1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
56
57 struct netpoll_info;
58 struct device;
59 struct phy_device;
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
68
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
71
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
75
76 /*
77 * Transmit return codes: transmit return codes originate from three different
78 * namespaces:
79 *
80 * - qdisc return codes
81 * - driver transmit return codes
82 * - errno values
83 *
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
91 */
92
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
98
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
107
108 enum netdev_tx {
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
112 };
113 typedef enum netdev_tx netdev_tx_t;
114
115 /*
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118 */
dev_xmit_complete(int rc)119 static inline bool dev_xmit_complete(int rc)
120 {
121 /*
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 */
127 if (likely(rc < NET_XMIT_MASK))
128 return true;
129
130 return false;
131 }
132
133 /*
134 * Compute the worst-case header length according to the protocols
135 * used.
136 */
137
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
143 # else
144 # define LL_MAX_HEADER 96
145 # endif
146 #else
147 # define LL_MAX_HEADER 32
148 #endif
149
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
153 #else
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
155 #endif
156
157 /*
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
160 */
161
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
186 };
187
188
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
191
192 #ifdef CONFIG_RPS
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
195 #endif
196
197 struct neighbour;
198 struct neigh_parms;
199 struct sk_buff;
200
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
210 bool global_use;
211 int sync_cnt;
212 int refcount;
213 int synced;
214 struct rcu_head rcu_head;
215 };
216
217 struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220 };
221
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236
237 struct hh_cache {
238 u16 hh_len;
239 u16 __pad;
240 seqlock_t hh_lock;
241
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
249 };
250
251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
252 * Alternative is:
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
255 *
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
258 */
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263
264 struct header_ops {
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
273 bool (*validate)(const char *ll_header, unsigned int len);
274 };
275
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
278 * code.
279 */
280
281 enum netdev_state_t {
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
287 };
288
289
290 /*
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
293 */
294 struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297 };
298 #define NETDEV_BOOT_SETUP_MAX 8
299
300 int __init netdev_boot_setup(char *str);
301
302 /*
303 * Structure for NAPI scheduling similar to tasklet but with weighting
304 */
305 struct napi_struct {
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-CPU poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
311 */
312 struct list_head poll_list;
313
314 unsigned long state;
315 int weight;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
320 int poll_owner;
321 #endif
322 struct net_device *dev;
323 struct sk_buff *gro_list;
324 struct sk_buff *skb;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
329 };
330
331 enum {
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
337 };
338
339 enum gro_result {
340 GRO_MERGED,
341 GRO_MERGED_FREE,
342 GRO_HELD,
343 GRO_NORMAL,
344 GRO_DROP,
345 };
346 typedef enum gro_result gro_result_t;
347
348 /*
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * further.
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
356 *
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
359 *
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
362 *
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
366 *
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
368 * do with the skb.
369 *
370 * If the rx_handler consumed the skb in some way, it should return
371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
372 * the skb to be delivered in some other way.
373 *
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
377 *
378 * If the rx_handler decides the skb should be ignored, it should return
379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
380 * are registered on exact device (ptype->dev == skb->dev).
381 *
382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
383 * delivered, it should return RX_HANDLER_PASS.
384 *
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
387 */
388
389 enum rx_handler_result {
390 RX_HANDLER_CONSUMED,
391 RX_HANDLER_ANOTHER,
392 RX_HANDLER_EXACT,
393 RX_HANDLER_PASS,
394 };
395 typedef enum rx_handler_result rx_handler_result_t;
396 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
397
398 void __napi_schedule(struct napi_struct *n);
399 void __napi_schedule_irqoff(struct napi_struct *n);
400
napi_disable_pending(struct napi_struct * n)401 static inline bool napi_disable_pending(struct napi_struct *n)
402 {
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
404 }
405
406 /**
407 * napi_schedule_prep - check if NAPI can be scheduled
408 * @n: NAPI context
409 *
410 * Test if NAPI routine is already running, and if not mark
411 * it as running. This is used as a condition variable to
412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
414 */
napi_schedule_prep(struct napi_struct * n)415 static inline bool napi_schedule_prep(struct napi_struct *n)
416 {
417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
419 }
420
421 /**
422 * napi_schedule - schedule NAPI poll
423 * @n: NAPI context
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
napi_schedule(struct napi_struct * n)428 static inline void napi_schedule(struct napi_struct *n)
429 {
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432 }
433
434 /**
435 * napi_schedule_irqoff - schedule NAPI poll
436 * @n: NAPI context
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
napi_schedule_irqoff(struct napi_struct * n)440 static inline void napi_schedule_irqoff(struct napi_struct *n)
441 {
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444 }
445
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)447 static inline bool napi_reschedule(struct napi_struct *napi)
448 {
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
451 return true;
452 }
453 return false;
454 }
455
456 void __napi_complete(struct napi_struct *n);
457 void napi_complete_done(struct napi_struct *n, int work_done);
458 /**
459 * napi_complete - NAPI processing complete
460 * @n: NAPI context
461 *
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
464 */
napi_complete(struct napi_struct * n)465 static inline void napi_complete(struct napi_struct *n)
466 {
467 return napi_complete_done(n, 0);
468 }
469
470 /**
471 * napi_hash_add - add a NAPI to global hashtable
472 * @napi: NAPI context
473 *
474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
476 * Note: This is normally automatically done from netif_napi_add(),
477 * so might disappear in a future Linux version.
478 */
479 void napi_hash_add(struct napi_struct *napi);
480
481 /**
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: NAPI context
484 *
485 * Warning: caller must observe RCU grace period
486 * before freeing memory containing @napi, if
487 * this function returns true.
488 * Note: core networking stack automatically calls it
489 * from netif_napi_del().
490 * Drivers might want to call this helper to combine all
491 * the needed RCU grace periods into a single one.
492 */
493 bool napi_hash_del(struct napi_struct *napi);
494
495 /**
496 * napi_disable - prevent NAPI from scheduling
497 * @n: NAPI context
498 *
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
501 */
502 void napi_disable(struct napi_struct *n);
503
504 /**
505 * napi_enable - enable NAPI scheduling
506 * @n: NAPI context
507 *
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
510 */
napi_enable(struct napi_struct * n)511 static inline void napi_enable(struct napi_struct *n)
512 {
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 smp_mb__before_atomic();
515 clear_bit(NAPI_STATE_SCHED, &n->state);
516 clear_bit(NAPI_STATE_NPSVC, &n->state);
517 }
518
519 /**
520 * napi_synchronize - wait until NAPI is not running
521 * @n: NAPI context
522 *
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
526 */
napi_synchronize(const struct napi_struct * n)527 static inline void napi_synchronize(const struct napi_struct *n)
528 {
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 msleep(1);
532 else
533 barrier();
534 }
535
536 enum netdev_queue_state_t {
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
539 __QUEUE_STATE_FROZEN,
540 };
541
542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
545
546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 QUEUE_STATE_FROZEN)
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 QUEUE_STATE_FROZEN)
551
552 /*
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 */
561
562 struct netdev_queue {
563 /*
564 * read-mostly part
565 */
566 struct net_device *dev;
567 struct Qdisc __rcu *qdisc;
568 struct Qdisc *qdisc_sleeping;
569 #ifdef CONFIG_SYSFS
570 struct kobject kobj;
571 #endif
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 int numa_node;
574 #endif
575 unsigned long tx_maxrate;
576 /*
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
579 */
580 unsigned long trans_timeout;
581 /*
582 * write-mostly part
583 */
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
585 int xmit_lock_owner;
586 /*
587 * Time (in jiffies) of last Tx
588 */
589 unsigned long trans_start;
590
591 unsigned long state;
592
593 #ifdef CONFIG_BQL
594 struct dql dql;
595 #endif
596 } ____cacheline_aligned_in_smp;
597
netdev_queue_numa_node_read(const struct netdev_queue * q)598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599 {
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 return q->numa_node;
602 #else
603 return NUMA_NO_NODE;
604 #endif
605 }
606
netdev_queue_numa_node_write(struct netdev_queue * q,int node)607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608 {
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 q->numa_node = node;
611 #endif
612 }
613
614 #ifdef CONFIG_RPS
615 /*
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
618 */
619 struct rps_map {
620 unsigned int len;
621 struct rcu_head rcu;
622 u16 cpus[0];
623 };
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
625
626 /*
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
630 */
631 struct rps_dev_flow {
632 u16 cpu;
633 u16 filter;
634 unsigned int last_qtail;
635 };
636 #define RPS_NO_FILTER 0xffff
637
638 /*
639 * The rps_dev_flow_table structure contains a table of flow mappings.
640 */
641 struct rps_dev_flow_table {
642 unsigned int mask;
643 struct rcu_head rcu;
644 struct rps_dev_flow flows[0];
645 };
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647 ((_num) * sizeof(struct rps_dev_flow)))
648
649 /*
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
654 * rps_cpu_mask is used to partition the space, depending on number of
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
657 * meaning we use 32-6=26 bits for the hash.
658 */
659 struct rps_sock_flow_table {
660 u32 mask;
661
662 u32 ents[0] ____cacheline_aligned_in_smp;
663 };
664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665
666 #define RPS_NO_CPU 0xffff
667
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 u32 hash)
673 {
674 if (table && hash) {
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
677
678 /* We only give a hint, preemption can change CPU under us */
679 val |= raw_smp_processor_id();
680
681 if (table->ents[index] != val)
682 table->ents[index] = val;
683 }
684 }
685
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 u16 filter_id);
689 #endif
690 #endif /* CONFIG_RPS */
691
692 /* This structure contains an instance of an RX queue. */
693 struct netdev_rx_queue {
694 #ifdef CONFIG_RPS
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
697 #endif
698 struct kobject kobj;
699 struct net_device *dev;
700 } ____cacheline_aligned_in_smp;
701
702 /*
703 * RX queue sysfs structures and functions.
704 */
705 struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
711 };
712
713 #ifdef CONFIG_XPS
714 /*
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
717 */
718 struct xps_map {
719 unsigned int len;
720 unsigned int alloc_len;
721 struct rcu_head rcu;
722 u16 queues[0];
723 };
724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
727
728 /*
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 */
731 struct xps_dev_maps {
732 struct rcu_head rcu;
733 struct xps_map __rcu *cpu_map[0];
734 };
735 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737 #endif /* CONFIG_XPS */
738
739 #define TC_MAX_QUEUE 16
740 #define TC_BITMASK 15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
743 u16 count;
744 u16 offset;
745 };
746
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 /*
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
751 */
752 struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
759 char model[256];
760 char model_description[256];
761 };
762 #endif
763
764 #define MAX_PHYS_ITEM_ID_LEN 32
765
766 /* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
768 */
769 struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 unsigned char id_len;
772 };
773
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
776 {
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
779 }
780
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
783
784 /* These structures hold the attributes of qdisc and classifiers
785 * that are being passed to the netdevice through the setup_tc op.
786 */
787 enum {
788 TC_SETUP_MQPRIO,
789 TC_SETUP_CLSU32,
790 TC_SETUP_CLSFLOWER,
791 TC_SETUP_MATCHALL,
792 TC_SETUP_CLSBPF,
793 };
794
795 struct tc_cls_u32_offload;
796
797 struct tc_to_netdev {
798 unsigned int type;
799 union {
800 u8 tc;
801 struct tc_cls_u32_offload *cls_u32;
802 struct tc_cls_flower_offload *cls_flower;
803 struct tc_cls_matchall_offload *cls_mall;
804 struct tc_cls_bpf_offload *cls_bpf;
805 };
806 };
807
808 /* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
810 */
811 enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
818 */
819 XDP_SETUP_PROG,
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
822 */
823 XDP_QUERY_PROG,
824 };
825
826 struct netdev_xdp {
827 enum xdp_netdev_command command;
828 union {
829 /* XDP_SETUP_PROG */
830 struct bpf_prog *prog;
831 /* XDP_QUERY_PROG */
832 bool prog_attached;
833 };
834 };
835
836 /*
837 * This structure defines the management hooks for network devices.
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
840 *
841 * int (*ndo_init)(struct net_device *dev);
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
846 *
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
850 *
851 * int (*ndo_open)(struct net_device *dev);
852 * This function is called when a network device transitions to the up
853 * state.
854 *
855 * int (*ndo_stop)(struct net_device *dev);
856 * This function is called when a network device transitions to the down
857 * state.
858 *
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
861 * Called when a packet needs to be transmitted.
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
866 * Required; cannot be NULL.
867 *
868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
877 *
878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
879 * void *accel_priv, select_queue_fallback_t fallback);
880 * Called to decide which queue to use when device supports multiple
881 * transmit queues.
882 *
883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
885 * changes to configuration when multicast or promiscuous is enabled.
886 *
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
889 * If driver handles unicast address filtering, it should set
890 * IFF_UNICAST_FLT in its priv_flags.
891 *
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
894 * needs to be changed. If this interface is not defined, the
895 * MAC address can not be changed.
896 *
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
899 *
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
903 * not supported error code.
904 *
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
907 * is retained for legacy reasons; new devices should use the bus
908 * interface (PCI) for low level management.
909 *
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
912 * of a device. If not defined, any request to change MTU will
913 * will return an error.
914 *
915 * void (*ndo_tx_timeout)(struct net_device *dev);
916 * Callback used when the transmitter has not made any progress
917 * for dev->watchdog ticks.
918 *
919 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
920 * struct rtnl_link_stats64 *storage);
921 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
922 * Called when a user wants to get the network device usage
923 * statistics. Drivers must do one of the following:
924 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
925 * rtnl_link_stats64 structure passed by the caller.
926 * 2. Define @ndo_get_stats to update a net_device_stats structure
927 * (which should normally be dev->stats) and return a pointer to
928 * it. The structure may be changed asynchronously only if each
929 * field is written atomically.
930 * 3. Update dev->stats asynchronously and atomically, and define
931 * neither operation.
932 *
933 * bool (*ndo_has_offload_stats)(int attr_id)
934 * Return true if this device supports offload stats of this attr_id.
935 *
936 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
937 * void *attr_data)
938 * Get statistics for offload operations by attr_id. Write it into the
939 * attr_data pointer.
940 *
941 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
942 * If device supports VLAN filtering this function is called when a
943 * VLAN id is registered.
944 *
945 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
946 * If device supports VLAN filtering this function is called when a
947 * VLAN id is unregistered.
948 *
949 * void (*ndo_poll_controller)(struct net_device *dev);
950 *
951 * SR-IOV management functions.
952 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
953 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
954 * u8 qos, __be16 proto);
955 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
956 * int max_tx_rate);
957 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
958 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
959 * int (*ndo_get_vf_config)(struct net_device *dev,
960 * int vf, struct ifla_vf_info *ivf);
961 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
962 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
963 * struct nlattr *port[]);
964 *
965 * Enable or disable the VF ability to query its RSS Redirection Table and
966 * Hash Key. This is needed since on some devices VF share this information
967 * with PF and querying it may introduce a theoretical security risk.
968 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
969 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
970 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
971 * Called to setup 'tc' number of traffic classes in the net device. This
972 * is always called from the stack with the rtnl lock held and netif tx
973 * queues stopped. This allows the netdevice to perform queue management
974 * safely.
975 *
976 * Fiber Channel over Ethernet (FCoE) offload functions.
977 * int (*ndo_fcoe_enable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to start using LLD for FCoE
979 * so the underlying device can perform whatever needed configuration or
980 * initialization to support acceleration of FCoE traffic.
981 *
982 * int (*ndo_fcoe_disable)(struct net_device *dev);
983 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
984 * so the underlying device can perform whatever needed clean-ups to
985 * stop supporting acceleration of FCoE traffic.
986 *
987 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
988 * struct scatterlist *sgl, unsigned int sgc);
989 * Called when the FCoE Initiator wants to initialize an I/O that
990 * is a possible candidate for Direct Data Placement (DDP). The LLD can
991 * perform necessary setup and returns 1 to indicate the device is set up
992 * successfully to perform DDP on this I/O, otherwise this returns 0.
993 *
994 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
995 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
996 * indicated by the FC exchange id 'xid', so the underlying device can
997 * clean up and reuse resources for later DDP requests.
998 *
999 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1000 * struct scatterlist *sgl, unsigned int sgc);
1001 * Called when the FCoE Target wants to initialize an I/O that
1002 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1003 * perform necessary setup and returns 1 to indicate the device is set up
1004 * successfully to perform DDP on this I/O, otherwise this returns 0.
1005 *
1006 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1007 * struct netdev_fcoe_hbainfo *hbainfo);
1008 * Called when the FCoE Protocol stack wants information on the underlying
1009 * device. This information is utilized by the FCoE protocol stack to
1010 * register attributes with Fiber Channel management service as per the
1011 * FC-GS Fabric Device Management Information(FDMI) specification.
1012 *
1013 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1014 * Called when the underlying device wants to override default World Wide
1015 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1016 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1017 * protocol stack to use.
1018 *
1019 * RFS acceleration.
1020 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1021 * u16 rxq_index, u32 flow_id);
1022 * Set hardware filter for RFS. rxq_index is the target queue index;
1023 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1024 * Return the filter ID on success, or a negative error code.
1025 *
1026 * Slave management functions (for bridge, bonding, etc).
1027 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1028 * Called to make another netdev an underling.
1029 *
1030 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1031 * Called to release previously enslaved netdev.
1032 *
1033 * Feature/offload setting functions.
1034 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1035 * netdev_features_t features);
1036 * Adjusts the requested feature flags according to device-specific
1037 * constraints, and returns the resulting flags. Must not modify
1038 * the device state.
1039 *
1040 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1041 * Called to update device configuration to new features. Passed
1042 * feature set might be less than what was returned by ndo_fix_features()).
1043 * Must return >0 or -errno if it changed dev->features itself.
1044 *
1045 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1046 * struct net_device *dev,
1047 * const unsigned char *addr, u16 vid, u16 flags)
1048 * Adds an FDB entry to dev for addr.
1049 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1050 * struct net_device *dev,
1051 * const unsigned char *addr, u16 vid)
1052 * Deletes the FDB entry from dev coresponding to addr.
1053 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1054 * struct net_device *dev, struct net_device *filter_dev,
1055 * int *idx)
1056 * Used to add FDB entries to dump requests. Implementers should add
1057 * entries to skb and update idx with the number of entries.
1058 *
1059 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1060 * u16 flags)
1061 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1062 * struct net_device *dev, u32 filter_mask,
1063 * int nlflags)
1064 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1065 * u16 flags);
1066 *
1067 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1068 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1069 * which do not represent real hardware may define this to allow their
1070 * userspace components to manage their virtual carrier state. Devices
1071 * that determine carrier state from physical hardware properties (eg
1072 * network cables) or protocol-dependent mechanisms (eg
1073 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1074 *
1075 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1076 * struct netdev_phys_item_id *ppid);
1077 * Called to get ID of physical port of this device. If driver does
1078 * not implement this, it is assumed that the hw is not able to have
1079 * multiple net devices on single physical port.
1080 *
1081 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1082 * struct udp_tunnel_info *ti);
1083 * Called by UDP tunnel to notify a driver about the UDP port and socket
1084 * address family that a UDP tunnel is listnening to. It is called only
1085 * when a new port starts listening. The operation is protected by the
1086 * RTNL.
1087 *
1088 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1089 * struct udp_tunnel_info *ti);
1090 * Called by UDP tunnel to notify the driver about a UDP port and socket
1091 * address family that the UDP tunnel is not listening to anymore. The
1092 * operation is protected by the RTNL.
1093 *
1094 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1095 * struct net_device *dev)
1096 * Called by upper layer devices to accelerate switching or other
1097 * station functionality into hardware. 'pdev is the lowerdev
1098 * to use for the offload and 'dev' is the net device that will
1099 * back the offload. Returns a pointer to the private structure
1100 * the upper layer will maintain.
1101 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1102 * Called by upper layer device to delete the station created
1103 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1104 * the station and priv is the structure returned by the add
1105 * operation.
1106 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1107 * struct net_device *dev,
1108 * void *priv);
1109 * Callback to use for xmit over the accelerated station. This
1110 * is used in place of ndo_start_xmit on accelerated net
1111 * devices.
1112 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1113 * int queue_index, u32 maxrate);
1114 * Called when a user wants to set a max-rate limitation of specific
1115 * TX queue.
1116 * int (*ndo_get_iflink)(const struct net_device *dev);
1117 * Called to get the iflink value of this device.
1118 * void (*ndo_change_proto_down)(struct net_device *dev,
1119 * bool proto_down);
1120 * This function is used to pass protocol port error state information
1121 * to the switch driver. The switch driver can react to the proto_down
1122 * by doing a phys down on the associated switch port.
1123 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1124 * This function is used to get egress tunnel information for given skb.
1125 * This is useful for retrieving outer tunnel header parameters while
1126 * sampling packet.
1127 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1128 * This function is used to specify the headroom that the skb must
1129 * consider when allocation skb during packet reception. Setting
1130 * appropriate rx headroom value allows avoiding skb head copy on
1131 * forward. Setting a negative value resets the rx headroom to the
1132 * default value.
1133 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1134 * This function is used to set or query state related to XDP on the
1135 * netdevice. See definition of enum xdp_netdev_command for details.
1136 *
1137 */
1138 struct net_device_ops {
1139 int (*ndo_init)(struct net_device *dev);
1140 void (*ndo_uninit)(struct net_device *dev);
1141 int (*ndo_open)(struct net_device *dev);
1142 int (*ndo_stop)(struct net_device *dev);
1143 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1144 struct net_device *dev);
1145 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1146 struct net_device *dev,
1147 netdev_features_t features);
1148 u16 (*ndo_select_queue)(struct net_device *dev,
1149 struct sk_buff *skb,
1150 void *accel_priv,
1151 select_queue_fallback_t fallback);
1152 void (*ndo_change_rx_flags)(struct net_device *dev,
1153 int flags);
1154 void (*ndo_set_rx_mode)(struct net_device *dev);
1155 int (*ndo_set_mac_address)(struct net_device *dev,
1156 void *addr);
1157 int (*ndo_validate_addr)(struct net_device *dev);
1158 int (*ndo_do_ioctl)(struct net_device *dev,
1159 struct ifreq *ifr, int cmd);
1160 int (*ndo_set_config)(struct net_device *dev,
1161 struct ifmap *map);
1162 int (*ndo_change_mtu)(struct net_device *dev,
1163 int new_mtu);
1164 int (*ndo_neigh_setup)(struct net_device *dev,
1165 struct neigh_parms *);
1166 void (*ndo_tx_timeout) (struct net_device *dev);
1167
1168 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1169 struct rtnl_link_stats64 *storage);
1170 bool (*ndo_has_offload_stats)(int attr_id);
1171 int (*ndo_get_offload_stats)(int attr_id,
1172 const struct net_device *dev,
1173 void *attr_data);
1174 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1175
1176 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1177 __be16 proto, u16 vid);
1178 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1179 __be16 proto, u16 vid);
1180 #ifdef CONFIG_NET_POLL_CONTROLLER
1181 void (*ndo_poll_controller)(struct net_device *dev);
1182 int (*ndo_netpoll_setup)(struct net_device *dev,
1183 struct netpoll_info *info);
1184 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1185 #endif
1186 #ifdef CONFIG_NET_RX_BUSY_POLL
1187 int (*ndo_busy_poll)(struct napi_struct *dev);
1188 #endif
1189 int (*ndo_set_vf_mac)(struct net_device *dev,
1190 int queue, u8 *mac);
1191 int (*ndo_set_vf_vlan)(struct net_device *dev,
1192 int queue, u16 vlan,
1193 u8 qos, __be16 proto);
1194 int (*ndo_set_vf_rate)(struct net_device *dev,
1195 int vf, int min_tx_rate,
1196 int max_tx_rate);
1197 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1198 int vf, bool setting);
1199 int (*ndo_set_vf_trust)(struct net_device *dev,
1200 int vf, bool setting);
1201 int (*ndo_get_vf_config)(struct net_device *dev,
1202 int vf,
1203 struct ifla_vf_info *ivf);
1204 int (*ndo_set_vf_link_state)(struct net_device *dev,
1205 int vf, int link_state);
1206 int (*ndo_get_vf_stats)(struct net_device *dev,
1207 int vf,
1208 struct ifla_vf_stats
1209 *vf_stats);
1210 int (*ndo_set_vf_port)(struct net_device *dev,
1211 int vf,
1212 struct nlattr *port[]);
1213 int (*ndo_get_vf_port)(struct net_device *dev,
1214 int vf, struct sk_buff *skb);
1215 int (*ndo_set_vf_guid)(struct net_device *dev,
1216 int vf, u64 guid,
1217 int guid_type);
1218 int (*ndo_set_vf_rss_query_en)(
1219 struct net_device *dev,
1220 int vf, bool setting);
1221 int (*ndo_setup_tc)(struct net_device *dev,
1222 u32 handle,
1223 __be16 protocol,
1224 struct tc_to_netdev *tc);
1225 #if IS_ENABLED(CONFIG_FCOE)
1226 int (*ndo_fcoe_enable)(struct net_device *dev);
1227 int (*ndo_fcoe_disable)(struct net_device *dev);
1228 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1229 u16 xid,
1230 struct scatterlist *sgl,
1231 unsigned int sgc);
1232 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1233 u16 xid);
1234 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1235 u16 xid,
1236 struct scatterlist *sgl,
1237 unsigned int sgc);
1238 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1239 struct netdev_fcoe_hbainfo *hbainfo);
1240 #endif
1241
1242 #if IS_ENABLED(CONFIG_LIBFCOE)
1243 #define NETDEV_FCOE_WWNN 0
1244 #define NETDEV_FCOE_WWPN 1
1245 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1246 u64 *wwn, int type);
1247 #endif
1248
1249 #ifdef CONFIG_RFS_ACCEL
1250 int (*ndo_rx_flow_steer)(struct net_device *dev,
1251 const struct sk_buff *skb,
1252 u16 rxq_index,
1253 u32 flow_id);
1254 #endif
1255 int (*ndo_add_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
1257 int (*ndo_del_slave)(struct net_device *dev,
1258 struct net_device *slave_dev);
1259 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1260 netdev_features_t features);
1261 int (*ndo_set_features)(struct net_device *dev,
1262 netdev_features_t features);
1263 int (*ndo_neigh_construct)(struct net_device *dev,
1264 struct neighbour *n);
1265 void (*ndo_neigh_destroy)(struct net_device *dev,
1266 struct neighbour *n);
1267
1268 int (*ndo_fdb_add)(struct ndmsg *ndm,
1269 struct nlattr *tb[],
1270 struct net_device *dev,
1271 const unsigned char *addr,
1272 u16 vid,
1273 u16 flags);
1274 int (*ndo_fdb_del)(struct ndmsg *ndm,
1275 struct nlattr *tb[],
1276 struct net_device *dev,
1277 const unsigned char *addr,
1278 u16 vid);
1279 int (*ndo_fdb_dump)(struct sk_buff *skb,
1280 struct netlink_callback *cb,
1281 struct net_device *dev,
1282 struct net_device *filter_dev,
1283 int *idx);
1284
1285 int (*ndo_bridge_setlink)(struct net_device *dev,
1286 struct nlmsghdr *nlh,
1287 u16 flags);
1288 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1289 u32 pid, u32 seq,
1290 struct net_device *dev,
1291 u32 filter_mask,
1292 int nlflags);
1293 int (*ndo_bridge_dellink)(struct net_device *dev,
1294 struct nlmsghdr *nlh,
1295 u16 flags);
1296 int (*ndo_change_carrier)(struct net_device *dev,
1297 bool new_carrier);
1298 int (*ndo_get_phys_port_id)(struct net_device *dev,
1299 struct netdev_phys_item_id *ppid);
1300 int (*ndo_get_phys_port_name)(struct net_device *dev,
1301 char *name, size_t len);
1302 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
1304 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1305 struct udp_tunnel_info *ti);
1306 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1307 struct net_device *dev);
1308 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1309 void *priv);
1310
1311 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1312 struct net_device *dev,
1313 void *priv);
1314 int (*ndo_get_lock_subclass)(struct net_device *dev);
1315 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1316 int queue_index,
1317 u32 maxrate);
1318 int (*ndo_get_iflink)(const struct net_device *dev);
1319 int (*ndo_change_proto_down)(struct net_device *dev,
1320 bool proto_down);
1321 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1322 struct sk_buff *skb);
1323 void (*ndo_set_rx_headroom)(struct net_device *dev,
1324 int needed_headroom);
1325 int (*ndo_xdp)(struct net_device *dev,
1326 struct netdev_xdp *xdp);
1327 };
1328
1329 /**
1330 * enum net_device_priv_flags - &struct net_device priv_flags
1331 *
1332 * These are the &struct net_device, they are only set internally
1333 * by drivers and used in the kernel. These flags are invisible to
1334 * userspace; this means that the order of these flags can change
1335 * during any kernel release.
1336 *
1337 * You should have a pretty good reason to be extending these flags.
1338 *
1339 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1340 * @IFF_EBRIDGE: Ethernet bridging device
1341 * @IFF_BONDING: bonding master or slave
1342 * @IFF_ISATAP: ISATAP interface (RFC4214)
1343 * @IFF_WAN_HDLC: WAN HDLC device
1344 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1345 * release skb->dst
1346 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1347 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1348 * @IFF_MACVLAN_PORT: device used as macvlan port
1349 * @IFF_BRIDGE_PORT: device used as bridge port
1350 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1351 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1352 * @IFF_UNICAST_FLT: Supports unicast filtering
1353 * @IFF_TEAM_PORT: device used as team port
1354 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1355 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1356 * change when it's running
1357 * @IFF_MACVLAN: Macvlan device
1358 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1359 * underlying stacked devices
1360 * @IFF_IPVLAN_MASTER: IPvlan master device
1361 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1362 * @IFF_L3MDEV_MASTER: device is an L3 master device
1363 * @IFF_NO_QUEUE: device can run without qdisc attached
1364 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1365 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1366 * @IFF_TEAM: device is a team device
1367 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1368 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1369 * entity (i.e. the master device for bridged veth)
1370 * @IFF_MACSEC: device is a MACsec device
1371 */
1372 enum netdev_priv_flags {
1373 IFF_802_1Q_VLAN = 1<<0,
1374 IFF_EBRIDGE = 1<<1,
1375 IFF_BONDING = 1<<2,
1376 IFF_ISATAP = 1<<3,
1377 IFF_WAN_HDLC = 1<<4,
1378 IFF_XMIT_DST_RELEASE = 1<<5,
1379 IFF_DONT_BRIDGE = 1<<6,
1380 IFF_DISABLE_NETPOLL = 1<<7,
1381 IFF_MACVLAN_PORT = 1<<8,
1382 IFF_BRIDGE_PORT = 1<<9,
1383 IFF_OVS_DATAPATH = 1<<10,
1384 IFF_TX_SKB_SHARING = 1<<11,
1385 IFF_UNICAST_FLT = 1<<12,
1386 IFF_TEAM_PORT = 1<<13,
1387 IFF_SUPP_NOFCS = 1<<14,
1388 IFF_LIVE_ADDR_CHANGE = 1<<15,
1389 IFF_MACVLAN = 1<<16,
1390 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1391 IFF_IPVLAN_MASTER = 1<<18,
1392 IFF_IPVLAN_SLAVE = 1<<19,
1393 IFF_L3MDEV_MASTER = 1<<20,
1394 IFF_NO_QUEUE = 1<<21,
1395 IFF_OPENVSWITCH = 1<<22,
1396 IFF_L3MDEV_SLAVE = 1<<23,
1397 IFF_TEAM = 1<<24,
1398 IFF_RXFH_CONFIGURED = 1<<25,
1399 IFF_PHONY_HEADROOM = 1<<26,
1400 IFF_MACSEC = 1<<27,
1401 };
1402
1403 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1404 #define IFF_EBRIDGE IFF_EBRIDGE
1405 #define IFF_BONDING IFF_BONDING
1406 #define IFF_ISATAP IFF_ISATAP
1407 #define IFF_WAN_HDLC IFF_WAN_HDLC
1408 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1409 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1410 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1411 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1412 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1413 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1414 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1415 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1416 #define IFF_TEAM_PORT IFF_TEAM_PORT
1417 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1418 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1419 #define IFF_MACVLAN IFF_MACVLAN
1420 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1421 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1422 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1423 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1424 #define IFF_NO_QUEUE IFF_NO_QUEUE
1425 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1426 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1427 #define IFF_TEAM IFF_TEAM
1428 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1429 #define IFF_MACSEC IFF_MACSEC
1430
1431 /**
1432 * struct net_device - The DEVICE structure.
1433 * Actually, this whole structure is a big mistake. It mixes I/O
1434 * data with strictly "high-level" data, and it has to know about
1435 * almost every data structure used in the INET module.
1436 *
1437 * @name: This is the first field of the "visible" part of this structure
1438 * (i.e. as seen by users in the "Space.c" file). It is the name
1439 * of the interface.
1440 *
1441 * @name_hlist: Device name hash chain, please keep it close to name[]
1442 * @ifalias: SNMP alias
1443 * @mem_end: Shared memory end
1444 * @mem_start: Shared memory start
1445 * @base_addr: Device I/O address
1446 * @irq: Device IRQ number
1447 *
1448 * @carrier_changes: Stats to monitor carrier on<->off transitions
1449 *
1450 * @state: Generic network queuing layer state, see netdev_state_t
1451 * @dev_list: The global list of network devices
1452 * @napi_list: List entry used for polling NAPI devices
1453 * @unreg_list: List entry when we are unregistering the
1454 * device; see the function unregister_netdev
1455 * @close_list: List entry used when we are closing the device
1456 * @ptype_all: Device-specific packet handlers for all protocols
1457 * @ptype_specific: Device-specific, protocol-specific packet handlers
1458 *
1459 * @adj_list: Directly linked devices, like slaves for bonding
1460 * @all_adj_list: All linked devices, *including* neighbours
1461 * @features: Currently active device features
1462 * @hw_features: User-changeable features
1463 *
1464 * @wanted_features: User-requested features
1465 * @vlan_features: Mask of features inheritable by VLAN devices
1466 *
1467 * @hw_enc_features: Mask of features inherited by encapsulating devices
1468 * This field indicates what encapsulation
1469 * offloads the hardware is capable of doing,
1470 * and drivers will need to set them appropriately.
1471 *
1472 * @mpls_features: Mask of features inheritable by MPLS
1473 *
1474 * @ifindex: interface index
1475 * @group: The group the device belongs to
1476 *
1477 * @stats: Statistics struct, which was left as a legacy, use
1478 * rtnl_link_stats64 instead
1479 *
1480 * @rx_dropped: Dropped packets by core network,
1481 * do not use this in drivers
1482 * @tx_dropped: Dropped packets by core network,
1483 * do not use this in drivers
1484 * @rx_nohandler: nohandler dropped packets by core network on
1485 * inactive devices, do not use this in drivers
1486 *
1487 * @wireless_handlers: List of functions to handle Wireless Extensions,
1488 * instead of ioctl,
1489 * see <net/iw_handler.h> for details.
1490 * @wireless_data: Instance data managed by the core of wireless extensions
1491 *
1492 * @netdev_ops: Includes several pointers to callbacks,
1493 * if one wants to override the ndo_*() functions
1494 * @ethtool_ops: Management operations
1495 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1496 * discovery handling. Necessary for e.g. 6LoWPAN.
1497 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1498 * of Layer 2 headers.
1499 *
1500 * @flags: Interface flags (a la BSD)
1501 * @priv_flags: Like 'flags' but invisible to userspace,
1502 * see if.h for the definitions
1503 * @gflags: Global flags ( kept as legacy )
1504 * @padded: How much padding added by alloc_netdev()
1505 * @operstate: RFC2863 operstate
1506 * @link_mode: Mapping policy to operstate
1507 * @if_port: Selectable AUI, TP, ...
1508 * @dma: DMA channel
1509 * @mtu: Interface MTU value
1510 * @type: Interface hardware type
1511 * @hard_header_len: Maximum hardware header length.
1512 * @min_header_len: Minimum hardware header length
1513 *
1514 * @needed_headroom: Extra headroom the hardware may need, but not in all
1515 * cases can this be guaranteed
1516 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1517 * cases can this be guaranteed. Some cases also use
1518 * LL_MAX_HEADER instead to allocate the skb
1519 *
1520 * interface address info:
1521 *
1522 * @perm_addr: Permanent hw address
1523 * @addr_assign_type: Hw address assignment type
1524 * @addr_len: Hardware address length
1525 * @neigh_priv_len: Used in neigh_alloc()
1526 * @dev_id: Used to differentiate devices that share
1527 * the same link layer address
1528 * @dev_port: Used to differentiate devices that share
1529 * the same function
1530 * @addr_list_lock: XXX: need comments on this one
1531 * @uc_promisc: Counter that indicates promiscuous mode
1532 * has been enabled due to the need to listen to
1533 * additional unicast addresses in a device that
1534 * does not implement ndo_set_rx_mode()
1535 * @uc: unicast mac addresses
1536 * @mc: multicast mac addresses
1537 * @dev_addrs: list of device hw addresses
1538 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1539 * @promiscuity: Number of times the NIC is told to work in
1540 * promiscuous mode; if it becomes 0 the NIC will
1541 * exit promiscuous mode
1542 * @allmulti: Counter, enables or disables allmulticast mode
1543 *
1544 * @vlan_info: VLAN info
1545 * @dsa_ptr: dsa specific data
1546 * @tipc_ptr: TIPC specific data
1547 * @atalk_ptr: AppleTalk link
1548 * @ip_ptr: IPv4 specific data
1549 * @dn_ptr: DECnet specific data
1550 * @ip6_ptr: IPv6 specific data
1551 * @ax25_ptr: AX.25 specific data
1552 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1553 *
1554 * @last_rx: Time of last Rx
1555 * @dev_addr: Hw address (before bcast,
1556 * because most packets are unicast)
1557 *
1558 * @_rx: Array of RX queues
1559 * @num_rx_queues: Number of RX queues
1560 * allocated at register_netdev() time
1561 * @real_num_rx_queues: Number of RX queues currently active in device
1562 *
1563 * @rx_handler: handler for received packets
1564 * @rx_handler_data: XXX: need comments on this one
1565 * @ingress_queue: XXX: need comments on this one
1566 * @broadcast: hw bcast address
1567 *
1568 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1569 * indexed by RX queue number. Assigned by driver.
1570 * This must only be set if the ndo_rx_flow_steer
1571 * operation is defined
1572 * @index_hlist: Device index hash chain
1573 *
1574 * @_tx: Array of TX queues
1575 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1576 * @real_num_tx_queues: Number of TX queues currently active in device
1577 * @qdisc: Root qdisc from userspace point of view
1578 * @tx_queue_len: Max frames per queue allowed
1579 * @tx_global_lock: XXX: need comments on this one
1580 *
1581 * @xps_maps: XXX: need comments on this one
1582 *
1583 * @watchdog_timeo: Represents the timeout that is used by
1584 * the watchdog (see dev_watchdog())
1585 * @watchdog_timer: List of timers
1586 *
1587 * @pcpu_refcnt: Number of references to this device
1588 * @todo_list: Delayed register/unregister
1589 * @link_watch_list: XXX: need comments on this one
1590 *
1591 * @reg_state: Register/unregister state machine
1592 * @dismantle: Device is going to be freed
1593 * @rtnl_link_state: This enum represents the phases of creating
1594 * a new link
1595 *
1596 * @destructor: Called from unregister,
1597 * can be used to call free_netdev
1598 * @npinfo: XXX: need comments on this one
1599 * @nd_net: Network namespace this network device is inside
1600 *
1601 * @ml_priv: Mid-layer private
1602 * @lstats: Loopback statistics
1603 * @tstats: Tunnel statistics
1604 * @dstats: Dummy statistics
1605 * @vstats: Virtual ethernet statistics
1606 *
1607 * @garp_port: GARP
1608 * @mrp_port: MRP
1609 *
1610 * @dev: Class/net/name entry
1611 * @sysfs_groups: Space for optional device, statistics and wireless
1612 * sysfs groups
1613 *
1614 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1615 * @rtnl_link_ops: Rtnl_link_ops
1616 *
1617 * @gso_max_size: Maximum size of generic segmentation offload
1618 * @gso_max_segs: Maximum number of segments that can be passed to the
1619 * NIC for GSO
1620 *
1621 * @dcbnl_ops: Data Center Bridging netlink ops
1622 * @num_tc: Number of traffic classes in the net device
1623 * @tc_to_txq: XXX: need comments on this one
1624 * @prio_tc_map: XXX: need comments on this one
1625 *
1626 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1627 *
1628 * @priomap: XXX: need comments on this one
1629 * @phydev: Physical device may attach itself
1630 * for hardware timestamping
1631 *
1632 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1633 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1634 *
1635 * @proto_down: protocol port state information can be sent to the
1636 * switch driver and used to set the phys state of the
1637 * switch port.
1638 *
1639 * FIXME: cleanup struct net_device such that network protocol info
1640 * moves out.
1641 */
1642
1643 struct net_device {
1644 char name[IFNAMSIZ];
1645 struct hlist_node name_hlist;
1646 char *ifalias;
1647 /*
1648 * I/O specific fields
1649 * FIXME: Merge these and struct ifmap into one
1650 */
1651 unsigned long mem_end;
1652 unsigned long mem_start;
1653 unsigned long base_addr;
1654 int irq;
1655
1656 atomic_t carrier_changes;
1657
1658 /*
1659 * Some hardware also needs these fields (state,dev_list,
1660 * napi_list,unreg_list,close_list) but they are not
1661 * part of the usual set specified in Space.c.
1662 */
1663
1664 unsigned long state;
1665
1666 struct list_head dev_list;
1667 struct list_head napi_list;
1668 struct list_head unreg_list;
1669 struct list_head close_list;
1670 struct list_head ptype_all;
1671 struct list_head ptype_specific;
1672
1673 struct {
1674 struct list_head upper;
1675 struct list_head lower;
1676 } adj_list;
1677
1678 struct {
1679 struct list_head upper;
1680 struct list_head lower;
1681 } all_adj_list;
1682
1683 netdev_features_t features;
1684 netdev_features_t hw_features;
1685 netdev_features_t wanted_features;
1686 netdev_features_t vlan_features;
1687 netdev_features_t hw_enc_features;
1688 netdev_features_t mpls_features;
1689 netdev_features_t gso_partial_features;
1690
1691 int ifindex;
1692 int group;
1693
1694 struct net_device_stats stats;
1695
1696 atomic_long_t rx_dropped;
1697 atomic_long_t tx_dropped;
1698 atomic_long_t rx_nohandler;
1699
1700 #ifdef CONFIG_WIRELESS_EXT
1701 const struct iw_handler_def *wireless_handlers;
1702 struct iw_public_data *wireless_data;
1703 #endif
1704 const struct net_device_ops *netdev_ops;
1705 const struct ethtool_ops *ethtool_ops;
1706 #ifdef CONFIG_NET_SWITCHDEV
1707 const struct switchdev_ops *switchdev_ops;
1708 #endif
1709 #ifdef CONFIG_NET_L3_MASTER_DEV
1710 const struct l3mdev_ops *l3mdev_ops;
1711 #endif
1712 #if IS_ENABLED(CONFIG_IPV6)
1713 const struct ndisc_ops *ndisc_ops;
1714 #endif
1715
1716 const struct header_ops *header_ops;
1717
1718 unsigned int flags;
1719 unsigned int priv_flags;
1720
1721 unsigned short gflags;
1722 unsigned short padded;
1723
1724 unsigned char operstate;
1725 unsigned char link_mode;
1726
1727 unsigned char if_port;
1728 unsigned char dma;
1729
1730 unsigned int mtu;
1731 unsigned short type;
1732 unsigned short hard_header_len;
1733 unsigned short min_header_len;
1734
1735 unsigned short needed_headroom;
1736 unsigned short needed_tailroom;
1737
1738 /* Interface address info. */
1739 unsigned char perm_addr[MAX_ADDR_LEN];
1740 unsigned char addr_assign_type;
1741 unsigned char addr_len;
1742 unsigned short neigh_priv_len;
1743 unsigned short dev_id;
1744 unsigned short dev_port;
1745 spinlock_t addr_list_lock;
1746 unsigned char name_assign_type;
1747 bool uc_promisc;
1748 struct netdev_hw_addr_list uc;
1749 struct netdev_hw_addr_list mc;
1750 struct netdev_hw_addr_list dev_addrs;
1751
1752 #ifdef CONFIG_SYSFS
1753 struct kset *queues_kset;
1754 #endif
1755 unsigned int promiscuity;
1756 unsigned int allmulti;
1757
1758
1759 /* Protocol-specific pointers */
1760
1761 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1762 struct vlan_info __rcu *vlan_info;
1763 #endif
1764 #if IS_ENABLED(CONFIG_NET_DSA)
1765 struct dsa_switch_tree *dsa_ptr;
1766 #endif
1767 #if IS_ENABLED(CONFIG_TIPC)
1768 struct tipc_bearer __rcu *tipc_ptr;
1769 #endif
1770 void *atalk_ptr;
1771 struct in_device __rcu *ip_ptr;
1772 struct dn_dev __rcu *dn_ptr;
1773 struct inet6_dev __rcu *ip6_ptr;
1774 void *ax25_ptr;
1775 struct wireless_dev *ieee80211_ptr;
1776 struct wpan_dev *ieee802154_ptr;
1777 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1778 struct mpls_dev __rcu *mpls_ptr;
1779 #endif
1780
1781 /*
1782 * Cache lines mostly used on receive path (including eth_type_trans())
1783 */
1784 unsigned long last_rx;
1785
1786 /* Interface address info used in eth_type_trans() */
1787 unsigned char *dev_addr;
1788
1789 #ifdef CONFIG_SYSFS
1790 struct netdev_rx_queue *_rx;
1791
1792 unsigned int num_rx_queues;
1793 unsigned int real_num_rx_queues;
1794 #endif
1795
1796 unsigned long gro_flush_timeout;
1797 rx_handler_func_t __rcu *rx_handler;
1798 void __rcu *rx_handler_data;
1799
1800 #ifdef CONFIG_NET_CLS_ACT
1801 struct tcf_proto __rcu *ingress_cl_list;
1802 #endif
1803 struct netdev_queue __rcu *ingress_queue;
1804 #ifdef CONFIG_NETFILTER_INGRESS
1805 struct nf_hook_entry __rcu *nf_hooks_ingress;
1806 #endif
1807
1808 unsigned char broadcast[MAX_ADDR_LEN];
1809 #ifdef CONFIG_RFS_ACCEL
1810 struct cpu_rmap *rx_cpu_rmap;
1811 #endif
1812 struct hlist_node index_hlist;
1813
1814 /*
1815 * Cache lines mostly used on transmit path
1816 */
1817 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1818 unsigned int num_tx_queues;
1819 unsigned int real_num_tx_queues;
1820 struct Qdisc *qdisc;
1821 #ifdef CONFIG_NET_SCHED
1822 DECLARE_HASHTABLE (qdisc_hash, 4);
1823 #endif
1824 unsigned long tx_queue_len;
1825 spinlock_t tx_global_lock;
1826 int watchdog_timeo;
1827
1828 #ifdef CONFIG_XPS
1829 struct xps_dev_maps __rcu *xps_maps;
1830 #endif
1831 #ifdef CONFIG_NET_CLS_ACT
1832 struct tcf_proto __rcu *egress_cl_list;
1833 #endif
1834
1835 /* These may be needed for future network-power-down code. */
1836 struct timer_list watchdog_timer;
1837
1838 int __percpu *pcpu_refcnt;
1839 struct list_head todo_list;
1840
1841 struct list_head link_watch_list;
1842
1843 enum { NETREG_UNINITIALIZED=0,
1844 NETREG_REGISTERED, /* completed register_netdevice */
1845 NETREG_UNREGISTERING, /* called unregister_netdevice */
1846 NETREG_UNREGISTERED, /* completed unregister todo */
1847 NETREG_RELEASED, /* called free_netdev */
1848 NETREG_DUMMY, /* dummy device for NAPI poll */
1849 } reg_state:8;
1850
1851 bool dismantle;
1852
1853 enum {
1854 RTNL_LINK_INITIALIZED,
1855 RTNL_LINK_INITIALIZING,
1856 } rtnl_link_state:16;
1857
1858 void (*destructor)(struct net_device *dev);
1859
1860 #ifdef CONFIG_NETPOLL
1861 struct netpoll_info __rcu *npinfo;
1862 #endif
1863
1864 possible_net_t nd_net;
1865
1866 /* mid-layer private */
1867 union {
1868 void *ml_priv;
1869 struct pcpu_lstats __percpu *lstats;
1870 struct pcpu_sw_netstats __percpu *tstats;
1871 struct pcpu_dstats __percpu *dstats;
1872 struct pcpu_vstats __percpu *vstats;
1873 };
1874
1875 struct garp_port __rcu *garp_port;
1876 struct mrp_port __rcu *mrp_port;
1877
1878 struct device dev;
1879 const struct attribute_group *sysfs_groups[4];
1880 const struct attribute_group *sysfs_rx_queue_group;
1881
1882 const struct rtnl_link_ops *rtnl_link_ops;
1883
1884 /* for setting kernel sock attribute on TCP connection setup */
1885 #define GSO_MAX_SIZE 65536
1886 unsigned int gso_max_size;
1887 #define GSO_MAX_SEGS 65535
1888 u16 gso_max_segs;
1889
1890 #ifdef CONFIG_DCB
1891 const struct dcbnl_rtnl_ops *dcbnl_ops;
1892 #endif
1893 u8 num_tc;
1894 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1895 u8 prio_tc_map[TC_BITMASK + 1];
1896
1897 #if IS_ENABLED(CONFIG_FCOE)
1898 unsigned int fcoe_ddp_xid;
1899 #endif
1900 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1901 struct netprio_map __rcu *priomap;
1902 #endif
1903 struct phy_device *phydev;
1904 struct lock_class_key *qdisc_tx_busylock;
1905 struct lock_class_key *qdisc_running_key;
1906 bool proto_down;
1907 };
1908 #define to_net_dev(d) container_of(d, struct net_device, dev)
1909
1910 #define NETDEV_ALIGN 32
1911
1912 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)1913 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1914 {
1915 return dev->prio_tc_map[prio & TC_BITMASK];
1916 }
1917
1918 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)1919 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1920 {
1921 if (tc >= dev->num_tc)
1922 return -EINVAL;
1923
1924 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1925 return 0;
1926 }
1927
1928 static inline
netdev_reset_tc(struct net_device * dev)1929 void netdev_reset_tc(struct net_device *dev)
1930 {
1931 dev->num_tc = 0;
1932 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1933 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1934 }
1935
1936 static inline
netdev_set_tc_queue(struct net_device * dev,u8 tc,u16 count,u16 offset)1937 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1938 {
1939 if (tc >= dev->num_tc)
1940 return -EINVAL;
1941
1942 dev->tc_to_txq[tc].count = count;
1943 dev->tc_to_txq[tc].offset = offset;
1944 return 0;
1945 }
1946
1947 static inline
netdev_set_num_tc(struct net_device * dev,u8 num_tc)1948 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1949 {
1950 if (num_tc > TC_MAX_QUEUE)
1951 return -EINVAL;
1952
1953 dev->num_tc = num_tc;
1954 return 0;
1955 }
1956
1957 static inline
netdev_get_num_tc(struct net_device * dev)1958 int netdev_get_num_tc(struct net_device *dev)
1959 {
1960 return dev->num_tc;
1961 }
1962
1963 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)1964 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1965 unsigned int index)
1966 {
1967 return &dev->_tx[index];
1968 }
1969
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)1970 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1971 const struct sk_buff *skb)
1972 {
1973 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1974 }
1975
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)1976 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1977 void (*f)(struct net_device *,
1978 struct netdev_queue *,
1979 void *),
1980 void *arg)
1981 {
1982 unsigned int i;
1983
1984 for (i = 0; i < dev->num_tx_queues; i++)
1985 f(dev, &dev->_tx[i], arg);
1986 }
1987
1988 #define netdev_lockdep_set_classes(dev) \
1989 { \
1990 static struct lock_class_key qdisc_tx_busylock_key; \
1991 static struct lock_class_key qdisc_running_key; \
1992 static struct lock_class_key qdisc_xmit_lock_key; \
1993 static struct lock_class_key dev_addr_list_lock_key; \
1994 unsigned int i; \
1995 \
1996 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1997 (dev)->qdisc_running_key = &qdisc_running_key; \
1998 lockdep_set_class(&(dev)->addr_list_lock, \
1999 &dev_addr_list_lock_key); \
2000 for (i = 0; i < (dev)->num_tx_queues; i++) \
2001 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2002 &qdisc_xmit_lock_key); \
2003 }
2004
2005 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2006 struct sk_buff *skb,
2007 void *accel_priv);
2008
2009 /* returns the headroom that the master device needs to take in account
2010 * when forwarding to this dev
2011 */
netdev_get_fwd_headroom(struct net_device * dev)2012 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2013 {
2014 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2015 }
2016
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2017 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2018 {
2019 if (dev->netdev_ops->ndo_set_rx_headroom)
2020 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2021 }
2022
2023 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2024 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2025 {
2026 netdev_set_rx_headroom(dev, -1);
2027 }
2028
2029 /*
2030 * Net namespace inlines
2031 */
2032 static inline
dev_net(const struct net_device * dev)2033 struct net *dev_net(const struct net_device *dev)
2034 {
2035 return read_pnet(&dev->nd_net);
2036 }
2037
2038 static inline
dev_net_set(struct net_device * dev,struct net * net)2039 void dev_net_set(struct net_device *dev, struct net *net)
2040 {
2041 write_pnet(&dev->nd_net, net);
2042 }
2043
netdev_uses_dsa(struct net_device * dev)2044 static inline bool netdev_uses_dsa(struct net_device *dev)
2045 {
2046 #if IS_ENABLED(CONFIG_NET_DSA)
2047 if (dev->dsa_ptr != NULL)
2048 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2049 #endif
2050 return false;
2051 }
2052
2053 /**
2054 * netdev_priv - access network device private data
2055 * @dev: network device
2056 *
2057 * Get network device private data
2058 */
netdev_priv(const struct net_device * dev)2059 static inline void *netdev_priv(const struct net_device *dev)
2060 {
2061 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2062 }
2063
2064 /* Set the sysfs physical device reference for the network logical device
2065 * if set prior to registration will cause a symlink during initialization.
2066 */
2067 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2068
2069 /* Set the sysfs device type for the network logical device to allow
2070 * fine-grained identification of different network device types. For
2071 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2072 */
2073 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2074
2075 /* Default NAPI poll() weight
2076 * Device drivers are strongly advised to not use bigger value
2077 */
2078 #define NAPI_POLL_WEIGHT 64
2079
2080 /**
2081 * netif_napi_add - initialize a NAPI context
2082 * @dev: network device
2083 * @napi: NAPI context
2084 * @poll: polling function
2085 * @weight: default weight
2086 *
2087 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2088 * *any* of the other NAPI-related functions.
2089 */
2090 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2091 int (*poll)(struct napi_struct *, int), int weight);
2092
2093 /**
2094 * netif_tx_napi_add - initialize a NAPI context
2095 * @dev: network device
2096 * @napi: NAPI context
2097 * @poll: polling function
2098 * @weight: default weight
2099 *
2100 * This variant of netif_napi_add() should be used from drivers using NAPI
2101 * to exclusively poll a TX queue.
2102 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2103 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2104 static inline void netif_tx_napi_add(struct net_device *dev,
2105 struct napi_struct *napi,
2106 int (*poll)(struct napi_struct *, int),
2107 int weight)
2108 {
2109 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2110 netif_napi_add(dev, napi, poll, weight);
2111 }
2112
2113 /**
2114 * netif_napi_del - remove a NAPI context
2115 * @napi: NAPI context
2116 *
2117 * netif_napi_del() removes a NAPI context from the network device NAPI list
2118 */
2119 void netif_napi_del(struct napi_struct *napi);
2120
2121 struct napi_gro_cb {
2122 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2123 void *frag0;
2124
2125 /* Length of frag0. */
2126 unsigned int frag0_len;
2127
2128 /* This indicates where we are processing relative to skb->data. */
2129 int data_offset;
2130
2131 /* This is non-zero if the packet cannot be merged with the new skb. */
2132 u16 flush;
2133
2134 /* Save the IP ID here and check when we get to the transport layer */
2135 u16 flush_id;
2136
2137 /* Number of segments aggregated. */
2138 u16 count;
2139
2140 /* Start offset for remote checksum offload */
2141 u16 gro_remcsum_start;
2142
2143 /* jiffies when first packet was created/queued */
2144 unsigned long age;
2145
2146 /* Used in ipv6_gro_receive() and foo-over-udp */
2147 u16 proto;
2148
2149 /* This is non-zero if the packet may be of the same flow. */
2150 u8 same_flow:1;
2151
2152 /* Used in tunnel GRO receive */
2153 u8 encap_mark:1;
2154
2155 /* GRO checksum is valid */
2156 u8 csum_valid:1;
2157
2158 /* Number of checksums via CHECKSUM_UNNECESSARY */
2159 u8 csum_cnt:3;
2160
2161 /* Free the skb? */
2162 u8 free:2;
2163 #define NAPI_GRO_FREE 1
2164 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2165
2166 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2167 u8 is_ipv6:1;
2168
2169 /* Used in GRE, set in fou/gue_gro_receive */
2170 u8 is_fou:1;
2171
2172 /* Used to determine if flush_id can be ignored */
2173 u8 is_atomic:1;
2174
2175 /* Number of gro_receive callbacks this packet already went through */
2176 u8 recursion_counter:4;
2177
2178 /* 1 bit hole */
2179
2180 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2181 __wsum csum;
2182
2183 /* used in skb_gro_receive() slow path */
2184 struct sk_buff *last;
2185 };
2186
2187 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2188
2189 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2190 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2191 {
2192 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2193 }
2194
2195 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
call_gro_receive(gro_receive_t cb,struct sk_buff ** head,struct sk_buff * skb)2196 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2197 struct sk_buff **head,
2198 struct sk_buff *skb)
2199 {
2200 if (unlikely(gro_recursion_inc_test(skb))) {
2201 NAPI_GRO_CB(skb)->flush |= 1;
2202 return NULL;
2203 }
2204
2205 return cb(head, skb);
2206 }
2207
2208 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2209 struct sk_buff *);
call_gro_receive_sk(gro_receive_sk_t cb,struct sock * sk,struct sk_buff ** head,struct sk_buff * skb)2210 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2211 struct sock *sk,
2212 struct sk_buff **head,
2213 struct sk_buff *skb)
2214 {
2215 if (unlikely(gro_recursion_inc_test(skb))) {
2216 NAPI_GRO_CB(skb)->flush |= 1;
2217 return NULL;
2218 }
2219
2220 return cb(sk, head, skb);
2221 }
2222
2223 struct packet_type {
2224 __be16 type; /* This is really htons(ether_type). */
2225 struct net_device *dev; /* NULL is wildcarded here */
2226 int (*func) (struct sk_buff *,
2227 struct net_device *,
2228 struct packet_type *,
2229 struct net_device *);
2230 bool (*id_match)(struct packet_type *ptype,
2231 struct sock *sk);
2232 void *af_packet_priv;
2233 struct list_head list;
2234 };
2235
2236 struct offload_callbacks {
2237 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2238 netdev_features_t features);
2239 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2240 struct sk_buff *skb);
2241 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2242 };
2243
2244 struct packet_offload {
2245 __be16 type; /* This is really htons(ether_type). */
2246 u16 priority;
2247 struct offload_callbacks callbacks;
2248 struct list_head list;
2249 };
2250
2251 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2252 struct pcpu_sw_netstats {
2253 u64 rx_packets;
2254 u64 rx_bytes;
2255 u64 tx_packets;
2256 u64 tx_bytes;
2257 struct u64_stats_sync syncp;
2258 };
2259
2260 #define __netdev_alloc_pcpu_stats(type, gfp) \
2261 ({ \
2262 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2263 if (pcpu_stats) { \
2264 int __cpu; \
2265 for_each_possible_cpu(__cpu) { \
2266 typeof(type) *stat; \
2267 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2268 u64_stats_init(&stat->syncp); \
2269 } \
2270 } \
2271 pcpu_stats; \
2272 })
2273
2274 #define netdev_alloc_pcpu_stats(type) \
2275 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2276
2277 enum netdev_lag_tx_type {
2278 NETDEV_LAG_TX_TYPE_UNKNOWN,
2279 NETDEV_LAG_TX_TYPE_RANDOM,
2280 NETDEV_LAG_TX_TYPE_BROADCAST,
2281 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2282 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2283 NETDEV_LAG_TX_TYPE_HASH,
2284 };
2285
2286 struct netdev_lag_upper_info {
2287 enum netdev_lag_tx_type tx_type;
2288 };
2289
2290 struct netdev_lag_lower_state_info {
2291 u8 link_up : 1,
2292 tx_enabled : 1;
2293 };
2294
2295 #include <linux/notifier.h>
2296
2297 /* netdevice notifier chain. Please remember to update the rtnetlink
2298 * notification exclusion list in rtnetlink_event() when adding new
2299 * types.
2300 */
2301 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2302 #define NETDEV_DOWN 0x0002
2303 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2304 detected a hardware crash and restarted
2305 - we can use this eg to kick tcp sessions
2306 once done */
2307 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2308 #define NETDEV_REGISTER 0x0005
2309 #define NETDEV_UNREGISTER 0x0006
2310 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2311 #define NETDEV_CHANGEADDR 0x0008
2312 #define NETDEV_GOING_DOWN 0x0009
2313 #define NETDEV_CHANGENAME 0x000A
2314 #define NETDEV_FEAT_CHANGE 0x000B
2315 #define NETDEV_BONDING_FAILOVER 0x000C
2316 #define NETDEV_PRE_UP 0x000D
2317 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2318 #define NETDEV_POST_TYPE_CHANGE 0x000F
2319 #define NETDEV_POST_INIT 0x0010
2320 #define NETDEV_UNREGISTER_FINAL 0x0011
2321 #define NETDEV_RELEASE 0x0012
2322 #define NETDEV_NOTIFY_PEERS 0x0013
2323 #define NETDEV_JOIN 0x0014
2324 #define NETDEV_CHANGEUPPER 0x0015
2325 #define NETDEV_RESEND_IGMP 0x0016
2326 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2327 #define NETDEV_CHANGEINFODATA 0x0018
2328 #define NETDEV_BONDING_INFO 0x0019
2329 #define NETDEV_PRECHANGEUPPER 0x001A
2330 #define NETDEV_CHANGELOWERSTATE 0x001B
2331 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2332 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2333
2334 int register_netdevice_notifier(struct notifier_block *nb);
2335 int unregister_netdevice_notifier(struct notifier_block *nb);
2336
2337 struct netdev_notifier_info {
2338 struct net_device *dev;
2339 };
2340
2341 struct netdev_notifier_change_info {
2342 struct netdev_notifier_info info; /* must be first */
2343 unsigned int flags_changed;
2344 };
2345
2346 struct netdev_notifier_changeupper_info {
2347 struct netdev_notifier_info info; /* must be first */
2348 struct net_device *upper_dev; /* new upper dev */
2349 bool master; /* is upper dev master */
2350 bool linking; /* is the notification for link or unlink */
2351 void *upper_info; /* upper dev info */
2352 };
2353
2354 struct netdev_notifier_changelowerstate_info {
2355 struct netdev_notifier_info info; /* must be first */
2356 void *lower_state_info; /* is lower dev state */
2357 };
2358
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2359 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2360 struct net_device *dev)
2361 {
2362 info->dev = dev;
2363 }
2364
2365 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2366 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2367 {
2368 return info->dev;
2369 }
2370
2371 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2372
2373
2374 extern rwlock_t dev_base_lock; /* Device list lock */
2375
2376 #define for_each_netdev(net, d) \
2377 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2378 #define for_each_netdev_reverse(net, d) \
2379 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2380 #define for_each_netdev_rcu(net, d) \
2381 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2382 #define for_each_netdev_safe(net, d, n) \
2383 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2384 #define for_each_netdev_continue(net, d) \
2385 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2386 #define for_each_netdev_continue_rcu(net, d) \
2387 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2388 #define for_each_netdev_in_bond_rcu(bond, slave) \
2389 for_each_netdev_rcu(&init_net, slave) \
2390 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2391 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2392
next_net_device(struct net_device * dev)2393 static inline struct net_device *next_net_device(struct net_device *dev)
2394 {
2395 struct list_head *lh;
2396 struct net *net;
2397
2398 net = dev_net(dev);
2399 lh = dev->dev_list.next;
2400 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2401 }
2402
next_net_device_rcu(struct net_device * dev)2403 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2404 {
2405 struct list_head *lh;
2406 struct net *net;
2407
2408 net = dev_net(dev);
2409 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2410 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2411 }
2412
first_net_device(struct net * net)2413 static inline struct net_device *first_net_device(struct net *net)
2414 {
2415 return list_empty(&net->dev_base_head) ? NULL :
2416 net_device_entry(net->dev_base_head.next);
2417 }
2418
first_net_device_rcu(struct net * net)2419 static inline struct net_device *first_net_device_rcu(struct net *net)
2420 {
2421 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2422
2423 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2424 }
2425
2426 int netdev_boot_setup_check(struct net_device *dev);
2427 unsigned long netdev_boot_base(const char *prefix, int unit);
2428 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2429 const char *hwaddr);
2430 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2431 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2432 void dev_add_pack(struct packet_type *pt);
2433 void dev_remove_pack(struct packet_type *pt);
2434 void __dev_remove_pack(struct packet_type *pt);
2435 void dev_add_offload(struct packet_offload *po);
2436 void dev_remove_offload(struct packet_offload *po);
2437
2438 int dev_get_iflink(const struct net_device *dev);
2439 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2440 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2441 unsigned short mask);
2442 struct net_device *dev_get_by_name(struct net *net, const char *name);
2443 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2444 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2445 int dev_alloc_name(struct net_device *dev, const char *name);
2446 int dev_open(struct net_device *dev);
2447 int dev_close(struct net_device *dev);
2448 int dev_close_many(struct list_head *head, bool unlink);
2449 void dev_disable_lro(struct net_device *dev);
2450 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2451 int dev_queue_xmit(struct sk_buff *skb);
2452 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2453 int register_netdevice(struct net_device *dev);
2454 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2455 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2456 static inline void unregister_netdevice(struct net_device *dev)
2457 {
2458 unregister_netdevice_queue(dev, NULL);
2459 }
2460
2461 int netdev_refcnt_read(const struct net_device *dev);
2462 void free_netdev(struct net_device *dev);
2463 void netdev_freemem(struct net_device *dev);
2464 void synchronize_net(void);
2465 int init_dummy_netdev(struct net_device *dev);
2466
2467 DECLARE_PER_CPU(int, xmit_recursion);
2468 #define XMIT_RECURSION_LIMIT 10
2469
dev_recursion_level(void)2470 static inline int dev_recursion_level(void)
2471 {
2472 return this_cpu_read(xmit_recursion);
2473 }
2474
2475 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2476 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2477 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2478 int netdev_get_name(struct net *net, char *name, int ifindex);
2479 int dev_restart(struct net_device *dev);
2480 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2481
skb_gro_offset(const struct sk_buff * skb)2482 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2483 {
2484 return NAPI_GRO_CB(skb)->data_offset;
2485 }
2486
skb_gro_len(const struct sk_buff * skb)2487 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2488 {
2489 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2490 }
2491
skb_gro_pull(struct sk_buff * skb,unsigned int len)2492 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2493 {
2494 NAPI_GRO_CB(skb)->data_offset += len;
2495 }
2496
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2497 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2498 unsigned int offset)
2499 {
2500 return NAPI_GRO_CB(skb)->frag0 + offset;
2501 }
2502
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2503 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2504 {
2505 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2506 }
2507
skb_gro_frag0_invalidate(struct sk_buff * skb)2508 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2509 {
2510 NAPI_GRO_CB(skb)->frag0 = NULL;
2511 NAPI_GRO_CB(skb)->frag0_len = 0;
2512 }
2513
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2514 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2515 unsigned int offset)
2516 {
2517 if (!pskb_may_pull(skb, hlen))
2518 return NULL;
2519
2520 skb_gro_frag0_invalidate(skb);
2521 return skb->data + offset;
2522 }
2523
skb_gro_network_header(struct sk_buff * skb)2524 static inline void *skb_gro_network_header(struct sk_buff *skb)
2525 {
2526 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2527 skb_network_offset(skb);
2528 }
2529
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2530 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2531 const void *start, unsigned int len)
2532 {
2533 if (NAPI_GRO_CB(skb)->csum_valid)
2534 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2535 csum_partial(start, len, 0));
2536 }
2537
2538 /* GRO checksum functions. These are logical equivalents of the normal
2539 * checksum functions (in skbuff.h) except that they operate on the GRO
2540 * offsets and fields in sk_buff.
2541 */
2542
2543 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2544
skb_at_gro_remcsum_start(struct sk_buff * skb)2545 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2546 {
2547 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2548 }
2549
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2550 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2551 bool zero_okay,
2552 __sum16 check)
2553 {
2554 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2555 skb_checksum_start_offset(skb) <
2556 skb_gro_offset(skb)) &&
2557 !skb_at_gro_remcsum_start(skb) &&
2558 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2559 (!zero_okay || check));
2560 }
2561
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2562 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2563 __wsum psum)
2564 {
2565 if (NAPI_GRO_CB(skb)->csum_valid &&
2566 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2567 return 0;
2568
2569 NAPI_GRO_CB(skb)->csum = psum;
2570
2571 return __skb_gro_checksum_complete(skb);
2572 }
2573
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)2574 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2575 {
2576 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2577 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2578 NAPI_GRO_CB(skb)->csum_cnt--;
2579 } else {
2580 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2581 * verified a new top level checksum or an encapsulated one
2582 * during GRO. This saves work if we fallback to normal path.
2583 */
2584 __skb_incr_checksum_unnecessary(skb);
2585 }
2586 }
2587
2588 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2589 compute_pseudo) \
2590 ({ \
2591 __sum16 __ret = 0; \
2592 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2593 __ret = __skb_gro_checksum_validate_complete(skb, \
2594 compute_pseudo(skb, proto)); \
2595 if (__ret) \
2596 __skb_mark_checksum_bad(skb); \
2597 else \
2598 skb_gro_incr_csum_unnecessary(skb); \
2599 __ret; \
2600 })
2601
2602 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2603 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2604
2605 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2606 compute_pseudo) \
2607 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2608
2609 #define skb_gro_checksum_simple_validate(skb) \
2610 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2611
__skb_gro_checksum_convert_check(struct sk_buff * skb)2612 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2613 {
2614 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2615 !NAPI_GRO_CB(skb)->csum_valid);
2616 }
2617
__skb_gro_checksum_convert(struct sk_buff * skb,__sum16 check,__wsum pseudo)2618 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2619 __sum16 check, __wsum pseudo)
2620 {
2621 NAPI_GRO_CB(skb)->csum = ~pseudo;
2622 NAPI_GRO_CB(skb)->csum_valid = 1;
2623 }
2624
2625 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2626 do { \
2627 if (__skb_gro_checksum_convert_check(skb)) \
2628 __skb_gro_checksum_convert(skb, check, \
2629 compute_pseudo(skb, proto)); \
2630 } while (0)
2631
2632 struct gro_remcsum {
2633 int offset;
2634 __wsum delta;
2635 };
2636
skb_gro_remcsum_init(struct gro_remcsum * grc)2637 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2638 {
2639 grc->offset = 0;
2640 grc->delta = 0;
2641 }
2642
skb_gro_remcsum_process(struct sk_buff * skb,void * ptr,unsigned int off,size_t hdrlen,int start,int offset,struct gro_remcsum * grc,bool nopartial)2643 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2644 unsigned int off, size_t hdrlen,
2645 int start, int offset,
2646 struct gro_remcsum *grc,
2647 bool nopartial)
2648 {
2649 __wsum delta;
2650 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2651
2652 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2653
2654 if (!nopartial) {
2655 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2656 return ptr;
2657 }
2658
2659 ptr = skb_gro_header_fast(skb, off);
2660 if (skb_gro_header_hard(skb, off + plen)) {
2661 ptr = skb_gro_header_slow(skb, off + plen, off);
2662 if (!ptr)
2663 return NULL;
2664 }
2665
2666 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2667 start, offset);
2668
2669 /* Adjust skb->csum since we changed the packet */
2670 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2671
2672 grc->offset = off + hdrlen + offset;
2673 grc->delta = delta;
2674
2675 return ptr;
2676 }
2677
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)2678 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2679 struct gro_remcsum *grc)
2680 {
2681 void *ptr;
2682 size_t plen = grc->offset + sizeof(u16);
2683
2684 if (!grc->delta)
2685 return;
2686
2687 ptr = skb_gro_header_fast(skb, grc->offset);
2688 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2689 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2690 if (!ptr)
2691 return;
2692 }
2693
2694 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2695 }
2696
2697 struct skb_csum_offl_spec {
2698 __u16 ipv4_okay:1,
2699 ipv6_okay:1,
2700 encap_okay:1,
2701 ip_options_okay:1,
2702 ext_hdrs_okay:1,
2703 tcp_okay:1,
2704 udp_okay:1,
2705 sctp_okay:1,
2706 vlan_okay:1,
2707 no_encapped_ipv6:1,
2708 no_not_encapped:1;
2709 };
2710
2711 bool __skb_csum_offload_chk(struct sk_buff *skb,
2712 const struct skb_csum_offl_spec *spec,
2713 bool *csum_encapped,
2714 bool csum_help);
2715
skb_csum_offload_chk(struct sk_buff * skb,const struct skb_csum_offl_spec * spec,bool * csum_encapped,bool csum_help)2716 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2717 const struct skb_csum_offl_spec *spec,
2718 bool *csum_encapped,
2719 bool csum_help)
2720 {
2721 if (skb->ip_summed != CHECKSUM_PARTIAL)
2722 return false;
2723
2724 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2725 }
2726
skb_csum_offload_chk_help(struct sk_buff * skb,const struct skb_csum_offl_spec * spec)2727 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2728 const struct skb_csum_offl_spec *spec)
2729 {
2730 bool csum_encapped;
2731
2732 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2733 }
2734
skb_csum_off_chk_help_cmn(struct sk_buff * skb)2735 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2736 {
2737 static const struct skb_csum_offl_spec csum_offl_spec = {
2738 .ipv4_okay = 1,
2739 .ip_options_okay = 1,
2740 .ipv6_okay = 1,
2741 .vlan_okay = 1,
2742 .tcp_okay = 1,
2743 .udp_okay = 1,
2744 };
2745
2746 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2747 }
2748
skb_csum_off_chk_help_cmn_v4_only(struct sk_buff * skb)2749 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2750 {
2751 static const struct skb_csum_offl_spec csum_offl_spec = {
2752 .ipv4_okay = 1,
2753 .ip_options_okay = 1,
2754 .tcp_okay = 1,
2755 .udp_okay = 1,
2756 .vlan_okay = 1,
2757 };
2758
2759 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2760 }
2761
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)2762 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2763 unsigned short type,
2764 const void *daddr, const void *saddr,
2765 unsigned int len)
2766 {
2767 if (!dev->header_ops || !dev->header_ops->create)
2768 return 0;
2769
2770 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2771 }
2772
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)2773 static inline int dev_parse_header(const struct sk_buff *skb,
2774 unsigned char *haddr)
2775 {
2776 const struct net_device *dev = skb->dev;
2777
2778 if (!dev->header_ops || !dev->header_ops->parse)
2779 return 0;
2780 return dev->header_ops->parse(skb, haddr);
2781 }
2782
2783 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)2784 static inline bool dev_validate_header(const struct net_device *dev,
2785 char *ll_header, int len)
2786 {
2787 if (likely(len >= dev->hard_header_len))
2788 return true;
2789 if (len < dev->min_header_len)
2790 return false;
2791
2792 if (capable(CAP_SYS_RAWIO)) {
2793 memset(ll_header + len, 0, dev->hard_header_len - len);
2794 return true;
2795 }
2796
2797 if (dev->header_ops && dev->header_ops->validate)
2798 return dev->header_ops->validate(ll_header, len);
2799
2800 return false;
2801 }
2802
2803 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2804 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
unregister_gifconf(unsigned int family)2805 static inline int unregister_gifconf(unsigned int family)
2806 {
2807 return register_gifconf(family, NULL);
2808 }
2809
2810 #ifdef CONFIG_NET_FLOW_LIMIT
2811 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2812 struct sd_flow_limit {
2813 u64 count;
2814 unsigned int num_buckets;
2815 unsigned int history_head;
2816 u16 history[FLOW_LIMIT_HISTORY];
2817 u8 buckets[];
2818 };
2819
2820 extern int netdev_flow_limit_table_len;
2821 #endif /* CONFIG_NET_FLOW_LIMIT */
2822
2823 /*
2824 * Incoming packets are placed on per-CPU queues
2825 */
2826 struct softnet_data {
2827 struct list_head poll_list;
2828 struct sk_buff_head process_queue;
2829
2830 /* stats */
2831 unsigned int processed;
2832 unsigned int time_squeeze;
2833 unsigned int received_rps;
2834 #ifdef CONFIG_RPS
2835 struct softnet_data *rps_ipi_list;
2836 #endif
2837 #ifdef CONFIG_NET_FLOW_LIMIT
2838 struct sd_flow_limit __rcu *flow_limit;
2839 #endif
2840 struct Qdisc *output_queue;
2841 struct Qdisc **output_queue_tailp;
2842 struct sk_buff *completion_queue;
2843
2844 #ifdef CONFIG_RPS
2845 /* input_queue_head should be written by cpu owning this struct,
2846 * and only read by other cpus. Worth using a cache line.
2847 */
2848 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2849
2850 /* Elements below can be accessed between CPUs for RPS/RFS */
2851 struct call_single_data csd ____cacheline_aligned_in_smp;
2852 struct softnet_data *rps_ipi_next;
2853 unsigned int cpu;
2854 unsigned int input_queue_tail;
2855 #endif
2856 unsigned int dropped;
2857 struct sk_buff_head input_pkt_queue;
2858 struct napi_struct backlog;
2859
2860 };
2861
input_queue_head_incr(struct softnet_data * sd)2862 static inline void input_queue_head_incr(struct softnet_data *sd)
2863 {
2864 #ifdef CONFIG_RPS
2865 sd->input_queue_head++;
2866 #endif
2867 }
2868
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)2869 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2870 unsigned int *qtail)
2871 {
2872 #ifdef CONFIG_RPS
2873 *qtail = ++sd->input_queue_tail;
2874 #endif
2875 }
2876
2877 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2878
2879 void __netif_schedule(struct Qdisc *q);
2880 void netif_schedule_queue(struct netdev_queue *txq);
2881
netif_tx_schedule_all(struct net_device * dev)2882 static inline void netif_tx_schedule_all(struct net_device *dev)
2883 {
2884 unsigned int i;
2885
2886 for (i = 0; i < dev->num_tx_queues; i++)
2887 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2888 }
2889
netif_tx_start_queue(struct netdev_queue * dev_queue)2890 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2891 {
2892 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2893 }
2894
2895 /**
2896 * netif_start_queue - allow transmit
2897 * @dev: network device
2898 *
2899 * Allow upper layers to call the device hard_start_xmit routine.
2900 */
netif_start_queue(struct net_device * dev)2901 static inline void netif_start_queue(struct net_device *dev)
2902 {
2903 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2904 }
2905
netif_tx_start_all_queues(struct net_device * dev)2906 static inline void netif_tx_start_all_queues(struct net_device *dev)
2907 {
2908 unsigned int i;
2909
2910 for (i = 0; i < dev->num_tx_queues; i++) {
2911 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2912 netif_tx_start_queue(txq);
2913 }
2914 }
2915
2916 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2917
2918 /**
2919 * netif_wake_queue - restart transmit
2920 * @dev: network device
2921 *
2922 * Allow upper layers to call the device hard_start_xmit routine.
2923 * Used for flow control when transmit resources are available.
2924 */
netif_wake_queue(struct net_device * dev)2925 static inline void netif_wake_queue(struct net_device *dev)
2926 {
2927 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2928 }
2929
netif_tx_wake_all_queues(struct net_device * dev)2930 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2931 {
2932 unsigned int i;
2933
2934 for (i = 0; i < dev->num_tx_queues; i++) {
2935 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2936 netif_tx_wake_queue(txq);
2937 }
2938 }
2939
netif_tx_stop_queue(struct netdev_queue * dev_queue)2940 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2941 {
2942 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2943 }
2944
2945 /**
2946 * netif_stop_queue - stop transmitted packets
2947 * @dev: network device
2948 *
2949 * Stop upper layers calling the device hard_start_xmit routine.
2950 * Used for flow control when transmit resources are unavailable.
2951 */
netif_stop_queue(struct net_device * dev)2952 static inline void netif_stop_queue(struct net_device *dev)
2953 {
2954 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2955 }
2956
2957 void netif_tx_stop_all_queues(struct net_device *dev);
2958
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)2959 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2960 {
2961 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2962 }
2963
2964 /**
2965 * netif_queue_stopped - test if transmit queue is flowblocked
2966 * @dev: network device
2967 *
2968 * Test if transmit queue on device is currently unable to send.
2969 */
netif_queue_stopped(const struct net_device * dev)2970 static inline bool netif_queue_stopped(const struct net_device *dev)
2971 {
2972 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2973 }
2974
netif_xmit_stopped(const struct netdev_queue * dev_queue)2975 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2976 {
2977 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2978 }
2979
2980 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)2981 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2982 {
2983 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2984 }
2985
2986 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)2987 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2988 {
2989 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2990 }
2991
2992 /**
2993 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2994 * @dev_queue: pointer to transmit queue
2995 *
2996 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2997 * to give appropriate hint to the CPU.
2998 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)2999 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3000 {
3001 #ifdef CONFIG_BQL
3002 prefetchw(&dev_queue->dql.num_queued);
3003 #endif
3004 }
3005
3006 /**
3007 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3008 * @dev_queue: pointer to transmit queue
3009 *
3010 * BQL enabled drivers might use this helper in their TX completion path,
3011 * to give appropriate hint to the CPU.
3012 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3013 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3014 {
3015 #ifdef CONFIG_BQL
3016 prefetchw(&dev_queue->dql.limit);
3017 #endif
3018 }
3019
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3020 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3021 unsigned int bytes)
3022 {
3023 #ifdef CONFIG_BQL
3024 dql_queued(&dev_queue->dql, bytes);
3025
3026 if (likely(dql_avail(&dev_queue->dql) >= 0))
3027 return;
3028
3029 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3030
3031 /*
3032 * The XOFF flag must be set before checking the dql_avail below,
3033 * because in netdev_tx_completed_queue we update the dql_completed
3034 * before checking the XOFF flag.
3035 */
3036 smp_mb();
3037
3038 /* check again in case another CPU has just made room avail */
3039 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3040 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3041 #endif
3042 }
3043
3044 /**
3045 * netdev_sent_queue - report the number of bytes queued to hardware
3046 * @dev: network device
3047 * @bytes: number of bytes queued to the hardware device queue
3048 *
3049 * Report the number of bytes queued for sending/completion to the network
3050 * device hardware queue. @bytes should be a good approximation and should
3051 * exactly match netdev_completed_queue() @bytes
3052 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3053 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3054 {
3055 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3056 }
3057
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3058 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3059 unsigned int pkts, unsigned int bytes)
3060 {
3061 #ifdef CONFIG_BQL
3062 if (unlikely(!bytes))
3063 return;
3064
3065 dql_completed(&dev_queue->dql, bytes);
3066
3067 /*
3068 * Without the memory barrier there is a small possiblity that
3069 * netdev_tx_sent_queue will miss the update and cause the queue to
3070 * be stopped forever
3071 */
3072 smp_mb();
3073
3074 if (dql_avail(&dev_queue->dql) < 0)
3075 return;
3076
3077 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3078 netif_schedule_queue(dev_queue);
3079 #endif
3080 }
3081
3082 /**
3083 * netdev_completed_queue - report bytes and packets completed by device
3084 * @dev: network device
3085 * @pkts: actual number of packets sent over the medium
3086 * @bytes: actual number of bytes sent over the medium
3087 *
3088 * Report the number of bytes and packets transmitted by the network device
3089 * hardware queue over the physical medium, @bytes must exactly match the
3090 * @bytes amount passed to netdev_sent_queue()
3091 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3092 static inline void netdev_completed_queue(struct net_device *dev,
3093 unsigned int pkts, unsigned int bytes)
3094 {
3095 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3096 }
3097
netdev_tx_reset_queue(struct netdev_queue * q)3098 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3099 {
3100 #ifdef CONFIG_BQL
3101 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3102 dql_reset(&q->dql);
3103 #endif
3104 }
3105
3106 /**
3107 * netdev_reset_queue - reset the packets and bytes count of a network device
3108 * @dev_queue: network device
3109 *
3110 * Reset the bytes and packet count of a network device and clear the
3111 * software flow control OFF bit for this network device
3112 */
netdev_reset_queue(struct net_device * dev_queue)3113 static inline void netdev_reset_queue(struct net_device *dev_queue)
3114 {
3115 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3116 }
3117
3118 /**
3119 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3120 * @dev: network device
3121 * @queue_index: given tx queue index
3122 *
3123 * Returns 0 if given tx queue index >= number of device tx queues,
3124 * otherwise returns the originally passed tx queue index.
3125 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3126 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3127 {
3128 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3129 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3130 dev->name, queue_index,
3131 dev->real_num_tx_queues);
3132 return 0;
3133 }
3134
3135 return queue_index;
3136 }
3137
3138 /**
3139 * netif_running - test if up
3140 * @dev: network device
3141 *
3142 * Test if the device has been brought up.
3143 */
netif_running(const struct net_device * dev)3144 static inline bool netif_running(const struct net_device *dev)
3145 {
3146 return test_bit(__LINK_STATE_START, &dev->state);
3147 }
3148
3149 /*
3150 * Routines to manage the subqueues on a device. We only need start,
3151 * stop, and a check if it's stopped. All other device management is
3152 * done at the overall netdevice level.
3153 * Also test the device if we're multiqueue.
3154 */
3155
3156 /**
3157 * netif_start_subqueue - allow sending packets on subqueue
3158 * @dev: network device
3159 * @queue_index: sub queue index
3160 *
3161 * Start individual transmit queue of a device with multiple transmit queues.
3162 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3163 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3164 {
3165 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3166
3167 netif_tx_start_queue(txq);
3168 }
3169
3170 /**
3171 * netif_stop_subqueue - stop sending packets on subqueue
3172 * @dev: network device
3173 * @queue_index: sub queue index
3174 *
3175 * Stop individual transmit queue of a device with multiple transmit queues.
3176 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3177 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3178 {
3179 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3180 netif_tx_stop_queue(txq);
3181 }
3182
3183 /**
3184 * netif_subqueue_stopped - test status of subqueue
3185 * @dev: network device
3186 * @queue_index: sub queue index
3187 *
3188 * Check individual transmit queue of a device with multiple transmit queues.
3189 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3190 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3191 u16 queue_index)
3192 {
3193 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3194
3195 return netif_tx_queue_stopped(txq);
3196 }
3197
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3198 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3199 struct sk_buff *skb)
3200 {
3201 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3202 }
3203
3204 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3205
3206 #ifdef CONFIG_XPS
3207 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3208 u16 index);
3209 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3210 static inline int netif_set_xps_queue(struct net_device *dev,
3211 const struct cpumask *mask,
3212 u16 index)
3213 {
3214 return 0;
3215 }
3216 #endif
3217
3218 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3219 unsigned int num_tx_queues);
3220
3221 /*
3222 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3223 * as a distribution range limit for the returned value.
3224 */
skb_tx_hash(const struct net_device * dev,struct sk_buff * skb)3225 static inline u16 skb_tx_hash(const struct net_device *dev,
3226 struct sk_buff *skb)
3227 {
3228 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3229 }
3230
3231 /**
3232 * netif_is_multiqueue - test if device has multiple transmit queues
3233 * @dev: network device
3234 *
3235 * Check if device has multiple transmit queues
3236 */
netif_is_multiqueue(const struct net_device * dev)3237 static inline bool netif_is_multiqueue(const struct net_device *dev)
3238 {
3239 return dev->num_tx_queues > 1;
3240 }
3241
3242 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3243
3244 #ifdef CONFIG_SYSFS
3245 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3246 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxq)3247 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3248 unsigned int rxq)
3249 {
3250 return 0;
3251 }
3252 #endif
3253
3254 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3255 static inline unsigned int get_netdev_rx_queue_index(
3256 struct netdev_rx_queue *queue)
3257 {
3258 struct net_device *dev = queue->dev;
3259 int index = queue - dev->_rx;
3260
3261 BUG_ON(index >= dev->num_rx_queues);
3262 return index;
3263 }
3264 #endif
3265
3266 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3267 int netif_get_num_default_rss_queues(void);
3268
3269 enum skb_free_reason {
3270 SKB_REASON_CONSUMED,
3271 SKB_REASON_DROPPED,
3272 };
3273
3274 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3275 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3276
3277 /*
3278 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3279 * interrupt context or with hardware interrupts being disabled.
3280 * (in_irq() || irqs_disabled())
3281 *
3282 * We provide four helpers that can be used in following contexts :
3283 *
3284 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3285 * replacing kfree_skb(skb)
3286 *
3287 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3288 * Typically used in place of consume_skb(skb) in TX completion path
3289 *
3290 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3291 * replacing kfree_skb(skb)
3292 *
3293 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3294 * and consumed a packet. Used in place of consume_skb(skb)
3295 */
dev_kfree_skb_irq(struct sk_buff * skb)3296 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3297 {
3298 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3299 }
3300
dev_consume_skb_irq(struct sk_buff * skb)3301 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3302 {
3303 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3304 }
3305
dev_kfree_skb_any(struct sk_buff * skb)3306 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3307 {
3308 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3309 }
3310
dev_consume_skb_any(struct sk_buff * skb)3311 static inline void dev_consume_skb_any(struct sk_buff *skb)
3312 {
3313 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3314 }
3315
3316 int netif_rx(struct sk_buff *skb);
3317 int netif_rx_ni(struct sk_buff *skb);
3318 int netif_receive_skb(struct sk_buff *skb);
3319 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3320 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3321 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3322 gro_result_t napi_gro_frags(struct napi_struct *napi);
3323 struct packet_offload *gro_find_receive_by_type(__be16 type);
3324 struct packet_offload *gro_find_complete_by_type(__be16 type);
3325
napi_free_frags(struct napi_struct * napi)3326 static inline void napi_free_frags(struct napi_struct *napi)
3327 {
3328 kfree_skb(napi->skb);
3329 napi->skb = NULL;
3330 }
3331
3332 bool netdev_is_rx_handler_busy(struct net_device *dev);
3333 int netdev_rx_handler_register(struct net_device *dev,
3334 rx_handler_func_t *rx_handler,
3335 void *rx_handler_data);
3336 void netdev_rx_handler_unregister(struct net_device *dev);
3337
3338 bool dev_valid_name(const char *name);
3339 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3340 int dev_ethtool(struct net *net, struct ifreq *);
3341 unsigned int dev_get_flags(const struct net_device *);
3342 int __dev_change_flags(struct net_device *, unsigned int flags);
3343 int dev_change_flags(struct net_device *, unsigned int);
3344 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3345 unsigned int gchanges);
3346 int dev_change_name(struct net_device *, const char *);
3347 int dev_set_alias(struct net_device *, const char *, size_t);
3348 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3349 int dev_set_mtu(struct net_device *, int);
3350 void dev_set_group(struct net_device *, int);
3351 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3352 int dev_change_carrier(struct net_device *, bool new_carrier);
3353 int dev_get_phys_port_id(struct net_device *dev,
3354 struct netdev_phys_item_id *ppid);
3355 int dev_get_phys_port_name(struct net_device *dev,
3356 char *name, size_t len);
3357 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3358 int dev_change_xdp_fd(struct net_device *dev, int fd);
3359 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3360 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3361 struct netdev_queue *txq, int *ret);
3362 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3363 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3364 bool is_skb_forwardable(const struct net_device *dev,
3365 const struct sk_buff *skb);
3366
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb)3367 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3368 struct sk_buff *skb)
3369 {
3370 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3371 unlikely(!is_skb_forwardable(dev, skb))) {
3372 atomic_long_inc(&dev->rx_dropped);
3373 kfree_skb(skb);
3374 return NET_RX_DROP;
3375 }
3376
3377 skb_scrub_packet(skb, true);
3378 skb->priority = 0;
3379 return 0;
3380 }
3381
3382 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3383
3384 extern int netdev_budget;
3385
3386 /* Called by rtnetlink.c:rtnl_unlock() */
3387 void netdev_run_todo(void);
3388
3389 /**
3390 * dev_put - release reference to device
3391 * @dev: network device
3392 *
3393 * Release reference to device to allow it to be freed.
3394 */
dev_put(struct net_device * dev)3395 static inline void dev_put(struct net_device *dev)
3396 {
3397 this_cpu_dec(*dev->pcpu_refcnt);
3398 }
3399
3400 /**
3401 * dev_hold - get reference to device
3402 * @dev: network device
3403 *
3404 * Hold reference to device to keep it from being freed.
3405 */
dev_hold(struct net_device * dev)3406 static inline void dev_hold(struct net_device *dev)
3407 {
3408 this_cpu_inc(*dev->pcpu_refcnt);
3409 }
3410
3411 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3412 * and _off may be called from IRQ context, but it is caller
3413 * who is responsible for serialization of these calls.
3414 *
3415 * The name carrier is inappropriate, these functions should really be
3416 * called netif_lowerlayer_*() because they represent the state of any
3417 * kind of lower layer not just hardware media.
3418 */
3419
3420 void linkwatch_init_dev(struct net_device *dev);
3421 void linkwatch_fire_event(struct net_device *dev);
3422 void linkwatch_forget_dev(struct net_device *dev);
3423
3424 /**
3425 * netif_carrier_ok - test if carrier present
3426 * @dev: network device
3427 *
3428 * Check if carrier is present on device
3429 */
netif_carrier_ok(const struct net_device * dev)3430 static inline bool netif_carrier_ok(const struct net_device *dev)
3431 {
3432 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3433 }
3434
3435 unsigned long dev_trans_start(struct net_device *dev);
3436
3437 void __netdev_watchdog_up(struct net_device *dev);
3438
3439 void netif_carrier_on(struct net_device *dev);
3440
3441 void netif_carrier_off(struct net_device *dev);
3442
3443 /**
3444 * netif_dormant_on - mark device as dormant.
3445 * @dev: network device
3446 *
3447 * Mark device as dormant (as per RFC2863).
3448 *
3449 * The dormant state indicates that the relevant interface is not
3450 * actually in a condition to pass packets (i.e., it is not 'up') but is
3451 * in a "pending" state, waiting for some external event. For "on-
3452 * demand" interfaces, this new state identifies the situation where the
3453 * interface is waiting for events to place it in the up state.
3454 */
netif_dormant_on(struct net_device * dev)3455 static inline void netif_dormant_on(struct net_device *dev)
3456 {
3457 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3458 linkwatch_fire_event(dev);
3459 }
3460
3461 /**
3462 * netif_dormant_off - set device as not dormant.
3463 * @dev: network device
3464 *
3465 * Device is not in dormant state.
3466 */
netif_dormant_off(struct net_device * dev)3467 static inline void netif_dormant_off(struct net_device *dev)
3468 {
3469 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3470 linkwatch_fire_event(dev);
3471 }
3472
3473 /**
3474 * netif_dormant - test if carrier present
3475 * @dev: network device
3476 *
3477 * Check if carrier is present on device
3478 */
netif_dormant(const struct net_device * dev)3479 static inline bool netif_dormant(const struct net_device *dev)
3480 {
3481 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3482 }
3483
3484
3485 /**
3486 * netif_oper_up - test if device is operational
3487 * @dev: network device
3488 *
3489 * Check if carrier is operational
3490 */
netif_oper_up(const struct net_device * dev)3491 static inline bool netif_oper_up(const struct net_device *dev)
3492 {
3493 return (dev->operstate == IF_OPER_UP ||
3494 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3495 }
3496
3497 /**
3498 * netif_device_present - is device available or removed
3499 * @dev: network device
3500 *
3501 * Check if device has not been removed from system.
3502 */
netif_device_present(struct net_device * dev)3503 static inline bool netif_device_present(struct net_device *dev)
3504 {
3505 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3506 }
3507
3508 void netif_device_detach(struct net_device *dev);
3509
3510 void netif_device_attach(struct net_device *dev);
3511
3512 /*
3513 * Network interface message level settings
3514 */
3515
3516 enum {
3517 NETIF_MSG_DRV = 0x0001,
3518 NETIF_MSG_PROBE = 0x0002,
3519 NETIF_MSG_LINK = 0x0004,
3520 NETIF_MSG_TIMER = 0x0008,
3521 NETIF_MSG_IFDOWN = 0x0010,
3522 NETIF_MSG_IFUP = 0x0020,
3523 NETIF_MSG_RX_ERR = 0x0040,
3524 NETIF_MSG_TX_ERR = 0x0080,
3525 NETIF_MSG_TX_QUEUED = 0x0100,
3526 NETIF_MSG_INTR = 0x0200,
3527 NETIF_MSG_TX_DONE = 0x0400,
3528 NETIF_MSG_RX_STATUS = 0x0800,
3529 NETIF_MSG_PKTDATA = 0x1000,
3530 NETIF_MSG_HW = 0x2000,
3531 NETIF_MSG_WOL = 0x4000,
3532 };
3533
3534 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3535 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3536 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3537 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3538 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3539 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3540 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3541 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3542 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3543 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3544 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3545 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3546 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3547 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3548 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3549
netif_msg_init(int debug_value,int default_msg_enable_bits)3550 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3551 {
3552 /* use default */
3553 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3554 return default_msg_enable_bits;
3555 if (debug_value == 0) /* no output */
3556 return 0;
3557 /* set low N bits */
3558 return (1 << debug_value) - 1;
3559 }
3560
__netif_tx_lock(struct netdev_queue * txq,int cpu)3561 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3562 {
3563 spin_lock(&txq->_xmit_lock);
3564 txq->xmit_lock_owner = cpu;
3565 }
3566
__netif_tx_lock_bh(struct netdev_queue * txq)3567 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3568 {
3569 spin_lock_bh(&txq->_xmit_lock);
3570 txq->xmit_lock_owner = smp_processor_id();
3571 }
3572
__netif_tx_trylock(struct netdev_queue * txq)3573 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3574 {
3575 bool ok = spin_trylock(&txq->_xmit_lock);
3576 if (likely(ok))
3577 txq->xmit_lock_owner = smp_processor_id();
3578 return ok;
3579 }
3580
__netif_tx_unlock(struct netdev_queue * txq)3581 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3582 {
3583 txq->xmit_lock_owner = -1;
3584 spin_unlock(&txq->_xmit_lock);
3585 }
3586
__netif_tx_unlock_bh(struct netdev_queue * txq)3587 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3588 {
3589 txq->xmit_lock_owner = -1;
3590 spin_unlock_bh(&txq->_xmit_lock);
3591 }
3592
txq_trans_update(struct netdev_queue * txq)3593 static inline void txq_trans_update(struct netdev_queue *txq)
3594 {
3595 if (txq->xmit_lock_owner != -1)
3596 txq->trans_start = jiffies;
3597 }
3598
3599 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)3600 static inline void netif_trans_update(struct net_device *dev)
3601 {
3602 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3603
3604 if (txq->trans_start != jiffies)
3605 txq->trans_start = jiffies;
3606 }
3607
3608 /**
3609 * netif_tx_lock - grab network device transmit lock
3610 * @dev: network device
3611 *
3612 * Get network device transmit lock
3613 */
netif_tx_lock(struct net_device * dev)3614 static inline void netif_tx_lock(struct net_device *dev)
3615 {
3616 unsigned int i;
3617 int cpu;
3618
3619 spin_lock(&dev->tx_global_lock);
3620 cpu = smp_processor_id();
3621 for (i = 0; i < dev->num_tx_queues; i++) {
3622 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3623
3624 /* We are the only thread of execution doing a
3625 * freeze, but we have to grab the _xmit_lock in
3626 * order to synchronize with threads which are in
3627 * the ->hard_start_xmit() handler and already
3628 * checked the frozen bit.
3629 */
3630 __netif_tx_lock(txq, cpu);
3631 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3632 __netif_tx_unlock(txq);
3633 }
3634 }
3635
netif_tx_lock_bh(struct net_device * dev)3636 static inline void netif_tx_lock_bh(struct net_device *dev)
3637 {
3638 local_bh_disable();
3639 netif_tx_lock(dev);
3640 }
3641
netif_tx_unlock(struct net_device * dev)3642 static inline void netif_tx_unlock(struct net_device *dev)
3643 {
3644 unsigned int i;
3645
3646 for (i = 0; i < dev->num_tx_queues; i++) {
3647 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3648
3649 /* No need to grab the _xmit_lock here. If the
3650 * queue is not stopped for another reason, we
3651 * force a schedule.
3652 */
3653 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3654 netif_schedule_queue(txq);
3655 }
3656 spin_unlock(&dev->tx_global_lock);
3657 }
3658
netif_tx_unlock_bh(struct net_device * dev)3659 static inline void netif_tx_unlock_bh(struct net_device *dev)
3660 {
3661 netif_tx_unlock(dev);
3662 local_bh_enable();
3663 }
3664
3665 #define HARD_TX_LOCK(dev, txq, cpu) { \
3666 if ((dev->features & NETIF_F_LLTX) == 0) { \
3667 __netif_tx_lock(txq, cpu); \
3668 } \
3669 }
3670
3671 #define HARD_TX_TRYLOCK(dev, txq) \
3672 (((dev->features & NETIF_F_LLTX) == 0) ? \
3673 __netif_tx_trylock(txq) : \
3674 true )
3675
3676 #define HARD_TX_UNLOCK(dev, txq) { \
3677 if ((dev->features & NETIF_F_LLTX) == 0) { \
3678 __netif_tx_unlock(txq); \
3679 } \
3680 }
3681
netif_tx_disable(struct net_device * dev)3682 static inline void netif_tx_disable(struct net_device *dev)
3683 {
3684 unsigned int i;
3685 int cpu;
3686
3687 local_bh_disable();
3688 cpu = smp_processor_id();
3689 for (i = 0; i < dev->num_tx_queues; i++) {
3690 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3691
3692 __netif_tx_lock(txq, cpu);
3693 netif_tx_stop_queue(txq);
3694 __netif_tx_unlock(txq);
3695 }
3696 local_bh_enable();
3697 }
3698
netif_addr_lock(struct net_device * dev)3699 static inline void netif_addr_lock(struct net_device *dev)
3700 {
3701 spin_lock(&dev->addr_list_lock);
3702 }
3703
netif_addr_lock_nested(struct net_device * dev)3704 static inline void netif_addr_lock_nested(struct net_device *dev)
3705 {
3706 int subclass = SINGLE_DEPTH_NESTING;
3707
3708 if (dev->netdev_ops->ndo_get_lock_subclass)
3709 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3710
3711 spin_lock_nested(&dev->addr_list_lock, subclass);
3712 }
3713
netif_addr_lock_bh(struct net_device * dev)3714 static inline void netif_addr_lock_bh(struct net_device *dev)
3715 {
3716 spin_lock_bh(&dev->addr_list_lock);
3717 }
3718
netif_addr_unlock(struct net_device * dev)3719 static inline void netif_addr_unlock(struct net_device *dev)
3720 {
3721 spin_unlock(&dev->addr_list_lock);
3722 }
3723
netif_addr_unlock_bh(struct net_device * dev)3724 static inline void netif_addr_unlock_bh(struct net_device *dev)
3725 {
3726 spin_unlock_bh(&dev->addr_list_lock);
3727 }
3728
3729 /*
3730 * dev_addrs walker. Should be used only for read access. Call with
3731 * rcu_read_lock held.
3732 */
3733 #define for_each_dev_addr(dev, ha) \
3734 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3735
3736 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3737
3738 void ether_setup(struct net_device *dev);
3739
3740 /* Support for loadable net-drivers */
3741 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3742 unsigned char name_assign_type,
3743 void (*setup)(struct net_device *),
3744 unsigned int txqs, unsigned int rxqs);
3745 int dev_get_valid_name(struct net *net, struct net_device *dev,
3746 const char *name);
3747
3748 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3749 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3750
3751 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3752 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3753 count)
3754
3755 int register_netdev(struct net_device *dev);
3756 void unregister_netdev(struct net_device *dev);
3757
3758 /* General hardware address lists handling functions */
3759 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3760 struct netdev_hw_addr_list *from_list, int addr_len);
3761 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3762 struct netdev_hw_addr_list *from_list, int addr_len);
3763 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3764 struct net_device *dev,
3765 int (*sync)(struct net_device *, const unsigned char *),
3766 int (*unsync)(struct net_device *,
3767 const unsigned char *));
3768 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3769 struct net_device *dev,
3770 int (*unsync)(struct net_device *,
3771 const unsigned char *));
3772 void __hw_addr_init(struct netdev_hw_addr_list *list);
3773
3774 /* Functions used for device addresses handling */
3775 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3776 unsigned char addr_type);
3777 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3778 unsigned char addr_type);
3779 void dev_addr_flush(struct net_device *dev);
3780 int dev_addr_init(struct net_device *dev);
3781
3782 /* Functions used for unicast addresses handling */
3783 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3784 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3785 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3786 int dev_uc_sync(struct net_device *to, struct net_device *from);
3787 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3788 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3789 void dev_uc_flush(struct net_device *dev);
3790 void dev_uc_init(struct net_device *dev);
3791
3792 /**
3793 * __dev_uc_sync - Synchonize device's unicast list
3794 * @dev: device to sync
3795 * @sync: function to call if address should be added
3796 * @unsync: function to call if address should be removed
3797 *
3798 * Add newly added addresses to the interface, and release
3799 * addresses that have been deleted.
3800 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))3801 static inline int __dev_uc_sync(struct net_device *dev,
3802 int (*sync)(struct net_device *,
3803 const unsigned char *),
3804 int (*unsync)(struct net_device *,
3805 const unsigned char *))
3806 {
3807 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3808 }
3809
3810 /**
3811 * __dev_uc_unsync - Remove synchronized addresses from device
3812 * @dev: device to sync
3813 * @unsync: function to call if address should be removed
3814 *
3815 * Remove all addresses that were added to the device by dev_uc_sync().
3816 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))3817 static inline void __dev_uc_unsync(struct net_device *dev,
3818 int (*unsync)(struct net_device *,
3819 const unsigned char *))
3820 {
3821 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3822 }
3823
3824 /* Functions used for multicast addresses handling */
3825 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3826 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3827 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3828 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3829 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3830 int dev_mc_sync(struct net_device *to, struct net_device *from);
3831 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3832 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3833 void dev_mc_flush(struct net_device *dev);
3834 void dev_mc_init(struct net_device *dev);
3835
3836 /**
3837 * __dev_mc_sync - Synchonize device's multicast list
3838 * @dev: device to sync
3839 * @sync: function to call if address should be added
3840 * @unsync: function to call if address should be removed
3841 *
3842 * Add newly added addresses to the interface, and release
3843 * addresses that have been deleted.
3844 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))3845 static inline int __dev_mc_sync(struct net_device *dev,
3846 int (*sync)(struct net_device *,
3847 const unsigned char *),
3848 int (*unsync)(struct net_device *,
3849 const unsigned char *))
3850 {
3851 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3852 }
3853
3854 /**
3855 * __dev_mc_unsync - Remove synchronized addresses from device
3856 * @dev: device to sync
3857 * @unsync: function to call if address should be removed
3858 *
3859 * Remove all addresses that were added to the device by dev_mc_sync().
3860 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))3861 static inline void __dev_mc_unsync(struct net_device *dev,
3862 int (*unsync)(struct net_device *,
3863 const unsigned char *))
3864 {
3865 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3866 }
3867
3868 /* Functions used for secondary unicast and multicast support */
3869 void dev_set_rx_mode(struct net_device *dev);
3870 void __dev_set_rx_mode(struct net_device *dev);
3871 int dev_set_promiscuity(struct net_device *dev, int inc);
3872 int dev_set_allmulti(struct net_device *dev, int inc);
3873 void netdev_state_change(struct net_device *dev);
3874 void netdev_notify_peers(struct net_device *dev);
3875 void netdev_features_change(struct net_device *dev);
3876 /* Load a device via the kmod */
3877 void dev_load(struct net *net, const char *name);
3878 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3879 struct rtnl_link_stats64 *storage);
3880 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3881 const struct net_device_stats *netdev_stats);
3882
3883 extern int netdev_max_backlog;
3884 extern int netdev_tstamp_prequeue;
3885 extern int weight_p;
3886
3887 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3888 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3889 struct list_head **iter);
3890 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3891 struct list_head **iter);
3892
3893 /* iterate through upper list, must be called under RCU read lock */
3894 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3895 for (iter = &(dev)->adj_list.upper, \
3896 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3897 updev; \
3898 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3899
3900 /* iterate through upper list, must be called under RCU read lock */
3901 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3902 for (iter = &(dev)->all_adj_list.upper, \
3903 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3904 updev; \
3905 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3906
3907 bool netdev_has_any_upper_dev(struct net_device *dev);
3908
3909 void *netdev_lower_get_next_private(struct net_device *dev,
3910 struct list_head **iter);
3911 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3912 struct list_head **iter);
3913
3914 #define netdev_for_each_lower_private(dev, priv, iter) \
3915 for (iter = (dev)->adj_list.lower.next, \
3916 priv = netdev_lower_get_next_private(dev, &(iter)); \
3917 priv; \
3918 priv = netdev_lower_get_next_private(dev, &(iter)))
3919
3920 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3921 for (iter = &(dev)->adj_list.lower, \
3922 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3923 priv; \
3924 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3925
3926 void *netdev_lower_get_next(struct net_device *dev,
3927 struct list_head **iter);
3928
3929 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3930 for (iter = (dev)->adj_list.lower.next, \
3931 ldev = netdev_lower_get_next(dev, &(iter)); \
3932 ldev; \
3933 ldev = netdev_lower_get_next(dev, &(iter)))
3934
3935 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3936 struct list_head **iter);
3937 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3938 struct list_head **iter);
3939
3940 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3941 for (iter = (dev)->all_adj_list.lower.next, \
3942 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3943 ldev; \
3944 ldev = netdev_all_lower_get_next(dev, &(iter)))
3945
3946 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3947 for (iter = &(dev)->all_adj_list.lower, \
3948 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3949 ldev; \
3950 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3951
3952 void *netdev_adjacent_get_private(struct list_head *adj_list);
3953 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3954 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3955 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3956 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3957 int netdev_master_upper_dev_link(struct net_device *dev,
3958 struct net_device *upper_dev,
3959 void *upper_priv, void *upper_info);
3960 void netdev_upper_dev_unlink(struct net_device *dev,
3961 struct net_device *upper_dev);
3962 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3963 void *netdev_lower_dev_get_private(struct net_device *dev,
3964 struct net_device *lower_dev);
3965 void netdev_lower_state_changed(struct net_device *lower_dev,
3966 void *lower_state_info);
3967 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3968 struct neighbour *n);
3969 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3970 struct neighbour *n);
3971
3972 /* RSS keys are 40 or 52 bytes long */
3973 #define NETDEV_RSS_KEY_LEN 52
3974 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3975 void netdev_rss_key_fill(void *buffer, size_t len);
3976
3977 int dev_get_nest_level(struct net_device *dev);
3978 int skb_checksum_help(struct sk_buff *skb);
3979 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3980 netdev_features_t features, bool tx_path);
3981 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3982 netdev_features_t features);
3983
3984 struct netdev_bonding_info {
3985 ifslave slave;
3986 ifbond master;
3987 };
3988
3989 struct netdev_notifier_bonding_info {
3990 struct netdev_notifier_info info; /* must be first */
3991 struct netdev_bonding_info bonding_info;
3992 };
3993
3994 void netdev_bonding_info_change(struct net_device *dev,
3995 struct netdev_bonding_info *bonding_info);
3996
3997 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)3998 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3999 {
4000 return __skb_gso_segment(skb, features, true);
4001 }
4002 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4003
can_checksum_protocol(netdev_features_t features,__be16 protocol)4004 static inline bool can_checksum_protocol(netdev_features_t features,
4005 __be16 protocol)
4006 {
4007 if (protocol == htons(ETH_P_FCOE))
4008 return !!(features & NETIF_F_FCOE_CRC);
4009
4010 /* Assume this is an IP checksum (not SCTP CRC) */
4011
4012 if (features & NETIF_F_HW_CSUM) {
4013 /* Can checksum everything */
4014 return true;
4015 }
4016
4017 switch (protocol) {
4018 case htons(ETH_P_IP):
4019 return !!(features & NETIF_F_IP_CSUM);
4020 case htons(ETH_P_IPV6):
4021 return !!(features & NETIF_F_IPV6_CSUM);
4022 default:
4023 return false;
4024 }
4025 }
4026
4027 /* Map an ethertype into IP protocol if possible */
eproto_to_ipproto(int eproto)4028 static inline int eproto_to_ipproto(int eproto)
4029 {
4030 switch (eproto) {
4031 case htons(ETH_P_IP):
4032 return IPPROTO_IP;
4033 case htons(ETH_P_IPV6):
4034 return IPPROTO_IPV6;
4035 default:
4036 return -1;
4037 }
4038 }
4039
4040 #ifdef CONFIG_BUG
4041 void netdev_rx_csum_fault(struct net_device *dev);
4042 #else
netdev_rx_csum_fault(struct net_device * dev)4043 static inline void netdev_rx_csum_fault(struct net_device *dev)
4044 {
4045 }
4046 #endif
4047 /* rx skb timestamps */
4048 void net_enable_timestamp(void);
4049 void net_disable_timestamp(void);
4050
4051 #ifdef CONFIG_PROC_FS
4052 int __init dev_proc_init(void);
4053 #else
4054 #define dev_proc_init() 0
4055 #endif
4056
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4057 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4058 struct sk_buff *skb, struct net_device *dev,
4059 bool more)
4060 {
4061 skb->xmit_more = more ? 1 : 0;
4062 return ops->ndo_start_xmit(skb, dev);
4063 }
4064
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4065 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4066 struct netdev_queue *txq, bool more)
4067 {
4068 const struct net_device_ops *ops = dev->netdev_ops;
4069 int rc;
4070
4071 rc = __netdev_start_xmit(ops, skb, dev, more);
4072 if (rc == NETDEV_TX_OK)
4073 txq_trans_update(txq);
4074
4075 return rc;
4076 }
4077
4078 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4079 const void *ns);
4080 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4081 const void *ns);
4082
netdev_class_create_file(struct class_attribute * class_attr)4083 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4084 {
4085 return netdev_class_create_file_ns(class_attr, NULL);
4086 }
4087
netdev_class_remove_file(struct class_attribute * class_attr)4088 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4089 {
4090 netdev_class_remove_file_ns(class_attr, NULL);
4091 }
4092
4093 extern struct kobj_ns_type_operations net_ns_type_operations;
4094
4095 const char *netdev_drivername(const struct net_device *dev);
4096
4097 void linkwatch_run_queue(void);
4098
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4099 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4100 netdev_features_t f2)
4101 {
4102 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4103 if (f1 & NETIF_F_HW_CSUM)
4104 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4105 else
4106 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4107 }
4108
4109 return f1 & f2;
4110 }
4111
netdev_get_wanted_features(struct net_device * dev)4112 static inline netdev_features_t netdev_get_wanted_features(
4113 struct net_device *dev)
4114 {
4115 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4116 }
4117 netdev_features_t netdev_increment_features(netdev_features_t all,
4118 netdev_features_t one, netdev_features_t mask);
4119
4120 /* Allow TSO being used on stacked device :
4121 * Performing the GSO segmentation before last device
4122 * is a performance improvement.
4123 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4124 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4125 netdev_features_t mask)
4126 {
4127 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4128 }
4129
4130 int __netdev_update_features(struct net_device *dev);
4131 void netdev_update_features(struct net_device *dev);
4132 void netdev_change_features(struct net_device *dev);
4133
4134 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4135 struct net_device *dev);
4136
4137 netdev_features_t passthru_features_check(struct sk_buff *skb,
4138 struct net_device *dev,
4139 netdev_features_t features);
4140 netdev_features_t netif_skb_features(struct sk_buff *skb);
4141
net_gso_ok(netdev_features_t features,int gso_type)4142 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4143 {
4144 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4145
4146 /* check flags correspondence */
4147 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4148 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4149 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4150 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4151 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4152 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4153 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4154 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4155 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4156 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4157 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4158 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4159 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4160 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4161 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4162 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4163
4164 return (features & feature) == feature;
4165 }
4166
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)4167 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4168 {
4169 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4170 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4171 }
4172
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)4173 static inline bool netif_needs_gso(struct sk_buff *skb,
4174 netdev_features_t features)
4175 {
4176 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4177 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4178 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4179 }
4180
netif_set_gso_max_size(struct net_device * dev,unsigned int size)4181 static inline void netif_set_gso_max_size(struct net_device *dev,
4182 unsigned int size)
4183 {
4184 dev->gso_max_size = size;
4185 }
4186
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)4187 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4188 int pulled_hlen, u16 mac_offset,
4189 int mac_len)
4190 {
4191 skb->protocol = protocol;
4192 skb->encapsulation = 1;
4193 skb_push(skb, pulled_hlen);
4194 skb_reset_transport_header(skb);
4195 skb->mac_header = mac_offset;
4196 skb->network_header = skb->mac_header + mac_len;
4197 skb->mac_len = mac_len;
4198 }
4199
netif_is_macsec(const struct net_device * dev)4200 static inline bool netif_is_macsec(const struct net_device *dev)
4201 {
4202 return dev->priv_flags & IFF_MACSEC;
4203 }
4204
netif_is_macvlan(const struct net_device * dev)4205 static inline bool netif_is_macvlan(const struct net_device *dev)
4206 {
4207 return dev->priv_flags & IFF_MACVLAN;
4208 }
4209
netif_is_macvlan_port(const struct net_device * dev)4210 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4211 {
4212 return dev->priv_flags & IFF_MACVLAN_PORT;
4213 }
4214
netif_is_ipvlan(const struct net_device * dev)4215 static inline bool netif_is_ipvlan(const struct net_device *dev)
4216 {
4217 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4218 }
4219
netif_is_ipvlan_port(const struct net_device * dev)4220 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4221 {
4222 return dev->priv_flags & IFF_IPVLAN_MASTER;
4223 }
4224
netif_is_bond_master(const struct net_device * dev)4225 static inline bool netif_is_bond_master(const struct net_device *dev)
4226 {
4227 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4228 }
4229
netif_is_bond_slave(const struct net_device * dev)4230 static inline bool netif_is_bond_slave(const struct net_device *dev)
4231 {
4232 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4233 }
4234
netif_supports_nofcs(struct net_device * dev)4235 static inline bool netif_supports_nofcs(struct net_device *dev)
4236 {
4237 return dev->priv_flags & IFF_SUPP_NOFCS;
4238 }
4239
netif_is_l3_master(const struct net_device * dev)4240 static inline bool netif_is_l3_master(const struct net_device *dev)
4241 {
4242 return dev->priv_flags & IFF_L3MDEV_MASTER;
4243 }
4244
netif_is_l3_slave(const struct net_device * dev)4245 static inline bool netif_is_l3_slave(const struct net_device *dev)
4246 {
4247 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4248 }
4249
netif_is_bridge_master(const struct net_device * dev)4250 static inline bool netif_is_bridge_master(const struct net_device *dev)
4251 {
4252 return dev->priv_flags & IFF_EBRIDGE;
4253 }
4254
netif_is_bridge_port(const struct net_device * dev)4255 static inline bool netif_is_bridge_port(const struct net_device *dev)
4256 {
4257 return dev->priv_flags & IFF_BRIDGE_PORT;
4258 }
4259
netif_is_ovs_master(const struct net_device * dev)4260 static inline bool netif_is_ovs_master(const struct net_device *dev)
4261 {
4262 return dev->priv_flags & IFF_OPENVSWITCH;
4263 }
4264
netif_is_team_master(const struct net_device * dev)4265 static inline bool netif_is_team_master(const struct net_device *dev)
4266 {
4267 return dev->priv_flags & IFF_TEAM;
4268 }
4269
netif_is_team_port(const struct net_device * dev)4270 static inline bool netif_is_team_port(const struct net_device *dev)
4271 {
4272 return dev->priv_flags & IFF_TEAM_PORT;
4273 }
4274
netif_is_lag_master(const struct net_device * dev)4275 static inline bool netif_is_lag_master(const struct net_device *dev)
4276 {
4277 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4278 }
4279
netif_is_lag_port(const struct net_device * dev)4280 static inline bool netif_is_lag_port(const struct net_device *dev)
4281 {
4282 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4283 }
4284
netif_is_rxfh_configured(const struct net_device * dev)4285 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4286 {
4287 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4288 }
4289
4290 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)4291 static inline void netif_keep_dst(struct net_device *dev)
4292 {
4293 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4294 }
4295
4296 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)4297 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4298 {
4299 /* TODO: reserve and use an additional IFF bit, if we get more users */
4300 return dev->priv_flags & IFF_MACSEC;
4301 }
4302
4303 extern struct pernet_operations __net_initdata loopback_net_ops;
4304
4305 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4306
4307 /* netdev_printk helpers, similar to dev_printk */
4308
netdev_name(const struct net_device * dev)4309 static inline const char *netdev_name(const struct net_device *dev)
4310 {
4311 if (!dev->name[0] || strchr(dev->name, '%'))
4312 return "(unnamed net_device)";
4313 return dev->name;
4314 }
4315
netdev_reg_state(const struct net_device * dev)4316 static inline const char *netdev_reg_state(const struct net_device *dev)
4317 {
4318 switch (dev->reg_state) {
4319 case NETREG_UNINITIALIZED: return " (uninitialized)";
4320 case NETREG_REGISTERED: return "";
4321 case NETREG_UNREGISTERING: return " (unregistering)";
4322 case NETREG_UNREGISTERED: return " (unregistered)";
4323 case NETREG_RELEASED: return " (released)";
4324 case NETREG_DUMMY: return " (dummy)";
4325 }
4326
4327 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4328 return " (unknown)";
4329 }
4330
4331 __printf(3, 4)
4332 void netdev_printk(const char *level, const struct net_device *dev,
4333 const char *format, ...);
4334 __printf(2, 3)
4335 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4336 __printf(2, 3)
4337 void netdev_alert(const struct net_device *dev, const char *format, ...);
4338 __printf(2, 3)
4339 void netdev_crit(const struct net_device *dev, const char *format, ...);
4340 __printf(2, 3)
4341 void netdev_err(const struct net_device *dev, const char *format, ...);
4342 __printf(2, 3)
4343 void netdev_warn(const struct net_device *dev, const char *format, ...);
4344 __printf(2, 3)
4345 void netdev_notice(const struct net_device *dev, const char *format, ...);
4346 __printf(2, 3)
4347 void netdev_info(const struct net_device *dev, const char *format, ...);
4348
4349 #define MODULE_ALIAS_NETDEV(device) \
4350 MODULE_ALIAS("netdev-" device)
4351
4352 #if defined(CONFIG_DYNAMIC_DEBUG)
4353 #define netdev_dbg(__dev, format, args...) \
4354 do { \
4355 dynamic_netdev_dbg(__dev, format, ##args); \
4356 } while (0)
4357 #elif defined(DEBUG)
4358 #define netdev_dbg(__dev, format, args...) \
4359 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4360 #else
4361 #define netdev_dbg(__dev, format, args...) \
4362 ({ \
4363 if (0) \
4364 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4365 })
4366 #endif
4367
4368 #if defined(VERBOSE_DEBUG)
4369 #define netdev_vdbg netdev_dbg
4370 #else
4371
4372 #define netdev_vdbg(dev, format, args...) \
4373 ({ \
4374 if (0) \
4375 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4376 0; \
4377 })
4378 #endif
4379
4380 /*
4381 * netdev_WARN() acts like dev_printk(), but with the key difference
4382 * of using a WARN/WARN_ON to get the message out, including the
4383 * file/line information and a backtrace.
4384 */
4385 #define netdev_WARN(dev, format, args...) \
4386 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4387 netdev_reg_state(dev), ##args)
4388
4389 /* netif printk helpers, similar to netdev_printk */
4390
4391 #define netif_printk(priv, type, level, dev, fmt, args...) \
4392 do { \
4393 if (netif_msg_##type(priv)) \
4394 netdev_printk(level, (dev), fmt, ##args); \
4395 } while (0)
4396
4397 #define netif_level(level, priv, type, dev, fmt, args...) \
4398 do { \
4399 if (netif_msg_##type(priv)) \
4400 netdev_##level(dev, fmt, ##args); \
4401 } while (0)
4402
4403 #define netif_emerg(priv, type, dev, fmt, args...) \
4404 netif_level(emerg, priv, type, dev, fmt, ##args)
4405 #define netif_alert(priv, type, dev, fmt, args...) \
4406 netif_level(alert, priv, type, dev, fmt, ##args)
4407 #define netif_crit(priv, type, dev, fmt, args...) \
4408 netif_level(crit, priv, type, dev, fmt, ##args)
4409 #define netif_err(priv, type, dev, fmt, args...) \
4410 netif_level(err, priv, type, dev, fmt, ##args)
4411 #define netif_warn(priv, type, dev, fmt, args...) \
4412 netif_level(warn, priv, type, dev, fmt, ##args)
4413 #define netif_notice(priv, type, dev, fmt, args...) \
4414 netif_level(notice, priv, type, dev, fmt, ##args)
4415 #define netif_info(priv, type, dev, fmt, args...) \
4416 netif_level(info, priv, type, dev, fmt, ##args)
4417
4418 #if defined(CONFIG_DYNAMIC_DEBUG)
4419 #define netif_dbg(priv, type, netdev, format, args...) \
4420 do { \
4421 if (netif_msg_##type(priv)) \
4422 dynamic_netdev_dbg(netdev, format, ##args); \
4423 } while (0)
4424 #elif defined(DEBUG)
4425 #define netif_dbg(priv, type, dev, format, args...) \
4426 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4427 #else
4428 #define netif_dbg(priv, type, dev, format, args...) \
4429 ({ \
4430 if (0) \
4431 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4432 0; \
4433 })
4434 #endif
4435
4436 #if defined(VERBOSE_DEBUG)
4437 #define netif_vdbg netif_dbg
4438 #else
4439 #define netif_vdbg(priv, type, dev, format, args...) \
4440 ({ \
4441 if (0) \
4442 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4443 0; \
4444 })
4445 #endif
4446
4447 /*
4448 * The list of packet types we will receive (as opposed to discard)
4449 * and the routines to invoke.
4450 *
4451 * Why 16. Because with 16 the only overlap we get on a hash of the
4452 * low nibble of the protocol value is RARP/SNAP/X.25.
4453 *
4454 * NOTE: That is no longer true with the addition of VLAN tags. Not
4455 * sure which should go first, but I bet it won't make much
4456 * difference if we are running VLANs. The good news is that
4457 * this protocol won't be in the list unless compiled in, so
4458 * the average user (w/out VLANs) will not be adversely affected.
4459 * --BLG
4460 *
4461 * 0800 IP
4462 * 8100 802.1Q VLAN
4463 * 0001 802.3
4464 * 0002 AX.25
4465 * 0004 802.2
4466 * 8035 RARP
4467 * 0005 SNAP
4468 * 0805 X.25
4469 * 0806 ARP
4470 * 8137 IPX
4471 * 0009 Localtalk
4472 * 86DD IPv6
4473 */
4474 #define PTYPE_HASH_SIZE (16)
4475 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4476
4477 #endif /* _LINUX_NETDEVICE_H */
4478