1885 * netif_napi_add() must be used to initialize a napi context prior to calling
1886 * *any* of the other napi related functions.
1887 */
1888 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1889 int (*poll)(struct napi_struct *, int), int weight);
1890
1891 /**
1892 * netif_napi_del - remove a napi context
1893 * @napi: napi context
1894 *
1895 * netif_napi_del() removes a napi context from the network device napi list
1896 */
1897 void netif_napi_del(struct napi_struct *napi);
1898
1899 struct napi_gro_cb {
1900 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1901 void *frag0;
1902
1903 /* Length of frag0. */
1904 unsigned int frag0_len;
1905
1906 /* This indicates where we are processing relative to skb->data. */
1907 int data_offset;
1908
1909 /* This is non-zero if the packet cannot be merged with the new skb. */
1910 u16 flush;
1911
1912 /* Save the IP ID here and check when we get to the transport layer */
1913 u16 flush_id;
1914
1915 /* Number of segments aggregated. */
1916 u16 count;
1917
1918 /* Start offset for remote checksum offload */
1919 u16 gro_remcsum_start;
1920
1921 /* jiffies when first packet was created/queued */
1922 unsigned long age;
1923
1924 /* Used in ipv6_gro_receive() and foo-over-udp */
1925 u16 proto;
1926
1927 /* This is non-zero if the packet may be of the same flow. */
1928 u8 same_flow:1;
1929
1930 /* Used in udp_gro_receive */
1931 u8 udp_mark:1;
1932
1933 /* GRO checksum is valid */
1934 u8 csum_valid:1;
1935
1936 /* Number of checksums via CHECKSUM_UNNECESSARY */
1937 u8 csum_cnt:3;
1938
1939 /* Free the skb? */
1940 u8 free:2;
1941 #define NAPI_GRO_FREE 1
1942 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1943
1944 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1945 u8 is_ipv6:1;
1946
1947 /* 7 bit hole */
1948
1949 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1950 __wsum csum;
1951
1952 /* used in skb_gro_receive() slow path */
1953 struct sk_buff *last;
1954 };
1955
1956 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1957
1958 struct packet_type {
1959 __be16 type; /* This is really htons(ether_type). */
1960 struct net_device *dev; /* NULL is wildcarded here */
1961 int (*func) (struct sk_buff *,
1962 struct net_device *,
1963 struct packet_type *,
1964 struct net_device *);
1965 bool (*id_match)(struct packet_type *ptype,
1966 struct sock *sk);
1967 void *af_packet_priv;
1968 struct list_head list;
1969 };
1970
1971 struct offload_callbacks {
1972 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1973 netdev_features_t features);
1974 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1975 struct sk_buff *skb);
1976 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1977 };
1978
1979 struct packet_offload {
1980 __be16 type; /* This is really htons(ether_type). */
1981 struct offload_callbacks callbacks;
1982 struct list_head list;
1983 };
1984
1985 struct udp_offload;
1986
1987 struct udp_offload_callbacks {
1988 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1989 struct sk_buff *skb,
1990 struct udp_offload *uoff);
1991 int (*gro_complete)(struct sk_buff *skb,
1992 int nhoff,
1993 struct udp_offload *uoff);
1994 };
1995
1996 struct udp_offload {
1997 __be16 port;
1998 u8 ipproto;
1999 struct udp_offload_callbacks callbacks;
2000 };
2001
2002 /* often modified stats are per cpu, other are shared (netdev->stats) */
2003 struct pcpu_sw_netstats {
2004 u64 rx_packets;
2005 u64 rx_bytes;
2006 u64 tx_packets;
2007 u64 tx_bytes;
2008 struct u64_stats_sync syncp;
2009 };
2010
2011 #define netdev_alloc_pcpu_stats(type) \
2012 ({ \
2013 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2014 if (pcpu_stats) { \
2015 int i; \
2016 for_each_possible_cpu(i) { \
2017 typeof(type) *stat; \
2018 stat = per_cpu_ptr(pcpu_stats, i); \
2019 u64_stats_init(&stat->syncp); \
2020 } \
2021 } \
2022 pcpu_stats; \
2023 })
2024
2025 #include <linux/notifier.h>
2026
2027 /* netdevice notifier chain. Please remember to update the rtnetlink
2028 * notification exclusion list in rtnetlink_event() when adding new
2029 * types.
2030 */
2031 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2032 #define NETDEV_DOWN 0x0002
2033 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2034 detected a hardware crash and restarted
2035 - we can use this eg to kick tcp sessions
2036 once done */
2037 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2038 #define NETDEV_REGISTER 0x0005
2039 #define NETDEV_UNREGISTER 0x0006
2040 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2041 #define NETDEV_CHANGEADDR 0x0008
2042 #define NETDEV_GOING_DOWN 0x0009
2043 #define NETDEV_CHANGENAME 0x000A
2044 #define NETDEV_FEAT_CHANGE 0x000B
2045 #define NETDEV_BONDING_FAILOVER 0x000C
2046 #define NETDEV_PRE_UP 0x000D
2047 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2048 #define NETDEV_POST_TYPE_CHANGE 0x000F
2049 #define NETDEV_POST_INIT 0x0010
2050 #define NETDEV_UNREGISTER_FINAL 0x0011
2051 #define NETDEV_RELEASE 0x0012
2052 #define NETDEV_NOTIFY_PEERS 0x0013
2053 #define NETDEV_JOIN 0x0014
2054 #define NETDEV_CHANGEUPPER 0x0015
2055 #define NETDEV_RESEND_IGMP 0x0016
2056 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2057 #define NETDEV_CHANGEINFODATA 0x0018
2058 #define NETDEV_BONDING_INFO 0x0019
2059
2060 int register_netdevice_notifier(struct notifier_block *nb);
2061 int unregister_netdevice_notifier(struct notifier_block *nb);
2062
2063 struct netdev_notifier_info {
2064 struct net_device *dev;
2065 };
2066
2067 struct netdev_notifier_change_info {
2068 struct netdev_notifier_info info; /* must be first */
2069 unsigned int flags_changed;
2070 };
2071
2072 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2073 struct net_device *dev)
2074 {
2075 info->dev = dev;
2076 }
2077
2078 static inline struct net_device *
2079 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2080 {
2081 return info->dev;
2082 }
2083
2084 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2085
2086
2087 extern rwlock_t dev_base_lock; /* Device list lock */
2088
2089 #define for_each_netdev(net, d) \
2090 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2091 #define for_each_netdev_reverse(net, d) \
2092 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2093 #define for_each_netdev_rcu(net, d) \
2094 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2095 #define for_each_netdev_safe(net, d, n) \
2096 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2097 #define for_each_netdev_continue(net, d) \
2098 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2099 #define for_each_netdev_continue_rcu(net, d) \
2100 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2101 #define for_each_netdev_in_bond_rcu(bond, slave) \
2102 for_each_netdev_rcu(&init_net, slave) \
2103 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2104 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2105
2106 static inline struct net_device *next_net_device(struct net_device *dev)
2107 {
2108 struct list_head *lh;
2109 struct net *net;
2110
2111 net = dev_net(dev);
2112 lh = dev->dev_list.next;
2113 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2114 }
2115
2116 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2117 {
2118 struct list_head *lh;
2119 struct net *net;
2120
2121 net = dev_net(dev);
2122 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2123 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2124 }
2125
2126 static inline struct net_device *first_net_device(struct net *net)
2127 {
2128 return list_empty(&net->dev_base_head) ? NULL :
2129 net_device_entry(net->dev_base_head.next);
2130 }
2131
2132 static inline struct net_device *first_net_device_rcu(struct net *net)
2133 {
2134 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2135
2136 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2137 }
2138
2139 int netdev_boot_setup_check(struct net_device *dev);
2140 unsigned long netdev_boot_base(const char *prefix, int unit);
2141 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2142 const char *hwaddr);
2143 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2144 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2145 void dev_add_pack(struct packet_type *pt);
2146 void dev_remove_pack(struct packet_type *pt);
2147 void __dev_remove_pack(struct packet_type *pt);
2148 void dev_add_offload(struct packet_offload *po);
2149 void dev_remove_offload(struct packet_offload *po);
2150
2151 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2152 unsigned short mask);
2153 struct net_device *dev_get_by_name(struct net *net, const char *name);
2154 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2155 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2156 int dev_alloc_name(struct net_device *dev, const char *name);
2157 int dev_open(struct net_device *dev);
2158 int dev_close(struct net_device *dev);
2159 int dev_close_many(struct list_head *head, bool unlink);
2160 void dev_disable_lro(struct net_device *dev);
2161 int dev_loopback_xmit(struct sk_buff *newskb);
2162 int dev_queue_xmit(struct sk_buff *skb);
2163 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2164 int register_netdevice(struct net_device *dev);
2165 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2166 void unregister_netdevice_many(struct list_head *head);
2167 static inline void unregister_netdevice(struct net_device *dev)
2168 {
2169 unregister_netdevice_queue(dev, NULL);
2170 }
2171
2172 int netdev_refcnt_read(const struct net_device *dev);
2173 void free_netdev(struct net_device *dev);
2174 void netdev_freemem(struct net_device *dev);
2175 void synchronize_net(void);
2176 int init_dummy_netdev(struct net_device *dev);
2177
2178 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2179 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2180 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2181 int netdev_get_name(struct net *net, char *name, int ifindex);
2182 int dev_restart(struct net_device *dev);
2183 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2184
2185 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2186 {
2187 return NAPI_GRO_CB(skb)->data_offset;
2188 }
2189
2190 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2191 {
2192 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2193 }
2194
2195 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2196 {
2197 NAPI_GRO_CB(skb)->data_offset += len;
2198 }
2199
2200 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2201 unsigned int offset)
2202 {
2203 return NAPI_GRO_CB(skb)->frag0 + offset;
2204 }
2205
2206 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2207 {
2208 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2209 }
2210
2211 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2212 unsigned int offset)
2213 {
2214 if (!pskb_may_pull(skb, hlen))
2215 return NULL;
2216
2217 NAPI_GRO_CB(skb)->frag0 = NULL;
2218 NAPI_GRO_CB(skb)->frag0_len = 0;
2219 return skb->data + offset;
2220 }
2221
2222 static inline void *skb_gro_network_header(struct sk_buff *skb)
2223 {
2224 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2225 skb_network_offset(skb);
2226 }
2227
2228 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2229 const void *start, unsigned int len)
2230 {
2231 if (NAPI_GRO_CB(skb)->csum_valid)
2232 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2233 csum_partial(start, len, 0));
2234 }
2235
2236 /* GRO checksum functions. These are logical equivalents of the normal
2237 * checksum functions (in skbuff.h) except that they operate on the GRO
2238 * offsets and fields in sk_buff.
2239 */
2240
2241 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2242
2243 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2244 {
2245 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2246 skb_gro_offset(skb));
2247 }
2248
2249 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2250 bool zero_okay,
2251 __sum16 check)
2252 {
2253 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2254 skb_checksum_start_offset(skb) <
2255 skb_gro_offset(skb)) &&
2256 !skb_at_gro_remcsum_start(skb) &&
2257 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2258 (!zero_okay || check));
2259 }
2260
2261 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2262 __wsum psum)
2263 {
2264 if (NAPI_GRO_CB(skb)->csum_valid &&
2265 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2266 return 0;
2267
2268 NAPI_GRO_CB(skb)->csum = psum;
2269
2270 return __skb_gro_checksum_complete(skb);
2271 }
2272
2273 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2274 {
2275 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2276 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2277 NAPI_GRO_CB(skb)->csum_cnt--;
2278 } else {
2279 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2280 * verified a new top level checksum or an encapsulated one
2281 * during GRO. This saves work if we fallback to normal path.
2282 */
2283 __skb_incr_checksum_unnecessary(skb);
2284 }
2285 }
2286
2287 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2288 compute_pseudo) \
2289 ({ \
2290 __sum16 __ret = 0; \
2291 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2292 __ret = __skb_gro_checksum_validate_complete(skb, \
2293 compute_pseudo(skb, proto)); \
2294 if (__ret) \
2295 __skb_mark_checksum_bad(skb); \
2296 else \
2297 skb_gro_incr_csum_unnecessary(skb); \
2298 __ret; \
2299 })
2300
2301 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2302 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2303
2304 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2305 compute_pseudo) \
2306 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2307
2308 #define skb_gro_checksum_simple_validate(skb) \
2309 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2310
2311 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2312 {
2313 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2314 !NAPI_GRO_CB(skb)->csum_valid);
2315 }
2316
2317 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2318 __sum16 check, __wsum pseudo)
2319 {
2320 NAPI_GRO_CB(skb)->csum = ~pseudo;
2321 NAPI_GRO_CB(skb)->csum_valid = 1;
2322 }
2323
2324 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2325 do { \
2326 if (__skb_gro_checksum_convert_check(skb)) \
2327 __skb_gro_checksum_convert(skb, check, \
2328 compute_pseudo(skb, proto)); \
2329 } while (0)
2330
2331 struct gro_remcsum {
2332 int offset;
2333 __wsum delta;
2334 };
2335
2336 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2337 {
2338 grc->offset = 0;
2339 grc->delta = 0;
2340 }
2341
2342 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2343 int start, int offset,
2344 struct gro_remcsum *grc,
2345 bool nopartial)
2346 {
2347 __wsum delta;
2348
2349 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2350
2351 if (!nopartial) {
2352 NAPI_GRO_CB(skb)->gro_remcsum_start =
2353 ((unsigned char *)ptr + start) - skb->head;
2354 return;
2355 }
2356
2357 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2358
2359 /* Adjust skb->csum since we changed the packet */
2360 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2361
2362 grc->offset = (ptr + offset) - (void *)skb->head;
2363 grc->delta = delta;
2364 }
2365
2366 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2367 struct gro_remcsum *grc)
2368 {
2369 if (!grc->delta)
2370 return;
2371
2372 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2373 }
2374
2375 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2376 unsigned short type,
2377 const void *daddr, const void *saddr,
2378 unsigned int len)
2379 {
2380 if (!dev->header_ops || !dev->header_ops->create)
2381 return 0;
2382
2383 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2384 }
2385
2386 static inline int dev_parse_header(const struct sk_buff *skb,
2387 unsigned char *haddr)
2388 {
2389 const struct net_device *dev = skb->dev;
2390
2391 if (!dev->header_ops || !dev->header_ops->parse)
2392 return 0;
2393 return dev->header_ops->parse(skb, haddr);
2394 }
2395
2396 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2397 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2398 static inline int unregister_gifconf(unsigned int family)
2399 {
2400 return register_gifconf(family, NULL);
2401 }
2402
2403 #ifdef CONFIG_NET_FLOW_LIMIT
2404 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2405 struct sd_flow_limit {
2406 u64 count;
2407 unsigned int num_buckets;
2408 unsigned int history_head;
2409 u16 history[FLOW_LIMIT_HISTORY];
2410 u8 buckets[];
2411 };
2412
2413 extern int netdev_flow_limit_table_len;
2414 #endif /* CONFIG_NET_FLOW_LIMIT */
2415
2416 /*
2417 * Incoming packets are placed on per-cpu queues
2418 */
2419 struct softnet_data {
2420 struct list_head poll_list;
2421 struct sk_buff_head process_queue;
2422
2423 /* stats */
2424 unsigned int processed;
2425 unsigned int time_squeeze;
2426 unsigned int cpu_collision;
2427 unsigned int received_rps;
2428 #ifdef CONFIG_RPS
2429 struct softnet_data *rps_ipi_list;
2430 #endif
2431 #ifdef CONFIG_NET_FLOW_LIMIT
2432 struct sd_flow_limit __rcu *flow_limit;
2433 #endif
2434 struct Qdisc *output_queue;
2435 struct Qdisc **output_queue_tailp;
2436 struct sk_buff *completion_queue;
2437
2438 #ifdef CONFIG_RPS
2439 /* Elements below can be accessed between CPUs for RPS */
2440 struct call_single_data csd ____cacheline_aligned_in_smp;
2441 struct softnet_data *rps_ipi_next;
2442 unsigned int cpu;
2443 unsigned int input_queue_head;
2444 unsigned int input_queue_tail;
2445 #endif
2446 unsigned int dropped;
2447 struct sk_buff_head input_pkt_queue;
2448 struct napi_struct backlog;
2449
2450 };
2451
2452 static inline void input_queue_head_incr(struct softnet_data *sd)
2453 {
2454 #ifdef CONFIG_RPS
2455 sd->input_queue_head++;
2456 #endif
2457 }
2458
2459 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2460 unsigned int *qtail)
2461 {
2462 #ifdef CONFIG_RPS
2463 *qtail = ++sd->input_queue_tail;
2464 #endif
2465 }
2466
2467 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2468
2469 void __netif_schedule(struct Qdisc *q);
2470 void netif_schedule_queue(struct netdev_queue *txq);
2471
2472 static inline void netif_tx_schedule_all(struct net_device *dev)
2473 {
2474 unsigned int i;
2475
2476 for (i = 0; i < dev->num_tx_queues; i++)
2477 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2478 }
2479
2480 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2481 {
2482 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2483 }
2484
2485 /**
2486 * netif_start_queue - allow transmit
2487 * @dev: network device
2488 *
2489 * Allow upper layers to call the device hard_start_xmit routine.
2490 */
2491 static inline void netif_start_queue(struct net_device *dev)
2492 {
2493 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2494 }
2495
2496 static inline void netif_tx_start_all_queues(struct net_device *dev)
2497 {
2498 unsigned int i;
2499
2500 for (i = 0; i < dev->num_tx_queues; i++) {
2501 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2502 netif_tx_start_queue(txq);
2503 }
2504 }
2505
2506 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2507
2508 /**
2509 * netif_wake_queue - restart transmit
2510 * @dev: network device
2511 *
2512 * Allow upper layers to call the device hard_start_xmit routine.
2513 * Used for flow control when transmit resources are available.
2514 */
2515 static inline void netif_wake_queue(struct net_device *dev)
2516 {
2517 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2518 }
2519
2520 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2521 {
2522 unsigned int i;
2523
2524 for (i = 0; i < dev->num_tx_queues; i++) {
2525 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2526 netif_tx_wake_queue(txq);
2527 }
2528 }
2529
2530 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2531 {
2532 if (WARN_ON(!dev_queue)) {
2533 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2534 return;
2535 }
2536 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2537 }
2538
2539 /**
2540 * netif_stop_queue - stop transmitted packets
2541 * @dev: network device
2542 *
2543 * Stop upper layers calling the device hard_start_xmit routine.
2544 * Used for flow control when transmit resources are unavailable.
2545 */
2546 static inline void netif_stop_queue(struct net_device *dev)
2547 {
2548 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2549 }
2550
2551 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2552 {
2553 unsigned int i;
2554
2555 for (i = 0; i < dev->num_tx_queues; i++) {
2556 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2557 netif_tx_stop_queue(txq);
2558 }
2559 }
2560
2561 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2562 {
2563 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2564 }
2565
2566 /**
2567 * netif_queue_stopped - test if transmit queue is flowblocked
2568 * @dev: network device
2569 *
2570 * Test if transmit queue on device is currently unable to send.
2571 */
2572 static inline bool netif_queue_stopped(const struct net_device *dev)
2573 {
2574 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2575 }
2576
2577 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2578 {
2579 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2580 }
2581
2582 static inline bool
2583 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2584 {
2585 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2586 }
2587
2588 static inline bool
2589 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2590 {
2591 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2592 }
2593
2594 /**
2595 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2596 * @dev_queue: pointer to transmit queue
2597 *
2598 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2599 * to give appropriate hint to the cpu.
2600 */
2601 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2602 {
2603 #ifdef CONFIG_BQL
2604 prefetchw(&dev_queue->dql.num_queued);
2605 #endif
2606 }
2607
2608 /**
2609 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2610 * @dev_queue: pointer to transmit queue
2611 *
2612 * BQL enabled drivers might use this helper in their TX completion path,
2613 * to give appropriate hint to the cpu.
2614 */
2615 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2616 {
2617 #ifdef CONFIG_BQL
2618 prefetchw(&dev_queue->dql.limit);
2619 #endif
2620 }
2621
2622 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2623 unsigned int bytes)
2624 {
2625 #ifdef CONFIG_BQL
2626 dql_queued(&dev_queue->dql, bytes);
2627
2628 if (likely(dql_avail(&dev_queue->dql) >= 0))
2629 return;
2630
2631 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2632
2633 /*
2634 * The XOFF flag must be set before checking the dql_avail below,
2635 * because in netdev_tx_completed_queue we update the dql_completed
2636 * before checking the XOFF flag.
2637 */
2638 smp_mb();
2639
2640 /* check again in case another CPU has just made room avail */
2641 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2642 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2643 #endif
2644 }
2645
2646 /**
2647 * netdev_sent_queue - report the number of bytes queued to hardware
2648 * @dev: network device
2649 * @bytes: number of bytes queued to the hardware device queue
2650 *
2651 * Report the number of bytes queued for sending/completion to the network
2652 * device hardware queue. @bytes should be a good approximation and should
2653 * exactly match netdev_completed_queue() @bytes
2654 */
2655 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2656 {
2657 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2658 }
2659
2660 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2661 unsigned int pkts, unsigned int bytes)
2662 {
2663 #ifdef CONFIG_BQL
2664 if (unlikely(!bytes))
2665 return;
2666
2667 dql_completed(&dev_queue->dql, bytes);
2668
2669 /*
2670 * Without the memory barrier there is a small possiblity that
2671 * netdev_tx_sent_queue will miss the update and cause the queue to
2672 * be stopped forever
2673 */
2674 smp_mb();
2675
2676 if (dql_avail(&dev_queue->dql) < 0)
2677 return;
2678
2679 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2680 netif_schedule_queue(dev_queue);
2681 #endif
2682 }
2683
2684 /**
2685 * netdev_completed_queue - report bytes and packets completed by device
2686 * @dev: network device
2687 * @pkts: actual number of packets sent over the medium
2688 * @bytes: actual number of bytes sent over the medium
2689 *
2690 * Report the number of bytes and packets transmitted by the network device
2691 * hardware queue over the physical medium, @bytes must exactly match the
2692 * @bytes amount passed to netdev_sent_queue()
2693 */
2694 static inline void netdev_completed_queue(struct net_device *dev,
2695 unsigned int pkts, unsigned int bytes)
2696 {
2697 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2698 }
2699
2700 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2701 {
2702 #ifdef CONFIG_BQL
2703 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2704 dql_reset(&q->dql);
2705 #endif
2706 }
2707
2708 /**
2709 * netdev_reset_queue - reset the packets and bytes count of a network device
2710 * @dev_queue: network device
2711 *
2712 * Reset the bytes and packet count of a network device and clear the
2713 * software flow control OFF bit for this network device
2714 */
2715 static inline void netdev_reset_queue(struct net_device *dev_queue)
2716 {
2717 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2718 }
2719
2720 /**
2721 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2722 * @dev: network device
2723 * @queue_index: given tx queue index
2724 *
2725 * Returns 0 if given tx queue index >= number of device tx queues,
2726 * otherwise returns the originally passed tx queue index.
2727 */
2728 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2729 {
2730 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2731 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2732 dev->name, queue_index,
2733 dev->real_num_tx_queues);
2734 return 0;
2735 }
2736
2737 return queue_index;
2738 }
2739
2740 /**
2741 * netif_running - test if up
2742 * @dev: network device
2743 *
2744 * Test if the device has been brought up.
2745 */
2746 static inline bool netif_running(const struct net_device *dev)
2747 {
2748 return test_bit(__LINK_STATE_START, &dev->state);
2749 }
2750
2751 /*
2752 * Routines to manage the subqueues on a device. We only need start
2753 * stop, and a check if it's stopped. All other device management is
2754 * done at the overall netdevice level.
2755 * Also test the device if we're multiqueue.
2756 */
2757
2758 /**
2759 * netif_start_subqueue - allow sending packets on subqueue
2760 * @dev: network device
2761 * @queue_index: sub queue index
2762 *
2763 * Start individual transmit queue of a device with multiple transmit queues.
2764 */
2765 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2766 {
2767 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2768
2769 netif_tx_start_queue(txq);
2770 }
2771
2772 /**
2773 * netif_stop_subqueue - stop sending packets on subqueue
2774 * @dev: network device
2775 * @queue_index: sub queue index
2776 *
2777 * Stop individual transmit queue of a device with multiple transmit queues.
2778 */
2779 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2780 {
2781 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2782 netif_tx_stop_queue(txq);
2783 }
2784
2785 /**
2786 * netif_subqueue_stopped - test status of subqueue
2787 * @dev: network device
2788 * @queue_index: sub queue index
2789 *
2790 * Check individual transmit queue of a device with multiple transmit queues.
2791 */
2792 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2793 u16 queue_index)
2794 {
2795 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2796
2797 return netif_tx_queue_stopped(txq);
2798 }
2799
2800 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2801 struct sk_buff *skb)
2802 {
2803 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2804 }
2805
2806 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2807
2808 #ifdef CONFIG_XPS
2809 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2810 u16 index);
2811 #else
2812 static inline int netif_set_xps_queue(struct net_device *dev,
2813 const struct cpumask *mask,
2814 u16 index)
2815 {
2816 return 0;
2817 }
2818 #endif
2819
2820 /*
2821 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2822 * as a distribution range limit for the returned value.
2823 */
2824 static inline u16 skb_tx_hash(const struct net_device *dev,
2825 struct sk_buff *skb)
2826 {
2827 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2828 }
2829
2830 /**
2831 * netif_is_multiqueue - test if device has multiple transmit queues
2832 * @dev: network device
2833 *
2834 * Check if device has multiple transmit queues
2835 */
2836 static inline bool netif_is_multiqueue(const struct net_device *dev)
2837 {
2838 return dev->num_tx_queues > 1;
2839 }
2840
2841 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2842
2843 #ifdef CONFIG_SYSFS
2844 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2845 #else
2846 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2847 unsigned int rxq)
2848 {
2849 return 0;
2850 }
2851 #endif
2852
2853 #ifdef CONFIG_SYSFS
2854 static inline unsigned int get_netdev_rx_queue_index(
2855 struct netdev_rx_queue *queue)
2856 {
2857 struct net_device *dev = queue->dev;
2858 int index = queue - dev->_rx;
2859
2860 BUG_ON(index >= dev->num_rx_queues);
2861 return index;
2862 }
2863 #endif
2864
2865 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2866 int netif_get_num_default_rss_queues(void);
2867
2868 enum skb_free_reason {
2869 SKB_REASON_CONSUMED,
2870 SKB_REASON_DROPPED,
2871 };
2872
2873 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2874 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2875
2876 /*
2877 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2878 * interrupt context or with hardware interrupts being disabled.
2879 * (in_irq() || irqs_disabled())
2880 *
2881 * We provide four helpers that can be used in following contexts :
2882 *
2883 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2884 * replacing kfree_skb(skb)
2885 *
2886 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2887 * Typically used in place of consume_skb(skb) in TX completion path
2888 *
2889 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2890 * replacing kfree_skb(skb)
2891 *
2892 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2893 * and consumed a packet. Used in place of consume_skb(skb)
2894 */
2895 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2896 {
2897 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2898 }
2899
2900 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2901 {
2902 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2903 }
2904
2905 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2906 {
2907 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2908 }
2909
2910 static inline void dev_consume_skb_any(struct sk_buff *skb)
2911 {
2912 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2913 }
2914
2915 int netif_rx(struct sk_buff *skb);
2916 int netif_rx_ni(struct sk_buff *skb);
2917 int netif_receive_skb(struct sk_buff *skb);
2918 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2919 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2920 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2921 gro_result_t napi_gro_frags(struct napi_struct *napi);
2922 struct packet_offload *gro_find_receive_by_type(__be16 type);
2923 struct packet_offload *gro_find_complete_by_type(__be16 type);
2924
2925 static inline void napi_free_frags(struct napi_struct *napi)
2926 {
2927 kfree_skb(napi->skb);
2928 napi->skb = NULL;
2929 }
2930
2931 int netdev_rx_handler_register(struct net_device *dev,
2932 rx_handler_func_t *rx_handler,
2933 void *rx_handler_data);
2934 void netdev_rx_handler_unregister(struct net_device *dev);
2935
2936 bool dev_valid_name(const char *name);
2937 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2938 int dev_ethtool(struct net *net, struct ifreq *);
2939 unsigned int dev_get_flags(const struct net_device *);
2940 int __dev_change_flags(struct net_device *, unsigned int flags);
2941 int dev_change_flags(struct net_device *, unsigned int);
2942 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2943 unsigned int gchanges);
2944 int dev_change_name(struct net_device *, const char *);
2945 int dev_set_alias(struct net_device *, const char *, size_t);
2946 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2947 int dev_set_mtu(struct net_device *, int);
2948 void dev_set_group(struct net_device *, int);
2949 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2950 int dev_change_carrier(struct net_device *, bool new_carrier);
2951 int dev_get_phys_port_id(struct net_device *dev,
2952 struct netdev_phys_item_id *ppid);
2953 int dev_get_phys_port_name(struct net_device *dev,
2954 char *name, size_t len);
2955 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2956 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2957 struct netdev_queue *txq, int *ret);
2958 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2959 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2960 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2961
2962 extern int netdev_budget;
2963
2964 /* Called by rtnetlink.c:rtnl_unlock() */
2965 void netdev_run_todo(void);
2966
2967 /**
2968 * dev_put - release reference to device
2969 * @dev: network device
2970 *
2971 * Release reference to device to allow it to be freed.
2972 */
2973 static inline void dev_put(struct net_device *dev)
2974 {
2975 this_cpu_dec(*dev->pcpu_refcnt);
2976 }
2977
2978 /**
2979 * dev_hold - get reference to device
2980 * @dev: network device
2981 *
2982 * Hold reference to device to keep it from being freed.
2983 */
2984 static inline void dev_hold(struct net_device *dev)
2985 {
2986 this_cpu_inc(*dev->pcpu_refcnt);
2987 }
2988
2989 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2990 * and _off may be called from IRQ context, but it is caller
2991 * who is responsible for serialization of these calls.
2992 *
2993 * The name carrier is inappropriate, these functions should really be
2994 * called netif_lowerlayer_*() because they represent the state of any
2995 * kind of lower layer not just hardware media.
2996 */
2997
2998 void linkwatch_init_dev(struct net_device *dev);
2999 void linkwatch_fire_event(struct net_device *dev);
3000 void linkwatch_forget_dev(struct net_device *dev);
3001
3002 /**
3003 * netif_carrier_ok - test if carrier present
3004 * @dev: network device
3005 *
3006 * Check if carrier is present on device
3007 */
3008 static inline bool netif_carrier_ok(const struct net_device *dev)
3009 {
3010 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3011 }
3012
3013 unsigned long dev_trans_start(struct net_device *dev);
3014
3015 void __netdev_watchdog_up(struct net_device *dev);
3016
3017 void netif_carrier_on(struct net_device *dev);
3018
3019 void netif_carrier_off(struct net_device *dev);
3020
3021 /**
3022 * netif_dormant_on - mark device as dormant.
3023 * @dev: network device
3024 *
3025 * Mark device as dormant (as per RFC2863).
3026 *
3027 * The dormant state indicates that the relevant interface is not
3028 * actually in a condition to pass packets (i.e., it is not 'up') but is
3029 * in a "pending" state, waiting for some external event. For "on-
3030 * demand" interfaces, this new state identifies the situation where the
3031 * interface is waiting for events to place it in the up state.
3032 *
3033 */
3034 static inline void netif_dormant_on(struct net_device *dev)
3035 {
3036 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3037 linkwatch_fire_event(dev);
3038 }
3039
3040 /**
3041 * netif_dormant_off - set device as not dormant.
3042 * @dev: network device
3043 *
3044 * Device is not in dormant state.
3045 */
3046 static inline void netif_dormant_off(struct net_device *dev)
3047 {
3048 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3049 linkwatch_fire_event(dev);
3050 }
3051
3052 /**
3053 * netif_dormant - test if carrier present
3054 * @dev: network device
3055 *
3056 * Check if carrier is present on device
3057 */
3058 static inline bool netif_dormant(const struct net_device *dev)
3059 {
3060 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3061 }
3062
3063
3064 /**
3065 * netif_oper_up - test if device is operational
3066 * @dev: network device
3067 *
3068 * Check if carrier is operational
3069 */
3070 static inline bool netif_oper_up(const struct net_device *dev)
3071 {
3072 return (dev->operstate == IF_OPER_UP ||
3073 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3074 }
3075
3076 /**
3077 * netif_device_present - is device available or removed
3078 * @dev: network device
3079 *
3080 * Check if device has not been removed from system.
3081 */
3082 static inline bool netif_device_present(struct net_device *dev)
3083 {
3084 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3085 }
3086
3087 void netif_device_detach(struct net_device *dev);
3088
3089 void netif_device_attach(struct net_device *dev);
3090
3091 /*
3092 * Network interface message level settings
3093 */
3094
3095 enum {
3096 NETIF_MSG_DRV = 0x0001,
3097 NETIF_MSG_PROBE = 0x0002,
3098 NETIF_MSG_LINK = 0x0004,
3099 NETIF_MSG_TIMER = 0x0008,
3100 NETIF_MSG_IFDOWN = 0x0010,
3101 NETIF_MSG_IFUP = 0x0020,
3102 NETIF_MSG_RX_ERR = 0x0040,
3103 NETIF_MSG_TX_ERR = 0x0080,
3104 NETIF_MSG_TX_QUEUED = 0x0100,
3105 NETIF_MSG_INTR = 0x0200,
3106 NETIF_MSG_TX_DONE = 0x0400,
3107 NETIF_MSG_RX_STATUS = 0x0800,
3108 NETIF_MSG_PKTDATA = 0x1000,
3109 NETIF_MSG_HW = 0x2000,
3110 NETIF_MSG_WOL = 0x4000,
3111 };
3112
3113 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3114 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3115 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3116 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3117 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3118 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3119 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3120 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3121 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3122 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3123 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3124 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3125 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3126 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3127 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3128
3129 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3130 {
3131 /* use default */
3132 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3133 return default_msg_enable_bits;
3134 if (debug_value == 0) /* no output */
3135 return 0;
3136 /* set low N bits */
3137 return (1 << debug_value) - 1;
3138 }
3139
3140 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3141 {
3142 spin_lock(&txq->_xmit_lock);
3143 txq->xmit_lock_owner = cpu;
3144 }
3145
3146 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3147 {
3148 spin_lock_bh(&txq->_xmit_lock);
3149 txq->xmit_lock_owner = smp_processor_id();
3150 }
3151
3152 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3153 {
3154 bool ok = spin_trylock(&txq->_xmit_lock);
3155 if (likely(ok))
3156 txq->xmit_lock_owner = smp_processor_id();
3157 return ok;
3158 }
3159
3160 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3161 {
3162 txq->xmit_lock_owner = -1;
3163 spin_unlock(&txq->_xmit_lock);
3164 }
3165
3166 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3167 {
3168 txq->xmit_lock_owner = -1;
3169 spin_unlock_bh(&txq->_xmit_lock);
3170 }
3171
3172 static inline void txq_trans_update(struct netdev_queue *txq)
3173 {
3174 if (txq->xmit_lock_owner != -1)
3175 txq->trans_start = jiffies;
3176 }
3177
3178 /**
3179 * netif_tx_lock - grab network device transmit lock
3180 * @dev: network device
3181 *
3182 * Get network device transmit lock
3183 */
3184 static inline void netif_tx_lock(struct net_device *dev)
3185 {
3186 unsigned int i;
3187 int cpu;
3188
3189 spin_lock(&dev->tx_global_lock);
3190 cpu = smp_processor_id();
3191 for (i = 0; i < dev->num_tx_queues; i++) {
3192 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3193
3194 /* We are the only thread of execution doing a
3195 * freeze, but we have to grab the _xmit_lock in
3196 * order to synchronize with threads which are in
3197 * the ->hard_start_xmit() handler and already
3198 * checked the frozen bit.
3199 */
3200 __netif_tx_lock(txq, cpu);
3201 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3202 __netif_tx_unlock(txq);
3203 }
3204 }
3205
3206 static inline void netif_tx_lock_bh(struct net_device *dev)
3207 {
3208 local_bh_disable();
3209 netif_tx_lock(dev);
3210 }
3211
3212 static inline void netif_tx_unlock(struct net_device *dev)
3213 {
3214 unsigned int i;
3215
3216 for (i = 0; i < dev->num_tx_queues; i++) {
3217 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3218
3219 /* No need to grab the _xmit_lock here. If the
3220 * queue is not stopped for another reason, we
3221 * force a schedule.
3222 */
3223 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3224 netif_schedule_queue(txq);
3225 }
3226 spin_unlock(&dev->tx_global_lock);
3227 }
3228
3229 static inline void netif_tx_unlock_bh(struct net_device *dev)
3230 {
3231 netif_tx_unlock(dev);
3232 local_bh_enable();
3233 }
3234
3235 #define HARD_TX_LOCK(dev, txq, cpu) { \
3236 if ((dev->features & NETIF_F_LLTX) == 0) { \
3237 __netif_tx_lock(txq, cpu); \
3238 } \
3239 }
3240
3241 #define HARD_TX_TRYLOCK(dev, txq) \
3242 (((dev->features & NETIF_F_LLTX) == 0) ? \
3243 __netif_tx_trylock(txq) : \
3244 true )
3245
3246 #define HARD_TX_UNLOCK(dev, txq) { \
3247 if ((dev->features & NETIF_F_LLTX) == 0) { \
3248 __netif_tx_unlock(txq); \
3249 } \
3250 }
3251
3252 static inline void netif_tx_disable(struct net_device *dev)
3253 {
3254 unsigned int i;
3255 int cpu;
3256
3257 local_bh_disable();
3258 cpu = smp_processor_id();
3259 for (i = 0; i < dev->num_tx_queues; i++) {
3260 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3261
3262 __netif_tx_lock(txq, cpu);
3263 netif_tx_stop_queue(txq);
3264 __netif_tx_unlock(txq);
3265 }
3266 local_bh_enable();
3267 }
3268
3269 static inline void netif_addr_lock(struct net_device *dev)
3270 {
3271 spin_lock(&dev->addr_list_lock);
3272 }
3273
3274 static inline void netif_addr_lock_nested(struct net_device *dev)
3275 {
3276 int subclass = SINGLE_DEPTH_NESTING;
3277
3278 if (dev->netdev_ops->ndo_get_lock_subclass)
3279 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3280
3281 spin_lock_nested(&dev->addr_list_lock, subclass);
3282 }
3283
3284 static inline void netif_addr_lock_bh(struct net_device *dev)
3285 {
3286 spin_lock_bh(&dev->addr_list_lock);
3287 }
3288
3289 static inline void netif_addr_unlock(struct net_device *dev)
3290 {
3291 spin_unlock(&dev->addr_list_lock);
3292 }
3293
3294 static inline void netif_addr_unlock_bh(struct net_device *dev)
3295 {
3296 spin_unlock_bh(&dev->addr_list_lock);
3297 }
3298
3299 /*
3300 * dev_addrs walker. Should be used only for read access. Call with
3301 * rcu_read_lock held.
3302 */
3303 #define for_each_dev_addr(dev, ha) \
3304 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3305
3306 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3307
3308 void ether_setup(struct net_device *dev);
3309
3310 /* Support for loadable net-drivers */
3311 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3312 unsigned char name_assign_type,
3313 void (*setup)(struct net_device *),
3314 unsigned int txqs, unsigned int rxqs);
3315 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3316 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3317
3318 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3319 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3320 count)
3321
3322 int register_netdev(struct net_device *dev);
3323 void unregister_netdev(struct net_device *dev);
3324
3325 /* General hardware address lists handling functions */
3326 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3327 struct netdev_hw_addr_list *from_list, int addr_len);
3328 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3329 struct netdev_hw_addr_list *from_list, int addr_len);
3330 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3331 struct net_device *dev,
3332 int (*sync)(struct net_device *, const unsigned char *),
3333 int (*unsync)(struct net_device *,
3334 const unsigned char *));
3335 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3336 struct net_device *dev,
3337 int (*unsync)(struct net_device *,
3338 const unsigned char *));
3339 void __hw_addr_init(struct netdev_hw_addr_list *list);
3340
3341 /* Functions used for device addresses handling */
3342 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3343 unsigned char addr_type);
3344 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3345 unsigned char addr_type);
3346 void dev_addr_flush(struct net_device *dev);
3347 int dev_addr_init(struct net_device *dev);
3348
3349 /* Functions used for unicast addresses handling */
3350 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3351 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3352 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3353 int dev_uc_sync(struct net_device *to, struct net_device *from);
3354 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3355 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3356 void dev_uc_flush(struct net_device *dev);
3357 void dev_uc_init(struct net_device *dev);
3358
3359 /**
3360 * __dev_uc_sync - Synchonize device's unicast list
3361 * @dev: device to sync
3362 * @sync: function to call if address should be added
3363 * @unsync: function to call if address should be removed
3364 *
3365 * Add newly added addresses to the interface, and release
3366 * addresses that have been deleted.
3367 **/
3368 static inline int __dev_uc_sync(struct net_device *dev,
3369 int (*sync)(struct net_device *,
3370 const unsigned char *),
3371 int (*unsync)(struct net_device *,
3372 const unsigned char *))
3373 {
3374 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3375 }
3376
3377 /**
3378 * __dev_uc_unsync - Remove synchronized addresses from device
3379 * @dev: device to sync
3380 * @unsync: function to call if address should be removed
3381 *
3382 * Remove all addresses that were added to the device by dev_uc_sync().
3383 **/
3384 static inline void __dev_uc_unsync(struct net_device *dev,
3385 int (*unsync)(struct net_device *,
3386 const unsigned char *))
3387 {
3388 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3389 }
3390
3391 /* Functions used for multicast addresses handling */
3392 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3393 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3394 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3395 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3396 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3397 int dev_mc_sync(struct net_device *to, struct net_device *from);
3398 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3399 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3400 void dev_mc_flush(struct net_device *dev);
3401 void dev_mc_init(struct net_device *dev);
3402
3403 /**
3404 * __dev_mc_sync - Synchonize device's multicast list
3405 * @dev: device to sync
3406 * @sync: function to call if address should be added
3407 * @unsync: function to call if address should be removed
3408 *
3409 * Add newly added addresses to the interface, and release
3410 * addresses that have been deleted.
3411 **/
3412 static inline int __dev_mc_sync(struct net_device *dev,
3413 int (*sync)(struct net_device *,
3414 const unsigned char *),
3415 int (*unsync)(struct net_device *,
3416 const unsigned char *))
3417 {
3418 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3419 }
3420
3421 /**
3422 * __dev_mc_unsync - Remove synchronized addresses from device
3423 * @dev: device to sync
3424 * @unsync: function to call if address should be removed
3425 *
3426 * Remove all addresses that were added to the device by dev_mc_sync().
3427 **/
3428 static inline void __dev_mc_unsync(struct net_device *dev,
3429 int (*unsync)(struct net_device *,
3430 const unsigned char *))
3431 {
3432 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3433 }
3434
3435 /* Functions used for secondary unicast and multicast support */
3436 void dev_set_rx_mode(struct net_device *dev);
3437 void __dev_set_rx_mode(struct net_device *dev);
3438 int dev_set_promiscuity(struct net_device *dev, int inc);
3439 int dev_set_allmulti(struct net_device *dev, int inc);
3440 void netdev_state_change(struct net_device *dev);
3441 void netdev_notify_peers(struct net_device *dev);
3442 void netdev_features_change(struct net_device *dev);
3443 /* Load a device via the kmod */
3444 void dev_load(struct net *net, const char *name);
3445 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3446 struct rtnl_link_stats64 *storage);
3447 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3448 const struct net_device_stats *netdev_stats);
3449
3450 extern int netdev_max_backlog;
3451 extern int netdev_tstamp_prequeue;
3452 extern int weight_p;
3453 extern int bpf_jit_enable;
3454
3455 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3456 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3457 struct list_head **iter);
3458 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3459 struct list_head **iter);
3460
3461 /* iterate through upper list, must be called under RCU read lock */
3462 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3463 for (iter = &(dev)->adj_list.upper, \
3464 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3465 updev; \
3466 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3467
3468 /* iterate through upper list, must be called under RCU read lock */
3469 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3470 for (iter = &(dev)->all_adj_list.upper, \
3471 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3472 updev; \
3473 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3474
3475 void *netdev_lower_get_next_private(struct net_device *dev,
3476 struct list_head **iter);
3477 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3478 struct list_head **iter);
3479
3480 #define netdev_for_each_lower_private(dev, priv, iter) \
3481 for (iter = (dev)->adj_list.lower.next, \
3482 priv = netdev_lower_get_next_private(dev, &(iter)); \
3483 priv; \
3484 priv = netdev_lower_get_next_private(dev, &(iter)))
3485
3486 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3487 for (iter = &(dev)->adj_list.lower, \
3488 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3489 priv; \
3490 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3491
3492 void *netdev_lower_get_next(struct net_device *dev,
3493 struct list_head **iter);
3494 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3495 for (iter = &(dev)->adj_list.lower, \
3496 ldev = netdev_lower_get_next(dev, &(iter)); \
3497 ldev; \
3498 ldev = netdev_lower_get_next(dev, &(iter)))
3499
3500 void *netdev_adjacent_get_private(struct list_head *adj_list);
3501 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3502 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3503 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3504 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3505 int netdev_master_upper_dev_link(struct net_device *dev,
3506 struct net_device *upper_dev);
3507 int netdev_master_upper_dev_link_private(struct net_device *dev,
3508 struct net_device *upper_dev,
3509 void *private);
3510 void netdev_upper_dev_unlink(struct net_device *dev,
3511 struct net_device *upper_dev);
3512 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3513 void *netdev_lower_dev_get_private(struct net_device *dev,
3514 struct net_device *lower_dev);
3515
3516 /* RSS keys are 40 or 52 bytes long */
3517 #define NETDEV_RSS_KEY_LEN 52
3518 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3519 void netdev_rss_key_fill(void *buffer, size_t len);
3520
3521 int dev_get_nest_level(struct net_device *dev,
3522 bool (*type_check)(struct net_device *dev));
3523 int skb_checksum_help(struct sk_buff *skb);
3524 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3525 netdev_features_t features, bool tx_path);
3526 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3527 netdev_features_t features);
3528
3529 struct netdev_bonding_info {
3530 ifslave slave;
3531 ifbond master;
3532 };
3533
3534 struct netdev_notifier_bonding_info {
3535 struct netdev_notifier_info info; /* must be first */
3536 struct netdev_bonding_info bonding_info;
3537 };
3538
3539 void netdev_bonding_info_change(struct net_device *dev,
3540 struct netdev_bonding_info *bonding_info);
3541
3542 static inline
3543 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3544 {
3545 return __skb_gso_segment(skb, features, true);
3546 }
3547 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3548
3549 static inline bool can_checksum_protocol(netdev_features_t features,
3550 __be16 protocol)
3551 {
3552 return ((features & NETIF_F_GEN_CSUM) ||
3553 ((features & NETIF_F_V4_CSUM) &&
3554 protocol == htons(ETH_P_IP)) ||
3555 ((features & NETIF_F_V6_CSUM) &&
3556 protocol == htons(ETH_P_IPV6)) ||
3557 ((features & NETIF_F_FCOE_CRC) &&
3558 protocol == htons(ETH_P_FCOE)));
3559 }
3560
3561 #ifdef CONFIG_BUG
3562 void netdev_rx_csum_fault(struct net_device *dev);
3563 #else
3564 static inline void netdev_rx_csum_fault(struct net_device *dev)
3565 {
3566 }
3567 #endif
3568 /* rx skb timestamps */
3569 void net_enable_timestamp(void);
3570 void net_disable_timestamp(void);
3571
3572 #ifdef CONFIG_PROC_FS
3573 int __init dev_proc_init(void);
3574 #else
3575 #define dev_proc_init() 0
3576 #endif
3577
3578 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3579 struct sk_buff *skb, struct net_device *dev,
3580 bool more)
3581 {
3582 skb->xmit_more = more ? 1 : 0;
3583 return ops->ndo_start_xmit(skb, dev);
3584 }
3585
3586 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3587 struct netdev_queue *txq, bool more)
3588 {
3589 const struct net_device_ops *ops = dev->netdev_ops;
3590 int rc;
3591
3592 rc = __netdev_start_xmit(ops, skb, dev, more);
3593 if (rc == NETDEV_TX_OK)
3594 txq_trans_update(txq);
3595
3596 return rc;
3597 }
3598
3599 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3600 const void *ns);
3601 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3602 const void *ns);
3603
3604 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3605 {
3606 return netdev_class_create_file_ns(class_attr, NULL);
3607 }
3608
3609 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3610 {
3611 netdev_class_remove_file_ns(class_attr, NULL);
3612 }
3613
3614 extern struct kobj_ns_type_operations net_ns_type_operations;
3615
3616 const char *netdev_drivername(const struct net_device *dev);
3617
3618 void linkwatch_run_queue(void);
3619
3620 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3621 netdev_features_t f2)
3622 {
3623 if (f1 & NETIF_F_GEN_CSUM)
3624 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3625 if (f2 & NETIF_F_GEN_CSUM)
3626 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3627 f1 &= f2;
3628 if (f1 & NETIF_F_GEN_CSUM)
3629 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3630
3631 return f1;
3632 }
3633
3634 static inline netdev_features_t netdev_get_wanted_features(
3635 struct net_device *dev)
3636 {
3637 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3638 }
3639 netdev_features_t netdev_increment_features(netdev_features_t all,
3640 netdev_features_t one, netdev_features_t mask);
3641
3642 /* Allow TSO being used on stacked device :
3643 * Performing the GSO segmentation before last device
3644 * is a performance improvement.
3645 */
3646 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3647 netdev_features_t mask)
3648 {
3649 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3650 }
3651
3652 int __netdev_update_features(struct net_device *dev);
3653 void netdev_update_features(struct net_device *dev);
3654 void netdev_change_features(struct net_device *dev);
3655
3656 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3657 struct net_device *dev);
3658
3659 netdev_features_t netif_skb_features(struct sk_buff *skb);
3660
3661 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3662 {
3663 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3664
3665 /* check flags correspondence */
3666 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3667 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3668 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3669 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3670 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3671 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3672 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3673 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3674 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3675 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3676 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3677 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3678 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3679
3680 return (features & feature) == feature;
3681 }
3682
3683 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3684 {
3685 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3686 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3687 }
3688
3689 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3690 netdev_features_t features)
3691 {
3692 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3693 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3694 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3695 }
3696
3697 static inline void netif_set_gso_max_size(struct net_device *dev,
3698 unsigned int size)
3699 {
3700 dev->gso_max_size = size;
3701 }
3702
3703 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3704 int pulled_hlen, u16 mac_offset,
3705 int mac_len)
3706 {
3707 skb->protocol = protocol;
3708 skb->encapsulation = 1;
3709 skb_push(skb, pulled_hlen);
3710 skb_reset_transport_header(skb);
3711 skb->mac_header = mac_offset;
3712 skb->network_header = skb->mac_header + mac_len;
3713 skb->mac_len = mac_len;
3714 }
3715
3716 static inline bool netif_is_macvlan(struct net_device *dev)
3717 {
3718 return dev->priv_flags & IFF_MACVLAN;
3719 }
3720
3721 static inline bool netif_is_macvlan_port(struct net_device *dev)
3722 {
3723 return dev->priv_flags & IFF_MACVLAN_PORT;
3724 }
3725
3726 static inline bool netif_is_ipvlan(struct net_device *dev)
3727 {
3728 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3729 }
3730
3731 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3732 {
3733 return dev->priv_flags & IFF_IPVLAN_MASTER;
3734 }
3735
3736 static inline bool netif_is_bond_master(struct net_device *dev)
3737 {
3738 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3739 }
3740
3741 static inline bool netif_is_bond_slave(struct net_device *dev)
3742 {
3743 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3744 }
3745
3746 static inline bool netif_supports_nofcs(struct net_device *dev)
3747 {
3748 return dev->priv_flags & IFF_SUPP_NOFCS;
3749 }
3750
3751 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3752 static inline void netif_keep_dst(struct net_device *dev)
3753 {
3754 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3755 }
3756
3757 extern struct pernet_operations __net_initdata loopback_net_ops;
3758
3759 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3760
3761 /* netdev_printk helpers, similar to dev_printk */
3762
3763 static inline const char *netdev_name(const struct net_device *dev)
3764 {
3765 if (!dev->name[0] || strchr(dev->name, '%'))
3766 return "(unnamed net_device)";
3767 return dev->name;
3768 }
3769
3770 static inline const char *netdev_reg_state(const struct net_device *dev)
3771 {
3772 switch (dev->reg_state) {
3773 case NETREG_UNINITIALIZED: return " (uninitialized)";
3774 case NETREG_REGISTERED: return "";
3775 case NETREG_UNREGISTERING: return " (unregistering)";
3776 case NETREG_UNREGISTERED: return " (unregistered)";
3777 case NETREG_RELEASED: return " (released)";
3778 case NETREG_DUMMY: return " (dummy)";
3779 }
3780
3781 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3782 return " (unknown)";
3783 }
3784
3785 __printf(3, 4)
3786 void netdev_printk(const char *level, const struct net_device *dev,
3787 const char *format, ...);
3788 __printf(2, 3)
3789 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3790 __printf(2, 3)
3791 void netdev_alert(const struct net_device *dev, const char *format, ...);
3792 __printf(2, 3)
3793 void netdev_crit(const struct net_device *dev, const char *format, ...);
3794 __printf(2, 3)
3795 void netdev_err(const struct net_device *dev, const char *format, ...);
3796 __printf(2, 3)
3797 void netdev_warn(const struct net_device *dev, const char *format, ...);
3798 __printf(2, 3)
3799 void netdev_notice(const struct net_device *dev, const char *format, ...);
3800 __printf(2, 3)
3801 void netdev_info(const struct net_device *dev, const char *format, ...);
3802
3803 #define MODULE_ALIAS_NETDEV(device) \
3804 MODULE_ALIAS("netdev-" device)
3805
3806 #if defined(CONFIG_DYNAMIC_DEBUG)
3807 #define netdev_dbg(__dev, format, args...) \
3808 do { \
3809 dynamic_netdev_dbg(__dev, format, ##args); \
3810 } while (0)
3811 #elif defined(DEBUG)
3812 #define netdev_dbg(__dev, format, args...) \
3813 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3814 #else
3815 #define netdev_dbg(__dev, format, args...) \
3816 ({ \
3817 if (0) \
3818 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3819 })
3820 #endif
3821
3822 #if defined(VERBOSE_DEBUG)
3823 #define netdev_vdbg netdev_dbg
3824 #else
3825
3826 #define netdev_vdbg(dev, format, args...) \
3827 ({ \
3828 if (0) \
3829 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3830 0; \
3831 })
3832 #endif
3833
3834 /*
3835 * netdev_WARN() acts like dev_printk(), but with the key difference
3836 * of using a WARN/WARN_ON to get the message out, including the
3837 * file/line information and a backtrace.
3838 */
3839 #define netdev_WARN(dev, format, args...) \
3840 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3841 netdev_reg_state(dev), ##args)
3842
3843 /* netif printk helpers, similar to netdev_printk */
3844
3845 #define netif_printk(priv, type, level, dev, fmt, args...) \
3846 do { \
3847 if (netif_msg_##type(priv)) \
3848 netdev_printk(level, (dev), fmt, ##args); \
3849 } while (0)
3850
3851 #define netif_level(level, priv, type, dev, fmt, args...) \
3852 do { \
3853 if (netif_msg_##type(priv)) \
3854 netdev_##level(dev, fmt, ##args); \
3855 } while (0)
3856
3857 #define netif_emerg(priv, type, dev, fmt, args...) \
3858 netif_level(emerg, priv, type, dev, fmt, ##args)
3859 #define netif_alert(priv, type, dev, fmt, args...) \
3860 netif_level(alert, priv, type, dev, fmt, ##args)
3861 #define netif_crit(priv, type, dev, fmt, args...) \
3862 netif_level(crit, priv, type, dev, fmt, ##args)
3863 #define netif_err(priv, type, dev, fmt, args...) \
3864 netif_level(err, priv, type, dev, fmt, ##args)
3865 #define netif_warn(priv, type, dev, fmt, args...) \
3866 netif_level(warn, priv, type, dev, fmt, ##args)
3867 #define netif_notice(priv, type, dev, fmt, args...) \
3868 netif_level(notice, priv, type, dev, fmt, ##args)
3869 #define netif_info(priv, type, dev, fmt, args...) \
3870 netif_level(info, priv, type, dev, fmt, ##args)
3871
3872 #if defined(CONFIG_DYNAMIC_DEBUG)
3873 #define netif_dbg(priv, type, netdev, format, args...) \
3874 do { \
3875 if (netif_msg_##type(priv)) \
3876 dynamic_netdev_dbg(netdev, format, ##args); \
3877 } while (0)
3878 #elif defined(DEBUG)
3879 #define netif_dbg(priv, type, dev, format, args...) \
3880 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3881 #else
3882 #define netif_dbg(priv, type, dev, format, args...) \
3883 ({ \
3884 if (0) \
3885 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3886 0; \
3887 })
3888 #endif
3889
3890 #if defined(VERBOSE_DEBUG)
3891 #define netif_vdbg netif_dbg
3892 #else
3893 #define netif_vdbg(priv, type, dev, format, args...) \
3894 ({ \
3895 if (0) \
3896 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3897 0; \
3898 })
3899 #endif
3900
3901 /*
3902 * The list of packet types we will receive (as opposed to discard)
3903 * and the routines to invoke.
3904 *
3905 * Why 16. Because with 16 the only overlap we get on a hash of the
3906 * low nibble of the protocol value is RARP/SNAP/X.25.
3907 *
3908 * NOTE: That is no longer true with the addition of VLAN tags. Not
3909 * sure which should go first, but I bet it won't make much
3910 * difference if we are running VLANs. The good news is that
3911 * this protocol won't be in the list unless compiled in, so
3912 * the average user (w/out VLANs) will not be adversely affected.
3913 * --BLG
3914 *
3915 * 0800 IP
3916 * 8100 802.1Q VLAN
3917 * 0001 802.3
3918 * 0002 AX.25
3919 * 0004 802.2
3920 * 8035 RARP
3921 * 0005 SNAP
3922 * 0805 X.25
3923 * 0806 ARP
3924 * 8137 IPX
3925 * 0009 Localtalk
3926 * 86DD IPv6
3927 */
3928 #define PTYPE_HASH_SIZE (16)
3929 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3930
3931 #endif /* _LINUX_NETDEVICE_H */
This page took 0.114423 seconds and 5 git commands to generate.