]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/netdevice.h
net: introduce lower state changed info structure for LAG lowers
[mirror_ubuntu-bionic-kernel.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
70
71 /*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
103
104 enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 };
110 typedef enum netdev_tx netdev_tx_t;
111
112 /*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116 static inline bool dev_xmit_complete(int rc)
117 {
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128 }
129
130 /*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135 #if defined(CONFIG_HYPERV_NET)
136 # define LL_MAX_HEADER 128
137 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
138 # if defined(CONFIG_MAC80211_MESH)
139 # define LL_MAX_HEADER 128
140 # else
141 # define LL_MAX_HEADER 96
142 # endif
143 #else
144 # define LL_MAX_HEADER 32
145 #endif
146
147 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
148 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
149 #define MAX_HEADER LL_MAX_HEADER
150 #else
151 #define MAX_HEADER (LL_MAX_HEADER + 48)
152 #endif
153
154 /*
155 * Old network device statistics. Fields are native words
156 * (unsigned long) so they can be read and written atomically.
157 */
158
159 struct net_device_stats {
160 unsigned long rx_packets;
161 unsigned long tx_packets;
162 unsigned long rx_bytes;
163 unsigned long tx_bytes;
164 unsigned long rx_errors;
165 unsigned long tx_errors;
166 unsigned long rx_dropped;
167 unsigned long tx_dropped;
168 unsigned long multicast;
169 unsigned long collisions;
170 unsigned long rx_length_errors;
171 unsigned long rx_over_errors;
172 unsigned long rx_crc_errors;
173 unsigned long rx_frame_errors;
174 unsigned long rx_fifo_errors;
175 unsigned long rx_missed_errors;
176 unsigned long tx_aborted_errors;
177 unsigned long tx_carrier_errors;
178 unsigned long tx_fifo_errors;
179 unsigned long tx_heartbeat_errors;
180 unsigned long tx_window_errors;
181 unsigned long rx_compressed;
182 unsigned long tx_compressed;
183 };
184
185
186 #include <linux/cache.h>
187 #include <linux/skbuff.h>
188
189 #ifdef CONFIG_RPS
190 #include <linux/static_key.h>
191 extern struct static_key rps_needed;
192 #endif
193
194 struct neighbour;
195 struct neigh_parms;
196 struct sk_buff;
197
198 struct netdev_hw_addr {
199 struct list_head list;
200 unsigned char addr[MAX_ADDR_LEN];
201 unsigned char type;
202 #define NETDEV_HW_ADDR_T_LAN 1
203 #define NETDEV_HW_ADDR_T_SAN 2
204 #define NETDEV_HW_ADDR_T_SLAVE 3
205 #define NETDEV_HW_ADDR_T_UNICAST 4
206 #define NETDEV_HW_ADDR_T_MULTICAST 5
207 bool global_use;
208 int sync_cnt;
209 int refcount;
210 int synced;
211 struct rcu_head rcu_head;
212 };
213
214 struct netdev_hw_addr_list {
215 struct list_head list;
216 int count;
217 };
218
219 #define netdev_hw_addr_list_count(l) ((l)->count)
220 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
221 #define netdev_hw_addr_list_for_each(ha, l) \
222 list_for_each_entry(ha, &(l)->list, list)
223
224 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
225 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
226 #define netdev_for_each_uc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
228
229 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
230 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
231 #define netdev_for_each_mc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
233
234 struct hh_cache {
235 u16 hh_len;
236 u16 __pad;
237 seqlock_t hh_lock;
238
239 /* cached hardware header; allow for machine alignment needs. */
240 #define HH_DATA_MOD 16
241 #define HH_DATA_OFF(__len) \
242 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
243 #define HH_DATA_ALIGN(__len) \
244 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
245 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 };
247
248 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
249 * Alternative is:
250 * dev->hard_header_len ? (dev->hard_header_len +
251 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
252 *
253 * We could use other alignment values, but we must maintain the
254 * relationship HH alignment <= LL alignment.
255 */
256 #define LL_RESERVED_SPACE(dev) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260
261 struct header_ops {
262 int (*create) (struct sk_buff *skb, struct net_device *dev,
263 unsigned short type, const void *daddr,
264 const void *saddr, unsigned int len);
265 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
266 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
267 void (*cache_update)(struct hh_cache *hh,
268 const struct net_device *dev,
269 const unsigned char *haddr);
270 };
271
272 /* These flag bits are private to the generic network queueing
273 * layer, they may not be explicitly referenced by any other
274 * code.
275 */
276
277 enum netdev_state_t {
278 __LINK_STATE_START,
279 __LINK_STATE_PRESENT,
280 __LINK_STATE_NOCARRIER,
281 __LINK_STATE_LINKWATCH_PENDING,
282 __LINK_STATE_DORMANT,
283 };
284
285
286 /*
287 * This structure holds at boot time configured netdevice settings. They
288 * are then used in the device probing.
289 */
290 struct netdev_boot_setup {
291 char name[IFNAMSIZ];
292 struct ifmap map;
293 };
294 #define NETDEV_BOOT_SETUP_MAX 8
295
296 int __init netdev_boot_setup(char *str);
297
298 /*
299 * Structure for NAPI scheduling similar to tasklet but with weighting
300 */
301 struct napi_struct {
302 /* The poll_list must only be managed by the entity which
303 * changes the state of the NAPI_STATE_SCHED bit. This means
304 * whoever atomically sets that bit can add this napi_struct
305 * to the per-cpu poll_list, and whoever clears that bit
306 * can remove from the list right before clearing the bit.
307 */
308 struct list_head poll_list;
309
310 unsigned long state;
311 int weight;
312 unsigned int gro_count;
313 int (*poll)(struct napi_struct *, int);
314 #ifdef CONFIG_NETPOLL
315 spinlock_t poll_lock;
316 int poll_owner;
317 #endif
318 struct net_device *dev;
319 struct sk_buff *gro_list;
320 struct sk_buff *skb;
321 struct hrtimer timer;
322 struct list_head dev_list;
323 struct hlist_node napi_hash_node;
324 unsigned int napi_id;
325 };
326
327 enum {
328 NAPI_STATE_SCHED, /* Poll is scheduled */
329 NAPI_STATE_DISABLE, /* Disable pending */
330 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
331 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
332 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
333 };
334
335 enum gro_result {
336 GRO_MERGED,
337 GRO_MERGED_FREE,
338 GRO_HELD,
339 GRO_NORMAL,
340 GRO_DROP,
341 };
342 typedef enum gro_result gro_result_t;
343
344 /*
345 * enum rx_handler_result - Possible return values for rx_handlers.
346 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
347 * further.
348 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
349 * case skb->dev was changed by rx_handler.
350 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
351 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
352 *
353 * rx_handlers are functions called from inside __netif_receive_skb(), to do
354 * special processing of the skb, prior to delivery to protocol handlers.
355 *
356 * Currently, a net_device can only have a single rx_handler registered. Trying
357 * to register a second rx_handler will return -EBUSY.
358 *
359 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
360 * To unregister a rx_handler on a net_device, use
361 * netdev_rx_handler_unregister().
362 *
363 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * do with the skb.
365 *
366 * If the rx_handler consumed to skb in some way, it should return
367 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
368 * the skb to be delivered in some other ways.
369 *
370 * If the rx_handler changed skb->dev, to divert the skb to another
371 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
372 * new device will be called if it exists.
373 *
374 * If the rx_handler consider the skb should be ignored, it should return
375 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
376 * are registered on exact device (ptype->dev == skb->dev).
377 *
378 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
379 * delivered, it should return RX_HANDLER_PASS.
380 *
381 * A device without a registered rx_handler will behave as if rx_handler
382 * returned RX_HANDLER_PASS.
383 */
384
385 enum rx_handler_result {
386 RX_HANDLER_CONSUMED,
387 RX_HANDLER_ANOTHER,
388 RX_HANDLER_EXACT,
389 RX_HANDLER_PASS,
390 };
391 typedef enum rx_handler_result rx_handler_result_t;
392 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
393
394 void __napi_schedule(struct napi_struct *n);
395 void __napi_schedule_irqoff(struct napi_struct *n);
396
397 static inline bool napi_disable_pending(struct napi_struct *n)
398 {
399 return test_bit(NAPI_STATE_DISABLE, &n->state);
400 }
401
402 /**
403 * napi_schedule_prep - check if napi can be scheduled
404 * @n: napi context
405 *
406 * Test if NAPI routine is already running, and if not mark
407 * it as running. This is used as a condition variable
408 * insure only one NAPI poll instance runs. We also make
409 * sure there is no pending NAPI disable.
410 */
411 static inline bool napi_schedule_prep(struct napi_struct *n)
412 {
413 return !napi_disable_pending(n) &&
414 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
415 }
416
417 /**
418 * napi_schedule - schedule NAPI poll
419 * @n: napi context
420 *
421 * Schedule NAPI poll routine to be called if it is not already
422 * running.
423 */
424 static inline void napi_schedule(struct napi_struct *n)
425 {
426 if (napi_schedule_prep(n))
427 __napi_schedule(n);
428 }
429
430 /**
431 * napi_schedule_irqoff - schedule NAPI poll
432 * @n: napi context
433 *
434 * Variant of napi_schedule(), assuming hard irqs are masked.
435 */
436 static inline void napi_schedule_irqoff(struct napi_struct *n)
437 {
438 if (napi_schedule_prep(n))
439 __napi_schedule_irqoff(n);
440 }
441
442 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
443 static inline bool napi_reschedule(struct napi_struct *napi)
444 {
445 if (napi_schedule_prep(napi)) {
446 __napi_schedule(napi);
447 return true;
448 }
449 return false;
450 }
451
452 void __napi_complete(struct napi_struct *n);
453 void napi_complete_done(struct napi_struct *n, int work_done);
454 /**
455 * napi_complete - NAPI processing complete
456 * @n: napi context
457 *
458 * Mark NAPI processing as complete.
459 * Consider using napi_complete_done() instead.
460 */
461 static inline void napi_complete(struct napi_struct *n)
462 {
463 return napi_complete_done(n, 0);
464 }
465
466 /**
467 * napi_hash_add - add a NAPI to global hashtable
468 * @napi: napi context
469 *
470 * generate a new napi_id and store a @napi under it in napi_hash
471 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
472 * Note: This is normally automatically done from netif_napi_add(),
473 * so might disappear in a future linux version.
474 */
475 void napi_hash_add(struct napi_struct *napi);
476
477 /**
478 * napi_hash_del - remove a NAPI from global table
479 * @napi: napi context
480 *
481 * Warning: caller must observe rcu grace period
482 * before freeing memory containing @napi, if
483 * this function returns true.
484 * Note: core networking stack automatically calls it
485 * from netif_napi_del()
486 * Drivers might want to call this helper to combine all
487 * the needed rcu grace periods into a single one.
488 */
489 bool napi_hash_del(struct napi_struct *napi);
490
491 /**
492 * napi_disable - prevent NAPI from scheduling
493 * @n: napi context
494 *
495 * Stop NAPI from being scheduled on this context.
496 * Waits till any outstanding processing completes.
497 */
498 void napi_disable(struct napi_struct *n);
499
500 /**
501 * napi_enable - enable NAPI scheduling
502 * @n: napi context
503 *
504 * Resume NAPI from being scheduled on this context.
505 * Must be paired with napi_disable.
506 */
507 static inline void napi_enable(struct napi_struct *n)
508 {
509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
510 smp_mb__before_atomic();
511 clear_bit(NAPI_STATE_SCHED, &n->state);
512 clear_bit(NAPI_STATE_NPSVC, &n->state);
513 }
514
515 #ifdef CONFIG_SMP
516 /**
517 * napi_synchronize - wait until NAPI is not running
518 * @n: napi context
519 *
520 * Wait until NAPI is done being scheduled on this context.
521 * Waits till any outstanding processing completes but
522 * does not disable future activations.
523 */
524 static inline void napi_synchronize(const struct napi_struct *n)
525 {
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
527 msleep(1);
528 }
529 #else
530 # define napi_synchronize(n) barrier()
531 #endif
532
533 enum netdev_queue_state_t {
534 __QUEUE_STATE_DRV_XOFF,
535 __QUEUE_STATE_STACK_XOFF,
536 __QUEUE_STATE_FROZEN,
537 };
538
539 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
540 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
542
543 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
545 QUEUE_STATE_FROZEN)
546 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 QUEUE_STATE_FROZEN)
548
549 /*
550 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
551 * netif_tx_* functions below are used to manipulate this flag. The
552 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
553 * queue independently. The netif_xmit_*stopped functions below are called
554 * to check if the queue has been stopped by the driver or stack (either
555 * of the XOFF bits are set in the state). Drivers should not need to call
556 * netif_xmit*stopped functions, they should only be using netif_tx_*.
557 */
558
559 struct netdev_queue {
560 /*
561 * read mostly part
562 */
563 struct net_device *dev;
564 struct Qdisc __rcu *qdisc;
565 struct Qdisc *qdisc_sleeping;
566 #ifdef CONFIG_SYSFS
567 struct kobject kobj;
568 #endif
569 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
570 int numa_node;
571 #endif
572 /*
573 * write mostly part
574 */
575 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 int xmit_lock_owner;
577 /*
578 * please use this field instead of dev->trans_start
579 */
580 unsigned long trans_start;
581
582 /*
583 * Number of TX timeouts for this queue
584 * (/sys/class/net/DEV/Q/trans_timeout)
585 */
586 unsigned long trans_timeout;
587
588 unsigned long state;
589
590 #ifdef CONFIG_BQL
591 struct dql dql;
592 #endif
593 unsigned long tx_maxrate;
594 } ____cacheline_aligned_in_smp;
595
596 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597 {
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 return q->numa_node;
600 #else
601 return NUMA_NO_NODE;
602 #endif
603 }
604
605 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606 {
607 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 q->numa_node = node;
609 #endif
610 }
611
612 #ifdef CONFIG_RPS
613 /*
614 * This structure holds an RPS map which can be of variable length. The
615 * map is an array of CPUs.
616 */
617 struct rps_map {
618 unsigned int len;
619 struct rcu_head rcu;
620 u16 cpus[0];
621 };
622 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
623
624 /*
625 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626 * tail pointer for that CPU's input queue at the time of last enqueue, and
627 * a hardware filter index.
628 */
629 struct rps_dev_flow {
630 u16 cpu;
631 u16 filter;
632 unsigned int last_qtail;
633 };
634 #define RPS_NO_FILTER 0xffff
635
636 /*
637 * The rps_dev_flow_table structure contains a table of flow mappings.
638 */
639 struct rps_dev_flow_table {
640 unsigned int mask;
641 struct rcu_head rcu;
642 struct rps_dev_flow flows[0];
643 };
644 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
645 ((_num) * sizeof(struct rps_dev_flow)))
646
647 /*
648 * The rps_sock_flow_table contains mappings of flows to the last CPU
649 * on which they were processed by the application (set in recvmsg).
650 * Each entry is a 32bit value. Upper part is the high order bits
651 * of flow hash, lower part is cpu number.
652 * rps_cpu_mask is used to partition the space, depending on number of
653 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
655 * meaning we use 32-6=26 bits for the hash.
656 */
657 struct rps_sock_flow_table {
658 u32 mask;
659
660 u32 ents[0] ____cacheline_aligned_in_smp;
661 };
662 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
663
664 #define RPS_NO_CPU 0xffff
665
666 extern u32 rps_cpu_mask;
667 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668
669 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 u32 hash)
671 {
672 if (table && hash) {
673 unsigned int index = hash & table->mask;
674 u32 val = hash & ~rps_cpu_mask;
675
676 /* We only give a hint, preemption can change cpu under us */
677 val |= raw_smp_processor_id();
678
679 if (table->ents[index] != val)
680 table->ents[index] = val;
681 }
682 }
683
684 #ifdef CONFIG_RFS_ACCEL
685 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 u16 filter_id);
687 #endif
688 #endif /* CONFIG_RPS */
689
690 /* This structure contains an instance of an RX queue. */
691 struct netdev_rx_queue {
692 #ifdef CONFIG_RPS
693 struct rps_map __rcu *rps_map;
694 struct rps_dev_flow_table __rcu *rps_flow_table;
695 #endif
696 struct kobject kobj;
697 struct net_device *dev;
698 } ____cacheline_aligned_in_smp;
699
700 /*
701 * RX queue sysfs structures and functions.
702 */
703 struct rx_queue_attribute {
704 struct attribute attr;
705 ssize_t (*show)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, char *buf);
707 ssize_t (*store)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, const char *buf, size_t len);
709 };
710
711 #ifdef CONFIG_XPS
712 /*
713 * This structure holds an XPS map which can be of variable length. The
714 * map is an array of queues.
715 */
716 struct xps_map {
717 unsigned int len;
718 unsigned int alloc_len;
719 struct rcu_head rcu;
720 u16 queues[0];
721 };
722 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
723 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724 - sizeof(struct xps_map)) / sizeof(u16))
725
726 /*
727 * This structure holds all XPS maps for device. Maps are indexed by CPU.
728 */
729 struct xps_dev_maps {
730 struct rcu_head rcu;
731 struct xps_map __rcu *cpu_map[0];
732 };
733 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
734 (nr_cpu_ids * sizeof(struct xps_map *)))
735 #endif /* CONFIG_XPS */
736
737 #define TC_MAX_QUEUE 16
738 #define TC_BITMASK 15
739 /* HW offloaded queuing disciplines txq count and offset maps */
740 struct netdev_tc_txq {
741 u16 count;
742 u16 offset;
743 };
744
745 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746 /*
747 * This structure is to hold information about the device
748 * configured to run FCoE protocol stack.
749 */
750 struct netdev_fcoe_hbainfo {
751 char manufacturer[64];
752 char serial_number[64];
753 char hardware_version[64];
754 char driver_version[64];
755 char optionrom_version[64];
756 char firmware_version[64];
757 char model[256];
758 char model_description[256];
759 };
760 #endif
761
762 #define MAX_PHYS_ITEM_ID_LEN 32
763
764 /* This structure holds a unique identifier to identify some
765 * physical item (port for example) used by a netdevice.
766 */
767 struct netdev_phys_item_id {
768 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
769 unsigned char id_len;
770 };
771
772 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 struct netdev_phys_item_id *b)
774 {
775 return a->id_len == b->id_len &&
776 memcmp(a->id, b->id, a->id_len) == 0;
777 }
778
779 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 struct sk_buff *skb);
781
782 /*
783 * This structure defines the management hooks for network devices.
784 * The following hooks can be defined; unless noted otherwise, they are
785 * optional and can be filled with a null pointer.
786 *
787 * int (*ndo_init)(struct net_device *dev);
788 * This function is called once when network device is registered.
789 * The network device can use this to any late stage initializaton
790 * or semantic validattion. It can fail with an error code which will
791 * be propogated back to register_netdev
792 *
793 * void (*ndo_uninit)(struct net_device *dev);
794 * This function is called when device is unregistered or when registration
795 * fails. It is not called if init fails.
796 *
797 * int (*ndo_open)(struct net_device *dev);
798 * This function is called when network device transistions to the up
799 * state.
800 *
801 * int (*ndo_stop)(struct net_device *dev);
802 * This function is called when network device transistions to the down
803 * state.
804 *
805 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
806 * struct net_device *dev);
807 * Called when a packet needs to be transmitted.
808 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
809 * the queue before that can happen; it's for obsolete devices and weird
810 * corner cases, but the stack really does a non-trivial amount
811 * of useless work if you return NETDEV_TX_BUSY.
812 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
813 * Required can not be NULL.
814 *
815 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
816 * void *accel_priv, select_queue_fallback_t fallback);
817 * Called to decide which queue to when device supports multiple
818 * transmit queues.
819 *
820 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
821 * This function is called to allow device receiver to make
822 * changes to configuration when multicast or promiscious is enabled.
823 *
824 * void (*ndo_set_rx_mode)(struct net_device *dev);
825 * This function is called device changes address list filtering.
826 * If driver handles unicast address filtering, it should set
827 * IFF_UNICAST_FLT to its priv_flags.
828 *
829 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
830 * This function is called when the Media Access Control address
831 * needs to be changed. If this interface is not defined, the
832 * mac address can not be changed.
833 *
834 * int (*ndo_validate_addr)(struct net_device *dev);
835 * Test if Media Access Control address is valid for the device.
836 *
837 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
838 * Called when a user request an ioctl which can't be handled by
839 * the generic interface code. If not defined ioctl's return
840 * not supported error code.
841 *
842 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
843 * Used to set network devices bus interface parameters. This interface
844 * is retained for legacy reason, new devices should use the bus
845 * interface (PCI) for low level management.
846 *
847 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
848 * Called when a user wants to change the Maximum Transfer Unit
849 * of a device. If not defined, any request to change MTU will
850 * will return an error.
851 *
852 * void (*ndo_tx_timeout)(struct net_device *dev);
853 * Callback uses when the transmitter has not made any progress
854 * for dev->watchdog ticks.
855 *
856 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
857 * struct rtnl_link_stats64 *storage);
858 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
859 * Called when a user wants to get the network device usage
860 * statistics. Drivers must do one of the following:
861 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
862 * rtnl_link_stats64 structure passed by the caller.
863 * 2. Define @ndo_get_stats to update a net_device_stats structure
864 * (which should normally be dev->stats) and return a pointer to
865 * it. The structure may be changed asynchronously only if each
866 * field is written atomically.
867 * 3. Update dev->stats asynchronously and atomically, and define
868 * neither operation.
869 *
870 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
871 * If device support VLAN filtering this function is called when a
872 * VLAN id is registered.
873 *
874 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
875 * If device support VLAN filtering this function is called when a
876 * VLAN id is unregistered.
877 *
878 * void (*ndo_poll_controller)(struct net_device *dev);
879 *
880 * SR-IOV management functions.
881 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
882 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
883 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
884 * int max_tx_rate);
885 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
886 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
887 * int (*ndo_get_vf_config)(struct net_device *dev,
888 * int vf, struct ifla_vf_info *ivf);
889 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
890 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
891 * struct nlattr *port[]);
892 *
893 * Enable or disable the VF ability to query its RSS Redirection Table and
894 * Hash Key. This is needed since on some devices VF share this information
895 * with PF and querying it may adduce a theoretical security risk.
896 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
897 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
898 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
899 * Called to setup 'tc' number of traffic classes in the net device. This
900 * is always called from the stack with the rtnl lock held and netif tx
901 * queues stopped. This allows the netdevice to perform queue management
902 * safely.
903 *
904 * Fiber Channel over Ethernet (FCoE) offload functions.
905 * int (*ndo_fcoe_enable)(struct net_device *dev);
906 * Called when the FCoE protocol stack wants to start using LLD for FCoE
907 * so the underlying device can perform whatever needed configuration or
908 * initialization to support acceleration of FCoE traffic.
909 *
910 * int (*ndo_fcoe_disable)(struct net_device *dev);
911 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
912 * so the underlying device can perform whatever needed clean-ups to
913 * stop supporting acceleration of FCoE traffic.
914 *
915 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
916 * struct scatterlist *sgl, unsigned int sgc);
917 * Called when the FCoE Initiator wants to initialize an I/O that
918 * is a possible candidate for Direct Data Placement (DDP). The LLD can
919 * perform necessary setup and returns 1 to indicate the device is set up
920 * successfully to perform DDP on this I/O, otherwise this returns 0.
921 *
922 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
923 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
924 * indicated by the FC exchange id 'xid', so the underlying device can
925 * clean up and reuse resources for later DDP requests.
926 *
927 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
928 * struct scatterlist *sgl, unsigned int sgc);
929 * Called when the FCoE Target wants to initialize an I/O that
930 * is a possible candidate for Direct Data Placement (DDP). The LLD can
931 * perform necessary setup and returns 1 to indicate the device is set up
932 * successfully to perform DDP on this I/O, otherwise this returns 0.
933 *
934 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
935 * struct netdev_fcoe_hbainfo *hbainfo);
936 * Called when the FCoE Protocol stack wants information on the underlying
937 * device. This information is utilized by the FCoE protocol stack to
938 * register attributes with Fiber Channel management service as per the
939 * FC-GS Fabric Device Management Information(FDMI) specification.
940 *
941 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
942 * Called when the underlying device wants to override default World Wide
943 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
944 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
945 * protocol stack to use.
946 *
947 * RFS acceleration.
948 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
949 * u16 rxq_index, u32 flow_id);
950 * Set hardware filter for RFS. rxq_index is the target queue index;
951 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
952 * Return the filter ID on success, or a negative error code.
953 *
954 * Slave management functions (for bridge, bonding, etc).
955 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
956 * Called to make another netdev an underling.
957 *
958 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
959 * Called to release previously enslaved netdev.
960 *
961 * Feature/offload setting functions.
962 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
963 * netdev_features_t features);
964 * Adjusts the requested feature flags according to device-specific
965 * constraints, and returns the resulting flags. Must not modify
966 * the device state.
967 *
968 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
969 * Called to update device configuration to new features. Passed
970 * feature set might be less than what was returned by ndo_fix_features()).
971 * Must return >0 or -errno if it changed dev->features itself.
972 *
973 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
974 * struct net_device *dev,
975 * const unsigned char *addr, u16 vid, u16 flags)
976 * Adds an FDB entry to dev for addr.
977 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
978 * struct net_device *dev,
979 * const unsigned char *addr, u16 vid)
980 * Deletes the FDB entry from dev coresponding to addr.
981 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
982 * struct net_device *dev, struct net_device *filter_dev,
983 * int idx)
984 * Used to add FDB entries to dump requests. Implementers should add
985 * entries to skb and update idx with the number of entries.
986 *
987 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
988 * u16 flags)
989 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
990 * struct net_device *dev, u32 filter_mask,
991 * int nlflags)
992 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
993 * u16 flags);
994 *
995 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
996 * Called to change device carrier. Soft-devices (like dummy, team, etc)
997 * which do not represent real hardware may define this to allow their
998 * userspace components to manage their virtual carrier state. Devices
999 * that determine carrier state from physical hardware properties (eg
1000 * network cables) or protocol-dependent mechanisms (eg
1001 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1002 *
1003 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1004 * struct netdev_phys_item_id *ppid);
1005 * Called to get ID of physical port of this device. If driver does
1006 * not implement this, it is assumed that the hw is not able to have
1007 * multiple net devices on single physical port.
1008 *
1009 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1010 * sa_family_t sa_family, __be16 port);
1011 * Called by vxlan to notiy a driver about the UDP port and socket
1012 * address family that vxlan is listnening to. It is called only when
1013 * a new port starts listening. The operation is protected by the
1014 * vxlan_net->sock_lock.
1015 *
1016 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1017 * sa_family_t sa_family, __be16 port);
1018 * Called by vxlan to notify the driver about a UDP port and socket
1019 * address family that vxlan is not listening to anymore. The operation
1020 * is protected by the vxlan_net->sock_lock.
1021 *
1022 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1023 * struct net_device *dev)
1024 * Called by upper layer devices to accelerate switching or other
1025 * station functionality into hardware. 'pdev is the lowerdev
1026 * to use for the offload and 'dev' is the net device that will
1027 * back the offload. Returns a pointer to the private structure
1028 * the upper layer will maintain.
1029 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1030 * Called by upper layer device to delete the station created
1031 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1032 * the station and priv is the structure returned by the add
1033 * operation.
1034 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1035 * struct net_device *dev,
1036 * void *priv);
1037 * Callback to use for xmit over the accelerated station. This
1038 * is used in place of ndo_start_xmit on accelerated net
1039 * devices.
1040 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1041 * struct net_device *dev
1042 * netdev_features_t features);
1043 * Called by core transmit path to determine if device is capable of
1044 * performing offload operations on a given packet. This is to give
1045 * the device an opportunity to implement any restrictions that cannot
1046 * be otherwise expressed by feature flags. The check is called with
1047 * the set of features that the stack has calculated and it returns
1048 * those the driver believes to be appropriate.
1049 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1050 * int queue_index, u32 maxrate);
1051 * Called when a user wants to set a max-rate limitation of specific
1052 * TX queue.
1053 * int (*ndo_get_iflink)(const struct net_device *dev);
1054 * Called to get the iflink value of this device.
1055 * void (*ndo_change_proto_down)(struct net_device *dev,
1056 * bool proto_down);
1057 * This function is used to pass protocol port error state information
1058 * to the switch driver. The switch driver can react to the proto_down
1059 * by doing a phys down on the associated switch port.
1060 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1061 * This function is used to get egress tunnel information for given skb.
1062 * This is useful for retrieving outer tunnel header parameters while
1063 * sampling packet.
1064 *
1065 */
1066 struct net_device_ops {
1067 int (*ndo_init)(struct net_device *dev);
1068 void (*ndo_uninit)(struct net_device *dev);
1069 int (*ndo_open)(struct net_device *dev);
1070 int (*ndo_stop)(struct net_device *dev);
1071 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1072 struct net_device *dev);
1073 u16 (*ndo_select_queue)(struct net_device *dev,
1074 struct sk_buff *skb,
1075 void *accel_priv,
1076 select_queue_fallback_t fallback);
1077 void (*ndo_change_rx_flags)(struct net_device *dev,
1078 int flags);
1079 void (*ndo_set_rx_mode)(struct net_device *dev);
1080 int (*ndo_set_mac_address)(struct net_device *dev,
1081 void *addr);
1082 int (*ndo_validate_addr)(struct net_device *dev);
1083 int (*ndo_do_ioctl)(struct net_device *dev,
1084 struct ifreq *ifr, int cmd);
1085 int (*ndo_set_config)(struct net_device *dev,
1086 struct ifmap *map);
1087 int (*ndo_change_mtu)(struct net_device *dev,
1088 int new_mtu);
1089 int (*ndo_neigh_setup)(struct net_device *dev,
1090 struct neigh_parms *);
1091 void (*ndo_tx_timeout) (struct net_device *dev);
1092
1093 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1094 struct rtnl_link_stats64 *storage);
1095 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1096
1097 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1098 __be16 proto, u16 vid);
1099 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1100 __be16 proto, u16 vid);
1101 #ifdef CONFIG_NET_POLL_CONTROLLER
1102 void (*ndo_poll_controller)(struct net_device *dev);
1103 int (*ndo_netpoll_setup)(struct net_device *dev,
1104 struct netpoll_info *info);
1105 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1106 #endif
1107 #ifdef CONFIG_NET_RX_BUSY_POLL
1108 int (*ndo_busy_poll)(struct napi_struct *dev);
1109 #endif
1110 int (*ndo_set_vf_mac)(struct net_device *dev,
1111 int queue, u8 *mac);
1112 int (*ndo_set_vf_vlan)(struct net_device *dev,
1113 int queue, u16 vlan, u8 qos);
1114 int (*ndo_set_vf_rate)(struct net_device *dev,
1115 int vf, int min_tx_rate,
1116 int max_tx_rate);
1117 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1118 int vf, bool setting);
1119 int (*ndo_set_vf_trust)(struct net_device *dev,
1120 int vf, bool setting);
1121 int (*ndo_get_vf_config)(struct net_device *dev,
1122 int vf,
1123 struct ifla_vf_info *ivf);
1124 int (*ndo_set_vf_link_state)(struct net_device *dev,
1125 int vf, int link_state);
1126 int (*ndo_get_vf_stats)(struct net_device *dev,
1127 int vf,
1128 struct ifla_vf_stats
1129 *vf_stats);
1130 int (*ndo_set_vf_port)(struct net_device *dev,
1131 int vf,
1132 struct nlattr *port[]);
1133 int (*ndo_get_vf_port)(struct net_device *dev,
1134 int vf, struct sk_buff *skb);
1135 int (*ndo_set_vf_rss_query_en)(
1136 struct net_device *dev,
1137 int vf, bool setting);
1138 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1139 #if IS_ENABLED(CONFIG_FCOE)
1140 int (*ndo_fcoe_enable)(struct net_device *dev);
1141 int (*ndo_fcoe_disable)(struct net_device *dev);
1142 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1143 u16 xid,
1144 struct scatterlist *sgl,
1145 unsigned int sgc);
1146 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1147 u16 xid);
1148 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1149 u16 xid,
1150 struct scatterlist *sgl,
1151 unsigned int sgc);
1152 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1153 struct netdev_fcoe_hbainfo *hbainfo);
1154 #endif
1155
1156 #if IS_ENABLED(CONFIG_LIBFCOE)
1157 #define NETDEV_FCOE_WWNN 0
1158 #define NETDEV_FCOE_WWPN 1
1159 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1160 u64 *wwn, int type);
1161 #endif
1162
1163 #ifdef CONFIG_RFS_ACCEL
1164 int (*ndo_rx_flow_steer)(struct net_device *dev,
1165 const struct sk_buff *skb,
1166 u16 rxq_index,
1167 u32 flow_id);
1168 #endif
1169 int (*ndo_add_slave)(struct net_device *dev,
1170 struct net_device *slave_dev);
1171 int (*ndo_del_slave)(struct net_device *dev,
1172 struct net_device *slave_dev);
1173 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1174 netdev_features_t features);
1175 int (*ndo_set_features)(struct net_device *dev,
1176 netdev_features_t features);
1177 int (*ndo_neigh_construct)(struct neighbour *n);
1178 void (*ndo_neigh_destroy)(struct neighbour *n);
1179
1180 int (*ndo_fdb_add)(struct ndmsg *ndm,
1181 struct nlattr *tb[],
1182 struct net_device *dev,
1183 const unsigned char *addr,
1184 u16 vid,
1185 u16 flags);
1186 int (*ndo_fdb_del)(struct ndmsg *ndm,
1187 struct nlattr *tb[],
1188 struct net_device *dev,
1189 const unsigned char *addr,
1190 u16 vid);
1191 int (*ndo_fdb_dump)(struct sk_buff *skb,
1192 struct netlink_callback *cb,
1193 struct net_device *dev,
1194 struct net_device *filter_dev,
1195 int idx);
1196
1197 int (*ndo_bridge_setlink)(struct net_device *dev,
1198 struct nlmsghdr *nlh,
1199 u16 flags);
1200 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1201 u32 pid, u32 seq,
1202 struct net_device *dev,
1203 u32 filter_mask,
1204 int nlflags);
1205 int (*ndo_bridge_dellink)(struct net_device *dev,
1206 struct nlmsghdr *nlh,
1207 u16 flags);
1208 int (*ndo_change_carrier)(struct net_device *dev,
1209 bool new_carrier);
1210 int (*ndo_get_phys_port_id)(struct net_device *dev,
1211 struct netdev_phys_item_id *ppid);
1212 int (*ndo_get_phys_port_name)(struct net_device *dev,
1213 char *name, size_t len);
1214 void (*ndo_add_vxlan_port)(struct net_device *dev,
1215 sa_family_t sa_family,
1216 __be16 port);
1217 void (*ndo_del_vxlan_port)(struct net_device *dev,
1218 sa_family_t sa_family,
1219 __be16 port);
1220
1221 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1222 struct net_device *dev);
1223 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1224 void *priv);
1225
1226 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1227 struct net_device *dev,
1228 void *priv);
1229 int (*ndo_get_lock_subclass)(struct net_device *dev);
1230 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1231 struct net_device *dev,
1232 netdev_features_t features);
1233 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1234 int queue_index,
1235 u32 maxrate);
1236 int (*ndo_get_iflink)(const struct net_device *dev);
1237 int (*ndo_change_proto_down)(struct net_device *dev,
1238 bool proto_down);
1239 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1240 struct sk_buff *skb);
1241 };
1242
1243 /**
1244 * enum net_device_priv_flags - &struct net_device priv_flags
1245 *
1246 * These are the &struct net_device, they are only set internally
1247 * by drivers and used in the kernel. These flags are invisible to
1248 * userspace, this means that the order of these flags can change
1249 * during any kernel release.
1250 *
1251 * You should have a pretty good reason to be extending these flags.
1252 *
1253 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1254 * @IFF_EBRIDGE: Ethernet bridging device
1255 * @IFF_BONDING: bonding master or slave
1256 * @IFF_ISATAP: ISATAP interface (RFC4214)
1257 * @IFF_WAN_HDLC: WAN HDLC device
1258 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1259 * release skb->dst
1260 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1261 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1262 * @IFF_MACVLAN_PORT: device used as macvlan port
1263 * @IFF_BRIDGE_PORT: device used as bridge port
1264 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1265 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1266 * @IFF_UNICAST_FLT: Supports unicast filtering
1267 * @IFF_TEAM_PORT: device used as team port
1268 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1269 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1270 * change when it's running
1271 * @IFF_MACVLAN: Macvlan device
1272 * @IFF_L3MDEV_MASTER: device is an L3 master device
1273 * @IFF_NO_QUEUE: device can run without qdisc attached
1274 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1275 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1276 * @IFF_TEAM: device is a team device
1277 */
1278 enum netdev_priv_flags {
1279 IFF_802_1Q_VLAN = 1<<0,
1280 IFF_EBRIDGE = 1<<1,
1281 IFF_BONDING = 1<<2,
1282 IFF_ISATAP = 1<<3,
1283 IFF_WAN_HDLC = 1<<4,
1284 IFF_XMIT_DST_RELEASE = 1<<5,
1285 IFF_DONT_BRIDGE = 1<<6,
1286 IFF_DISABLE_NETPOLL = 1<<7,
1287 IFF_MACVLAN_PORT = 1<<8,
1288 IFF_BRIDGE_PORT = 1<<9,
1289 IFF_OVS_DATAPATH = 1<<10,
1290 IFF_TX_SKB_SHARING = 1<<11,
1291 IFF_UNICAST_FLT = 1<<12,
1292 IFF_TEAM_PORT = 1<<13,
1293 IFF_SUPP_NOFCS = 1<<14,
1294 IFF_LIVE_ADDR_CHANGE = 1<<15,
1295 IFF_MACVLAN = 1<<16,
1296 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1297 IFF_IPVLAN_MASTER = 1<<18,
1298 IFF_IPVLAN_SLAVE = 1<<19,
1299 IFF_L3MDEV_MASTER = 1<<20,
1300 IFF_NO_QUEUE = 1<<21,
1301 IFF_OPENVSWITCH = 1<<22,
1302 IFF_L3MDEV_SLAVE = 1<<23,
1303 IFF_TEAM = 1<<24,
1304 };
1305
1306 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1307 #define IFF_EBRIDGE IFF_EBRIDGE
1308 #define IFF_BONDING IFF_BONDING
1309 #define IFF_ISATAP IFF_ISATAP
1310 #define IFF_WAN_HDLC IFF_WAN_HDLC
1311 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1312 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1313 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1314 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1315 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1316 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1317 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1318 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1319 #define IFF_TEAM_PORT IFF_TEAM_PORT
1320 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1321 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1322 #define IFF_MACVLAN IFF_MACVLAN
1323 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1324 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1325 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1326 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1327 #define IFF_NO_QUEUE IFF_NO_QUEUE
1328 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1329 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1330 #define IFF_TEAM IFF_TEAM
1331
1332 /**
1333 * struct net_device - The DEVICE structure.
1334 * Actually, this whole structure is a big mistake. It mixes I/O
1335 * data with strictly "high-level" data, and it has to know about
1336 * almost every data structure used in the INET module.
1337 *
1338 * @name: This is the first field of the "visible" part of this structure
1339 * (i.e. as seen by users in the "Space.c" file). It is the name
1340 * of the interface.
1341 *
1342 * @name_hlist: Device name hash chain, please keep it close to name[]
1343 * @ifalias: SNMP alias
1344 * @mem_end: Shared memory end
1345 * @mem_start: Shared memory start
1346 * @base_addr: Device I/O address
1347 * @irq: Device IRQ number
1348 *
1349 * @carrier_changes: Stats to monitor carrier on<->off transitions
1350 *
1351 * @state: Generic network queuing layer state, see netdev_state_t
1352 * @dev_list: The global list of network devices
1353 * @napi_list: List entry, that is used for polling napi devices
1354 * @unreg_list: List entry, that is used, when we are unregistering the
1355 * device, see the function unregister_netdev
1356 * @close_list: List entry, that is used, when we are closing the device
1357 *
1358 * @adj_list: Directly linked devices, like slaves for bonding
1359 * @all_adj_list: All linked devices, *including* neighbours
1360 * @features: Currently active device features
1361 * @hw_features: User-changeable features
1362 *
1363 * @wanted_features: User-requested features
1364 * @vlan_features: Mask of features inheritable by VLAN devices
1365 *
1366 * @hw_enc_features: Mask of features inherited by encapsulating devices
1367 * This field indicates what encapsulation
1368 * offloads the hardware is capable of doing,
1369 * and drivers will need to set them appropriately.
1370 *
1371 * @mpls_features: Mask of features inheritable by MPLS
1372 *
1373 * @ifindex: interface index
1374 * @group: The group, that the device belongs to
1375 *
1376 * @stats: Statistics struct, which was left as a legacy, use
1377 * rtnl_link_stats64 instead
1378 *
1379 * @rx_dropped: Dropped packets by core network,
1380 * do not use this in drivers
1381 * @tx_dropped: Dropped packets by core network,
1382 * do not use this in drivers
1383 *
1384 * @wireless_handlers: List of functions to handle Wireless Extensions,
1385 * instead of ioctl,
1386 * see <net/iw_handler.h> for details.
1387 * @wireless_data: Instance data managed by the core of wireless extensions
1388 *
1389 * @netdev_ops: Includes several pointers to callbacks,
1390 * if one wants to override the ndo_*() functions
1391 * @ethtool_ops: Management operations
1392 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1393 * of Layer 2 headers.
1394 *
1395 * @flags: Interface flags (a la BSD)
1396 * @priv_flags: Like 'flags' but invisible to userspace,
1397 * see if.h for the definitions
1398 * @gflags: Global flags ( kept as legacy )
1399 * @padded: How much padding added by alloc_netdev()
1400 * @operstate: RFC2863 operstate
1401 * @link_mode: Mapping policy to operstate
1402 * @if_port: Selectable AUI, TP, ...
1403 * @dma: DMA channel
1404 * @mtu: Interface MTU value
1405 * @type: Interface hardware type
1406 * @hard_header_len: Hardware header length
1407 *
1408 * @needed_headroom: Extra headroom the hardware may need, but not in all
1409 * cases can this be guaranteed
1410 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1411 * cases can this be guaranteed. Some cases also use
1412 * LL_MAX_HEADER instead to allocate the skb
1413 *
1414 * interface address info:
1415 *
1416 * @perm_addr: Permanent hw address
1417 * @addr_assign_type: Hw address assignment type
1418 * @addr_len: Hardware address length
1419 * @neigh_priv_len; Used in neigh_alloc(),
1420 * initialized only in atm/clip.c
1421 * @dev_id: Used to differentiate devices that share
1422 * the same link layer address
1423 * @dev_port: Used to differentiate devices that share
1424 * the same function
1425 * @addr_list_lock: XXX: need comments on this one
1426 * @uc_promisc: Counter, that indicates, that promiscuous mode
1427 * has been enabled due to the need to listen to
1428 * additional unicast addresses in a device that
1429 * does not implement ndo_set_rx_mode()
1430 * @uc: unicast mac addresses
1431 * @mc: multicast mac addresses
1432 * @dev_addrs: list of device hw addresses
1433 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1434 * @promiscuity: Number of times, the NIC is told to work in
1435 * Promiscuous mode, if it becomes 0 the NIC will
1436 * exit from working in Promiscuous mode
1437 * @allmulti: Counter, enables or disables allmulticast mode
1438 *
1439 * @vlan_info: VLAN info
1440 * @dsa_ptr: dsa specific data
1441 * @tipc_ptr: TIPC specific data
1442 * @atalk_ptr: AppleTalk link
1443 * @ip_ptr: IPv4 specific data
1444 * @dn_ptr: DECnet specific data
1445 * @ip6_ptr: IPv6 specific data
1446 * @ax25_ptr: AX.25 specific data
1447 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1448 *
1449 * @last_rx: Time of last Rx
1450 * @dev_addr: Hw address (before bcast,
1451 * because most packets are unicast)
1452 *
1453 * @_rx: Array of RX queues
1454 * @num_rx_queues: Number of RX queues
1455 * allocated at register_netdev() time
1456 * @real_num_rx_queues: Number of RX queues currently active in device
1457 *
1458 * @rx_handler: handler for received packets
1459 * @rx_handler_data: XXX: need comments on this one
1460 * @ingress_queue: XXX: need comments on this one
1461 * @broadcast: hw bcast address
1462 *
1463 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1464 * indexed by RX queue number. Assigned by driver.
1465 * This must only be set if the ndo_rx_flow_steer
1466 * operation is defined
1467 * @index_hlist: Device index hash chain
1468 *
1469 * @_tx: Array of TX queues
1470 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1471 * @real_num_tx_queues: Number of TX queues currently active in device
1472 * @qdisc: Root qdisc from userspace point of view
1473 * @tx_queue_len: Max frames per queue allowed
1474 * @tx_global_lock: XXX: need comments on this one
1475 *
1476 * @xps_maps: XXX: need comments on this one
1477 *
1478 * @offload_fwd_mark: Offload device fwding mark
1479 *
1480 * @trans_start: Time (in jiffies) of last Tx
1481 * @watchdog_timeo: Represents the timeout that is used by
1482 * the watchdog ( see dev_watchdog() )
1483 * @watchdog_timer: List of timers
1484 *
1485 * @pcpu_refcnt: Number of references to this device
1486 * @todo_list: Delayed register/unregister
1487 * @link_watch_list: XXX: need comments on this one
1488 *
1489 * @reg_state: Register/unregister state machine
1490 * @dismantle: Device is going to be freed
1491 * @rtnl_link_state: This enum represents the phases of creating
1492 * a new link
1493 *
1494 * @destructor: Called from unregister,
1495 * can be used to call free_netdev
1496 * @npinfo: XXX: need comments on this one
1497 * @nd_net: Network namespace this network device is inside
1498 *
1499 * @ml_priv: Mid-layer private
1500 * @lstats: Loopback statistics
1501 * @tstats: Tunnel statistics
1502 * @dstats: Dummy statistics
1503 * @vstats: Virtual ethernet statistics
1504 *
1505 * @garp_port: GARP
1506 * @mrp_port: MRP
1507 *
1508 * @dev: Class/net/name entry
1509 * @sysfs_groups: Space for optional device, statistics and wireless
1510 * sysfs groups
1511 *
1512 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1513 * @rtnl_link_ops: Rtnl_link_ops
1514 *
1515 * @gso_max_size: Maximum size of generic segmentation offload
1516 * @gso_max_segs: Maximum number of segments that can be passed to the
1517 * NIC for GSO
1518 * @gso_min_segs: Minimum number of segments that can be passed to the
1519 * NIC for GSO
1520 *
1521 * @dcbnl_ops: Data Center Bridging netlink ops
1522 * @num_tc: Number of traffic classes in the net device
1523 * @tc_to_txq: XXX: need comments on this one
1524 * @prio_tc_map XXX: need comments on this one
1525 *
1526 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1527 *
1528 * @priomap: XXX: need comments on this one
1529 * @phydev: Physical device may attach itself
1530 * for hardware timestamping
1531 *
1532 * @qdisc_tx_busylock: XXX: need comments on this one
1533 *
1534 * @proto_down: protocol port state information can be sent to the
1535 * switch driver and used to set the phys state of the
1536 * switch port.
1537 *
1538 * FIXME: cleanup struct net_device such that network protocol info
1539 * moves out.
1540 */
1541
1542 struct net_device {
1543 char name[IFNAMSIZ];
1544 struct hlist_node name_hlist;
1545 char *ifalias;
1546 /*
1547 * I/O specific fields
1548 * FIXME: Merge these and struct ifmap into one
1549 */
1550 unsigned long mem_end;
1551 unsigned long mem_start;
1552 unsigned long base_addr;
1553 int irq;
1554
1555 atomic_t carrier_changes;
1556
1557 /*
1558 * Some hardware also needs these fields (state,dev_list,
1559 * napi_list,unreg_list,close_list) but they are not
1560 * part of the usual set specified in Space.c.
1561 */
1562
1563 unsigned long state;
1564
1565 struct list_head dev_list;
1566 struct list_head napi_list;
1567 struct list_head unreg_list;
1568 struct list_head close_list;
1569 struct list_head ptype_all;
1570 struct list_head ptype_specific;
1571
1572 struct {
1573 struct list_head upper;
1574 struct list_head lower;
1575 } adj_list;
1576
1577 struct {
1578 struct list_head upper;
1579 struct list_head lower;
1580 } all_adj_list;
1581
1582 netdev_features_t features;
1583 netdev_features_t hw_features;
1584 netdev_features_t wanted_features;
1585 netdev_features_t vlan_features;
1586 netdev_features_t hw_enc_features;
1587 netdev_features_t mpls_features;
1588
1589 int ifindex;
1590 int group;
1591
1592 struct net_device_stats stats;
1593
1594 atomic_long_t rx_dropped;
1595 atomic_long_t tx_dropped;
1596
1597 #ifdef CONFIG_WIRELESS_EXT
1598 const struct iw_handler_def * wireless_handlers;
1599 struct iw_public_data * wireless_data;
1600 #endif
1601 const struct net_device_ops *netdev_ops;
1602 const struct ethtool_ops *ethtool_ops;
1603 #ifdef CONFIG_NET_SWITCHDEV
1604 const struct switchdev_ops *switchdev_ops;
1605 #endif
1606 #ifdef CONFIG_NET_L3_MASTER_DEV
1607 const struct l3mdev_ops *l3mdev_ops;
1608 #endif
1609
1610 const struct header_ops *header_ops;
1611
1612 unsigned int flags;
1613 unsigned int priv_flags;
1614
1615 unsigned short gflags;
1616 unsigned short padded;
1617
1618 unsigned char operstate;
1619 unsigned char link_mode;
1620
1621 unsigned char if_port;
1622 unsigned char dma;
1623
1624 unsigned int mtu;
1625 unsigned short type;
1626 unsigned short hard_header_len;
1627
1628 unsigned short needed_headroom;
1629 unsigned short needed_tailroom;
1630
1631 /* Interface address info. */
1632 unsigned char perm_addr[MAX_ADDR_LEN];
1633 unsigned char addr_assign_type;
1634 unsigned char addr_len;
1635 unsigned short neigh_priv_len;
1636 unsigned short dev_id;
1637 unsigned short dev_port;
1638 spinlock_t addr_list_lock;
1639 unsigned char name_assign_type;
1640 bool uc_promisc;
1641 struct netdev_hw_addr_list uc;
1642 struct netdev_hw_addr_list mc;
1643 struct netdev_hw_addr_list dev_addrs;
1644
1645 #ifdef CONFIG_SYSFS
1646 struct kset *queues_kset;
1647 #endif
1648 unsigned int promiscuity;
1649 unsigned int allmulti;
1650
1651
1652 /* Protocol specific pointers */
1653
1654 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1655 struct vlan_info __rcu *vlan_info;
1656 #endif
1657 #if IS_ENABLED(CONFIG_NET_DSA)
1658 struct dsa_switch_tree *dsa_ptr;
1659 #endif
1660 #if IS_ENABLED(CONFIG_TIPC)
1661 struct tipc_bearer __rcu *tipc_ptr;
1662 #endif
1663 void *atalk_ptr;
1664 struct in_device __rcu *ip_ptr;
1665 struct dn_dev __rcu *dn_ptr;
1666 struct inet6_dev __rcu *ip6_ptr;
1667 void *ax25_ptr;
1668 struct wireless_dev *ieee80211_ptr;
1669 struct wpan_dev *ieee802154_ptr;
1670 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1671 struct mpls_dev __rcu *mpls_ptr;
1672 #endif
1673
1674 /*
1675 * Cache lines mostly used on receive path (including eth_type_trans())
1676 */
1677 unsigned long last_rx;
1678
1679 /* Interface address info used in eth_type_trans() */
1680 unsigned char *dev_addr;
1681
1682
1683 #ifdef CONFIG_SYSFS
1684 struct netdev_rx_queue *_rx;
1685
1686 unsigned int num_rx_queues;
1687 unsigned int real_num_rx_queues;
1688
1689 #endif
1690
1691 unsigned long gro_flush_timeout;
1692 rx_handler_func_t __rcu *rx_handler;
1693 void __rcu *rx_handler_data;
1694
1695 #ifdef CONFIG_NET_CLS_ACT
1696 struct tcf_proto __rcu *ingress_cl_list;
1697 #endif
1698 struct netdev_queue __rcu *ingress_queue;
1699 #ifdef CONFIG_NETFILTER_INGRESS
1700 struct list_head nf_hooks_ingress;
1701 #endif
1702
1703 unsigned char broadcast[MAX_ADDR_LEN];
1704 #ifdef CONFIG_RFS_ACCEL
1705 struct cpu_rmap *rx_cpu_rmap;
1706 #endif
1707 struct hlist_node index_hlist;
1708
1709 /*
1710 * Cache lines mostly used on transmit path
1711 */
1712 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1713 unsigned int num_tx_queues;
1714 unsigned int real_num_tx_queues;
1715 struct Qdisc *qdisc;
1716 unsigned long tx_queue_len;
1717 spinlock_t tx_global_lock;
1718 int watchdog_timeo;
1719
1720 #ifdef CONFIG_XPS
1721 struct xps_dev_maps __rcu *xps_maps;
1722 #endif
1723
1724 #ifdef CONFIG_NET_SWITCHDEV
1725 u32 offload_fwd_mark;
1726 #endif
1727
1728 /* These may be needed for future network-power-down code. */
1729
1730 /*
1731 * trans_start here is expensive for high speed devices on SMP,
1732 * please use netdev_queue->trans_start instead.
1733 */
1734 unsigned long trans_start;
1735
1736 struct timer_list watchdog_timer;
1737
1738 int __percpu *pcpu_refcnt;
1739 struct list_head todo_list;
1740
1741 struct list_head link_watch_list;
1742
1743 enum { NETREG_UNINITIALIZED=0,
1744 NETREG_REGISTERED, /* completed register_netdevice */
1745 NETREG_UNREGISTERING, /* called unregister_netdevice */
1746 NETREG_UNREGISTERED, /* completed unregister todo */
1747 NETREG_RELEASED, /* called free_netdev */
1748 NETREG_DUMMY, /* dummy device for NAPI poll */
1749 } reg_state:8;
1750
1751 bool dismantle;
1752
1753 enum {
1754 RTNL_LINK_INITIALIZED,
1755 RTNL_LINK_INITIALIZING,
1756 } rtnl_link_state:16;
1757
1758 void (*destructor)(struct net_device *dev);
1759
1760 #ifdef CONFIG_NETPOLL
1761 struct netpoll_info __rcu *npinfo;
1762 #endif
1763
1764 possible_net_t nd_net;
1765
1766 /* mid-layer private */
1767 union {
1768 void *ml_priv;
1769 struct pcpu_lstats __percpu *lstats;
1770 struct pcpu_sw_netstats __percpu *tstats;
1771 struct pcpu_dstats __percpu *dstats;
1772 struct pcpu_vstats __percpu *vstats;
1773 };
1774
1775 struct garp_port __rcu *garp_port;
1776 struct mrp_port __rcu *mrp_port;
1777
1778 struct device dev;
1779 const struct attribute_group *sysfs_groups[4];
1780 const struct attribute_group *sysfs_rx_queue_group;
1781
1782 const struct rtnl_link_ops *rtnl_link_ops;
1783
1784 /* for setting kernel sock attribute on TCP connection setup */
1785 #define GSO_MAX_SIZE 65536
1786 unsigned int gso_max_size;
1787 #define GSO_MAX_SEGS 65535
1788 u16 gso_max_segs;
1789 u16 gso_min_segs;
1790 #ifdef CONFIG_DCB
1791 const struct dcbnl_rtnl_ops *dcbnl_ops;
1792 #endif
1793 u8 num_tc;
1794 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1795 u8 prio_tc_map[TC_BITMASK + 1];
1796
1797 #if IS_ENABLED(CONFIG_FCOE)
1798 unsigned int fcoe_ddp_xid;
1799 #endif
1800 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1801 struct netprio_map __rcu *priomap;
1802 #endif
1803 struct phy_device *phydev;
1804 struct lock_class_key *qdisc_tx_busylock;
1805 bool proto_down;
1806 };
1807 #define to_net_dev(d) container_of(d, struct net_device, dev)
1808
1809 #define NETDEV_ALIGN 32
1810
1811 static inline
1812 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1813 {
1814 return dev->prio_tc_map[prio & TC_BITMASK];
1815 }
1816
1817 static inline
1818 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1819 {
1820 if (tc >= dev->num_tc)
1821 return -EINVAL;
1822
1823 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1824 return 0;
1825 }
1826
1827 static inline
1828 void netdev_reset_tc(struct net_device *dev)
1829 {
1830 dev->num_tc = 0;
1831 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1832 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1833 }
1834
1835 static inline
1836 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1837 {
1838 if (tc >= dev->num_tc)
1839 return -EINVAL;
1840
1841 dev->tc_to_txq[tc].count = count;
1842 dev->tc_to_txq[tc].offset = offset;
1843 return 0;
1844 }
1845
1846 static inline
1847 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1848 {
1849 if (num_tc > TC_MAX_QUEUE)
1850 return -EINVAL;
1851
1852 dev->num_tc = num_tc;
1853 return 0;
1854 }
1855
1856 static inline
1857 int netdev_get_num_tc(struct net_device *dev)
1858 {
1859 return dev->num_tc;
1860 }
1861
1862 static inline
1863 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1864 unsigned int index)
1865 {
1866 return &dev->_tx[index];
1867 }
1868
1869 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1870 const struct sk_buff *skb)
1871 {
1872 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1873 }
1874
1875 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1876 void (*f)(struct net_device *,
1877 struct netdev_queue *,
1878 void *),
1879 void *arg)
1880 {
1881 unsigned int i;
1882
1883 for (i = 0; i < dev->num_tx_queues; i++)
1884 f(dev, &dev->_tx[i], arg);
1885 }
1886
1887 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1888 struct sk_buff *skb,
1889 void *accel_priv);
1890
1891 /*
1892 * Net namespace inlines
1893 */
1894 static inline
1895 struct net *dev_net(const struct net_device *dev)
1896 {
1897 return read_pnet(&dev->nd_net);
1898 }
1899
1900 static inline
1901 void dev_net_set(struct net_device *dev, struct net *net)
1902 {
1903 write_pnet(&dev->nd_net, net);
1904 }
1905
1906 static inline bool netdev_uses_dsa(struct net_device *dev)
1907 {
1908 #if IS_ENABLED(CONFIG_NET_DSA)
1909 if (dev->dsa_ptr != NULL)
1910 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1911 #endif
1912 return false;
1913 }
1914
1915 /**
1916 * netdev_priv - access network device private data
1917 * @dev: network device
1918 *
1919 * Get network device private data
1920 */
1921 static inline void *netdev_priv(const struct net_device *dev)
1922 {
1923 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1924 }
1925
1926 /* Set the sysfs physical device reference for the network logical device
1927 * if set prior to registration will cause a symlink during initialization.
1928 */
1929 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1930
1931 /* Set the sysfs device type for the network logical device to allow
1932 * fine-grained identification of different network device types. For
1933 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1934 */
1935 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1936
1937 /* Default NAPI poll() weight
1938 * Device drivers are strongly advised to not use bigger value
1939 */
1940 #define NAPI_POLL_WEIGHT 64
1941
1942 /**
1943 * netif_napi_add - initialize a napi context
1944 * @dev: network device
1945 * @napi: napi context
1946 * @poll: polling function
1947 * @weight: default weight
1948 *
1949 * netif_napi_add() must be used to initialize a napi context prior to calling
1950 * *any* of the other napi related functions.
1951 */
1952 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1953 int (*poll)(struct napi_struct *, int), int weight);
1954
1955 /**
1956 * netif_tx_napi_add - initialize a napi context
1957 * @dev: network device
1958 * @napi: napi context
1959 * @poll: polling function
1960 * @weight: default weight
1961 *
1962 * This variant of netif_napi_add() should be used from drivers using NAPI
1963 * to exclusively poll a TX queue.
1964 * This will avoid we add it into napi_hash[], thus polluting this hash table.
1965 */
1966 static inline void netif_tx_napi_add(struct net_device *dev,
1967 struct napi_struct *napi,
1968 int (*poll)(struct napi_struct *, int),
1969 int weight)
1970 {
1971 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
1972 netif_napi_add(dev, napi, poll, weight);
1973 }
1974
1975 /**
1976 * netif_napi_del - remove a napi context
1977 * @napi: napi context
1978 *
1979 * netif_napi_del() removes a napi context from the network device napi list
1980 */
1981 void netif_napi_del(struct napi_struct *napi);
1982
1983 struct napi_gro_cb {
1984 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1985 void *frag0;
1986
1987 /* Length of frag0. */
1988 unsigned int frag0_len;
1989
1990 /* This indicates where we are processing relative to skb->data. */
1991 int data_offset;
1992
1993 /* This is non-zero if the packet cannot be merged with the new skb. */
1994 u16 flush;
1995
1996 /* Save the IP ID here and check when we get to the transport layer */
1997 u16 flush_id;
1998
1999 /* Number of segments aggregated. */
2000 u16 count;
2001
2002 /* Start offset for remote checksum offload */
2003 u16 gro_remcsum_start;
2004
2005 /* jiffies when first packet was created/queued */
2006 unsigned long age;
2007
2008 /* Used in ipv6_gro_receive() and foo-over-udp */
2009 u16 proto;
2010
2011 /* This is non-zero if the packet may be of the same flow. */
2012 u8 same_flow:1;
2013
2014 /* Used in udp_gro_receive */
2015 u8 udp_mark:1;
2016
2017 /* GRO checksum is valid */
2018 u8 csum_valid:1;
2019
2020 /* Number of checksums via CHECKSUM_UNNECESSARY */
2021 u8 csum_cnt:3;
2022
2023 /* Free the skb? */
2024 u8 free:2;
2025 #define NAPI_GRO_FREE 1
2026 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2027
2028 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2029 u8 is_ipv6:1;
2030
2031 /* 7 bit hole */
2032
2033 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2034 __wsum csum;
2035
2036 /* used in skb_gro_receive() slow path */
2037 struct sk_buff *last;
2038 };
2039
2040 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2041
2042 struct packet_type {
2043 __be16 type; /* This is really htons(ether_type). */
2044 struct net_device *dev; /* NULL is wildcarded here */
2045 int (*func) (struct sk_buff *,
2046 struct net_device *,
2047 struct packet_type *,
2048 struct net_device *);
2049 bool (*id_match)(struct packet_type *ptype,
2050 struct sock *sk);
2051 void *af_packet_priv;
2052 struct list_head list;
2053 };
2054
2055 struct offload_callbacks {
2056 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2057 netdev_features_t features);
2058 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2059 struct sk_buff *skb);
2060 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2061 };
2062
2063 struct packet_offload {
2064 __be16 type; /* This is really htons(ether_type). */
2065 u16 priority;
2066 struct offload_callbacks callbacks;
2067 struct list_head list;
2068 };
2069
2070 struct udp_offload;
2071
2072 struct udp_offload_callbacks {
2073 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2074 struct sk_buff *skb,
2075 struct udp_offload *uoff);
2076 int (*gro_complete)(struct sk_buff *skb,
2077 int nhoff,
2078 struct udp_offload *uoff);
2079 };
2080
2081 struct udp_offload {
2082 __be16 port;
2083 u8 ipproto;
2084 struct udp_offload_callbacks callbacks;
2085 };
2086
2087 /* often modified stats are per cpu, other are shared (netdev->stats) */
2088 struct pcpu_sw_netstats {
2089 u64 rx_packets;
2090 u64 rx_bytes;
2091 u64 tx_packets;
2092 u64 tx_bytes;
2093 struct u64_stats_sync syncp;
2094 };
2095
2096 #define __netdev_alloc_pcpu_stats(type, gfp) \
2097 ({ \
2098 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2099 if (pcpu_stats) { \
2100 int __cpu; \
2101 for_each_possible_cpu(__cpu) { \
2102 typeof(type) *stat; \
2103 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2104 u64_stats_init(&stat->syncp); \
2105 } \
2106 } \
2107 pcpu_stats; \
2108 })
2109
2110 #define netdev_alloc_pcpu_stats(type) \
2111 __netdev_alloc_pcpu_stats(type, GFP_KERNEL);
2112
2113 enum netdev_lag_tx_type {
2114 NETDEV_LAG_TX_TYPE_UNKNOWN,
2115 NETDEV_LAG_TX_TYPE_RANDOM,
2116 NETDEV_LAG_TX_TYPE_BROADCAST,
2117 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2118 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2119 NETDEV_LAG_TX_TYPE_HASH,
2120 };
2121
2122 struct netdev_lag_upper_info {
2123 enum netdev_lag_tx_type tx_type;
2124 };
2125
2126 struct netdev_lag_lower_state_info {
2127 u8 link_up : 1,
2128 tx_enabled : 1;
2129 };
2130
2131 #include <linux/notifier.h>
2132
2133 /* netdevice notifier chain. Please remember to update the rtnetlink
2134 * notification exclusion list in rtnetlink_event() when adding new
2135 * types.
2136 */
2137 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2138 #define NETDEV_DOWN 0x0002
2139 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2140 detected a hardware crash and restarted
2141 - we can use this eg to kick tcp sessions
2142 once done */
2143 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2144 #define NETDEV_REGISTER 0x0005
2145 #define NETDEV_UNREGISTER 0x0006
2146 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2147 #define NETDEV_CHANGEADDR 0x0008
2148 #define NETDEV_GOING_DOWN 0x0009
2149 #define NETDEV_CHANGENAME 0x000A
2150 #define NETDEV_FEAT_CHANGE 0x000B
2151 #define NETDEV_BONDING_FAILOVER 0x000C
2152 #define NETDEV_PRE_UP 0x000D
2153 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2154 #define NETDEV_POST_TYPE_CHANGE 0x000F
2155 #define NETDEV_POST_INIT 0x0010
2156 #define NETDEV_UNREGISTER_FINAL 0x0011
2157 #define NETDEV_RELEASE 0x0012
2158 #define NETDEV_NOTIFY_PEERS 0x0013
2159 #define NETDEV_JOIN 0x0014
2160 #define NETDEV_CHANGEUPPER 0x0015
2161 #define NETDEV_RESEND_IGMP 0x0016
2162 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2163 #define NETDEV_CHANGEINFODATA 0x0018
2164 #define NETDEV_BONDING_INFO 0x0019
2165 #define NETDEV_PRECHANGEUPPER 0x001A
2166 #define NETDEV_CHANGELOWERSTATE 0x001B
2167
2168 int register_netdevice_notifier(struct notifier_block *nb);
2169 int unregister_netdevice_notifier(struct notifier_block *nb);
2170
2171 struct netdev_notifier_info {
2172 struct net_device *dev;
2173 };
2174
2175 struct netdev_notifier_change_info {
2176 struct netdev_notifier_info info; /* must be first */
2177 unsigned int flags_changed;
2178 };
2179
2180 struct netdev_notifier_changeupper_info {
2181 struct netdev_notifier_info info; /* must be first */
2182 struct net_device *upper_dev; /* new upper dev */
2183 bool master; /* is upper dev master */
2184 bool linking; /* is the nofication for link or unlink */
2185 void *upper_info; /* upper dev info */
2186 };
2187
2188 struct netdev_notifier_changelowerstate_info {
2189 struct netdev_notifier_info info; /* must be first */
2190 void *lower_state_info; /* is lower dev state */
2191 };
2192
2193 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2194 struct net_device *dev)
2195 {
2196 info->dev = dev;
2197 }
2198
2199 static inline struct net_device *
2200 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2201 {
2202 return info->dev;
2203 }
2204
2205 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2206
2207
2208 extern rwlock_t dev_base_lock; /* Device list lock */
2209
2210 #define for_each_netdev(net, d) \
2211 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2212 #define for_each_netdev_reverse(net, d) \
2213 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2214 #define for_each_netdev_rcu(net, d) \
2215 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2216 #define for_each_netdev_safe(net, d, n) \
2217 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2218 #define for_each_netdev_continue(net, d) \
2219 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2220 #define for_each_netdev_continue_rcu(net, d) \
2221 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2222 #define for_each_netdev_in_bond_rcu(bond, slave) \
2223 for_each_netdev_rcu(&init_net, slave) \
2224 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2225 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2226
2227 static inline struct net_device *next_net_device(struct net_device *dev)
2228 {
2229 struct list_head *lh;
2230 struct net *net;
2231
2232 net = dev_net(dev);
2233 lh = dev->dev_list.next;
2234 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2235 }
2236
2237 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2238 {
2239 struct list_head *lh;
2240 struct net *net;
2241
2242 net = dev_net(dev);
2243 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2244 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2245 }
2246
2247 static inline struct net_device *first_net_device(struct net *net)
2248 {
2249 return list_empty(&net->dev_base_head) ? NULL :
2250 net_device_entry(net->dev_base_head.next);
2251 }
2252
2253 static inline struct net_device *first_net_device_rcu(struct net *net)
2254 {
2255 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2256
2257 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2258 }
2259
2260 int netdev_boot_setup_check(struct net_device *dev);
2261 unsigned long netdev_boot_base(const char *prefix, int unit);
2262 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2263 const char *hwaddr);
2264 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2265 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2266 void dev_add_pack(struct packet_type *pt);
2267 void dev_remove_pack(struct packet_type *pt);
2268 void __dev_remove_pack(struct packet_type *pt);
2269 void dev_add_offload(struct packet_offload *po);
2270 void dev_remove_offload(struct packet_offload *po);
2271
2272 int dev_get_iflink(const struct net_device *dev);
2273 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2274 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2275 unsigned short mask);
2276 struct net_device *dev_get_by_name(struct net *net, const char *name);
2277 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2278 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2279 int dev_alloc_name(struct net_device *dev, const char *name);
2280 int dev_open(struct net_device *dev);
2281 int dev_close(struct net_device *dev);
2282 int dev_close_many(struct list_head *head, bool unlink);
2283 void dev_disable_lro(struct net_device *dev);
2284 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2285 int dev_queue_xmit(struct sk_buff *skb);
2286 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2287 int register_netdevice(struct net_device *dev);
2288 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2289 void unregister_netdevice_many(struct list_head *head);
2290 static inline void unregister_netdevice(struct net_device *dev)
2291 {
2292 unregister_netdevice_queue(dev, NULL);
2293 }
2294
2295 int netdev_refcnt_read(const struct net_device *dev);
2296 void free_netdev(struct net_device *dev);
2297 void netdev_freemem(struct net_device *dev);
2298 void synchronize_net(void);
2299 int init_dummy_netdev(struct net_device *dev);
2300
2301 DECLARE_PER_CPU(int, xmit_recursion);
2302 static inline int dev_recursion_level(void)
2303 {
2304 return this_cpu_read(xmit_recursion);
2305 }
2306
2307 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2308 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2309 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2310 int netdev_get_name(struct net *net, char *name, int ifindex);
2311 int dev_restart(struct net_device *dev);
2312 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2313
2314 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2315 {
2316 return NAPI_GRO_CB(skb)->data_offset;
2317 }
2318
2319 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2320 {
2321 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2322 }
2323
2324 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2325 {
2326 NAPI_GRO_CB(skb)->data_offset += len;
2327 }
2328
2329 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2330 unsigned int offset)
2331 {
2332 return NAPI_GRO_CB(skb)->frag0 + offset;
2333 }
2334
2335 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2336 {
2337 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2338 }
2339
2340 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2341 unsigned int offset)
2342 {
2343 if (!pskb_may_pull(skb, hlen))
2344 return NULL;
2345
2346 NAPI_GRO_CB(skb)->frag0 = NULL;
2347 NAPI_GRO_CB(skb)->frag0_len = 0;
2348 return skb->data + offset;
2349 }
2350
2351 static inline void *skb_gro_network_header(struct sk_buff *skb)
2352 {
2353 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2354 skb_network_offset(skb);
2355 }
2356
2357 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2358 const void *start, unsigned int len)
2359 {
2360 if (NAPI_GRO_CB(skb)->csum_valid)
2361 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2362 csum_partial(start, len, 0));
2363 }
2364
2365 /* GRO checksum functions. These are logical equivalents of the normal
2366 * checksum functions (in skbuff.h) except that they operate on the GRO
2367 * offsets and fields in sk_buff.
2368 */
2369
2370 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2371
2372 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2373 {
2374 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2375 }
2376
2377 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2378 bool zero_okay,
2379 __sum16 check)
2380 {
2381 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2382 skb_checksum_start_offset(skb) <
2383 skb_gro_offset(skb)) &&
2384 !skb_at_gro_remcsum_start(skb) &&
2385 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2386 (!zero_okay || check));
2387 }
2388
2389 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2390 __wsum psum)
2391 {
2392 if (NAPI_GRO_CB(skb)->csum_valid &&
2393 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2394 return 0;
2395
2396 NAPI_GRO_CB(skb)->csum = psum;
2397
2398 return __skb_gro_checksum_complete(skb);
2399 }
2400
2401 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2402 {
2403 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2404 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2405 NAPI_GRO_CB(skb)->csum_cnt--;
2406 } else {
2407 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2408 * verified a new top level checksum or an encapsulated one
2409 * during GRO. This saves work if we fallback to normal path.
2410 */
2411 __skb_incr_checksum_unnecessary(skb);
2412 }
2413 }
2414
2415 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2416 compute_pseudo) \
2417 ({ \
2418 __sum16 __ret = 0; \
2419 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2420 __ret = __skb_gro_checksum_validate_complete(skb, \
2421 compute_pseudo(skb, proto)); \
2422 if (__ret) \
2423 __skb_mark_checksum_bad(skb); \
2424 else \
2425 skb_gro_incr_csum_unnecessary(skb); \
2426 __ret; \
2427 })
2428
2429 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2430 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2431
2432 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2433 compute_pseudo) \
2434 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2435
2436 #define skb_gro_checksum_simple_validate(skb) \
2437 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2438
2439 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2440 {
2441 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2442 !NAPI_GRO_CB(skb)->csum_valid);
2443 }
2444
2445 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2446 __sum16 check, __wsum pseudo)
2447 {
2448 NAPI_GRO_CB(skb)->csum = ~pseudo;
2449 NAPI_GRO_CB(skb)->csum_valid = 1;
2450 }
2451
2452 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2453 do { \
2454 if (__skb_gro_checksum_convert_check(skb)) \
2455 __skb_gro_checksum_convert(skb, check, \
2456 compute_pseudo(skb, proto)); \
2457 } while (0)
2458
2459 struct gro_remcsum {
2460 int offset;
2461 __wsum delta;
2462 };
2463
2464 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2465 {
2466 grc->offset = 0;
2467 grc->delta = 0;
2468 }
2469
2470 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2471 unsigned int off, size_t hdrlen,
2472 int start, int offset,
2473 struct gro_remcsum *grc,
2474 bool nopartial)
2475 {
2476 __wsum delta;
2477 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2478
2479 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2480
2481 if (!nopartial) {
2482 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2483 return ptr;
2484 }
2485
2486 ptr = skb_gro_header_fast(skb, off);
2487 if (skb_gro_header_hard(skb, off + plen)) {
2488 ptr = skb_gro_header_slow(skb, off + plen, off);
2489 if (!ptr)
2490 return NULL;
2491 }
2492
2493 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2494 start, offset);
2495
2496 /* Adjust skb->csum since we changed the packet */
2497 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2498
2499 grc->offset = off + hdrlen + offset;
2500 grc->delta = delta;
2501
2502 return ptr;
2503 }
2504
2505 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2506 struct gro_remcsum *grc)
2507 {
2508 void *ptr;
2509 size_t plen = grc->offset + sizeof(u16);
2510
2511 if (!grc->delta)
2512 return;
2513
2514 ptr = skb_gro_header_fast(skb, grc->offset);
2515 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2516 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2517 if (!ptr)
2518 return;
2519 }
2520
2521 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2522 }
2523
2524 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2525 unsigned short type,
2526 const void *daddr, const void *saddr,
2527 unsigned int len)
2528 {
2529 if (!dev->header_ops || !dev->header_ops->create)
2530 return 0;
2531
2532 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2533 }
2534
2535 static inline int dev_parse_header(const struct sk_buff *skb,
2536 unsigned char *haddr)
2537 {
2538 const struct net_device *dev = skb->dev;
2539
2540 if (!dev->header_ops || !dev->header_ops->parse)
2541 return 0;
2542 return dev->header_ops->parse(skb, haddr);
2543 }
2544
2545 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2546 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2547 static inline int unregister_gifconf(unsigned int family)
2548 {
2549 return register_gifconf(family, NULL);
2550 }
2551
2552 #ifdef CONFIG_NET_FLOW_LIMIT
2553 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2554 struct sd_flow_limit {
2555 u64 count;
2556 unsigned int num_buckets;
2557 unsigned int history_head;
2558 u16 history[FLOW_LIMIT_HISTORY];
2559 u8 buckets[];
2560 };
2561
2562 extern int netdev_flow_limit_table_len;
2563 #endif /* CONFIG_NET_FLOW_LIMIT */
2564
2565 /*
2566 * Incoming packets are placed on per-cpu queues
2567 */
2568 struct softnet_data {
2569 struct list_head poll_list;
2570 struct sk_buff_head process_queue;
2571
2572 /* stats */
2573 unsigned int processed;
2574 unsigned int time_squeeze;
2575 unsigned int cpu_collision;
2576 unsigned int received_rps;
2577 #ifdef CONFIG_RPS
2578 struct softnet_data *rps_ipi_list;
2579 #endif
2580 #ifdef CONFIG_NET_FLOW_LIMIT
2581 struct sd_flow_limit __rcu *flow_limit;
2582 #endif
2583 struct Qdisc *output_queue;
2584 struct Qdisc **output_queue_tailp;
2585 struct sk_buff *completion_queue;
2586
2587 #ifdef CONFIG_RPS
2588 /* Elements below can be accessed between CPUs for RPS */
2589 struct call_single_data csd ____cacheline_aligned_in_smp;
2590 struct softnet_data *rps_ipi_next;
2591 unsigned int cpu;
2592 unsigned int input_queue_head;
2593 unsigned int input_queue_tail;
2594 #endif
2595 unsigned int dropped;
2596 struct sk_buff_head input_pkt_queue;
2597 struct napi_struct backlog;
2598
2599 };
2600
2601 static inline void input_queue_head_incr(struct softnet_data *sd)
2602 {
2603 #ifdef CONFIG_RPS
2604 sd->input_queue_head++;
2605 #endif
2606 }
2607
2608 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2609 unsigned int *qtail)
2610 {
2611 #ifdef CONFIG_RPS
2612 *qtail = ++sd->input_queue_tail;
2613 #endif
2614 }
2615
2616 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2617
2618 void __netif_schedule(struct Qdisc *q);
2619 void netif_schedule_queue(struct netdev_queue *txq);
2620
2621 static inline void netif_tx_schedule_all(struct net_device *dev)
2622 {
2623 unsigned int i;
2624
2625 for (i = 0; i < dev->num_tx_queues; i++)
2626 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2627 }
2628
2629 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2630 {
2631 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2632 }
2633
2634 /**
2635 * netif_start_queue - allow transmit
2636 * @dev: network device
2637 *
2638 * Allow upper layers to call the device hard_start_xmit routine.
2639 */
2640 static inline void netif_start_queue(struct net_device *dev)
2641 {
2642 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2643 }
2644
2645 static inline void netif_tx_start_all_queues(struct net_device *dev)
2646 {
2647 unsigned int i;
2648
2649 for (i = 0; i < dev->num_tx_queues; i++) {
2650 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2651 netif_tx_start_queue(txq);
2652 }
2653 }
2654
2655 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2656
2657 /**
2658 * netif_wake_queue - restart transmit
2659 * @dev: network device
2660 *
2661 * Allow upper layers to call the device hard_start_xmit routine.
2662 * Used for flow control when transmit resources are available.
2663 */
2664 static inline void netif_wake_queue(struct net_device *dev)
2665 {
2666 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2667 }
2668
2669 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2670 {
2671 unsigned int i;
2672
2673 for (i = 0; i < dev->num_tx_queues; i++) {
2674 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2675 netif_tx_wake_queue(txq);
2676 }
2677 }
2678
2679 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2680 {
2681 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2682 }
2683
2684 /**
2685 * netif_stop_queue - stop transmitted packets
2686 * @dev: network device
2687 *
2688 * Stop upper layers calling the device hard_start_xmit routine.
2689 * Used for flow control when transmit resources are unavailable.
2690 */
2691 static inline void netif_stop_queue(struct net_device *dev)
2692 {
2693 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2694 }
2695
2696 void netif_tx_stop_all_queues(struct net_device *dev);
2697
2698 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2699 {
2700 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2701 }
2702
2703 /**
2704 * netif_queue_stopped - test if transmit queue is flowblocked
2705 * @dev: network device
2706 *
2707 * Test if transmit queue on device is currently unable to send.
2708 */
2709 static inline bool netif_queue_stopped(const struct net_device *dev)
2710 {
2711 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2712 }
2713
2714 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2715 {
2716 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2717 }
2718
2719 static inline bool
2720 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2721 {
2722 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2723 }
2724
2725 static inline bool
2726 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2727 {
2728 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2729 }
2730
2731 /**
2732 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2733 * @dev_queue: pointer to transmit queue
2734 *
2735 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2736 * to give appropriate hint to the cpu.
2737 */
2738 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2739 {
2740 #ifdef CONFIG_BQL
2741 prefetchw(&dev_queue->dql.num_queued);
2742 #endif
2743 }
2744
2745 /**
2746 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2747 * @dev_queue: pointer to transmit queue
2748 *
2749 * BQL enabled drivers might use this helper in their TX completion path,
2750 * to give appropriate hint to the cpu.
2751 */
2752 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2753 {
2754 #ifdef CONFIG_BQL
2755 prefetchw(&dev_queue->dql.limit);
2756 #endif
2757 }
2758
2759 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2760 unsigned int bytes)
2761 {
2762 #ifdef CONFIG_BQL
2763 dql_queued(&dev_queue->dql, bytes);
2764
2765 if (likely(dql_avail(&dev_queue->dql) >= 0))
2766 return;
2767
2768 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2769
2770 /*
2771 * The XOFF flag must be set before checking the dql_avail below,
2772 * because in netdev_tx_completed_queue we update the dql_completed
2773 * before checking the XOFF flag.
2774 */
2775 smp_mb();
2776
2777 /* check again in case another CPU has just made room avail */
2778 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2779 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2780 #endif
2781 }
2782
2783 /**
2784 * netdev_sent_queue - report the number of bytes queued to hardware
2785 * @dev: network device
2786 * @bytes: number of bytes queued to the hardware device queue
2787 *
2788 * Report the number of bytes queued for sending/completion to the network
2789 * device hardware queue. @bytes should be a good approximation and should
2790 * exactly match netdev_completed_queue() @bytes
2791 */
2792 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2793 {
2794 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2795 }
2796
2797 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2798 unsigned int pkts, unsigned int bytes)
2799 {
2800 #ifdef CONFIG_BQL
2801 if (unlikely(!bytes))
2802 return;
2803
2804 dql_completed(&dev_queue->dql, bytes);
2805
2806 /*
2807 * Without the memory barrier there is a small possiblity that
2808 * netdev_tx_sent_queue will miss the update and cause the queue to
2809 * be stopped forever
2810 */
2811 smp_mb();
2812
2813 if (dql_avail(&dev_queue->dql) < 0)
2814 return;
2815
2816 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2817 netif_schedule_queue(dev_queue);
2818 #endif
2819 }
2820
2821 /**
2822 * netdev_completed_queue - report bytes and packets completed by device
2823 * @dev: network device
2824 * @pkts: actual number of packets sent over the medium
2825 * @bytes: actual number of bytes sent over the medium
2826 *
2827 * Report the number of bytes and packets transmitted by the network device
2828 * hardware queue over the physical medium, @bytes must exactly match the
2829 * @bytes amount passed to netdev_sent_queue()
2830 */
2831 static inline void netdev_completed_queue(struct net_device *dev,
2832 unsigned int pkts, unsigned int bytes)
2833 {
2834 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2835 }
2836
2837 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2838 {
2839 #ifdef CONFIG_BQL
2840 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2841 dql_reset(&q->dql);
2842 #endif
2843 }
2844
2845 /**
2846 * netdev_reset_queue - reset the packets and bytes count of a network device
2847 * @dev_queue: network device
2848 *
2849 * Reset the bytes and packet count of a network device and clear the
2850 * software flow control OFF bit for this network device
2851 */
2852 static inline void netdev_reset_queue(struct net_device *dev_queue)
2853 {
2854 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2855 }
2856
2857 /**
2858 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2859 * @dev: network device
2860 * @queue_index: given tx queue index
2861 *
2862 * Returns 0 if given tx queue index >= number of device tx queues,
2863 * otherwise returns the originally passed tx queue index.
2864 */
2865 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2866 {
2867 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2868 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2869 dev->name, queue_index,
2870 dev->real_num_tx_queues);
2871 return 0;
2872 }
2873
2874 return queue_index;
2875 }
2876
2877 /**
2878 * netif_running - test if up
2879 * @dev: network device
2880 *
2881 * Test if the device has been brought up.
2882 */
2883 static inline bool netif_running(const struct net_device *dev)
2884 {
2885 return test_bit(__LINK_STATE_START, &dev->state);
2886 }
2887
2888 /*
2889 * Routines to manage the subqueues on a device. We only need start
2890 * stop, and a check if it's stopped. All other device management is
2891 * done at the overall netdevice level.
2892 * Also test the device if we're multiqueue.
2893 */
2894
2895 /**
2896 * netif_start_subqueue - allow sending packets on subqueue
2897 * @dev: network device
2898 * @queue_index: sub queue index
2899 *
2900 * Start individual transmit queue of a device with multiple transmit queues.
2901 */
2902 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2903 {
2904 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2905
2906 netif_tx_start_queue(txq);
2907 }
2908
2909 /**
2910 * netif_stop_subqueue - stop sending packets on subqueue
2911 * @dev: network device
2912 * @queue_index: sub queue index
2913 *
2914 * Stop individual transmit queue of a device with multiple transmit queues.
2915 */
2916 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2917 {
2918 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2919 netif_tx_stop_queue(txq);
2920 }
2921
2922 /**
2923 * netif_subqueue_stopped - test status of subqueue
2924 * @dev: network device
2925 * @queue_index: sub queue index
2926 *
2927 * Check individual transmit queue of a device with multiple transmit queues.
2928 */
2929 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2930 u16 queue_index)
2931 {
2932 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2933
2934 return netif_tx_queue_stopped(txq);
2935 }
2936
2937 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2938 struct sk_buff *skb)
2939 {
2940 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2941 }
2942
2943 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2944
2945 #ifdef CONFIG_XPS
2946 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2947 u16 index);
2948 #else
2949 static inline int netif_set_xps_queue(struct net_device *dev,
2950 const struct cpumask *mask,
2951 u16 index)
2952 {
2953 return 0;
2954 }
2955 #endif
2956
2957 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2958 unsigned int num_tx_queues);
2959
2960 /*
2961 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2962 * as a distribution range limit for the returned value.
2963 */
2964 static inline u16 skb_tx_hash(const struct net_device *dev,
2965 struct sk_buff *skb)
2966 {
2967 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2968 }
2969
2970 /**
2971 * netif_is_multiqueue - test if device has multiple transmit queues
2972 * @dev: network device
2973 *
2974 * Check if device has multiple transmit queues
2975 */
2976 static inline bool netif_is_multiqueue(const struct net_device *dev)
2977 {
2978 return dev->num_tx_queues > 1;
2979 }
2980
2981 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2982
2983 #ifdef CONFIG_SYSFS
2984 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2985 #else
2986 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2987 unsigned int rxq)
2988 {
2989 return 0;
2990 }
2991 #endif
2992
2993 #ifdef CONFIG_SYSFS
2994 static inline unsigned int get_netdev_rx_queue_index(
2995 struct netdev_rx_queue *queue)
2996 {
2997 struct net_device *dev = queue->dev;
2998 int index = queue - dev->_rx;
2999
3000 BUG_ON(index >= dev->num_rx_queues);
3001 return index;
3002 }
3003 #endif
3004
3005 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3006 int netif_get_num_default_rss_queues(void);
3007
3008 enum skb_free_reason {
3009 SKB_REASON_CONSUMED,
3010 SKB_REASON_DROPPED,
3011 };
3012
3013 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3014 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3015
3016 /*
3017 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3018 * interrupt context or with hardware interrupts being disabled.
3019 * (in_irq() || irqs_disabled())
3020 *
3021 * We provide four helpers that can be used in following contexts :
3022 *
3023 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3024 * replacing kfree_skb(skb)
3025 *
3026 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3027 * Typically used in place of consume_skb(skb) in TX completion path
3028 *
3029 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3030 * replacing kfree_skb(skb)
3031 *
3032 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3033 * and consumed a packet. Used in place of consume_skb(skb)
3034 */
3035 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3036 {
3037 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3038 }
3039
3040 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3041 {
3042 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3043 }
3044
3045 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3046 {
3047 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3048 }
3049
3050 static inline void dev_consume_skb_any(struct sk_buff *skb)
3051 {
3052 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3053 }
3054
3055 int netif_rx(struct sk_buff *skb);
3056 int netif_rx_ni(struct sk_buff *skb);
3057 int netif_receive_skb(struct sk_buff *skb);
3058 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3059 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3060 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3061 gro_result_t napi_gro_frags(struct napi_struct *napi);
3062 struct packet_offload *gro_find_receive_by_type(__be16 type);
3063 struct packet_offload *gro_find_complete_by_type(__be16 type);
3064
3065 static inline void napi_free_frags(struct napi_struct *napi)
3066 {
3067 kfree_skb(napi->skb);
3068 napi->skb = NULL;
3069 }
3070
3071 int netdev_rx_handler_register(struct net_device *dev,
3072 rx_handler_func_t *rx_handler,
3073 void *rx_handler_data);
3074 void netdev_rx_handler_unregister(struct net_device *dev);
3075
3076 bool dev_valid_name(const char *name);
3077 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3078 int dev_ethtool(struct net *net, struct ifreq *);
3079 unsigned int dev_get_flags(const struct net_device *);
3080 int __dev_change_flags(struct net_device *, unsigned int flags);
3081 int dev_change_flags(struct net_device *, unsigned int);
3082 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3083 unsigned int gchanges);
3084 int dev_change_name(struct net_device *, const char *);
3085 int dev_set_alias(struct net_device *, const char *, size_t);
3086 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3087 int dev_set_mtu(struct net_device *, int);
3088 void dev_set_group(struct net_device *, int);
3089 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3090 int dev_change_carrier(struct net_device *, bool new_carrier);
3091 int dev_get_phys_port_id(struct net_device *dev,
3092 struct netdev_phys_item_id *ppid);
3093 int dev_get_phys_port_name(struct net_device *dev,
3094 char *name, size_t len);
3095 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3096 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3097 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3098 struct netdev_queue *txq, int *ret);
3099 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3100 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3101 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3102
3103 extern int netdev_budget;
3104
3105 /* Called by rtnetlink.c:rtnl_unlock() */
3106 void netdev_run_todo(void);
3107
3108 /**
3109 * dev_put - release reference to device
3110 * @dev: network device
3111 *
3112 * Release reference to device to allow it to be freed.
3113 */
3114 static inline void dev_put(struct net_device *dev)
3115 {
3116 this_cpu_dec(*dev->pcpu_refcnt);
3117 }
3118
3119 /**
3120 * dev_hold - get reference to device
3121 * @dev: network device
3122 *
3123 * Hold reference to device to keep it from being freed.
3124 */
3125 static inline void dev_hold(struct net_device *dev)
3126 {
3127 this_cpu_inc(*dev->pcpu_refcnt);
3128 }
3129
3130 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3131 * and _off may be called from IRQ context, but it is caller
3132 * who is responsible for serialization of these calls.
3133 *
3134 * The name carrier is inappropriate, these functions should really be
3135 * called netif_lowerlayer_*() because they represent the state of any
3136 * kind of lower layer not just hardware media.
3137 */
3138
3139 void linkwatch_init_dev(struct net_device *dev);
3140 void linkwatch_fire_event(struct net_device *dev);
3141 void linkwatch_forget_dev(struct net_device *dev);
3142
3143 /**
3144 * netif_carrier_ok - test if carrier present
3145 * @dev: network device
3146 *
3147 * Check if carrier is present on device
3148 */
3149 static inline bool netif_carrier_ok(const struct net_device *dev)
3150 {
3151 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3152 }
3153
3154 unsigned long dev_trans_start(struct net_device *dev);
3155
3156 void __netdev_watchdog_up(struct net_device *dev);
3157
3158 void netif_carrier_on(struct net_device *dev);
3159
3160 void netif_carrier_off(struct net_device *dev);
3161
3162 /**
3163 * netif_dormant_on - mark device as dormant.
3164 * @dev: network device
3165 *
3166 * Mark device as dormant (as per RFC2863).
3167 *
3168 * The dormant state indicates that the relevant interface is not
3169 * actually in a condition to pass packets (i.e., it is not 'up') but is
3170 * in a "pending" state, waiting for some external event. For "on-
3171 * demand" interfaces, this new state identifies the situation where the
3172 * interface is waiting for events to place it in the up state.
3173 *
3174 */
3175 static inline void netif_dormant_on(struct net_device *dev)
3176 {
3177 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3178 linkwatch_fire_event(dev);
3179 }
3180
3181 /**
3182 * netif_dormant_off - set device as not dormant.
3183 * @dev: network device
3184 *
3185 * Device is not in dormant state.
3186 */
3187 static inline void netif_dormant_off(struct net_device *dev)
3188 {
3189 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3190 linkwatch_fire_event(dev);
3191 }
3192
3193 /**
3194 * netif_dormant - test if carrier present
3195 * @dev: network device
3196 *
3197 * Check if carrier is present on device
3198 */
3199 static inline bool netif_dormant(const struct net_device *dev)
3200 {
3201 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3202 }
3203
3204
3205 /**
3206 * netif_oper_up - test if device is operational
3207 * @dev: network device
3208 *
3209 * Check if carrier is operational
3210 */
3211 static inline bool netif_oper_up(const struct net_device *dev)
3212 {
3213 return (dev->operstate == IF_OPER_UP ||
3214 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3215 }
3216
3217 /**
3218 * netif_device_present - is device available or removed
3219 * @dev: network device
3220 *
3221 * Check if device has not been removed from system.
3222 */
3223 static inline bool netif_device_present(struct net_device *dev)
3224 {
3225 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3226 }
3227
3228 void netif_device_detach(struct net_device *dev);
3229
3230 void netif_device_attach(struct net_device *dev);
3231
3232 /*
3233 * Network interface message level settings
3234 */
3235
3236 enum {
3237 NETIF_MSG_DRV = 0x0001,
3238 NETIF_MSG_PROBE = 0x0002,
3239 NETIF_MSG_LINK = 0x0004,
3240 NETIF_MSG_TIMER = 0x0008,
3241 NETIF_MSG_IFDOWN = 0x0010,
3242 NETIF_MSG_IFUP = 0x0020,
3243 NETIF_MSG_RX_ERR = 0x0040,
3244 NETIF_MSG_TX_ERR = 0x0080,
3245 NETIF_MSG_TX_QUEUED = 0x0100,
3246 NETIF_MSG_INTR = 0x0200,
3247 NETIF_MSG_TX_DONE = 0x0400,
3248 NETIF_MSG_RX_STATUS = 0x0800,
3249 NETIF_MSG_PKTDATA = 0x1000,
3250 NETIF_MSG_HW = 0x2000,
3251 NETIF_MSG_WOL = 0x4000,
3252 };
3253
3254 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3255 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3256 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3257 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3258 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3259 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3260 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3261 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3262 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3263 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3264 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3265 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3266 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3267 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3268 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3269
3270 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3271 {
3272 /* use default */
3273 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3274 return default_msg_enable_bits;
3275 if (debug_value == 0) /* no output */
3276 return 0;
3277 /* set low N bits */
3278 return (1 << debug_value) - 1;
3279 }
3280
3281 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3282 {
3283 spin_lock(&txq->_xmit_lock);
3284 txq->xmit_lock_owner = cpu;
3285 }
3286
3287 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3288 {
3289 spin_lock_bh(&txq->_xmit_lock);
3290 txq->xmit_lock_owner = smp_processor_id();
3291 }
3292
3293 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3294 {
3295 bool ok = spin_trylock(&txq->_xmit_lock);
3296 if (likely(ok))
3297 txq->xmit_lock_owner = smp_processor_id();
3298 return ok;
3299 }
3300
3301 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3302 {
3303 txq->xmit_lock_owner = -1;
3304 spin_unlock(&txq->_xmit_lock);
3305 }
3306
3307 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3308 {
3309 txq->xmit_lock_owner = -1;
3310 spin_unlock_bh(&txq->_xmit_lock);
3311 }
3312
3313 static inline void txq_trans_update(struct netdev_queue *txq)
3314 {
3315 if (txq->xmit_lock_owner != -1)
3316 txq->trans_start = jiffies;
3317 }
3318
3319 /**
3320 * netif_tx_lock - grab network device transmit lock
3321 * @dev: network device
3322 *
3323 * Get network device transmit lock
3324 */
3325 static inline void netif_tx_lock(struct net_device *dev)
3326 {
3327 unsigned int i;
3328 int cpu;
3329
3330 spin_lock(&dev->tx_global_lock);
3331 cpu = smp_processor_id();
3332 for (i = 0; i < dev->num_tx_queues; i++) {
3333 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3334
3335 /* We are the only thread of execution doing a
3336 * freeze, but we have to grab the _xmit_lock in
3337 * order to synchronize with threads which are in
3338 * the ->hard_start_xmit() handler and already
3339 * checked the frozen bit.
3340 */
3341 __netif_tx_lock(txq, cpu);
3342 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3343 __netif_tx_unlock(txq);
3344 }
3345 }
3346
3347 static inline void netif_tx_lock_bh(struct net_device *dev)
3348 {
3349 local_bh_disable();
3350 netif_tx_lock(dev);
3351 }
3352
3353 static inline void netif_tx_unlock(struct net_device *dev)
3354 {
3355 unsigned int i;
3356
3357 for (i = 0; i < dev->num_tx_queues; i++) {
3358 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3359
3360 /* No need to grab the _xmit_lock here. If the
3361 * queue is not stopped for another reason, we
3362 * force a schedule.
3363 */
3364 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3365 netif_schedule_queue(txq);
3366 }
3367 spin_unlock(&dev->tx_global_lock);
3368 }
3369
3370 static inline void netif_tx_unlock_bh(struct net_device *dev)
3371 {
3372 netif_tx_unlock(dev);
3373 local_bh_enable();
3374 }
3375
3376 #define HARD_TX_LOCK(dev, txq, cpu) { \
3377 if ((dev->features & NETIF_F_LLTX) == 0) { \
3378 __netif_tx_lock(txq, cpu); \
3379 } \
3380 }
3381
3382 #define HARD_TX_TRYLOCK(dev, txq) \
3383 (((dev->features & NETIF_F_LLTX) == 0) ? \
3384 __netif_tx_trylock(txq) : \
3385 true )
3386
3387 #define HARD_TX_UNLOCK(dev, txq) { \
3388 if ((dev->features & NETIF_F_LLTX) == 0) { \
3389 __netif_tx_unlock(txq); \
3390 } \
3391 }
3392
3393 static inline void netif_tx_disable(struct net_device *dev)
3394 {
3395 unsigned int i;
3396 int cpu;
3397
3398 local_bh_disable();
3399 cpu = smp_processor_id();
3400 for (i = 0; i < dev->num_tx_queues; i++) {
3401 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3402
3403 __netif_tx_lock(txq, cpu);
3404 netif_tx_stop_queue(txq);
3405 __netif_tx_unlock(txq);
3406 }
3407 local_bh_enable();
3408 }
3409
3410 static inline void netif_addr_lock(struct net_device *dev)
3411 {
3412 spin_lock(&dev->addr_list_lock);
3413 }
3414
3415 static inline void netif_addr_lock_nested(struct net_device *dev)
3416 {
3417 int subclass = SINGLE_DEPTH_NESTING;
3418
3419 if (dev->netdev_ops->ndo_get_lock_subclass)
3420 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3421
3422 spin_lock_nested(&dev->addr_list_lock, subclass);
3423 }
3424
3425 static inline void netif_addr_lock_bh(struct net_device *dev)
3426 {
3427 spin_lock_bh(&dev->addr_list_lock);
3428 }
3429
3430 static inline void netif_addr_unlock(struct net_device *dev)
3431 {
3432 spin_unlock(&dev->addr_list_lock);
3433 }
3434
3435 static inline void netif_addr_unlock_bh(struct net_device *dev)
3436 {
3437 spin_unlock_bh(&dev->addr_list_lock);
3438 }
3439
3440 /*
3441 * dev_addrs walker. Should be used only for read access. Call with
3442 * rcu_read_lock held.
3443 */
3444 #define for_each_dev_addr(dev, ha) \
3445 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3446
3447 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3448
3449 void ether_setup(struct net_device *dev);
3450
3451 /* Support for loadable net-drivers */
3452 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3453 unsigned char name_assign_type,
3454 void (*setup)(struct net_device *),
3455 unsigned int txqs, unsigned int rxqs);
3456 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3457 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3458
3459 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3460 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3461 count)
3462
3463 int register_netdev(struct net_device *dev);
3464 void unregister_netdev(struct net_device *dev);
3465
3466 /* General hardware address lists handling functions */
3467 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3468 struct netdev_hw_addr_list *from_list, int addr_len);
3469 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3470 struct netdev_hw_addr_list *from_list, int addr_len);
3471 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3472 struct net_device *dev,
3473 int (*sync)(struct net_device *, const unsigned char *),
3474 int (*unsync)(struct net_device *,
3475 const unsigned char *));
3476 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3477 struct net_device *dev,
3478 int (*unsync)(struct net_device *,
3479 const unsigned char *));
3480 void __hw_addr_init(struct netdev_hw_addr_list *list);
3481
3482 /* Functions used for device addresses handling */
3483 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3484 unsigned char addr_type);
3485 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3486 unsigned char addr_type);
3487 void dev_addr_flush(struct net_device *dev);
3488 int dev_addr_init(struct net_device *dev);
3489
3490 /* Functions used for unicast addresses handling */
3491 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3492 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3493 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3494 int dev_uc_sync(struct net_device *to, struct net_device *from);
3495 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3496 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3497 void dev_uc_flush(struct net_device *dev);
3498 void dev_uc_init(struct net_device *dev);
3499
3500 /**
3501 * __dev_uc_sync - Synchonize device's unicast list
3502 * @dev: device to sync
3503 * @sync: function to call if address should be added
3504 * @unsync: function to call if address should be removed
3505 *
3506 * Add newly added addresses to the interface, and release
3507 * addresses that have been deleted.
3508 **/
3509 static inline int __dev_uc_sync(struct net_device *dev,
3510 int (*sync)(struct net_device *,
3511 const unsigned char *),
3512 int (*unsync)(struct net_device *,
3513 const unsigned char *))
3514 {
3515 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3516 }
3517
3518 /**
3519 * __dev_uc_unsync - Remove synchronized addresses from device
3520 * @dev: device to sync
3521 * @unsync: function to call if address should be removed
3522 *
3523 * Remove all addresses that were added to the device by dev_uc_sync().
3524 **/
3525 static inline void __dev_uc_unsync(struct net_device *dev,
3526 int (*unsync)(struct net_device *,
3527 const unsigned char *))
3528 {
3529 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3530 }
3531
3532 /* Functions used for multicast addresses handling */
3533 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3534 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3535 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3536 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3537 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3538 int dev_mc_sync(struct net_device *to, struct net_device *from);
3539 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3540 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3541 void dev_mc_flush(struct net_device *dev);
3542 void dev_mc_init(struct net_device *dev);
3543
3544 /**
3545 * __dev_mc_sync - Synchonize device's multicast list
3546 * @dev: device to sync
3547 * @sync: function to call if address should be added
3548 * @unsync: function to call if address should be removed
3549 *
3550 * Add newly added addresses to the interface, and release
3551 * addresses that have been deleted.
3552 **/
3553 static inline int __dev_mc_sync(struct net_device *dev,
3554 int (*sync)(struct net_device *,
3555 const unsigned char *),
3556 int (*unsync)(struct net_device *,
3557 const unsigned char *))
3558 {
3559 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3560 }
3561
3562 /**
3563 * __dev_mc_unsync - Remove synchronized addresses from device
3564 * @dev: device to sync
3565 * @unsync: function to call if address should be removed
3566 *
3567 * Remove all addresses that were added to the device by dev_mc_sync().
3568 **/
3569 static inline void __dev_mc_unsync(struct net_device *dev,
3570 int (*unsync)(struct net_device *,
3571 const unsigned char *))
3572 {
3573 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3574 }
3575
3576 /* Functions used for secondary unicast and multicast support */
3577 void dev_set_rx_mode(struct net_device *dev);
3578 void __dev_set_rx_mode(struct net_device *dev);
3579 int dev_set_promiscuity(struct net_device *dev, int inc);
3580 int dev_set_allmulti(struct net_device *dev, int inc);
3581 void netdev_state_change(struct net_device *dev);
3582 void netdev_notify_peers(struct net_device *dev);
3583 void netdev_features_change(struct net_device *dev);
3584 /* Load a device via the kmod */
3585 void dev_load(struct net *net, const char *name);
3586 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3587 struct rtnl_link_stats64 *storage);
3588 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3589 const struct net_device_stats *netdev_stats);
3590
3591 extern int netdev_max_backlog;
3592 extern int netdev_tstamp_prequeue;
3593 extern int weight_p;
3594 extern int bpf_jit_enable;
3595
3596 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3597 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3598 struct list_head **iter);
3599 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3600 struct list_head **iter);
3601
3602 /* iterate through upper list, must be called under RCU read lock */
3603 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3604 for (iter = &(dev)->adj_list.upper, \
3605 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3606 updev; \
3607 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3608
3609 /* iterate through upper list, must be called under RCU read lock */
3610 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3611 for (iter = &(dev)->all_adj_list.upper, \
3612 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3613 updev; \
3614 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3615
3616 void *netdev_lower_get_next_private(struct net_device *dev,
3617 struct list_head **iter);
3618 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3619 struct list_head **iter);
3620
3621 #define netdev_for_each_lower_private(dev, priv, iter) \
3622 for (iter = (dev)->adj_list.lower.next, \
3623 priv = netdev_lower_get_next_private(dev, &(iter)); \
3624 priv; \
3625 priv = netdev_lower_get_next_private(dev, &(iter)))
3626
3627 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3628 for (iter = &(dev)->adj_list.lower, \
3629 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3630 priv; \
3631 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3632
3633 void *netdev_lower_get_next(struct net_device *dev,
3634 struct list_head **iter);
3635 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3636 for (iter = &(dev)->adj_list.lower, \
3637 ldev = netdev_lower_get_next(dev, &(iter)); \
3638 ldev; \
3639 ldev = netdev_lower_get_next(dev, &(iter)))
3640
3641 void *netdev_adjacent_get_private(struct list_head *adj_list);
3642 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3643 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3644 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3645 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3646 int netdev_master_upper_dev_link(struct net_device *dev,
3647 struct net_device *upper_dev,
3648 void *upper_priv, void *upper_info);
3649 void netdev_upper_dev_unlink(struct net_device *dev,
3650 struct net_device *upper_dev);
3651 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3652 void *netdev_lower_dev_get_private(struct net_device *dev,
3653 struct net_device *lower_dev);
3654 void netdev_lower_state_changed(struct net_device *lower_dev,
3655 void *lower_state_info);
3656
3657 /* RSS keys are 40 or 52 bytes long */
3658 #define NETDEV_RSS_KEY_LEN 52
3659 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3660 void netdev_rss_key_fill(void *buffer, size_t len);
3661
3662 int dev_get_nest_level(struct net_device *dev,
3663 bool (*type_check)(struct net_device *dev));
3664 int skb_checksum_help(struct sk_buff *skb);
3665 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3666 netdev_features_t features, bool tx_path);
3667 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3668 netdev_features_t features);
3669
3670 struct netdev_bonding_info {
3671 ifslave slave;
3672 ifbond master;
3673 };
3674
3675 struct netdev_notifier_bonding_info {
3676 struct netdev_notifier_info info; /* must be first */
3677 struct netdev_bonding_info bonding_info;
3678 };
3679
3680 void netdev_bonding_info_change(struct net_device *dev,
3681 struct netdev_bonding_info *bonding_info);
3682
3683 static inline
3684 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3685 {
3686 return __skb_gso_segment(skb, features, true);
3687 }
3688 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3689
3690 static inline bool can_checksum_protocol(netdev_features_t features,
3691 __be16 protocol)
3692 {
3693 return ((features & NETIF_F_GEN_CSUM) ||
3694 ((features & NETIF_F_V4_CSUM) &&
3695 protocol == htons(ETH_P_IP)) ||
3696 ((features & NETIF_F_V6_CSUM) &&
3697 protocol == htons(ETH_P_IPV6)) ||
3698 ((features & NETIF_F_FCOE_CRC) &&
3699 protocol == htons(ETH_P_FCOE)));
3700 }
3701
3702 #ifdef CONFIG_BUG
3703 void netdev_rx_csum_fault(struct net_device *dev);
3704 #else
3705 static inline void netdev_rx_csum_fault(struct net_device *dev)
3706 {
3707 }
3708 #endif
3709 /* rx skb timestamps */
3710 void net_enable_timestamp(void);
3711 void net_disable_timestamp(void);
3712
3713 #ifdef CONFIG_PROC_FS
3714 int __init dev_proc_init(void);
3715 #else
3716 #define dev_proc_init() 0
3717 #endif
3718
3719 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3720 struct sk_buff *skb, struct net_device *dev,
3721 bool more)
3722 {
3723 skb->xmit_more = more ? 1 : 0;
3724 return ops->ndo_start_xmit(skb, dev);
3725 }
3726
3727 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3728 struct netdev_queue *txq, bool more)
3729 {
3730 const struct net_device_ops *ops = dev->netdev_ops;
3731 int rc;
3732
3733 rc = __netdev_start_xmit(ops, skb, dev, more);
3734 if (rc == NETDEV_TX_OK)
3735 txq_trans_update(txq);
3736
3737 return rc;
3738 }
3739
3740 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3741 const void *ns);
3742 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3743 const void *ns);
3744
3745 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3746 {
3747 return netdev_class_create_file_ns(class_attr, NULL);
3748 }
3749
3750 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3751 {
3752 netdev_class_remove_file_ns(class_attr, NULL);
3753 }
3754
3755 extern struct kobj_ns_type_operations net_ns_type_operations;
3756
3757 const char *netdev_drivername(const struct net_device *dev);
3758
3759 void linkwatch_run_queue(void);
3760
3761 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3762 netdev_features_t f2)
3763 {
3764 if (f1 & NETIF_F_GEN_CSUM)
3765 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3766 if (f2 & NETIF_F_GEN_CSUM)
3767 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3768 f1 &= f2;
3769 if (f1 & NETIF_F_GEN_CSUM)
3770 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3771
3772 return f1;
3773 }
3774
3775 static inline netdev_features_t netdev_get_wanted_features(
3776 struct net_device *dev)
3777 {
3778 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3779 }
3780 netdev_features_t netdev_increment_features(netdev_features_t all,
3781 netdev_features_t one, netdev_features_t mask);
3782
3783 /* Allow TSO being used on stacked device :
3784 * Performing the GSO segmentation before last device
3785 * is a performance improvement.
3786 */
3787 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3788 netdev_features_t mask)
3789 {
3790 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3791 }
3792
3793 int __netdev_update_features(struct net_device *dev);
3794 void netdev_update_features(struct net_device *dev);
3795 void netdev_change_features(struct net_device *dev);
3796
3797 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3798 struct net_device *dev);
3799
3800 netdev_features_t passthru_features_check(struct sk_buff *skb,
3801 struct net_device *dev,
3802 netdev_features_t features);
3803 netdev_features_t netif_skb_features(struct sk_buff *skb);
3804
3805 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3806 {
3807 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3808
3809 /* check flags correspondence */
3810 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3811 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3812 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3813 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3814 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3815 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3816 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3817 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3818 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3819 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3820 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3821 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3822 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3823
3824 return (features & feature) == feature;
3825 }
3826
3827 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3828 {
3829 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3830 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3831 }
3832
3833 static inline bool netif_needs_gso(struct sk_buff *skb,
3834 netdev_features_t features)
3835 {
3836 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3837 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3838 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3839 }
3840
3841 static inline void netif_set_gso_max_size(struct net_device *dev,
3842 unsigned int size)
3843 {
3844 dev->gso_max_size = size;
3845 }
3846
3847 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3848 int pulled_hlen, u16 mac_offset,
3849 int mac_len)
3850 {
3851 skb->protocol = protocol;
3852 skb->encapsulation = 1;
3853 skb_push(skb, pulled_hlen);
3854 skb_reset_transport_header(skb);
3855 skb->mac_header = mac_offset;
3856 skb->network_header = skb->mac_header + mac_len;
3857 skb->mac_len = mac_len;
3858 }
3859
3860 static inline bool netif_is_macvlan(struct net_device *dev)
3861 {
3862 return dev->priv_flags & IFF_MACVLAN;
3863 }
3864
3865 static inline bool netif_is_macvlan_port(struct net_device *dev)
3866 {
3867 return dev->priv_flags & IFF_MACVLAN_PORT;
3868 }
3869
3870 static inline bool netif_is_ipvlan(struct net_device *dev)
3871 {
3872 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3873 }
3874
3875 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3876 {
3877 return dev->priv_flags & IFF_IPVLAN_MASTER;
3878 }
3879
3880 static inline bool netif_is_bond_master(struct net_device *dev)
3881 {
3882 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3883 }
3884
3885 static inline bool netif_is_bond_slave(struct net_device *dev)
3886 {
3887 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3888 }
3889
3890 static inline bool netif_supports_nofcs(struct net_device *dev)
3891 {
3892 return dev->priv_flags & IFF_SUPP_NOFCS;
3893 }
3894
3895 static inline bool netif_is_l3_master(const struct net_device *dev)
3896 {
3897 return dev->priv_flags & IFF_L3MDEV_MASTER;
3898 }
3899
3900 static inline bool netif_is_l3_slave(const struct net_device *dev)
3901 {
3902 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3903 }
3904
3905 static inline bool netif_is_bridge_master(const struct net_device *dev)
3906 {
3907 return dev->priv_flags & IFF_EBRIDGE;
3908 }
3909
3910 static inline bool netif_is_bridge_port(const struct net_device *dev)
3911 {
3912 return dev->priv_flags & IFF_BRIDGE_PORT;
3913 }
3914
3915 static inline bool netif_is_ovs_master(const struct net_device *dev)
3916 {
3917 return dev->priv_flags & IFF_OPENVSWITCH;
3918 }
3919
3920 static inline bool netif_is_team_master(struct net_device *dev)
3921 {
3922 return dev->priv_flags & IFF_TEAM;
3923 }
3924
3925 static inline bool netif_is_team_port(struct net_device *dev)
3926 {
3927 return dev->priv_flags & IFF_TEAM_PORT;
3928 }
3929
3930 static inline bool netif_is_lag_master(struct net_device *dev)
3931 {
3932 return netif_is_bond_master(dev) || netif_is_team_master(dev);
3933 }
3934
3935 static inline bool netif_is_lag_port(struct net_device *dev)
3936 {
3937 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
3938 }
3939
3940 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3941 static inline void netif_keep_dst(struct net_device *dev)
3942 {
3943 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3944 }
3945
3946 extern struct pernet_operations __net_initdata loopback_net_ops;
3947
3948 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3949
3950 /* netdev_printk helpers, similar to dev_printk */
3951
3952 static inline const char *netdev_name(const struct net_device *dev)
3953 {
3954 if (!dev->name[0] || strchr(dev->name, '%'))
3955 return "(unnamed net_device)";
3956 return dev->name;
3957 }
3958
3959 static inline const char *netdev_reg_state(const struct net_device *dev)
3960 {
3961 switch (dev->reg_state) {
3962 case NETREG_UNINITIALIZED: return " (uninitialized)";
3963 case NETREG_REGISTERED: return "";
3964 case NETREG_UNREGISTERING: return " (unregistering)";
3965 case NETREG_UNREGISTERED: return " (unregistered)";
3966 case NETREG_RELEASED: return " (released)";
3967 case NETREG_DUMMY: return " (dummy)";
3968 }
3969
3970 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3971 return " (unknown)";
3972 }
3973
3974 __printf(3, 4)
3975 void netdev_printk(const char *level, const struct net_device *dev,
3976 const char *format, ...);
3977 __printf(2, 3)
3978 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3979 __printf(2, 3)
3980 void netdev_alert(const struct net_device *dev, const char *format, ...);
3981 __printf(2, 3)
3982 void netdev_crit(const struct net_device *dev, const char *format, ...);
3983 __printf(2, 3)
3984 void netdev_err(const struct net_device *dev, const char *format, ...);
3985 __printf(2, 3)
3986 void netdev_warn(const struct net_device *dev, const char *format, ...);
3987 __printf(2, 3)
3988 void netdev_notice(const struct net_device *dev, const char *format, ...);
3989 __printf(2, 3)
3990 void netdev_info(const struct net_device *dev, const char *format, ...);
3991
3992 #define MODULE_ALIAS_NETDEV(device) \
3993 MODULE_ALIAS("netdev-" device)
3994
3995 #if defined(CONFIG_DYNAMIC_DEBUG)
3996 #define netdev_dbg(__dev, format, args...) \
3997 do { \
3998 dynamic_netdev_dbg(__dev, format, ##args); \
3999 } while (0)
4000 #elif defined(DEBUG)
4001 #define netdev_dbg(__dev, format, args...) \
4002 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4003 #else
4004 #define netdev_dbg(__dev, format, args...) \
4005 ({ \
4006 if (0) \
4007 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4008 })
4009 #endif
4010
4011 #if defined(VERBOSE_DEBUG)
4012 #define netdev_vdbg netdev_dbg
4013 #else
4014
4015 #define netdev_vdbg(dev, format, args...) \
4016 ({ \
4017 if (0) \
4018 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4019 0; \
4020 })
4021 #endif
4022
4023 /*
4024 * netdev_WARN() acts like dev_printk(), but with the key difference
4025 * of using a WARN/WARN_ON to get the message out, including the
4026 * file/line information and a backtrace.
4027 */
4028 #define netdev_WARN(dev, format, args...) \
4029 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4030 netdev_reg_state(dev), ##args)
4031
4032 /* netif printk helpers, similar to netdev_printk */
4033
4034 #define netif_printk(priv, type, level, dev, fmt, args...) \
4035 do { \
4036 if (netif_msg_##type(priv)) \
4037 netdev_printk(level, (dev), fmt, ##args); \
4038 } while (0)
4039
4040 #define netif_level(level, priv, type, dev, fmt, args...) \
4041 do { \
4042 if (netif_msg_##type(priv)) \
4043 netdev_##level(dev, fmt, ##args); \
4044 } while (0)
4045
4046 #define netif_emerg(priv, type, dev, fmt, args...) \
4047 netif_level(emerg, priv, type, dev, fmt, ##args)
4048 #define netif_alert(priv, type, dev, fmt, args...) \
4049 netif_level(alert, priv, type, dev, fmt, ##args)
4050 #define netif_crit(priv, type, dev, fmt, args...) \
4051 netif_level(crit, priv, type, dev, fmt, ##args)
4052 #define netif_err(priv, type, dev, fmt, args...) \
4053 netif_level(err, priv, type, dev, fmt, ##args)
4054 #define netif_warn(priv, type, dev, fmt, args...) \
4055 netif_level(warn, priv, type, dev, fmt, ##args)
4056 #define netif_notice(priv, type, dev, fmt, args...) \
4057 netif_level(notice, priv, type, dev, fmt, ##args)
4058 #define netif_info(priv, type, dev, fmt, args...) \
4059 netif_level(info, priv, type, dev, fmt, ##args)
4060
4061 #if defined(CONFIG_DYNAMIC_DEBUG)
4062 #define netif_dbg(priv, type, netdev, format, args...) \
4063 do { \
4064 if (netif_msg_##type(priv)) \
4065 dynamic_netdev_dbg(netdev, format, ##args); \
4066 } while (0)
4067 #elif defined(DEBUG)
4068 #define netif_dbg(priv, type, dev, format, args...) \
4069 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4070 #else
4071 #define netif_dbg(priv, type, dev, format, args...) \
4072 ({ \
4073 if (0) \
4074 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4075 0; \
4076 })
4077 #endif
4078
4079 #if defined(VERBOSE_DEBUG)
4080 #define netif_vdbg netif_dbg
4081 #else
4082 #define netif_vdbg(priv, type, dev, format, args...) \
4083 ({ \
4084 if (0) \
4085 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4086 0; \
4087 })
4088 #endif
4089
4090 /*
4091 * The list of packet types we will receive (as opposed to discard)
4092 * and the routines to invoke.
4093 *
4094 * Why 16. Because with 16 the only overlap we get on a hash of the
4095 * low nibble of the protocol value is RARP/SNAP/X.25.
4096 *
4097 * NOTE: That is no longer true with the addition of VLAN tags. Not
4098 * sure which should go first, but I bet it won't make much
4099 * difference if we are running VLANs. The good news is that
4100 * this protocol won't be in the list unless compiled in, so
4101 * the average user (w/out VLANs) will not be adversely affected.
4102 * --BLG
4103 *
4104 * 0800 IP
4105 * 8100 802.1Q VLAN
4106 * 0001 802.3
4107 * 0002 AX.25
4108 * 0004 802.2
4109 * 8035 RARP
4110 * 0005 SNAP
4111 * 0805 X.25
4112 * 0806 ARP
4113 * 8137 IPX
4114 * 0009 Localtalk
4115 * 86DD IPv6
4116 */
4117 #define PTYPE_HASH_SIZE (16)
4118 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4119
4120 #endif /* _LINUX_NETDEVICE_H */