]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/netdevice.h
net: sched: convert qdisc linked list to hashtable
[mirror_ubuntu-artful-kernel.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
56
57 struct netpoll_info;
58 struct device;
59 struct phy_device;
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
68
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
71
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
75
76 /*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
91 */
92
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
98
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
107
108 enum netdev_tx {
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
112 };
113 typedef enum netdev_tx netdev_tx_t;
114
115 /*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
119 static inline bool dev_xmit_complete(int rc)
120 {
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131 }
132
133 /*
134 * Compute the worst-case header length according to the protocols
135 * used.
136 */
137
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
143 # else
144 # define LL_MAX_HEADER 96
145 # endif
146 #else
147 # define LL_MAX_HEADER 32
148 #endif
149
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
153 #else
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
155 #endif
156
157 /*
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
160 */
161
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186 };
187
188
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
191
192 #ifdef CONFIG_RPS
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
195 #endif
196
197 struct neighbour;
198 struct neigh_parms;
199 struct sk_buff;
200
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
210 bool global_use;
211 int sync_cnt;
212 int refcount;
213 int synced;
214 struct rcu_head rcu_head;
215 };
216
217 struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220 };
221
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236
237 struct hh_cache {
238 u16 hh_len;
239 u16 __pad;
240 seqlock_t hh_lock;
241
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249 };
250
251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263
264 struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
273 bool (*validate)(const char *ll_header, unsigned int len);
274 };
275
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
278 * code.
279 */
280
281 enum netdev_state_t {
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
287 };
288
289
290 /*
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
293 */
294 struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297 };
298 #define NETDEV_BOOT_SETUP_MAX 8
299
300 int __init netdev_boot_setup(char *str);
301
302 /*
303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305 struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-CPU poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
320 int poll_owner;
321 #endif
322 struct net_device *dev;
323 struct sk_buff *gro_list;
324 struct sk_buff *skb;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
329 };
330
331 enum {
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
337 };
338
339 enum gro_result {
340 GRO_MERGED,
341 GRO_MERGED_FREE,
342 GRO_HELD,
343 GRO_NORMAL,
344 GRO_DROP,
345 };
346 typedef enum gro_result gro_result_t;
347
348 /*
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * further.
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
356 *
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
359 *
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
362 *
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
366 *
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
368 * do with the skb.
369 *
370 * If the rx_handler consumed the skb in some way, it should return
371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
372 * the skb to be delivered in some other way.
373 *
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
377 *
378 * If the rx_handler decides the skb should be ignored, it should return
379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
380 * are registered on exact device (ptype->dev == skb->dev).
381 *
382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
383 * delivered, it should return RX_HANDLER_PASS.
384 *
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
387 */
388
389 enum rx_handler_result {
390 RX_HANDLER_CONSUMED,
391 RX_HANDLER_ANOTHER,
392 RX_HANDLER_EXACT,
393 RX_HANDLER_PASS,
394 };
395 typedef enum rx_handler_result rx_handler_result_t;
396 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
397
398 void __napi_schedule(struct napi_struct *n);
399 void __napi_schedule_irqoff(struct napi_struct *n);
400
401 static inline bool napi_disable_pending(struct napi_struct *n)
402 {
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
404 }
405
406 /**
407 * napi_schedule_prep - check if NAPI can be scheduled
408 * @n: NAPI context
409 *
410 * Test if NAPI routine is already running, and if not mark
411 * it as running. This is used as a condition variable to
412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
414 */
415 static inline bool napi_schedule_prep(struct napi_struct *n)
416 {
417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
419 }
420
421 /**
422 * napi_schedule - schedule NAPI poll
423 * @n: NAPI context
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428 static inline void napi_schedule(struct napi_struct *n)
429 {
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432 }
433
434 /**
435 * napi_schedule_irqoff - schedule NAPI poll
436 * @n: NAPI context
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
441 {
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444 }
445
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct *napi)
448 {
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
451 return true;
452 }
453 return false;
454 }
455
456 void __napi_complete(struct napi_struct *n);
457 void napi_complete_done(struct napi_struct *n, int work_done);
458 /**
459 * napi_complete - NAPI processing complete
460 * @n: NAPI context
461 *
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
464 */
465 static inline void napi_complete(struct napi_struct *n)
466 {
467 return napi_complete_done(n, 0);
468 }
469
470 /**
471 * napi_hash_add - add a NAPI to global hashtable
472 * @napi: NAPI context
473 *
474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
476 * Note: This is normally automatically done from netif_napi_add(),
477 * so might disappear in a future Linux version.
478 */
479 void napi_hash_add(struct napi_struct *napi);
480
481 /**
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: NAPI context
484 *
485 * Warning: caller must observe RCU grace period
486 * before freeing memory containing @napi, if
487 * this function returns true.
488 * Note: core networking stack automatically calls it
489 * from netif_napi_del().
490 * Drivers might want to call this helper to combine all
491 * the needed RCU grace periods into a single one.
492 */
493 bool napi_hash_del(struct napi_struct *napi);
494
495 /**
496 * napi_disable - prevent NAPI from scheduling
497 * @n: NAPI context
498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
502 void napi_disable(struct napi_struct *n);
503
504 /**
505 * napi_enable - enable NAPI scheduling
506 * @n: NAPI context
507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
511 static inline void napi_enable(struct napi_struct *n)
512 {
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 smp_mb__before_atomic();
515 clear_bit(NAPI_STATE_SCHED, &n->state);
516 clear_bit(NAPI_STATE_NPSVC, &n->state);
517 }
518
519 /**
520 * napi_synchronize - wait until NAPI is not running
521 * @n: NAPI context
522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
527 static inline void napi_synchronize(const struct napi_struct *n)
528 {
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
534 }
535
536 enum netdev_queue_state_t {
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
539 __QUEUE_STATE_FROZEN,
540 };
541
542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
552 /*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
561
562 struct netdev_queue {
563 /*
564 * read-mostly part
565 */
566 struct net_device *dev;
567 struct Qdisc __rcu *qdisc;
568 struct Qdisc *qdisc_sleeping;
569 #ifdef CONFIG_SYSFS
570 struct kobject kobj;
571 #endif
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574 #endif
575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
581 /*
582 * write-mostly part
583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
586 /*
587 * Time (in jiffies) of last Tx
588 */
589 unsigned long trans_start;
590
591 unsigned long state;
592
593 #ifdef CONFIG_BQL
594 struct dql dql;
595 #endif
596 } ____cacheline_aligned_in_smp;
597
598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599 {
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602 #else
603 return NUMA_NO_NODE;
604 #endif
605 }
606
607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608 {
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611 #endif
612 }
613
614 #ifdef CONFIG_RPS
615 /*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619 struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623 };
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
625
626 /*
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
630 */
631 struct rps_dev_flow {
632 u16 cpu;
633 u16 filter;
634 unsigned int last_qtail;
635 };
636 #define RPS_NO_FILTER 0xffff
637
638 /*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641 struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
644 struct rps_dev_flow flows[0];
645 };
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647 ((_num) * sizeof(struct rps_dev_flow)))
648
649 /*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
654 * rps_cpu_mask is used to partition the space, depending on number of
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
657 * meaning we use 32-6=26 bits for the hash.
658 */
659 struct rps_sock_flow_table {
660 u32 mask;
661
662 u32 ents[0] ____cacheline_aligned_in_smp;
663 };
664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665
666 #define RPS_NO_CPU 0xffff
667
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673 {
674 if (table && hash) {
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
677
678 /* We only give a hint, preemption can change CPU under us */
679 val |= raw_smp_processor_id();
680
681 if (table->ents[index] != val)
682 table->ents[index] = val;
683 }
684 }
685
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
689 #endif
690 #endif /* CONFIG_RPS */
691
692 /* This structure contains an instance of an RX queue. */
693 struct netdev_rx_queue {
694 #ifdef CONFIG_RPS
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
697 #endif
698 struct kobject kobj;
699 struct net_device *dev;
700 } ____cacheline_aligned_in_smp;
701
702 /*
703 * RX queue sysfs structures and functions.
704 */
705 struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711 };
712
713 #ifdef CONFIG_XPS
714 /*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718 struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723 };
724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
727
728 /*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731 struct xps_dev_maps {
732 struct rcu_head rcu;
733 struct xps_map __rcu *cpu_map[0];
734 };
735 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737 #endif /* CONFIG_XPS */
738
739 #define TC_MAX_QUEUE 16
740 #define TC_BITMASK 15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745 };
746
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 /*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752 struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761 };
762 #endif
763
764 #define MAX_PHYS_ITEM_ID_LEN 32
765
766 /* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
768 */
769 struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 unsigned char id_len;
772 };
773
774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776 {
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779 }
780
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
784 /* These structures hold the attributes of qdisc and classifiers
785 * that are being passed to the netdevice through the setup_tc op.
786 */
787 enum {
788 TC_SETUP_MQPRIO,
789 TC_SETUP_CLSU32,
790 TC_SETUP_CLSFLOWER,
791 TC_SETUP_MATCHALL,
792 };
793
794 struct tc_cls_u32_offload;
795
796 struct tc_to_netdev {
797 unsigned int type;
798 union {
799 u8 tc;
800 struct tc_cls_u32_offload *cls_u32;
801 struct tc_cls_flower_offload *cls_flower;
802 struct tc_cls_matchall_offload *cls_mall;
803 };
804 };
805
806 /* These structures hold the attributes of xdp state that are being passed
807 * to the netdevice through the xdp op.
808 */
809 enum xdp_netdev_command {
810 /* Set or clear a bpf program used in the earliest stages of packet
811 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
812 * is responsible for calling bpf_prog_put on any old progs that are
813 * stored. In case of error, the callee need not release the new prog
814 * reference, but on success it takes ownership and must bpf_prog_put
815 * when it is no longer used.
816 */
817 XDP_SETUP_PROG,
818 /* Check if a bpf program is set on the device. The callee should
819 * return true if a program is currently attached and running.
820 */
821 XDP_QUERY_PROG,
822 };
823
824 struct netdev_xdp {
825 enum xdp_netdev_command command;
826 union {
827 /* XDP_SETUP_PROG */
828 struct bpf_prog *prog;
829 /* XDP_QUERY_PROG */
830 bool prog_attached;
831 };
832 };
833
834 /*
835 * This structure defines the management hooks for network devices.
836 * The following hooks can be defined; unless noted otherwise, they are
837 * optional and can be filled with a null pointer.
838 *
839 * int (*ndo_init)(struct net_device *dev);
840 * This function is called once when a network device is registered.
841 * The network device can use this for any late stage initialization
842 * or semantic validation. It can fail with an error code which will
843 * be propagated back to register_netdev.
844 *
845 * void (*ndo_uninit)(struct net_device *dev);
846 * This function is called when device is unregistered or when registration
847 * fails. It is not called if init fails.
848 *
849 * int (*ndo_open)(struct net_device *dev);
850 * This function is called when a network device transitions to the up
851 * state.
852 *
853 * int (*ndo_stop)(struct net_device *dev);
854 * This function is called when a network device transitions to the down
855 * state.
856 *
857 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
858 * struct net_device *dev);
859 * Called when a packet needs to be transmitted.
860 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
861 * the queue before that can happen; it's for obsolete devices and weird
862 * corner cases, but the stack really does a non-trivial amount
863 * of useless work if you return NETDEV_TX_BUSY.
864 * Required; cannot be NULL.
865 *
866 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
867 * netdev_features_t features);
868 * Adjusts the requested feature flags according to device-specific
869 * constraints, and returns the resulting flags. Must not modify
870 * the device state.
871 *
872 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
873 * void *accel_priv, select_queue_fallback_t fallback);
874 * Called to decide which queue to use when device supports multiple
875 * transmit queues.
876 *
877 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
878 * This function is called to allow device receiver to make
879 * changes to configuration when multicast or promiscuous is enabled.
880 *
881 * void (*ndo_set_rx_mode)(struct net_device *dev);
882 * This function is called device changes address list filtering.
883 * If driver handles unicast address filtering, it should set
884 * IFF_UNICAST_FLT in its priv_flags.
885 *
886 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
887 * This function is called when the Media Access Control address
888 * needs to be changed. If this interface is not defined, the
889 * MAC address can not be changed.
890 *
891 * int (*ndo_validate_addr)(struct net_device *dev);
892 * Test if Media Access Control address is valid for the device.
893 *
894 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
895 * Called when a user requests an ioctl which can't be handled by
896 * the generic interface code. If not defined ioctls return
897 * not supported error code.
898 *
899 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
900 * Used to set network devices bus interface parameters. This interface
901 * is retained for legacy reasons; new devices should use the bus
902 * interface (PCI) for low level management.
903 *
904 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
905 * Called when a user wants to change the Maximum Transfer Unit
906 * of a device. If not defined, any request to change MTU will
907 * will return an error.
908 *
909 * void (*ndo_tx_timeout)(struct net_device *dev);
910 * Callback used when the transmitter has not made any progress
911 * for dev->watchdog ticks.
912 *
913 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
914 * struct rtnl_link_stats64 *storage);
915 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
916 * Called when a user wants to get the network device usage
917 * statistics. Drivers must do one of the following:
918 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
919 * rtnl_link_stats64 structure passed by the caller.
920 * 2. Define @ndo_get_stats to update a net_device_stats structure
921 * (which should normally be dev->stats) and return a pointer to
922 * it. The structure may be changed asynchronously only if each
923 * field is written atomically.
924 * 3. Update dev->stats asynchronously and atomically, and define
925 * neither operation.
926 *
927 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
928 * If device supports VLAN filtering this function is called when a
929 * VLAN id is registered.
930 *
931 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
932 * If device supports VLAN filtering this function is called when a
933 * VLAN id is unregistered.
934 *
935 * void (*ndo_poll_controller)(struct net_device *dev);
936 *
937 * SR-IOV management functions.
938 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
939 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
940 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
941 * int max_tx_rate);
942 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
943 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
944 * int (*ndo_get_vf_config)(struct net_device *dev,
945 * int vf, struct ifla_vf_info *ivf);
946 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
947 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
948 * struct nlattr *port[]);
949 *
950 * Enable or disable the VF ability to query its RSS Redirection Table and
951 * Hash Key. This is needed since on some devices VF share this information
952 * with PF and querying it may introduce a theoretical security risk.
953 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
954 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
955 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
956 * Called to setup 'tc' number of traffic classes in the net device. This
957 * is always called from the stack with the rtnl lock held and netif tx
958 * queues stopped. This allows the netdevice to perform queue management
959 * safely.
960 *
961 * Fiber Channel over Ethernet (FCoE) offload functions.
962 * int (*ndo_fcoe_enable)(struct net_device *dev);
963 * Called when the FCoE protocol stack wants to start using LLD for FCoE
964 * so the underlying device can perform whatever needed configuration or
965 * initialization to support acceleration of FCoE traffic.
966 *
967 * int (*ndo_fcoe_disable)(struct net_device *dev);
968 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
969 * so the underlying device can perform whatever needed clean-ups to
970 * stop supporting acceleration of FCoE traffic.
971 *
972 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
973 * struct scatterlist *sgl, unsigned int sgc);
974 * Called when the FCoE Initiator wants to initialize an I/O that
975 * is a possible candidate for Direct Data Placement (DDP). The LLD can
976 * perform necessary setup and returns 1 to indicate the device is set up
977 * successfully to perform DDP on this I/O, otherwise this returns 0.
978 *
979 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
980 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
981 * indicated by the FC exchange id 'xid', so the underlying device can
982 * clean up and reuse resources for later DDP requests.
983 *
984 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
985 * struct scatterlist *sgl, unsigned int sgc);
986 * Called when the FCoE Target wants to initialize an I/O that
987 * is a possible candidate for Direct Data Placement (DDP). The LLD can
988 * perform necessary setup and returns 1 to indicate the device is set up
989 * successfully to perform DDP on this I/O, otherwise this returns 0.
990 *
991 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
992 * struct netdev_fcoe_hbainfo *hbainfo);
993 * Called when the FCoE Protocol stack wants information on the underlying
994 * device. This information is utilized by the FCoE protocol stack to
995 * register attributes with Fiber Channel management service as per the
996 * FC-GS Fabric Device Management Information(FDMI) specification.
997 *
998 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
999 * Called when the underlying device wants to override default World Wide
1000 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1001 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1002 * protocol stack to use.
1003 *
1004 * RFS acceleration.
1005 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1006 * u16 rxq_index, u32 flow_id);
1007 * Set hardware filter for RFS. rxq_index is the target queue index;
1008 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1009 * Return the filter ID on success, or a negative error code.
1010 *
1011 * Slave management functions (for bridge, bonding, etc).
1012 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1013 * Called to make another netdev an underling.
1014 *
1015 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1016 * Called to release previously enslaved netdev.
1017 *
1018 * Feature/offload setting functions.
1019 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1020 * Called to update device configuration to new features. Passed
1021 * feature set might be less than what was returned by ndo_fix_features()).
1022 * Must return >0 or -errno if it changed dev->features itself.
1023 *
1024 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1025 * struct net_device *dev,
1026 * const unsigned char *addr, u16 vid, u16 flags)
1027 * Adds an FDB entry to dev for addr.
1028 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1029 * struct net_device *dev,
1030 * const unsigned char *addr, u16 vid)
1031 * Deletes the FDB entry from dev coresponding to addr.
1032 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1033 * struct net_device *dev, struct net_device *filter_dev,
1034 * int idx)
1035 * Used to add FDB entries to dump requests. Implementers should add
1036 * entries to skb and update idx with the number of entries.
1037 *
1038 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1039 * u16 flags)
1040 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1041 * struct net_device *dev, u32 filter_mask,
1042 * int nlflags)
1043 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1044 * u16 flags);
1045 *
1046 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1047 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1048 * which do not represent real hardware may define this to allow their
1049 * userspace components to manage their virtual carrier state. Devices
1050 * that determine carrier state from physical hardware properties (eg
1051 * network cables) or protocol-dependent mechanisms (eg
1052 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1053 *
1054 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1055 * struct netdev_phys_item_id *ppid);
1056 * Called to get ID of physical port of this device. If driver does
1057 * not implement this, it is assumed that the hw is not able to have
1058 * multiple net devices on single physical port.
1059 *
1060 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1061 * struct udp_tunnel_info *ti);
1062 * Called by UDP tunnel to notify a driver about the UDP port and socket
1063 * address family that a UDP tunnel is listnening to. It is called only
1064 * when a new port starts listening. The operation is protected by the
1065 * RTNL.
1066 *
1067 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1068 * struct udp_tunnel_info *ti);
1069 * Called by UDP tunnel to notify the driver about a UDP port and socket
1070 * address family that the UDP tunnel is not listening to anymore. The
1071 * operation is protected by the RTNL.
1072 *
1073 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1074 * struct net_device *dev)
1075 * Called by upper layer devices to accelerate switching or other
1076 * station functionality into hardware. 'pdev is the lowerdev
1077 * to use for the offload and 'dev' is the net device that will
1078 * back the offload. Returns a pointer to the private structure
1079 * the upper layer will maintain.
1080 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1081 * Called by upper layer device to delete the station created
1082 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1083 * the station and priv is the structure returned by the add
1084 * operation.
1085 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1086 * struct net_device *dev,
1087 * void *priv);
1088 * Callback to use for xmit over the accelerated station. This
1089 * is used in place of ndo_start_xmit on accelerated net
1090 * devices.
1091 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1092 * struct net_device *dev
1093 * netdev_features_t features);
1094 * Called by core transmit path to determine if device is capable of
1095 * performing offload operations on a given packet. This is to give
1096 * the device an opportunity to implement any restrictions that cannot
1097 * be otherwise expressed by feature flags. The check is called with
1098 * the set of features that the stack has calculated and it returns
1099 * those the driver believes to be appropriate.
1100 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1101 * int queue_index, u32 maxrate);
1102 * Called when a user wants to set a max-rate limitation of specific
1103 * TX queue.
1104 * int (*ndo_get_iflink)(const struct net_device *dev);
1105 * Called to get the iflink value of this device.
1106 * void (*ndo_change_proto_down)(struct net_device *dev,
1107 * bool proto_down);
1108 * This function is used to pass protocol port error state information
1109 * to the switch driver. The switch driver can react to the proto_down
1110 * by doing a phys down on the associated switch port.
1111 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1112 * This function is used to get egress tunnel information for given skb.
1113 * This is useful for retrieving outer tunnel header parameters while
1114 * sampling packet.
1115 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1116 * This function is used to specify the headroom that the skb must
1117 * consider when allocation skb during packet reception. Setting
1118 * appropriate rx headroom value allows avoiding skb head copy on
1119 * forward. Setting a negative value resets the rx headroom to the
1120 * default value.
1121 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1122 * This function is used to set or query state related to XDP on the
1123 * netdevice. See definition of enum xdp_netdev_command for details.
1124 *
1125 */
1126 struct net_device_ops {
1127 int (*ndo_init)(struct net_device *dev);
1128 void (*ndo_uninit)(struct net_device *dev);
1129 int (*ndo_open)(struct net_device *dev);
1130 int (*ndo_stop)(struct net_device *dev);
1131 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1132 struct net_device *dev);
1133 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1134 struct net_device *dev,
1135 netdev_features_t features);
1136 u16 (*ndo_select_queue)(struct net_device *dev,
1137 struct sk_buff *skb,
1138 void *accel_priv,
1139 select_queue_fallback_t fallback);
1140 void (*ndo_change_rx_flags)(struct net_device *dev,
1141 int flags);
1142 void (*ndo_set_rx_mode)(struct net_device *dev);
1143 int (*ndo_set_mac_address)(struct net_device *dev,
1144 void *addr);
1145 int (*ndo_validate_addr)(struct net_device *dev);
1146 int (*ndo_do_ioctl)(struct net_device *dev,
1147 struct ifreq *ifr, int cmd);
1148 int (*ndo_set_config)(struct net_device *dev,
1149 struct ifmap *map);
1150 int (*ndo_change_mtu)(struct net_device *dev,
1151 int new_mtu);
1152 int (*ndo_neigh_setup)(struct net_device *dev,
1153 struct neigh_parms *);
1154 void (*ndo_tx_timeout) (struct net_device *dev);
1155
1156 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1157 struct rtnl_link_stats64 *storage);
1158 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1159
1160 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1161 __be16 proto, u16 vid);
1162 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1163 __be16 proto, u16 vid);
1164 #ifdef CONFIG_NET_POLL_CONTROLLER
1165 void (*ndo_poll_controller)(struct net_device *dev);
1166 int (*ndo_netpoll_setup)(struct net_device *dev,
1167 struct netpoll_info *info);
1168 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1169 #endif
1170 #ifdef CONFIG_NET_RX_BUSY_POLL
1171 int (*ndo_busy_poll)(struct napi_struct *dev);
1172 #endif
1173 int (*ndo_set_vf_mac)(struct net_device *dev,
1174 int queue, u8 *mac);
1175 int (*ndo_set_vf_vlan)(struct net_device *dev,
1176 int queue, u16 vlan, u8 qos);
1177 int (*ndo_set_vf_rate)(struct net_device *dev,
1178 int vf, int min_tx_rate,
1179 int max_tx_rate);
1180 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1181 int vf, bool setting);
1182 int (*ndo_set_vf_trust)(struct net_device *dev,
1183 int vf, bool setting);
1184 int (*ndo_get_vf_config)(struct net_device *dev,
1185 int vf,
1186 struct ifla_vf_info *ivf);
1187 int (*ndo_set_vf_link_state)(struct net_device *dev,
1188 int vf, int link_state);
1189 int (*ndo_get_vf_stats)(struct net_device *dev,
1190 int vf,
1191 struct ifla_vf_stats
1192 *vf_stats);
1193 int (*ndo_set_vf_port)(struct net_device *dev,
1194 int vf,
1195 struct nlattr *port[]);
1196 int (*ndo_get_vf_port)(struct net_device *dev,
1197 int vf, struct sk_buff *skb);
1198 int (*ndo_set_vf_guid)(struct net_device *dev,
1199 int vf, u64 guid,
1200 int guid_type);
1201 int (*ndo_set_vf_rss_query_en)(
1202 struct net_device *dev,
1203 int vf, bool setting);
1204 int (*ndo_setup_tc)(struct net_device *dev,
1205 u32 handle,
1206 __be16 protocol,
1207 struct tc_to_netdev *tc);
1208 #if IS_ENABLED(CONFIG_FCOE)
1209 int (*ndo_fcoe_enable)(struct net_device *dev);
1210 int (*ndo_fcoe_disable)(struct net_device *dev);
1211 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1212 u16 xid,
1213 struct scatterlist *sgl,
1214 unsigned int sgc);
1215 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1216 u16 xid);
1217 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1218 u16 xid,
1219 struct scatterlist *sgl,
1220 unsigned int sgc);
1221 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1222 struct netdev_fcoe_hbainfo *hbainfo);
1223 #endif
1224
1225 #if IS_ENABLED(CONFIG_LIBFCOE)
1226 #define NETDEV_FCOE_WWNN 0
1227 #define NETDEV_FCOE_WWPN 1
1228 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1229 u64 *wwn, int type);
1230 #endif
1231
1232 #ifdef CONFIG_RFS_ACCEL
1233 int (*ndo_rx_flow_steer)(struct net_device *dev,
1234 const struct sk_buff *skb,
1235 u16 rxq_index,
1236 u32 flow_id);
1237 #endif
1238 int (*ndo_add_slave)(struct net_device *dev,
1239 struct net_device *slave_dev);
1240 int (*ndo_del_slave)(struct net_device *dev,
1241 struct net_device *slave_dev);
1242 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1243 netdev_features_t features);
1244 int (*ndo_set_features)(struct net_device *dev,
1245 netdev_features_t features);
1246 int (*ndo_neigh_construct)(struct net_device *dev,
1247 struct neighbour *n);
1248 void (*ndo_neigh_destroy)(struct net_device *dev,
1249 struct neighbour *n);
1250
1251 int (*ndo_fdb_add)(struct ndmsg *ndm,
1252 struct nlattr *tb[],
1253 struct net_device *dev,
1254 const unsigned char *addr,
1255 u16 vid,
1256 u16 flags);
1257 int (*ndo_fdb_del)(struct ndmsg *ndm,
1258 struct nlattr *tb[],
1259 struct net_device *dev,
1260 const unsigned char *addr,
1261 u16 vid);
1262 int (*ndo_fdb_dump)(struct sk_buff *skb,
1263 struct netlink_callback *cb,
1264 struct net_device *dev,
1265 struct net_device *filter_dev,
1266 int idx);
1267
1268 int (*ndo_bridge_setlink)(struct net_device *dev,
1269 struct nlmsghdr *nlh,
1270 u16 flags);
1271 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1272 u32 pid, u32 seq,
1273 struct net_device *dev,
1274 u32 filter_mask,
1275 int nlflags);
1276 int (*ndo_bridge_dellink)(struct net_device *dev,
1277 struct nlmsghdr *nlh,
1278 u16 flags);
1279 int (*ndo_change_carrier)(struct net_device *dev,
1280 bool new_carrier);
1281 int (*ndo_get_phys_port_id)(struct net_device *dev,
1282 struct netdev_phys_item_id *ppid);
1283 int (*ndo_get_phys_port_name)(struct net_device *dev,
1284 char *name, size_t len);
1285 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1286 struct udp_tunnel_info *ti);
1287 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1288 struct udp_tunnel_info *ti);
1289 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1290 struct net_device *dev);
1291 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1292 void *priv);
1293
1294 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1295 struct net_device *dev,
1296 void *priv);
1297 int (*ndo_get_lock_subclass)(struct net_device *dev);
1298 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1299 int queue_index,
1300 u32 maxrate);
1301 int (*ndo_get_iflink)(const struct net_device *dev);
1302 int (*ndo_change_proto_down)(struct net_device *dev,
1303 bool proto_down);
1304 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1305 struct sk_buff *skb);
1306 void (*ndo_set_rx_headroom)(struct net_device *dev,
1307 int needed_headroom);
1308 int (*ndo_xdp)(struct net_device *dev,
1309 struct netdev_xdp *xdp);
1310 };
1311
1312 /**
1313 * enum net_device_priv_flags - &struct net_device priv_flags
1314 *
1315 * These are the &struct net_device, they are only set internally
1316 * by drivers and used in the kernel. These flags are invisible to
1317 * userspace; this means that the order of these flags can change
1318 * during any kernel release.
1319 *
1320 * You should have a pretty good reason to be extending these flags.
1321 *
1322 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1323 * @IFF_EBRIDGE: Ethernet bridging device
1324 * @IFF_BONDING: bonding master or slave
1325 * @IFF_ISATAP: ISATAP interface (RFC4214)
1326 * @IFF_WAN_HDLC: WAN HDLC device
1327 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1328 * release skb->dst
1329 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1330 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1331 * @IFF_MACVLAN_PORT: device used as macvlan port
1332 * @IFF_BRIDGE_PORT: device used as bridge port
1333 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1334 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1335 * @IFF_UNICAST_FLT: Supports unicast filtering
1336 * @IFF_TEAM_PORT: device used as team port
1337 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1338 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1339 * change when it's running
1340 * @IFF_MACVLAN: Macvlan device
1341 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1342 * underlying stacked devices
1343 * @IFF_IPVLAN_MASTER: IPvlan master device
1344 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1345 * @IFF_L3MDEV_MASTER: device is an L3 master device
1346 * @IFF_NO_QUEUE: device can run without qdisc attached
1347 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1348 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1349 * @IFF_TEAM: device is a team device
1350 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1351 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1352 * entity (i.e. the master device for bridged veth)
1353 * @IFF_MACSEC: device is a MACsec device
1354 */
1355 enum netdev_priv_flags {
1356 IFF_802_1Q_VLAN = 1<<0,
1357 IFF_EBRIDGE = 1<<1,
1358 IFF_BONDING = 1<<2,
1359 IFF_ISATAP = 1<<3,
1360 IFF_WAN_HDLC = 1<<4,
1361 IFF_XMIT_DST_RELEASE = 1<<5,
1362 IFF_DONT_BRIDGE = 1<<6,
1363 IFF_DISABLE_NETPOLL = 1<<7,
1364 IFF_MACVLAN_PORT = 1<<8,
1365 IFF_BRIDGE_PORT = 1<<9,
1366 IFF_OVS_DATAPATH = 1<<10,
1367 IFF_TX_SKB_SHARING = 1<<11,
1368 IFF_UNICAST_FLT = 1<<12,
1369 IFF_TEAM_PORT = 1<<13,
1370 IFF_SUPP_NOFCS = 1<<14,
1371 IFF_LIVE_ADDR_CHANGE = 1<<15,
1372 IFF_MACVLAN = 1<<16,
1373 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1374 IFF_IPVLAN_MASTER = 1<<18,
1375 IFF_IPVLAN_SLAVE = 1<<19,
1376 IFF_L3MDEV_MASTER = 1<<20,
1377 IFF_NO_QUEUE = 1<<21,
1378 IFF_OPENVSWITCH = 1<<22,
1379 IFF_L3MDEV_SLAVE = 1<<23,
1380 IFF_TEAM = 1<<24,
1381 IFF_RXFH_CONFIGURED = 1<<25,
1382 IFF_PHONY_HEADROOM = 1<<26,
1383 IFF_MACSEC = 1<<27,
1384 };
1385
1386 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1387 #define IFF_EBRIDGE IFF_EBRIDGE
1388 #define IFF_BONDING IFF_BONDING
1389 #define IFF_ISATAP IFF_ISATAP
1390 #define IFF_WAN_HDLC IFF_WAN_HDLC
1391 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1392 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1393 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1394 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1395 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1396 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1397 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1398 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1399 #define IFF_TEAM_PORT IFF_TEAM_PORT
1400 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1401 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1402 #define IFF_MACVLAN IFF_MACVLAN
1403 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1404 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1405 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1406 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1407 #define IFF_NO_QUEUE IFF_NO_QUEUE
1408 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1409 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1410 #define IFF_TEAM IFF_TEAM
1411 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1412 #define IFF_MACSEC IFF_MACSEC
1413
1414 /**
1415 * struct net_device - The DEVICE structure.
1416 * Actually, this whole structure is a big mistake. It mixes I/O
1417 * data with strictly "high-level" data, and it has to know about
1418 * almost every data structure used in the INET module.
1419 *
1420 * @name: This is the first field of the "visible" part of this structure
1421 * (i.e. as seen by users in the "Space.c" file). It is the name
1422 * of the interface.
1423 *
1424 * @name_hlist: Device name hash chain, please keep it close to name[]
1425 * @ifalias: SNMP alias
1426 * @mem_end: Shared memory end
1427 * @mem_start: Shared memory start
1428 * @base_addr: Device I/O address
1429 * @irq: Device IRQ number
1430 *
1431 * @carrier_changes: Stats to monitor carrier on<->off transitions
1432 *
1433 * @state: Generic network queuing layer state, see netdev_state_t
1434 * @dev_list: The global list of network devices
1435 * @napi_list: List entry used for polling NAPI devices
1436 * @unreg_list: List entry when we are unregistering the
1437 * device; see the function unregister_netdev
1438 * @close_list: List entry used when we are closing the device
1439 * @ptype_all: Device-specific packet handlers for all protocols
1440 * @ptype_specific: Device-specific, protocol-specific packet handlers
1441 *
1442 * @adj_list: Directly linked devices, like slaves for bonding
1443 * @all_adj_list: All linked devices, *including* neighbours
1444 * @features: Currently active device features
1445 * @hw_features: User-changeable features
1446 *
1447 * @wanted_features: User-requested features
1448 * @vlan_features: Mask of features inheritable by VLAN devices
1449 *
1450 * @hw_enc_features: Mask of features inherited by encapsulating devices
1451 * This field indicates what encapsulation
1452 * offloads the hardware is capable of doing,
1453 * and drivers will need to set them appropriately.
1454 *
1455 * @mpls_features: Mask of features inheritable by MPLS
1456 *
1457 * @ifindex: interface index
1458 * @group: The group the device belongs to
1459 *
1460 * @stats: Statistics struct, which was left as a legacy, use
1461 * rtnl_link_stats64 instead
1462 *
1463 * @rx_dropped: Dropped packets by core network,
1464 * do not use this in drivers
1465 * @tx_dropped: Dropped packets by core network,
1466 * do not use this in drivers
1467 * @rx_nohandler: nohandler dropped packets by core network on
1468 * inactive devices, do not use this in drivers
1469 *
1470 * @wireless_handlers: List of functions to handle Wireless Extensions,
1471 * instead of ioctl,
1472 * see <net/iw_handler.h> for details.
1473 * @wireless_data: Instance data managed by the core of wireless extensions
1474 *
1475 * @netdev_ops: Includes several pointers to callbacks,
1476 * if one wants to override the ndo_*() functions
1477 * @ethtool_ops: Management operations
1478 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1479 * discovery handling. Necessary for e.g. 6LoWPAN.
1480 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1481 * of Layer 2 headers.
1482 *
1483 * @flags: Interface flags (a la BSD)
1484 * @priv_flags: Like 'flags' but invisible to userspace,
1485 * see if.h for the definitions
1486 * @gflags: Global flags ( kept as legacy )
1487 * @padded: How much padding added by alloc_netdev()
1488 * @operstate: RFC2863 operstate
1489 * @link_mode: Mapping policy to operstate
1490 * @if_port: Selectable AUI, TP, ...
1491 * @dma: DMA channel
1492 * @mtu: Interface MTU value
1493 * @type: Interface hardware type
1494 * @hard_header_len: Maximum hardware header length.
1495 *
1496 * @needed_headroom: Extra headroom the hardware may need, but not in all
1497 * cases can this be guaranteed
1498 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1499 * cases can this be guaranteed. Some cases also use
1500 * LL_MAX_HEADER instead to allocate the skb
1501 *
1502 * interface address info:
1503 *
1504 * @perm_addr: Permanent hw address
1505 * @addr_assign_type: Hw address assignment type
1506 * @addr_len: Hardware address length
1507 * @neigh_priv_len: Used in neigh_alloc()
1508 * @dev_id: Used to differentiate devices that share
1509 * the same link layer address
1510 * @dev_port: Used to differentiate devices that share
1511 * the same function
1512 * @addr_list_lock: XXX: need comments on this one
1513 * @uc_promisc: Counter that indicates promiscuous mode
1514 * has been enabled due to the need to listen to
1515 * additional unicast addresses in a device that
1516 * does not implement ndo_set_rx_mode()
1517 * @uc: unicast mac addresses
1518 * @mc: multicast mac addresses
1519 * @dev_addrs: list of device hw addresses
1520 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1521 * @promiscuity: Number of times the NIC is told to work in
1522 * promiscuous mode; if it becomes 0 the NIC will
1523 * exit promiscuous mode
1524 * @allmulti: Counter, enables or disables allmulticast mode
1525 *
1526 * @vlan_info: VLAN info
1527 * @dsa_ptr: dsa specific data
1528 * @tipc_ptr: TIPC specific data
1529 * @atalk_ptr: AppleTalk link
1530 * @ip_ptr: IPv4 specific data
1531 * @dn_ptr: DECnet specific data
1532 * @ip6_ptr: IPv6 specific data
1533 * @ax25_ptr: AX.25 specific data
1534 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1535 *
1536 * @last_rx: Time of last Rx
1537 * @dev_addr: Hw address (before bcast,
1538 * because most packets are unicast)
1539 *
1540 * @_rx: Array of RX queues
1541 * @num_rx_queues: Number of RX queues
1542 * allocated at register_netdev() time
1543 * @real_num_rx_queues: Number of RX queues currently active in device
1544 *
1545 * @rx_handler: handler for received packets
1546 * @rx_handler_data: XXX: need comments on this one
1547 * @ingress_queue: XXX: need comments on this one
1548 * @broadcast: hw bcast address
1549 *
1550 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1551 * indexed by RX queue number. Assigned by driver.
1552 * This must only be set if the ndo_rx_flow_steer
1553 * operation is defined
1554 * @index_hlist: Device index hash chain
1555 *
1556 * @_tx: Array of TX queues
1557 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1558 * @real_num_tx_queues: Number of TX queues currently active in device
1559 * @qdisc: Root qdisc from userspace point of view
1560 * @tx_queue_len: Max frames per queue allowed
1561 * @tx_global_lock: XXX: need comments on this one
1562 *
1563 * @xps_maps: XXX: need comments on this one
1564 *
1565 * @offload_fwd_mark: Offload device fwding mark
1566 *
1567 * @watchdog_timeo: Represents the timeout that is used by
1568 * the watchdog (see dev_watchdog())
1569 * @watchdog_timer: List of timers
1570 *
1571 * @pcpu_refcnt: Number of references to this device
1572 * @todo_list: Delayed register/unregister
1573 * @link_watch_list: XXX: need comments on this one
1574 *
1575 * @reg_state: Register/unregister state machine
1576 * @dismantle: Device is going to be freed
1577 * @rtnl_link_state: This enum represents the phases of creating
1578 * a new link
1579 *
1580 * @destructor: Called from unregister,
1581 * can be used to call free_netdev
1582 * @npinfo: XXX: need comments on this one
1583 * @nd_net: Network namespace this network device is inside
1584 *
1585 * @ml_priv: Mid-layer private
1586 * @lstats: Loopback statistics
1587 * @tstats: Tunnel statistics
1588 * @dstats: Dummy statistics
1589 * @vstats: Virtual ethernet statistics
1590 *
1591 * @garp_port: GARP
1592 * @mrp_port: MRP
1593 *
1594 * @dev: Class/net/name entry
1595 * @sysfs_groups: Space for optional device, statistics and wireless
1596 * sysfs groups
1597 *
1598 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1599 * @rtnl_link_ops: Rtnl_link_ops
1600 *
1601 * @gso_max_size: Maximum size of generic segmentation offload
1602 * @gso_max_segs: Maximum number of segments that can be passed to the
1603 * NIC for GSO
1604 *
1605 * @dcbnl_ops: Data Center Bridging netlink ops
1606 * @num_tc: Number of traffic classes in the net device
1607 * @tc_to_txq: XXX: need comments on this one
1608 * @prio_tc_map XXX: need comments on this one
1609 *
1610 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1611 *
1612 * @priomap: XXX: need comments on this one
1613 * @phydev: Physical device may attach itself
1614 * for hardware timestamping
1615 *
1616 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1617 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1618 *
1619 * @proto_down: protocol port state information can be sent to the
1620 * switch driver and used to set the phys state of the
1621 * switch port.
1622 *
1623 * FIXME: cleanup struct net_device such that network protocol info
1624 * moves out.
1625 */
1626
1627 struct net_device {
1628 char name[IFNAMSIZ];
1629 struct hlist_node name_hlist;
1630 char *ifalias;
1631 /*
1632 * I/O specific fields
1633 * FIXME: Merge these and struct ifmap into one
1634 */
1635 unsigned long mem_end;
1636 unsigned long mem_start;
1637 unsigned long base_addr;
1638 int irq;
1639
1640 atomic_t carrier_changes;
1641
1642 /*
1643 * Some hardware also needs these fields (state,dev_list,
1644 * napi_list,unreg_list,close_list) but they are not
1645 * part of the usual set specified in Space.c.
1646 */
1647
1648 unsigned long state;
1649
1650 struct list_head dev_list;
1651 struct list_head napi_list;
1652 struct list_head unreg_list;
1653 struct list_head close_list;
1654 struct list_head ptype_all;
1655 struct list_head ptype_specific;
1656
1657 struct {
1658 struct list_head upper;
1659 struct list_head lower;
1660 } adj_list;
1661
1662 struct {
1663 struct list_head upper;
1664 struct list_head lower;
1665 } all_adj_list;
1666
1667 netdev_features_t features;
1668 netdev_features_t hw_features;
1669 netdev_features_t wanted_features;
1670 netdev_features_t vlan_features;
1671 netdev_features_t hw_enc_features;
1672 netdev_features_t mpls_features;
1673 netdev_features_t gso_partial_features;
1674
1675 int ifindex;
1676 int group;
1677
1678 struct net_device_stats stats;
1679
1680 atomic_long_t rx_dropped;
1681 atomic_long_t tx_dropped;
1682 atomic_long_t rx_nohandler;
1683
1684 #ifdef CONFIG_WIRELESS_EXT
1685 const struct iw_handler_def *wireless_handlers;
1686 struct iw_public_data *wireless_data;
1687 #endif
1688 const struct net_device_ops *netdev_ops;
1689 const struct ethtool_ops *ethtool_ops;
1690 #ifdef CONFIG_NET_SWITCHDEV
1691 const struct switchdev_ops *switchdev_ops;
1692 #endif
1693 #ifdef CONFIG_NET_L3_MASTER_DEV
1694 const struct l3mdev_ops *l3mdev_ops;
1695 #endif
1696 #if IS_ENABLED(CONFIG_IPV6)
1697 const struct ndisc_ops *ndisc_ops;
1698 #endif
1699
1700 const struct header_ops *header_ops;
1701
1702 unsigned int flags;
1703 unsigned int priv_flags;
1704
1705 unsigned short gflags;
1706 unsigned short padded;
1707
1708 unsigned char operstate;
1709 unsigned char link_mode;
1710
1711 unsigned char if_port;
1712 unsigned char dma;
1713
1714 unsigned int mtu;
1715 unsigned short type;
1716 unsigned short hard_header_len;
1717
1718 unsigned short needed_headroom;
1719 unsigned short needed_tailroom;
1720
1721 /* Interface address info. */
1722 unsigned char perm_addr[MAX_ADDR_LEN];
1723 unsigned char addr_assign_type;
1724 unsigned char addr_len;
1725 unsigned short neigh_priv_len;
1726 unsigned short dev_id;
1727 unsigned short dev_port;
1728 spinlock_t addr_list_lock;
1729 unsigned char name_assign_type;
1730 bool uc_promisc;
1731 struct netdev_hw_addr_list uc;
1732 struct netdev_hw_addr_list mc;
1733 struct netdev_hw_addr_list dev_addrs;
1734
1735 #ifdef CONFIG_SYSFS
1736 struct kset *queues_kset;
1737 #endif
1738 unsigned int promiscuity;
1739 unsigned int allmulti;
1740
1741
1742 /* Protocol-specific pointers */
1743
1744 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1745 struct vlan_info __rcu *vlan_info;
1746 #endif
1747 #if IS_ENABLED(CONFIG_NET_DSA)
1748 struct dsa_switch_tree *dsa_ptr;
1749 #endif
1750 #if IS_ENABLED(CONFIG_TIPC)
1751 struct tipc_bearer __rcu *tipc_ptr;
1752 #endif
1753 void *atalk_ptr;
1754 struct in_device __rcu *ip_ptr;
1755 struct dn_dev __rcu *dn_ptr;
1756 struct inet6_dev __rcu *ip6_ptr;
1757 void *ax25_ptr;
1758 struct wireless_dev *ieee80211_ptr;
1759 struct wpan_dev *ieee802154_ptr;
1760 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1761 struct mpls_dev __rcu *mpls_ptr;
1762 #endif
1763
1764 /*
1765 * Cache lines mostly used on receive path (including eth_type_trans())
1766 */
1767 unsigned long last_rx;
1768
1769 /* Interface address info used in eth_type_trans() */
1770 unsigned char *dev_addr;
1771
1772 #ifdef CONFIG_SYSFS
1773 struct netdev_rx_queue *_rx;
1774
1775 unsigned int num_rx_queues;
1776 unsigned int real_num_rx_queues;
1777 #endif
1778
1779 unsigned long gro_flush_timeout;
1780 rx_handler_func_t __rcu *rx_handler;
1781 void __rcu *rx_handler_data;
1782
1783 #ifdef CONFIG_NET_CLS_ACT
1784 struct tcf_proto __rcu *ingress_cl_list;
1785 #endif
1786 struct netdev_queue __rcu *ingress_queue;
1787 #ifdef CONFIG_NETFILTER_INGRESS
1788 struct list_head nf_hooks_ingress;
1789 #endif
1790
1791 unsigned char broadcast[MAX_ADDR_LEN];
1792 #ifdef CONFIG_RFS_ACCEL
1793 struct cpu_rmap *rx_cpu_rmap;
1794 #endif
1795 struct hlist_node index_hlist;
1796
1797 /*
1798 * Cache lines mostly used on transmit path
1799 */
1800 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1801 unsigned int num_tx_queues;
1802 unsigned int real_num_tx_queues;
1803 struct Qdisc *qdisc;
1804 #ifdef CONFIG_NET_SCHED
1805 DECLARE_HASHTABLE (qdisc_hash, 4);
1806 #endif
1807 unsigned long tx_queue_len;
1808 spinlock_t tx_global_lock;
1809 int watchdog_timeo;
1810
1811 #ifdef CONFIG_XPS
1812 struct xps_dev_maps __rcu *xps_maps;
1813 #endif
1814 #ifdef CONFIG_NET_CLS_ACT
1815 struct tcf_proto __rcu *egress_cl_list;
1816 #endif
1817 #ifdef CONFIG_NET_SWITCHDEV
1818 u32 offload_fwd_mark;
1819 #endif
1820
1821 /* These may be needed for future network-power-down code. */
1822 struct timer_list watchdog_timer;
1823
1824 int __percpu *pcpu_refcnt;
1825 struct list_head todo_list;
1826
1827 struct list_head link_watch_list;
1828
1829 enum { NETREG_UNINITIALIZED=0,
1830 NETREG_REGISTERED, /* completed register_netdevice */
1831 NETREG_UNREGISTERING, /* called unregister_netdevice */
1832 NETREG_UNREGISTERED, /* completed unregister todo */
1833 NETREG_RELEASED, /* called free_netdev */
1834 NETREG_DUMMY, /* dummy device for NAPI poll */
1835 } reg_state:8;
1836
1837 bool dismantle;
1838
1839 enum {
1840 RTNL_LINK_INITIALIZED,
1841 RTNL_LINK_INITIALIZING,
1842 } rtnl_link_state:16;
1843
1844 void (*destructor)(struct net_device *dev);
1845
1846 #ifdef CONFIG_NETPOLL
1847 struct netpoll_info __rcu *npinfo;
1848 #endif
1849
1850 possible_net_t nd_net;
1851
1852 /* mid-layer private */
1853 union {
1854 void *ml_priv;
1855 struct pcpu_lstats __percpu *lstats;
1856 struct pcpu_sw_netstats __percpu *tstats;
1857 struct pcpu_dstats __percpu *dstats;
1858 struct pcpu_vstats __percpu *vstats;
1859 };
1860
1861 struct garp_port __rcu *garp_port;
1862 struct mrp_port __rcu *mrp_port;
1863
1864 struct device dev;
1865 const struct attribute_group *sysfs_groups[4];
1866 const struct attribute_group *sysfs_rx_queue_group;
1867
1868 const struct rtnl_link_ops *rtnl_link_ops;
1869
1870 /* for setting kernel sock attribute on TCP connection setup */
1871 #define GSO_MAX_SIZE 65536
1872 unsigned int gso_max_size;
1873 #define GSO_MAX_SEGS 65535
1874 u16 gso_max_segs;
1875
1876 #ifdef CONFIG_DCB
1877 const struct dcbnl_rtnl_ops *dcbnl_ops;
1878 #endif
1879 u8 num_tc;
1880 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1881 u8 prio_tc_map[TC_BITMASK + 1];
1882
1883 #if IS_ENABLED(CONFIG_FCOE)
1884 unsigned int fcoe_ddp_xid;
1885 #endif
1886 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1887 struct netprio_map __rcu *priomap;
1888 #endif
1889 struct phy_device *phydev;
1890 struct lock_class_key *qdisc_tx_busylock;
1891 struct lock_class_key *qdisc_running_key;
1892 bool proto_down;
1893 };
1894 #define to_net_dev(d) container_of(d, struct net_device, dev)
1895
1896 #define NETDEV_ALIGN 32
1897
1898 static inline
1899 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1900 {
1901 return dev->prio_tc_map[prio & TC_BITMASK];
1902 }
1903
1904 static inline
1905 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1906 {
1907 if (tc >= dev->num_tc)
1908 return -EINVAL;
1909
1910 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1911 return 0;
1912 }
1913
1914 static inline
1915 void netdev_reset_tc(struct net_device *dev)
1916 {
1917 dev->num_tc = 0;
1918 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1919 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1920 }
1921
1922 static inline
1923 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1924 {
1925 if (tc >= dev->num_tc)
1926 return -EINVAL;
1927
1928 dev->tc_to_txq[tc].count = count;
1929 dev->tc_to_txq[tc].offset = offset;
1930 return 0;
1931 }
1932
1933 static inline
1934 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1935 {
1936 if (num_tc > TC_MAX_QUEUE)
1937 return -EINVAL;
1938
1939 dev->num_tc = num_tc;
1940 return 0;
1941 }
1942
1943 static inline
1944 int netdev_get_num_tc(struct net_device *dev)
1945 {
1946 return dev->num_tc;
1947 }
1948
1949 static inline
1950 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1951 unsigned int index)
1952 {
1953 return &dev->_tx[index];
1954 }
1955
1956 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1957 const struct sk_buff *skb)
1958 {
1959 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1960 }
1961
1962 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1963 void (*f)(struct net_device *,
1964 struct netdev_queue *,
1965 void *),
1966 void *arg)
1967 {
1968 unsigned int i;
1969
1970 for (i = 0; i < dev->num_tx_queues; i++)
1971 f(dev, &dev->_tx[i], arg);
1972 }
1973
1974 #define netdev_lockdep_set_classes(dev) \
1975 { \
1976 static struct lock_class_key qdisc_tx_busylock_key; \
1977 static struct lock_class_key qdisc_running_key; \
1978 static struct lock_class_key qdisc_xmit_lock_key; \
1979 static struct lock_class_key dev_addr_list_lock_key; \
1980 unsigned int i; \
1981 \
1982 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1983 (dev)->qdisc_running_key = &qdisc_running_key; \
1984 lockdep_set_class(&(dev)->addr_list_lock, \
1985 &dev_addr_list_lock_key); \
1986 for (i = 0; i < (dev)->num_tx_queues; i++) \
1987 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1988 &qdisc_xmit_lock_key); \
1989 }
1990
1991 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1992 struct sk_buff *skb,
1993 void *accel_priv);
1994
1995 /* returns the headroom that the master device needs to take in account
1996 * when forwarding to this dev
1997 */
1998 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1999 {
2000 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2001 }
2002
2003 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2004 {
2005 if (dev->netdev_ops->ndo_set_rx_headroom)
2006 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2007 }
2008
2009 /* set the device rx headroom to the dev's default */
2010 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2011 {
2012 netdev_set_rx_headroom(dev, -1);
2013 }
2014
2015 /*
2016 * Net namespace inlines
2017 */
2018 static inline
2019 struct net *dev_net(const struct net_device *dev)
2020 {
2021 return read_pnet(&dev->nd_net);
2022 }
2023
2024 static inline
2025 void dev_net_set(struct net_device *dev, struct net *net)
2026 {
2027 write_pnet(&dev->nd_net, net);
2028 }
2029
2030 static inline bool netdev_uses_dsa(struct net_device *dev)
2031 {
2032 #if IS_ENABLED(CONFIG_NET_DSA)
2033 if (dev->dsa_ptr != NULL)
2034 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2035 #endif
2036 return false;
2037 }
2038
2039 /**
2040 * netdev_priv - access network device private data
2041 * @dev: network device
2042 *
2043 * Get network device private data
2044 */
2045 static inline void *netdev_priv(const struct net_device *dev)
2046 {
2047 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2048 }
2049
2050 /* Set the sysfs physical device reference for the network logical device
2051 * if set prior to registration will cause a symlink during initialization.
2052 */
2053 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2054
2055 /* Set the sysfs device type for the network logical device to allow
2056 * fine-grained identification of different network device types. For
2057 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2058 */
2059 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2060
2061 /* Default NAPI poll() weight
2062 * Device drivers are strongly advised to not use bigger value
2063 */
2064 #define NAPI_POLL_WEIGHT 64
2065
2066 /**
2067 * netif_napi_add - initialize a NAPI context
2068 * @dev: network device
2069 * @napi: NAPI context
2070 * @poll: polling function
2071 * @weight: default weight
2072 *
2073 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2074 * *any* of the other NAPI-related functions.
2075 */
2076 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2077 int (*poll)(struct napi_struct *, int), int weight);
2078
2079 /**
2080 * netif_tx_napi_add - initialize a NAPI context
2081 * @dev: network device
2082 * @napi: NAPI context
2083 * @poll: polling function
2084 * @weight: default weight
2085 *
2086 * This variant of netif_napi_add() should be used from drivers using NAPI
2087 * to exclusively poll a TX queue.
2088 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2089 */
2090 static inline void netif_tx_napi_add(struct net_device *dev,
2091 struct napi_struct *napi,
2092 int (*poll)(struct napi_struct *, int),
2093 int weight)
2094 {
2095 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2096 netif_napi_add(dev, napi, poll, weight);
2097 }
2098
2099 /**
2100 * netif_napi_del - remove a NAPI context
2101 * @napi: NAPI context
2102 *
2103 * netif_napi_del() removes a NAPI context from the network device NAPI list
2104 */
2105 void netif_napi_del(struct napi_struct *napi);
2106
2107 struct napi_gro_cb {
2108 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2109 void *frag0;
2110
2111 /* Length of frag0. */
2112 unsigned int frag0_len;
2113
2114 /* This indicates where we are processing relative to skb->data. */
2115 int data_offset;
2116
2117 /* This is non-zero if the packet cannot be merged with the new skb. */
2118 u16 flush;
2119
2120 /* Save the IP ID here and check when we get to the transport layer */
2121 u16 flush_id;
2122
2123 /* Number of segments aggregated. */
2124 u16 count;
2125
2126 /* Start offset for remote checksum offload */
2127 u16 gro_remcsum_start;
2128
2129 /* jiffies when first packet was created/queued */
2130 unsigned long age;
2131
2132 /* Used in ipv6_gro_receive() and foo-over-udp */
2133 u16 proto;
2134
2135 /* This is non-zero if the packet may be of the same flow. */
2136 u8 same_flow:1;
2137
2138 /* Used in tunnel GRO receive */
2139 u8 encap_mark:1;
2140
2141 /* GRO checksum is valid */
2142 u8 csum_valid:1;
2143
2144 /* Number of checksums via CHECKSUM_UNNECESSARY */
2145 u8 csum_cnt:3;
2146
2147 /* Free the skb? */
2148 u8 free:2;
2149 #define NAPI_GRO_FREE 1
2150 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2151
2152 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2153 u8 is_ipv6:1;
2154
2155 /* Used in GRE, set in fou/gue_gro_receive */
2156 u8 is_fou:1;
2157
2158 /* Used to determine if flush_id can be ignored */
2159 u8 is_atomic:1;
2160
2161 /* 5 bit hole */
2162
2163 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2164 __wsum csum;
2165
2166 /* used in skb_gro_receive() slow path */
2167 struct sk_buff *last;
2168 };
2169
2170 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2171
2172 struct packet_type {
2173 __be16 type; /* This is really htons(ether_type). */
2174 struct net_device *dev; /* NULL is wildcarded here */
2175 int (*func) (struct sk_buff *,
2176 struct net_device *,
2177 struct packet_type *,
2178 struct net_device *);
2179 bool (*id_match)(struct packet_type *ptype,
2180 struct sock *sk);
2181 void *af_packet_priv;
2182 struct list_head list;
2183 };
2184
2185 struct offload_callbacks {
2186 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2187 netdev_features_t features);
2188 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2189 struct sk_buff *skb);
2190 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2191 };
2192
2193 struct packet_offload {
2194 __be16 type; /* This is really htons(ether_type). */
2195 u16 priority;
2196 struct offload_callbacks callbacks;
2197 struct list_head list;
2198 };
2199
2200 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2201 struct pcpu_sw_netstats {
2202 u64 rx_packets;
2203 u64 rx_bytes;
2204 u64 tx_packets;
2205 u64 tx_bytes;
2206 struct u64_stats_sync syncp;
2207 };
2208
2209 #define __netdev_alloc_pcpu_stats(type, gfp) \
2210 ({ \
2211 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2212 if (pcpu_stats) { \
2213 int __cpu; \
2214 for_each_possible_cpu(__cpu) { \
2215 typeof(type) *stat; \
2216 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2217 u64_stats_init(&stat->syncp); \
2218 } \
2219 } \
2220 pcpu_stats; \
2221 })
2222
2223 #define netdev_alloc_pcpu_stats(type) \
2224 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2225
2226 enum netdev_lag_tx_type {
2227 NETDEV_LAG_TX_TYPE_UNKNOWN,
2228 NETDEV_LAG_TX_TYPE_RANDOM,
2229 NETDEV_LAG_TX_TYPE_BROADCAST,
2230 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2231 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2232 NETDEV_LAG_TX_TYPE_HASH,
2233 };
2234
2235 struct netdev_lag_upper_info {
2236 enum netdev_lag_tx_type tx_type;
2237 };
2238
2239 struct netdev_lag_lower_state_info {
2240 u8 link_up : 1,
2241 tx_enabled : 1;
2242 };
2243
2244 #include <linux/notifier.h>
2245
2246 /* netdevice notifier chain. Please remember to update the rtnetlink
2247 * notification exclusion list in rtnetlink_event() when adding new
2248 * types.
2249 */
2250 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2251 #define NETDEV_DOWN 0x0002
2252 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2253 detected a hardware crash and restarted
2254 - we can use this eg to kick tcp sessions
2255 once done */
2256 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2257 #define NETDEV_REGISTER 0x0005
2258 #define NETDEV_UNREGISTER 0x0006
2259 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2260 #define NETDEV_CHANGEADDR 0x0008
2261 #define NETDEV_GOING_DOWN 0x0009
2262 #define NETDEV_CHANGENAME 0x000A
2263 #define NETDEV_FEAT_CHANGE 0x000B
2264 #define NETDEV_BONDING_FAILOVER 0x000C
2265 #define NETDEV_PRE_UP 0x000D
2266 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2267 #define NETDEV_POST_TYPE_CHANGE 0x000F
2268 #define NETDEV_POST_INIT 0x0010
2269 #define NETDEV_UNREGISTER_FINAL 0x0011
2270 #define NETDEV_RELEASE 0x0012
2271 #define NETDEV_NOTIFY_PEERS 0x0013
2272 #define NETDEV_JOIN 0x0014
2273 #define NETDEV_CHANGEUPPER 0x0015
2274 #define NETDEV_RESEND_IGMP 0x0016
2275 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2276 #define NETDEV_CHANGEINFODATA 0x0018
2277 #define NETDEV_BONDING_INFO 0x0019
2278 #define NETDEV_PRECHANGEUPPER 0x001A
2279 #define NETDEV_CHANGELOWERSTATE 0x001B
2280 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2281 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2282
2283 int register_netdevice_notifier(struct notifier_block *nb);
2284 int unregister_netdevice_notifier(struct notifier_block *nb);
2285
2286 struct netdev_notifier_info {
2287 struct net_device *dev;
2288 };
2289
2290 struct netdev_notifier_change_info {
2291 struct netdev_notifier_info info; /* must be first */
2292 unsigned int flags_changed;
2293 };
2294
2295 struct netdev_notifier_changeupper_info {
2296 struct netdev_notifier_info info; /* must be first */
2297 struct net_device *upper_dev; /* new upper dev */
2298 bool master; /* is upper dev master */
2299 bool linking; /* is the notification for link or unlink */
2300 void *upper_info; /* upper dev info */
2301 };
2302
2303 struct netdev_notifier_changelowerstate_info {
2304 struct netdev_notifier_info info; /* must be first */
2305 void *lower_state_info; /* is lower dev state */
2306 };
2307
2308 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2309 struct net_device *dev)
2310 {
2311 info->dev = dev;
2312 }
2313
2314 static inline struct net_device *
2315 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2316 {
2317 return info->dev;
2318 }
2319
2320 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2321
2322
2323 extern rwlock_t dev_base_lock; /* Device list lock */
2324
2325 #define for_each_netdev(net, d) \
2326 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2327 #define for_each_netdev_reverse(net, d) \
2328 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2329 #define for_each_netdev_rcu(net, d) \
2330 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2331 #define for_each_netdev_safe(net, d, n) \
2332 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2333 #define for_each_netdev_continue(net, d) \
2334 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2335 #define for_each_netdev_continue_rcu(net, d) \
2336 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2337 #define for_each_netdev_in_bond_rcu(bond, slave) \
2338 for_each_netdev_rcu(&init_net, slave) \
2339 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2340 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2341
2342 static inline struct net_device *next_net_device(struct net_device *dev)
2343 {
2344 struct list_head *lh;
2345 struct net *net;
2346
2347 net = dev_net(dev);
2348 lh = dev->dev_list.next;
2349 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2350 }
2351
2352 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2353 {
2354 struct list_head *lh;
2355 struct net *net;
2356
2357 net = dev_net(dev);
2358 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2359 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2360 }
2361
2362 static inline struct net_device *first_net_device(struct net *net)
2363 {
2364 return list_empty(&net->dev_base_head) ? NULL :
2365 net_device_entry(net->dev_base_head.next);
2366 }
2367
2368 static inline struct net_device *first_net_device_rcu(struct net *net)
2369 {
2370 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2371
2372 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2373 }
2374
2375 int netdev_boot_setup_check(struct net_device *dev);
2376 unsigned long netdev_boot_base(const char *prefix, int unit);
2377 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2378 const char *hwaddr);
2379 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2380 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2381 void dev_add_pack(struct packet_type *pt);
2382 void dev_remove_pack(struct packet_type *pt);
2383 void __dev_remove_pack(struct packet_type *pt);
2384 void dev_add_offload(struct packet_offload *po);
2385 void dev_remove_offload(struct packet_offload *po);
2386
2387 int dev_get_iflink(const struct net_device *dev);
2388 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2389 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2390 unsigned short mask);
2391 struct net_device *dev_get_by_name(struct net *net, const char *name);
2392 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2393 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2394 int dev_alloc_name(struct net_device *dev, const char *name);
2395 int dev_open(struct net_device *dev);
2396 int dev_close(struct net_device *dev);
2397 int dev_close_many(struct list_head *head, bool unlink);
2398 void dev_disable_lro(struct net_device *dev);
2399 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2400 int dev_queue_xmit(struct sk_buff *skb);
2401 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2402 int register_netdevice(struct net_device *dev);
2403 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2404 void unregister_netdevice_many(struct list_head *head);
2405 static inline void unregister_netdevice(struct net_device *dev)
2406 {
2407 unregister_netdevice_queue(dev, NULL);
2408 }
2409
2410 int netdev_refcnt_read(const struct net_device *dev);
2411 void free_netdev(struct net_device *dev);
2412 void netdev_freemem(struct net_device *dev);
2413 void synchronize_net(void);
2414 int init_dummy_netdev(struct net_device *dev);
2415
2416 DECLARE_PER_CPU(int, xmit_recursion);
2417 #define XMIT_RECURSION_LIMIT 10
2418
2419 static inline int dev_recursion_level(void)
2420 {
2421 return this_cpu_read(xmit_recursion);
2422 }
2423
2424 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2425 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2426 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2427 int netdev_get_name(struct net *net, char *name, int ifindex);
2428 int dev_restart(struct net_device *dev);
2429 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2430
2431 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2432 {
2433 return NAPI_GRO_CB(skb)->data_offset;
2434 }
2435
2436 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2437 {
2438 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2439 }
2440
2441 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2442 {
2443 NAPI_GRO_CB(skb)->data_offset += len;
2444 }
2445
2446 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2447 unsigned int offset)
2448 {
2449 return NAPI_GRO_CB(skb)->frag0 + offset;
2450 }
2451
2452 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2453 {
2454 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2455 }
2456
2457 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2458 unsigned int offset)
2459 {
2460 if (!pskb_may_pull(skb, hlen))
2461 return NULL;
2462
2463 NAPI_GRO_CB(skb)->frag0 = NULL;
2464 NAPI_GRO_CB(skb)->frag0_len = 0;
2465 return skb->data + offset;
2466 }
2467
2468 static inline void *skb_gro_network_header(struct sk_buff *skb)
2469 {
2470 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2471 skb_network_offset(skb);
2472 }
2473
2474 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2475 const void *start, unsigned int len)
2476 {
2477 if (NAPI_GRO_CB(skb)->csum_valid)
2478 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2479 csum_partial(start, len, 0));
2480 }
2481
2482 /* GRO checksum functions. These are logical equivalents of the normal
2483 * checksum functions (in skbuff.h) except that they operate on the GRO
2484 * offsets and fields in sk_buff.
2485 */
2486
2487 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2488
2489 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2490 {
2491 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2492 }
2493
2494 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2495 bool zero_okay,
2496 __sum16 check)
2497 {
2498 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2499 skb_checksum_start_offset(skb) <
2500 skb_gro_offset(skb)) &&
2501 !skb_at_gro_remcsum_start(skb) &&
2502 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2503 (!zero_okay || check));
2504 }
2505
2506 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2507 __wsum psum)
2508 {
2509 if (NAPI_GRO_CB(skb)->csum_valid &&
2510 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2511 return 0;
2512
2513 NAPI_GRO_CB(skb)->csum = psum;
2514
2515 return __skb_gro_checksum_complete(skb);
2516 }
2517
2518 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2519 {
2520 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2521 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2522 NAPI_GRO_CB(skb)->csum_cnt--;
2523 } else {
2524 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2525 * verified a new top level checksum or an encapsulated one
2526 * during GRO. This saves work if we fallback to normal path.
2527 */
2528 __skb_incr_checksum_unnecessary(skb);
2529 }
2530 }
2531
2532 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2533 compute_pseudo) \
2534 ({ \
2535 __sum16 __ret = 0; \
2536 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2537 __ret = __skb_gro_checksum_validate_complete(skb, \
2538 compute_pseudo(skb, proto)); \
2539 if (__ret) \
2540 __skb_mark_checksum_bad(skb); \
2541 else \
2542 skb_gro_incr_csum_unnecessary(skb); \
2543 __ret; \
2544 })
2545
2546 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2547 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2548
2549 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2550 compute_pseudo) \
2551 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2552
2553 #define skb_gro_checksum_simple_validate(skb) \
2554 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2555
2556 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2557 {
2558 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2559 !NAPI_GRO_CB(skb)->csum_valid);
2560 }
2561
2562 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2563 __sum16 check, __wsum pseudo)
2564 {
2565 NAPI_GRO_CB(skb)->csum = ~pseudo;
2566 NAPI_GRO_CB(skb)->csum_valid = 1;
2567 }
2568
2569 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2570 do { \
2571 if (__skb_gro_checksum_convert_check(skb)) \
2572 __skb_gro_checksum_convert(skb, check, \
2573 compute_pseudo(skb, proto)); \
2574 } while (0)
2575
2576 struct gro_remcsum {
2577 int offset;
2578 __wsum delta;
2579 };
2580
2581 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2582 {
2583 grc->offset = 0;
2584 grc->delta = 0;
2585 }
2586
2587 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2588 unsigned int off, size_t hdrlen,
2589 int start, int offset,
2590 struct gro_remcsum *grc,
2591 bool nopartial)
2592 {
2593 __wsum delta;
2594 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2595
2596 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2597
2598 if (!nopartial) {
2599 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2600 return ptr;
2601 }
2602
2603 ptr = skb_gro_header_fast(skb, off);
2604 if (skb_gro_header_hard(skb, off + plen)) {
2605 ptr = skb_gro_header_slow(skb, off + plen, off);
2606 if (!ptr)
2607 return NULL;
2608 }
2609
2610 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2611 start, offset);
2612
2613 /* Adjust skb->csum since we changed the packet */
2614 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2615
2616 grc->offset = off + hdrlen + offset;
2617 grc->delta = delta;
2618
2619 return ptr;
2620 }
2621
2622 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2623 struct gro_remcsum *grc)
2624 {
2625 void *ptr;
2626 size_t plen = grc->offset + sizeof(u16);
2627
2628 if (!grc->delta)
2629 return;
2630
2631 ptr = skb_gro_header_fast(skb, grc->offset);
2632 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2633 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2634 if (!ptr)
2635 return;
2636 }
2637
2638 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2639 }
2640
2641 struct skb_csum_offl_spec {
2642 __u16 ipv4_okay:1,
2643 ipv6_okay:1,
2644 encap_okay:1,
2645 ip_options_okay:1,
2646 ext_hdrs_okay:1,
2647 tcp_okay:1,
2648 udp_okay:1,
2649 sctp_okay:1,
2650 vlan_okay:1,
2651 no_encapped_ipv6:1,
2652 no_not_encapped:1;
2653 };
2654
2655 bool __skb_csum_offload_chk(struct sk_buff *skb,
2656 const struct skb_csum_offl_spec *spec,
2657 bool *csum_encapped,
2658 bool csum_help);
2659
2660 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2661 const struct skb_csum_offl_spec *spec,
2662 bool *csum_encapped,
2663 bool csum_help)
2664 {
2665 if (skb->ip_summed != CHECKSUM_PARTIAL)
2666 return false;
2667
2668 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2669 }
2670
2671 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2672 const struct skb_csum_offl_spec *spec)
2673 {
2674 bool csum_encapped;
2675
2676 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2677 }
2678
2679 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2680 {
2681 static const struct skb_csum_offl_spec csum_offl_spec = {
2682 .ipv4_okay = 1,
2683 .ip_options_okay = 1,
2684 .ipv6_okay = 1,
2685 .vlan_okay = 1,
2686 .tcp_okay = 1,
2687 .udp_okay = 1,
2688 };
2689
2690 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2691 }
2692
2693 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2694 {
2695 static const struct skb_csum_offl_spec csum_offl_spec = {
2696 .ipv4_okay = 1,
2697 .ip_options_okay = 1,
2698 .tcp_okay = 1,
2699 .udp_okay = 1,
2700 .vlan_okay = 1,
2701 };
2702
2703 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2704 }
2705
2706 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2707 unsigned short type,
2708 const void *daddr, const void *saddr,
2709 unsigned int len)
2710 {
2711 if (!dev->header_ops || !dev->header_ops->create)
2712 return 0;
2713
2714 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2715 }
2716
2717 static inline int dev_parse_header(const struct sk_buff *skb,
2718 unsigned char *haddr)
2719 {
2720 const struct net_device *dev = skb->dev;
2721
2722 if (!dev->header_ops || !dev->header_ops->parse)
2723 return 0;
2724 return dev->header_ops->parse(skb, haddr);
2725 }
2726
2727 /* ll_header must have at least hard_header_len allocated */
2728 static inline bool dev_validate_header(const struct net_device *dev,
2729 char *ll_header, int len)
2730 {
2731 if (likely(len >= dev->hard_header_len))
2732 return true;
2733
2734 if (capable(CAP_SYS_RAWIO)) {
2735 memset(ll_header + len, 0, dev->hard_header_len - len);
2736 return true;
2737 }
2738
2739 if (dev->header_ops && dev->header_ops->validate)
2740 return dev->header_ops->validate(ll_header, len);
2741
2742 return false;
2743 }
2744
2745 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2746 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2747 static inline int unregister_gifconf(unsigned int family)
2748 {
2749 return register_gifconf(family, NULL);
2750 }
2751
2752 #ifdef CONFIG_NET_FLOW_LIMIT
2753 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2754 struct sd_flow_limit {
2755 u64 count;
2756 unsigned int num_buckets;
2757 unsigned int history_head;
2758 u16 history[FLOW_LIMIT_HISTORY];
2759 u8 buckets[];
2760 };
2761
2762 extern int netdev_flow_limit_table_len;
2763 #endif /* CONFIG_NET_FLOW_LIMIT */
2764
2765 /*
2766 * Incoming packets are placed on per-CPU queues
2767 */
2768 struct softnet_data {
2769 struct list_head poll_list;
2770 struct sk_buff_head process_queue;
2771
2772 /* stats */
2773 unsigned int processed;
2774 unsigned int time_squeeze;
2775 unsigned int received_rps;
2776 #ifdef CONFIG_RPS
2777 struct softnet_data *rps_ipi_list;
2778 #endif
2779 #ifdef CONFIG_NET_FLOW_LIMIT
2780 struct sd_flow_limit __rcu *flow_limit;
2781 #endif
2782 struct Qdisc *output_queue;
2783 struct Qdisc **output_queue_tailp;
2784 struct sk_buff *completion_queue;
2785
2786 #ifdef CONFIG_RPS
2787 /* input_queue_head should be written by cpu owning this struct,
2788 * and only read by other cpus. Worth using a cache line.
2789 */
2790 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2791
2792 /* Elements below can be accessed between CPUs for RPS/RFS */
2793 struct call_single_data csd ____cacheline_aligned_in_smp;
2794 struct softnet_data *rps_ipi_next;
2795 unsigned int cpu;
2796 unsigned int input_queue_tail;
2797 #endif
2798 unsigned int dropped;
2799 struct sk_buff_head input_pkt_queue;
2800 struct napi_struct backlog;
2801
2802 };
2803
2804 static inline void input_queue_head_incr(struct softnet_data *sd)
2805 {
2806 #ifdef CONFIG_RPS
2807 sd->input_queue_head++;
2808 #endif
2809 }
2810
2811 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2812 unsigned int *qtail)
2813 {
2814 #ifdef CONFIG_RPS
2815 *qtail = ++sd->input_queue_tail;
2816 #endif
2817 }
2818
2819 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2820
2821 void __netif_schedule(struct Qdisc *q);
2822 void netif_schedule_queue(struct netdev_queue *txq);
2823
2824 static inline void netif_tx_schedule_all(struct net_device *dev)
2825 {
2826 unsigned int i;
2827
2828 for (i = 0; i < dev->num_tx_queues; i++)
2829 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2830 }
2831
2832 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2833 {
2834 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2835 }
2836
2837 /**
2838 * netif_start_queue - allow transmit
2839 * @dev: network device
2840 *
2841 * Allow upper layers to call the device hard_start_xmit routine.
2842 */
2843 static inline void netif_start_queue(struct net_device *dev)
2844 {
2845 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2846 }
2847
2848 static inline void netif_tx_start_all_queues(struct net_device *dev)
2849 {
2850 unsigned int i;
2851
2852 for (i = 0; i < dev->num_tx_queues; i++) {
2853 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2854 netif_tx_start_queue(txq);
2855 }
2856 }
2857
2858 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2859
2860 /**
2861 * netif_wake_queue - restart transmit
2862 * @dev: network device
2863 *
2864 * Allow upper layers to call the device hard_start_xmit routine.
2865 * Used for flow control when transmit resources are available.
2866 */
2867 static inline void netif_wake_queue(struct net_device *dev)
2868 {
2869 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2870 }
2871
2872 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2873 {
2874 unsigned int i;
2875
2876 for (i = 0; i < dev->num_tx_queues; i++) {
2877 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2878 netif_tx_wake_queue(txq);
2879 }
2880 }
2881
2882 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2883 {
2884 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2885 }
2886
2887 /**
2888 * netif_stop_queue - stop transmitted packets
2889 * @dev: network device
2890 *
2891 * Stop upper layers calling the device hard_start_xmit routine.
2892 * Used for flow control when transmit resources are unavailable.
2893 */
2894 static inline void netif_stop_queue(struct net_device *dev)
2895 {
2896 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2897 }
2898
2899 void netif_tx_stop_all_queues(struct net_device *dev);
2900
2901 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2902 {
2903 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2904 }
2905
2906 /**
2907 * netif_queue_stopped - test if transmit queue is flowblocked
2908 * @dev: network device
2909 *
2910 * Test if transmit queue on device is currently unable to send.
2911 */
2912 static inline bool netif_queue_stopped(const struct net_device *dev)
2913 {
2914 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2915 }
2916
2917 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2918 {
2919 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2920 }
2921
2922 static inline bool
2923 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2924 {
2925 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2926 }
2927
2928 static inline bool
2929 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2930 {
2931 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2932 }
2933
2934 /**
2935 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2936 * @dev_queue: pointer to transmit queue
2937 *
2938 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2939 * to give appropriate hint to the CPU.
2940 */
2941 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2942 {
2943 #ifdef CONFIG_BQL
2944 prefetchw(&dev_queue->dql.num_queued);
2945 #endif
2946 }
2947
2948 /**
2949 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2950 * @dev_queue: pointer to transmit queue
2951 *
2952 * BQL enabled drivers might use this helper in their TX completion path,
2953 * to give appropriate hint to the CPU.
2954 */
2955 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2956 {
2957 #ifdef CONFIG_BQL
2958 prefetchw(&dev_queue->dql.limit);
2959 #endif
2960 }
2961
2962 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2963 unsigned int bytes)
2964 {
2965 #ifdef CONFIG_BQL
2966 dql_queued(&dev_queue->dql, bytes);
2967
2968 if (likely(dql_avail(&dev_queue->dql) >= 0))
2969 return;
2970
2971 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2972
2973 /*
2974 * The XOFF flag must be set before checking the dql_avail below,
2975 * because in netdev_tx_completed_queue we update the dql_completed
2976 * before checking the XOFF flag.
2977 */
2978 smp_mb();
2979
2980 /* check again in case another CPU has just made room avail */
2981 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2982 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2983 #endif
2984 }
2985
2986 /**
2987 * netdev_sent_queue - report the number of bytes queued to hardware
2988 * @dev: network device
2989 * @bytes: number of bytes queued to the hardware device queue
2990 *
2991 * Report the number of bytes queued for sending/completion to the network
2992 * device hardware queue. @bytes should be a good approximation and should
2993 * exactly match netdev_completed_queue() @bytes
2994 */
2995 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2996 {
2997 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2998 }
2999
3000 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3001 unsigned int pkts, unsigned int bytes)
3002 {
3003 #ifdef CONFIG_BQL
3004 if (unlikely(!bytes))
3005 return;
3006
3007 dql_completed(&dev_queue->dql, bytes);
3008
3009 /*
3010 * Without the memory barrier there is a small possiblity that
3011 * netdev_tx_sent_queue will miss the update and cause the queue to
3012 * be stopped forever
3013 */
3014 smp_mb();
3015
3016 if (dql_avail(&dev_queue->dql) < 0)
3017 return;
3018
3019 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3020 netif_schedule_queue(dev_queue);
3021 #endif
3022 }
3023
3024 /**
3025 * netdev_completed_queue - report bytes and packets completed by device
3026 * @dev: network device
3027 * @pkts: actual number of packets sent over the medium
3028 * @bytes: actual number of bytes sent over the medium
3029 *
3030 * Report the number of bytes and packets transmitted by the network device
3031 * hardware queue over the physical medium, @bytes must exactly match the
3032 * @bytes amount passed to netdev_sent_queue()
3033 */
3034 static inline void netdev_completed_queue(struct net_device *dev,
3035 unsigned int pkts, unsigned int bytes)
3036 {
3037 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3038 }
3039
3040 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3041 {
3042 #ifdef CONFIG_BQL
3043 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3044 dql_reset(&q->dql);
3045 #endif
3046 }
3047
3048 /**
3049 * netdev_reset_queue - reset the packets and bytes count of a network device
3050 * @dev_queue: network device
3051 *
3052 * Reset the bytes and packet count of a network device and clear the
3053 * software flow control OFF bit for this network device
3054 */
3055 static inline void netdev_reset_queue(struct net_device *dev_queue)
3056 {
3057 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3058 }
3059
3060 /**
3061 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3062 * @dev: network device
3063 * @queue_index: given tx queue index
3064 *
3065 * Returns 0 if given tx queue index >= number of device tx queues,
3066 * otherwise returns the originally passed tx queue index.
3067 */
3068 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3069 {
3070 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3071 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3072 dev->name, queue_index,
3073 dev->real_num_tx_queues);
3074 return 0;
3075 }
3076
3077 return queue_index;
3078 }
3079
3080 /**
3081 * netif_running - test if up
3082 * @dev: network device
3083 *
3084 * Test if the device has been brought up.
3085 */
3086 static inline bool netif_running(const struct net_device *dev)
3087 {
3088 return test_bit(__LINK_STATE_START, &dev->state);
3089 }
3090
3091 /*
3092 * Routines to manage the subqueues on a device. We only need start,
3093 * stop, and a check if it's stopped. All other device management is
3094 * done at the overall netdevice level.
3095 * Also test the device if we're multiqueue.
3096 */
3097
3098 /**
3099 * netif_start_subqueue - allow sending packets on subqueue
3100 * @dev: network device
3101 * @queue_index: sub queue index
3102 *
3103 * Start individual transmit queue of a device with multiple transmit queues.
3104 */
3105 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3106 {
3107 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3108
3109 netif_tx_start_queue(txq);
3110 }
3111
3112 /**
3113 * netif_stop_subqueue - stop sending packets on subqueue
3114 * @dev: network device
3115 * @queue_index: sub queue index
3116 *
3117 * Stop individual transmit queue of a device with multiple transmit queues.
3118 */
3119 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3120 {
3121 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3122 netif_tx_stop_queue(txq);
3123 }
3124
3125 /**
3126 * netif_subqueue_stopped - test status of subqueue
3127 * @dev: network device
3128 * @queue_index: sub queue index
3129 *
3130 * Check individual transmit queue of a device with multiple transmit queues.
3131 */
3132 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3133 u16 queue_index)
3134 {
3135 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3136
3137 return netif_tx_queue_stopped(txq);
3138 }
3139
3140 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3141 struct sk_buff *skb)
3142 {
3143 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3144 }
3145
3146 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3147
3148 #ifdef CONFIG_XPS
3149 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3150 u16 index);
3151 #else
3152 static inline int netif_set_xps_queue(struct net_device *dev,
3153 const struct cpumask *mask,
3154 u16 index)
3155 {
3156 return 0;
3157 }
3158 #endif
3159
3160 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3161 unsigned int num_tx_queues);
3162
3163 /*
3164 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3165 * as a distribution range limit for the returned value.
3166 */
3167 static inline u16 skb_tx_hash(const struct net_device *dev,
3168 struct sk_buff *skb)
3169 {
3170 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3171 }
3172
3173 /**
3174 * netif_is_multiqueue - test if device has multiple transmit queues
3175 * @dev: network device
3176 *
3177 * Check if device has multiple transmit queues
3178 */
3179 static inline bool netif_is_multiqueue(const struct net_device *dev)
3180 {
3181 return dev->num_tx_queues > 1;
3182 }
3183
3184 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3185
3186 #ifdef CONFIG_SYSFS
3187 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3188 #else
3189 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3190 unsigned int rxq)
3191 {
3192 return 0;
3193 }
3194 #endif
3195
3196 #ifdef CONFIG_SYSFS
3197 static inline unsigned int get_netdev_rx_queue_index(
3198 struct netdev_rx_queue *queue)
3199 {
3200 struct net_device *dev = queue->dev;
3201 int index = queue - dev->_rx;
3202
3203 BUG_ON(index >= dev->num_rx_queues);
3204 return index;
3205 }
3206 #endif
3207
3208 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3209 int netif_get_num_default_rss_queues(void);
3210
3211 enum skb_free_reason {
3212 SKB_REASON_CONSUMED,
3213 SKB_REASON_DROPPED,
3214 };
3215
3216 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3217 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3218
3219 /*
3220 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3221 * interrupt context or with hardware interrupts being disabled.
3222 * (in_irq() || irqs_disabled())
3223 *
3224 * We provide four helpers that can be used in following contexts :
3225 *
3226 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3227 * replacing kfree_skb(skb)
3228 *
3229 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3230 * Typically used in place of consume_skb(skb) in TX completion path
3231 *
3232 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3233 * replacing kfree_skb(skb)
3234 *
3235 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3236 * and consumed a packet. Used in place of consume_skb(skb)
3237 */
3238 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3239 {
3240 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3241 }
3242
3243 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3244 {
3245 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3246 }
3247
3248 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3249 {
3250 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3251 }
3252
3253 static inline void dev_consume_skb_any(struct sk_buff *skb)
3254 {
3255 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3256 }
3257
3258 int netif_rx(struct sk_buff *skb);
3259 int netif_rx_ni(struct sk_buff *skb);
3260 int netif_receive_skb(struct sk_buff *skb);
3261 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3262 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3263 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3264 gro_result_t napi_gro_frags(struct napi_struct *napi);
3265 struct packet_offload *gro_find_receive_by_type(__be16 type);
3266 struct packet_offload *gro_find_complete_by_type(__be16 type);
3267
3268 static inline void napi_free_frags(struct napi_struct *napi)
3269 {
3270 kfree_skb(napi->skb);
3271 napi->skb = NULL;
3272 }
3273
3274 int netdev_rx_handler_register(struct net_device *dev,
3275 rx_handler_func_t *rx_handler,
3276 void *rx_handler_data);
3277 void netdev_rx_handler_unregister(struct net_device *dev);
3278
3279 bool dev_valid_name(const char *name);
3280 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3281 int dev_ethtool(struct net *net, struct ifreq *);
3282 unsigned int dev_get_flags(const struct net_device *);
3283 int __dev_change_flags(struct net_device *, unsigned int flags);
3284 int dev_change_flags(struct net_device *, unsigned int);
3285 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3286 unsigned int gchanges);
3287 int dev_change_name(struct net_device *, const char *);
3288 int dev_set_alias(struct net_device *, const char *, size_t);
3289 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3290 int dev_set_mtu(struct net_device *, int);
3291 void dev_set_group(struct net_device *, int);
3292 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3293 int dev_change_carrier(struct net_device *, bool new_carrier);
3294 int dev_get_phys_port_id(struct net_device *dev,
3295 struct netdev_phys_item_id *ppid);
3296 int dev_get_phys_port_name(struct net_device *dev,
3297 char *name, size_t len);
3298 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3299 int dev_change_xdp_fd(struct net_device *dev, int fd);
3300 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3301 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3302 struct netdev_queue *txq, int *ret);
3303 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3304 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3305 bool is_skb_forwardable(const struct net_device *dev,
3306 const struct sk_buff *skb);
3307
3308 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3309
3310 extern int netdev_budget;
3311
3312 /* Called by rtnetlink.c:rtnl_unlock() */
3313 void netdev_run_todo(void);
3314
3315 /**
3316 * dev_put - release reference to device
3317 * @dev: network device
3318 *
3319 * Release reference to device to allow it to be freed.
3320 */
3321 static inline void dev_put(struct net_device *dev)
3322 {
3323 this_cpu_dec(*dev->pcpu_refcnt);
3324 }
3325
3326 /**
3327 * dev_hold - get reference to device
3328 * @dev: network device
3329 *
3330 * Hold reference to device to keep it from being freed.
3331 */
3332 static inline void dev_hold(struct net_device *dev)
3333 {
3334 this_cpu_inc(*dev->pcpu_refcnt);
3335 }
3336
3337 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3338 * and _off may be called from IRQ context, but it is caller
3339 * who is responsible for serialization of these calls.
3340 *
3341 * The name carrier is inappropriate, these functions should really be
3342 * called netif_lowerlayer_*() because they represent the state of any
3343 * kind of lower layer not just hardware media.
3344 */
3345
3346 void linkwatch_init_dev(struct net_device *dev);
3347 void linkwatch_fire_event(struct net_device *dev);
3348 void linkwatch_forget_dev(struct net_device *dev);
3349
3350 /**
3351 * netif_carrier_ok - test if carrier present
3352 * @dev: network device
3353 *
3354 * Check if carrier is present on device
3355 */
3356 static inline bool netif_carrier_ok(const struct net_device *dev)
3357 {
3358 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3359 }
3360
3361 unsigned long dev_trans_start(struct net_device *dev);
3362
3363 void __netdev_watchdog_up(struct net_device *dev);
3364
3365 void netif_carrier_on(struct net_device *dev);
3366
3367 void netif_carrier_off(struct net_device *dev);
3368
3369 /**
3370 * netif_dormant_on - mark device as dormant.
3371 * @dev: network device
3372 *
3373 * Mark device as dormant (as per RFC2863).
3374 *
3375 * The dormant state indicates that the relevant interface is not
3376 * actually in a condition to pass packets (i.e., it is not 'up') but is
3377 * in a "pending" state, waiting for some external event. For "on-
3378 * demand" interfaces, this new state identifies the situation where the
3379 * interface is waiting for events to place it in the up state.
3380 */
3381 static inline void netif_dormant_on(struct net_device *dev)
3382 {
3383 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3384 linkwatch_fire_event(dev);
3385 }
3386
3387 /**
3388 * netif_dormant_off - set device as not dormant.
3389 * @dev: network device
3390 *
3391 * Device is not in dormant state.
3392 */
3393 static inline void netif_dormant_off(struct net_device *dev)
3394 {
3395 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3396 linkwatch_fire_event(dev);
3397 }
3398
3399 /**
3400 * netif_dormant - test if carrier present
3401 * @dev: network device
3402 *
3403 * Check if carrier is present on device
3404 */
3405 static inline bool netif_dormant(const struct net_device *dev)
3406 {
3407 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3408 }
3409
3410
3411 /**
3412 * netif_oper_up - test if device is operational
3413 * @dev: network device
3414 *
3415 * Check if carrier is operational
3416 */
3417 static inline bool netif_oper_up(const struct net_device *dev)
3418 {
3419 return (dev->operstate == IF_OPER_UP ||
3420 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3421 }
3422
3423 /**
3424 * netif_device_present - is device available or removed
3425 * @dev: network device
3426 *
3427 * Check if device has not been removed from system.
3428 */
3429 static inline bool netif_device_present(struct net_device *dev)
3430 {
3431 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3432 }
3433
3434 void netif_device_detach(struct net_device *dev);
3435
3436 void netif_device_attach(struct net_device *dev);
3437
3438 /*
3439 * Network interface message level settings
3440 */
3441
3442 enum {
3443 NETIF_MSG_DRV = 0x0001,
3444 NETIF_MSG_PROBE = 0x0002,
3445 NETIF_MSG_LINK = 0x0004,
3446 NETIF_MSG_TIMER = 0x0008,
3447 NETIF_MSG_IFDOWN = 0x0010,
3448 NETIF_MSG_IFUP = 0x0020,
3449 NETIF_MSG_RX_ERR = 0x0040,
3450 NETIF_MSG_TX_ERR = 0x0080,
3451 NETIF_MSG_TX_QUEUED = 0x0100,
3452 NETIF_MSG_INTR = 0x0200,
3453 NETIF_MSG_TX_DONE = 0x0400,
3454 NETIF_MSG_RX_STATUS = 0x0800,
3455 NETIF_MSG_PKTDATA = 0x1000,
3456 NETIF_MSG_HW = 0x2000,
3457 NETIF_MSG_WOL = 0x4000,
3458 };
3459
3460 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3461 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3462 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3463 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3464 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3465 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3466 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3467 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3468 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3469 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3470 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3471 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3472 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3473 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3474 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3475
3476 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3477 {
3478 /* use default */
3479 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3480 return default_msg_enable_bits;
3481 if (debug_value == 0) /* no output */
3482 return 0;
3483 /* set low N bits */
3484 return (1 << debug_value) - 1;
3485 }
3486
3487 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3488 {
3489 spin_lock(&txq->_xmit_lock);
3490 txq->xmit_lock_owner = cpu;
3491 }
3492
3493 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3494 {
3495 spin_lock_bh(&txq->_xmit_lock);
3496 txq->xmit_lock_owner = smp_processor_id();
3497 }
3498
3499 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3500 {
3501 bool ok = spin_trylock(&txq->_xmit_lock);
3502 if (likely(ok))
3503 txq->xmit_lock_owner = smp_processor_id();
3504 return ok;
3505 }
3506
3507 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3508 {
3509 txq->xmit_lock_owner = -1;
3510 spin_unlock(&txq->_xmit_lock);
3511 }
3512
3513 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3514 {
3515 txq->xmit_lock_owner = -1;
3516 spin_unlock_bh(&txq->_xmit_lock);
3517 }
3518
3519 static inline void txq_trans_update(struct netdev_queue *txq)
3520 {
3521 if (txq->xmit_lock_owner != -1)
3522 txq->trans_start = jiffies;
3523 }
3524
3525 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3526 static inline void netif_trans_update(struct net_device *dev)
3527 {
3528 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3529
3530 if (txq->trans_start != jiffies)
3531 txq->trans_start = jiffies;
3532 }
3533
3534 /**
3535 * netif_tx_lock - grab network device transmit lock
3536 * @dev: network device
3537 *
3538 * Get network device transmit lock
3539 */
3540 static inline void netif_tx_lock(struct net_device *dev)
3541 {
3542 unsigned int i;
3543 int cpu;
3544
3545 spin_lock(&dev->tx_global_lock);
3546 cpu = smp_processor_id();
3547 for (i = 0; i < dev->num_tx_queues; i++) {
3548 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3549
3550 /* We are the only thread of execution doing a
3551 * freeze, but we have to grab the _xmit_lock in
3552 * order to synchronize with threads which are in
3553 * the ->hard_start_xmit() handler and already
3554 * checked the frozen bit.
3555 */
3556 __netif_tx_lock(txq, cpu);
3557 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3558 __netif_tx_unlock(txq);
3559 }
3560 }
3561
3562 static inline void netif_tx_lock_bh(struct net_device *dev)
3563 {
3564 local_bh_disable();
3565 netif_tx_lock(dev);
3566 }
3567
3568 static inline void netif_tx_unlock(struct net_device *dev)
3569 {
3570 unsigned int i;
3571
3572 for (i = 0; i < dev->num_tx_queues; i++) {
3573 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3574
3575 /* No need to grab the _xmit_lock here. If the
3576 * queue is not stopped for another reason, we
3577 * force a schedule.
3578 */
3579 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 netif_schedule_queue(txq);
3581 }
3582 spin_unlock(&dev->tx_global_lock);
3583 }
3584
3585 static inline void netif_tx_unlock_bh(struct net_device *dev)
3586 {
3587 netif_tx_unlock(dev);
3588 local_bh_enable();
3589 }
3590
3591 #define HARD_TX_LOCK(dev, txq, cpu) { \
3592 if ((dev->features & NETIF_F_LLTX) == 0) { \
3593 __netif_tx_lock(txq, cpu); \
3594 } \
3595 }
3596
3597 #define HARD_TX_TRYLOCK(dev, txq) \
3598 (((dev->features & NETIF_F_LLTX) == 0) ? \
3599 __netif_tx_trylock(txq) : \
3600 true )
3601
3602 #define HARD_TX_UNLOCK(dev, txq) { \
3603 if ((dev->features & NETIF_F_LLTX) == 0) { \
3604 __netif_tx_unlock(txq); \
3605 } \
3606 }
3607
3608 static inline void netif_tx_disable(struct net_device *dev)
3609 {
3610 unsigned int i;
3611 int cpu;
3612
3613 local_bh_disable();
3614 cpu = smp_processor_id();
3615 for (i = 0; i < dev->num_tx_queues; i++) {
3616 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3617
3618 __netif_tx_lock(txq, cpu);
3619 netif_tx_stop_queue(txq);
3620 __netif_tx_unlock(txq);
3621 }
3622 local_bh_enable();
3623 }
3624
3625 static inline void netif_addr_lock(struct net_device *dev)
3626 {
3627 spin_lock(&dev->addr_list_lock);
3628 }
3629
3630 static inline void netif_addr_lock_nested(struct net_device *dev)
3631 {
3632 int subclass = SINGLE_DEPTH_NESTING;
3633
3634 if (dev->netdev_ops->ndo_get_lock_subclass)
3635 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3636
3637 spin_lock_nested(&dev->addr_list_lock, subclass);
3638 }
3639
3640 static inline void netif_addr_lock_bh(struct net_device *dev)
3641 {
3642 spin_lock_bh(&dev->addr_list_lock);
3643 }
3644
3645 static inline void netif_addr_unlock(struct net_device *dev)
3646 {
3647 spin_unlock(&dev->addr_list_lock);
3648 }
3649
3650 static inline void netif_addr_unlock_bh(struct net_device *dev)
3651 {
3652 spin_unlock_bh(&dev->addr_list_lock);
3653 }
3654
3655 /*
3656 * dev_addrs walker. Should be used only for read access. Call with
3657 * rcu_read_lock held.
3658 */
3659 #define for_each_dev_addr(dev, ha) \
3660 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3661
3662 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3663
3664 void ether_setup(struct net_device *dev);
3665
3666 /* Support for loadable net-drivers */
3667 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3668 unsigned char name_assign_type,
3669 void (*setup)(struct net_device *),
3670 unsigned int txqs, unsigned int rxqs);
3671 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3672 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3673
3674 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3675 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3676 count)
3677
3678 int register_netdev(struct net_device *dev);
3679 void unregister_netdev(struct net_device *dev);
3680
3681 /* General hardware address lists handling functions */
3682 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3683 struct netdev_hw_addr_list *from_list, int addr_len);
3684 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3685 struct netdev_hw_addr_list *from_list, int addr_len);
3686 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3687 struct net_device *dev,
3688 int (*sync)(struct net_device *, const unsigned char *),
3689 int (*unsync)(struct net_device *,
3690 const unsigned char *));
3691 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3692 struct net_device *dev,
3693 int (*unsync)(struct net_device *,
3694 const unsigned char *));
3695 void __hw_addr_init(struct netdev_hw_addr_list *list);
3696
3697 /* Functions used for device addresses handling */
3698 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3699 unsigned char addr_type);
3700 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3701 unsigned char addr_type);
3702 void dev_addr_flush(struct net_device *dev);
3703 int dev_addr_init(struct net_device *dev);
3704
3705 /* Functions used for unicast addresses handling */
3706 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3707 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3708 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3709 int dev_uc_sync(struct net_device *to, struct net_device *from);
3710 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3711 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3712 void dev_uc_flush(struct net_device *dev);
3713 void dev_uc_init(struct net_device *dev);
3714
3715 /**
3716 * __dev_uc_sync - Synchonize device's unicast list
3717 * @dev: device to sync
3718 * @sync: function to call if address should be added
3719 * @unsync: function to call if address should be removed
3720 *
3721 * Add newly added addresses to the interface, and release
3722 * addresses that have been deleted.
3723 */
3724 static inline int __dev_uc_sync(struct net_device *dev,
3725 int (*sync)(struct net_device *,
3726 const unsigned char *),
3727 int (*unsync)(struct net_device *,
3728 const unsigned char *))
3729 {
3730 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3731 }
3732
3733 /**
3734 * __dev_uc_unsync - Remove synchronized addresses from device
3735 * @dev: device to sync
3736 * @unsync: function to call if address should be removed
3737 *
3738 * Remove all addresses that were added to the device by dev_uc_sync().
3739 */
3740 static inline void __dev_uc_unsync(struct net_device *dev,
3741 int (*unsync)(struct net_device *,
3742 const unsigned char *))
3743 {
3744 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3745 }
3746
3747 /* Functions used for multicast addresses handling */
3748 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3749 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3750 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3751 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3752 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3753 int dev_mc_sync(struct net_device *to, struct net_device *from);
3754 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3755 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3756 void dev_mc_flush(struct net_device *dev);
3757 void dev_mc_init(struct net_device *dev);
3758
3759 /**
3760 * __dev_mc_sync - Synchonize device's multicast list
3761 * @dev: device to sync
3762 * @sync: function to call if address should be added
3763 * @unsync: function to call if address should be removed
3764 *
3765 * Add newly added addresses to the interface, and release
3766 * addresses that have been deleted.
3767 */
3768 static inline int __dev_mc_sync(struct net_device *dev,
3769 int (*sync)(struct net_device *,
3770 const unsigned char *),
3771 int (*unsync)(struct net_device *,
3772 const unsigned char *))
3773 {
3774 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3775 }
3776
3777 /**
3778 * __dev_mc_unsync - Remove synchronized addresses from device
3779 * @dev: device to sync
3780 * @unsync: function to call if address should be removed
3781 *
3782 * Remove all addresses that were added to the device by dev_mc_sync().
3783 */
3784 static inline void __dev_mc_unsync(struct net_device *dev,
3785 int (*unsync)(struct net_device *,
3786 const unsigned char *))
3787 {
3788 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3789 }
3790
3791 /* Functions used for secondary unicast and multicast support */
3792 void dev_set_rx_mode(struct net_device *dev);
3793 void __dev_set_rx_mode(struct net_device *dev);
3794 int dev_set_promiscuity(struct net_device *dev, int inc);
3795 int dev_set_allmulti(struct net_device *dev, int inc);
3796 void netdev_state_change(struct net_device *dev);
3797 void netdev_notify_peers(struct net_device *dev);
3798 void netdev_features_change(struct net_device *dev);
3799 /* Load a device via the kmod */
3800 void dev_load(struct net *net, const char *name);
3801 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3802 struct rtnl_link_stats64 *storage);
3803 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3804 const struct net_device_stats *netdev_stats);
3805
3806 extern int netdev_max_backlog;
3807 extern int netdev_tstamp_prequeue;
3808 extern int weight_p;
3809
3810 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3811 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3812 struct list_head **iter);
3813 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3814 struct list_head **iter);
3815
3816 /* iterate through upper list, must be called under RCU read lock */
3817 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3818 for (iter = &(dev)->adj_list.upper, \
3819 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3820 updev; \
3821 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3822
3823 /* iterate through upper list, must be called under RCU read lock */
3824 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3825 for (iter = &(dev)->all_adj_list.upper, \
3826 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3827 updev; \
3828 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3829
3830 void *netdev_lower_get_next_private(struct net_device *dev,
3831 struct list_head **iter);
3832 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3833 struct list_head **iter);
3834
3835 #define netdev_for_each_lower_private(dev, priv, iter) \
3836 for (iter = (dev)->adj_list.lower.next, \
3837 priv = netdev_lower_get_next_private(dev, &(iter)); \
3838 priv; \
3839 priv = netdev_lower_get_next_private(dev, &(iter)))
3840
3841 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3842 for (iter = &(dev)->adj_list.lower, \
3843 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3844 priv; \
3845 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3846
3847 void *netdev_lower_get_next(struct net_device *dev,
3848 struct list_head **iter);
3849
3850 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3851 for (iter = (dev)->adj_list.lower.next, \
3852 ldev = netdev_lower_get_next(dev, &(iter)); \
3853 ldev; \
3854 ldev = netdev_lower_get_next(dev, &(iter)))
3855
3856 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3857 struct list_head **iter);
3858 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3859 struct list_head **iter);
3860
3861 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3862 for (iter = (dev)->all_adj_list.lower.next, \
3863 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3864 ldev; \
3865 ldev = netdev_all_lower_get_next(dev, &(iter)))
3866
3867 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3868 for (iter = (dev)->all_adj_list.lower.next, \
3869 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3870 ldev; \
3871 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3872
3873 void *netdev_adjacent_get_private(struct list_head *adj_list);
3874 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3875 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3876 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3877 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3878 int netdev_master_upper_dev_link(struct net_device *dev,
3879 struct net_device *upper_dev,
3880 void *upper_priv, void *upper_info);
3881 void netdev_upper_dev_unlink(struct net_device *dev,
3882 struct net_device *upper_dev);
3883 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3884 void *netdev_lower_dev_get_private(struct net_device *dev,
3885 struct net_device *lower_dev);
3886 void netdev_lower_state_changed(struct net_device *lower_dev,
3887 void *lower_state_info);
3888 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3889 struct neighbour *n);
3890 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3891 struct neighbour *n);
3892
3893 /* RSS keys are 40 or 52 bytes long */
3894 #define NETDEV_RSS_KEY_LEN 52
3895 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3896 void netdev_rss_key_fill(void *buffer, size_t len);
3897
3898 int dev_get_nest_level(struct net_device *dev,
3899 bool (*type_check)(const struct net_device *dev));
3900 int skb_checksum_help(struct sk_buff *skb);
3901 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3902 netdev_features_t features, bool tx_path);
3903 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3904 netdev_features_t features);
3905
3906 struct netdev_bonding_info {
3907 ifslave slave;
3908 ifbond master;
3909 };
3910
3911 struct netdev_notifier_bonding_info {
3912 struct netdev_notifier_info info; /* must be first */
3913 struct netdev_bonding_info bonding_info;
3914 };
3915
3916 void netdev_bonding_info_change(struct net_device *dev,
3917 struct netdev_bonding_info *bonding_info);
3918
3919 static inline
3920 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3921 {
3922 return __skb_gso_segment(skb, features, true);
3923 }
3924 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3925
3926 static inline bool can_checksum_protocol(netdev_features_t features,
3927 __be16 protocol)
3928 {
3929 if (protocol == htons(ETH_P_FCOE))
3930 return !!(features & NETIF_F_FCOE_CRC);
3931
3932 /* Assume this is an IP checksum (not SCTP CRC) */
3933
3934 if (features & NETIF_F_HW_CSUM) {
3935 /* Can checksum everything */
3936 return true;
3937 }
3938
3939 switch (protocol) {
3940 case htons(ETH_P_IP):
3941 return !!(features & NETIF_F_IP_CSUM);
3942 case htons(ETH_P_IPV6):
3943 return !!(features & NETIF_F_IPV6_CSUM);
3944 default:
3945 return false;
3946 }
3947 }
3948
3949 /* Map an ethertype into IP protocol if possible */
3950 static inline int eproto_to_ipproto(int eproto)
3951 {
3952 switch (eproto) {
3953 case htons(ETH_P_IP):
3954 return IPPROTO_IP;
3955 case htons(ETH_P_IPV6):
3956 return IPPROTO_IPV6;
3957 default:
3958 return -1;
3959 }
3960 }
3961
3962 #ifdef CONFIG_BUG
3963 void netdev_rx_csum_fault(struct net_device *dev);
3964 #else
3965 static inline void netdev_rx_csum_fault(struct net_device *dev)
3966 {
3967 }
3968 #endif
3969 /* rx skb timestamps */
3970 void net_enable_timestamp(void);
3971 void net_disable_timestamp(void);
3972
3973 #ifdef CONFIG_PROC_FS
3974 int __init dev_proc_init(void);
3975 #else
3976 #define dev_proc_init() 0
3977 #endif
3978
3979 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3980 struct sk_buff *skb, struct net_device *dev,
3981 bool more)
3982 {
3983 skb->xmit_more = more ? 1 : 0;
3984 return ops->ndo_start_xmit(skb, dev);
3985 }
3986
3987 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3988 struct netdev_queue *txq, bool more)
3989 {
3990 const struct net_device_ops *ops = dev->netdev_ops;
3991 int rc;
3992
3993 rc = __netdev_start_xmit(ops, skb, dev, more);
3994 if (rc == NETDEV_TX_OK)
3995 txq_trans_update(txq);
3996
3997 return rc;
3998 }
3999
4000 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4001 const void *ns);
4002 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4003 const void *ns);
4004
4005 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4006 {
4007 return netdev_class_create_file_ns(class_attr, NULL);
4008 }
4009
4010 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4011 {
4012 netdev_class_remove_file_ns(class_attr, NULL);
4013 }
4014
4015 extern struct kobj_ns_type_operations net_ns_type_operations;
4016
4017 const char *netdev_drivername(const struct net_device *dev);
4018
4019 void linkwatch_run_queue(void);
4020
4021 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4022 netdev_features_t f2)
4023 {
4024 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4025 if (f1 & NETIF_F_HW_CSUM)
4026 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4027 else
4028 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4029 }
4030
4031 return f1 & f2;
4032 }
4033
4034 static inline netdev_features_t netdev_get_wanted_features(
4035 struct net_device *dev)
4036 {
4037 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4038 }
4039 netdev_features_t netdev_increment_features(netdev_features_t all,
4040 netdev_features_t one, netdev_features_t mask);
4041
4042 /* Allow TSO being used on stacked device :
4043 * Performing the GSO segmentation before last device
4044 * is a performance improvement.
4045 */
4046 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4047 netdev_features_t mask)
4048 {
4049 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4050 }
4051
4052 int __netdev_update_features(struct net_device *dev);
4053 void netdev_update_features(struct net_device *dev);
4054 void netdev_change_features(struct net_device *dev);
4055
4056 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4057 struct net_device *dev);
4058
4059 netdev_features_t passthru_features_check(struct sk_buff *skb,
4060 struct net_device *dev,
4061 netdev_features_t features);
4062 netdev_features_t netif_skb_features(struct sk_buff *skb);
4063
4064 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4065 {
4066 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4067
4068 /* check flags correspondence */
4069 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4070 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4071 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4072 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4073 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4074 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4075 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4076 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4077 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4078 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4079 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4080 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4081 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4082 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4083 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4084 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4085
4086 return (features & feature) == feature;
4087 }
4088
4089 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4090 {
4091 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4092 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4093 }
4094
4095 static inline bool netif_needs_gso(struct sk_buff *skb,
4096 netdev_features_t features)
4097 {
4098 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4099 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4100 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4101 }
4102
4103 static inline void netif_set_gso_max_size(struct net_device *dev,
4104 unsigned int size)
4105 {
4106 dev->gso_max_size = size;
4107 }
4108
4109 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4110 int pulled_hlen, u16 mac_offset,
4111 int mac_len)
4112 {
4113 skb->protocol = protocol;
4114 skb->encapsulation = 1;
4115 skb_push(skb, pulled_hlen);
4116 skb_reset_transport_header(skb);
4117 skb->mac_header = mac_offset;
4118 skb->network_header = skb->mac_header + mac_len;
4119 skb->mac_len = mac_len;
4120 }
4121
4122 static inline bool netif_is_macsec(const struct net_device *dev)
4123 {
4124 return dev->priv_flags & IFF_MACSEC;
4125 }
4126
4127 static inline bool netif_is_macvlan(const struct net_device *dev)
4128 {
4129 return dev->priv_flags & IFF_MACVLAN;
4130 }
4131
4132 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4133 {
4134 return dev->priv_flags & IFF_MACVLAN_PORT;
4135 }
4136
4137 static inline bool netif_is_ipvlan(const struct net_device *dev)
4138 {
4139 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4140 }
4141
4142 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4143 {
4144 return dev->priv_flags & IFF_IPVLAN_MASTER;
4145 }
4146
4147 static inline bool netif_is_bond_master(const struct net_device *dev)
4148 {
4149 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4150 }
4151
4152 static inline bool netif_is_bond_slave(const struct net_device *dev)
4153 {
4154 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4155 }
4156
4157 static inline bool netif_supports_nofcs(struct net_device *dev)
4158 {
4159 return dev->priv_flags & IFF_SUPP_NOFCS;
4160 }
4161
4162 static inline bool netif_is_l3_master(const struct net_device *dev)
4163 {
4164 return dev->priv_flags & IFF_L3MDEV_MASTER;
4165 }
4166
4167 static inline bool netif_is_l3_slave(const struct net_device *dev)
4168 {
4169 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4170 }
4171
4172 static inline bool netif_is_bridge_master(const struct net_device *dev)
4173 {
4174 return dev->priv_flags & IFF_EBRIDGE;
4175 }
4176
4177 static inline bool netif_is_bridge_port(const struct net_device *dev)
4178 {
4179 return dev->priv_flags & IFF_BRIDGE_PORT;
4180 }
4181
4182 static inline bool netif_is_ovs_master(const struct net_device *dev)
4183 {
4184 return dev->priv_flags & IFF_OPENVSWITCH;
4185 }
4186
4187 static inline bool netif_is_team_master(const struct net_device *dev)
4188 {
4189 return dev->priv_flags & IFF_TEAM;
4190 }
4191
4192 static inline bool netif_is_team_port(const struct net_device *dev)
4193 {
4194 return dev->priv_flags & IFF_TEAM_PORT;
4195 }
4196
4197 static inline bool netif_is_lag_master(const struct net_device *dev)
4198 {
4199 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4200 }
4201
4202 static inline bool netif_is_lag_port(const struct net_device *dev)
4203 {
4204 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4205 }
4206
4207 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4208 {
4209 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4210 }
4211
4212 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4213 static inline void netif_keep_dst(struct net_device *dev)
4214 {
4215 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4216 }
4217
4218 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4219 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4220 {
4221 /* TODO: reserve and use an additional IFF bit, if we get more users */
4222 return dev->priv_flags & IFF_MACSEC;
4223 }
4224
4225 extern struct pernet_operations __net_initdata loopback_net_ops;
4226
4227 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4228
4229 /* netdev_printk helpers, similar to dev_printk */
4230
4231 static inline const char *netdev_name(const struct net_device *dev)
4232 {
4233 if (!dev->name[0] || strchr(dev->name, '%'))
4234 return "(unnamed net_device)";
4235 return dev->name;
4236 }
4237
4238 static inline const char *netdev_reg_state(const struct net_device *dev)
4239 {
4240 switch (dev->reg_state) {
4241 case NETREG_UNINITIALIZED: return " (uninitialized)";
4242 case NETREG_REGISTERED: return "";
4243 case NETREG_UNREGISTERING: return " (unregistering)";
4244 case NETREG_UNREGISTERED: return " (unregistered)";
4245 case NETREG_RELEASED: return " (released)";
4246 case NETREG_DUMMY: return " (dummy)";
4247 }
4248
4249 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4250 return " (unknown)";
4251 }
4252
4253 __printf(3, 4)
4254 void netdev_printk(const char *level, const struct net_device *dev,
4255 const char *format, ...);
4256 __printf(2, 3)
4257 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4258 __printf(2, 3)
4259 void netdev_alert(const struct net_device *dev, const char *format, ...);
4260 __printf(2, 3)
4261 void netdev_crit(const struct net_device *dev, const char *format, ...);
4262 __printf(2, 3)
4263 void netdev_err(const struct net_device *dev, const char *format, ...);
4264 __printf(2, 3)
4265 void netdev_warn(const struct net_device *dev, const char *format, ...);
4266 __printf(2, 3)
4267 void netdev_notice(const struct net_device *dev, const char *format, ...);
4268 __printf(2, 3)
4269 void netdev_info(const struct net_device *dev, const char *format, ...);
4270
4271 #define MODULE_ALIAS_NETDEV(device) \
4272 MODULE_ALIAS("netdev-" device)
4273
4274 #if defined(CONFIG_DYNAMIC_DEBUG)
4275 #define netdev_dbg(__dev, format, args...) \
4276 do { \
4277 dynamic_netdev_dbg(__dev, format, ##args); \
4278 } while (0)
4279 #elif defined(DEBUG)
4280 #define netdev_dbg(__dev, format, args...) \
4281 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4282 #else
4283 #define netdev_dbg(__dev, format, args...) \
4284 ({ \
4285 if (0) \
4286 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4287 })
4288 #endif
4289
4290 #if defined(VERBOSE_DEBUG)
4291 #define netdev_vdbg netdev_dbg
4292 #else
4293
4294 #define netdev_vdbg(dev, format, args...) \
4295 ({ \
4296 if (0) \
4297 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4298 0; \
4299 })
4300 #endif
4301
4302 /*
4303 * netdev_WARN() acts like dev_printk(), but with the key difference
4304 * of using a WARN/WARN_ON to get the message out, including the
4305 * file/line information and a backtrace.
4306 */
4307 #define netdev_WARN(dev, format, args...) \
4308 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4309 netdev_reg_state(dev), ##args)
4310
4311 /* netif printk helpers, similar to netdev_printk */
4312
4313 #define netif_printk(priv, type, level, dev, fmt, args...) \
4314 do { \
4315 if (netif_msg_##type(priv)) \
4316 netdev_printk(level, (dev), fmt, ##args); \
4317 } while (0)
4318
4319 #define netif_level(level, priv, type, dev, fmt, args...) \
4320 do { \
4321 if (netif_msg_##type(priv)) \
4322 netdev_##level(dev, fmt, ##args); \
4323 } while (0)
4324
4325 #define netif_emerg(priv, type, dev, fmt, args...) \
4326 netif_level(emerg, priv, type, dev, fmt, ##args)
4327 #define netif_alert(priv, type, dev, fmt, args...) \
4328 netif_level(alert, priv, type, dev, fmt, ##args)
4329 #define netif_crit(priv, type, dev, fmt, args...) \
4330 netif_level(crit, priv, type, dev, fmt, ##args)
4331 #define netif_err(priv, type, dev, fmt, args...) \
4332 netif_level(err, priv, type, dev, fmt, ##args)
4333 #define netif_warn(priv, type, dev, fmt, args...) \
4334 netif_level(warn, priv, type, dev, fmt, ##args)
4335 #define netif_notice(priv, type, dev, fmt, args...) \
4336 netif_level(notice, priv, type, dev, fmt, ##args)
4337 #define netif_info(priv, type, dev, fmt, args...) \
4338 netif_level(info, priv, type, dev, fmt, ##args)
4339
4340 #if defined(CONFIG_DYNAMIC_DEBUG)
4341 #define netif_dbg(priv, type, netdev, format, args...) \
4342 do { \
4343 if (netif_msg_##type(priv)) \
4344 dynamic_netdev_dbg(netdev, format, ##args); \
4345 } while (0)
4346 #elif defined(DEBUG)
4347 #define netif_dbg(priv, type, dev, format, args...) \
4348 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4349 #else
4350 #define netif_dbg(priv, type, dev, format, args...) \
4351 ({ \
4352 if (0) \
4353 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4354 0; \
4355 })
4356 #endif
4357
4358 #if defined(VERBOSE_DEBUG)
4359 #define netif_vdbg netif_dbg
4360 #else
4361 #define netif_vdbg(priv, type, dev, format, args...) \
4362 ({ \
4363 if (0) \
4364 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4365 0; \
4366 })
4367 #endif
4368
4369 /*
4370 * The list of packet types we will receive (as opposed to discard)
4371 * and the routines to invoke.
4372 *
4373 * Why 16. Because with 16 the only overlap we get on a hash of the
4374 * low nibble of the protocol value is RARP/SNAP/X.25.
4375 *
4376 * NOTE: That is no longer true with the addition of VLAN tags. Not
4377 * sure which should go first, but I bet it won't make much
4378 * difference if we are running VLANs. The good news is that
4379 * this protocol won't be in the list unless compiled in, so
4380 * the average user (w/out VLANs) will not be adversely affected.
4381 * --BLG
4382 *
4383 * 0800 IP
4384 * 8100 802.1Q VLAN
4385 * 0001 802.3
4386 * 0002 AX.25
4387 * 0004 802.2
4388 * 8035 RARP
4389 * 0005 SNAP
4390 * 0805 X.25
4391 * 0806 ARP
4392 * 8137 IPX
4393 * 0009 Localtalk
4394 * 86DD IPv6
4395 */
4396 #define PTYPE_HASH_SIZE (16)
4397 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4398
4399 #endif /* _LINUX_NETDEVICE_H */