1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
70
71 /*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
103
104 enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 };
110 typedef enum netdev_tx netdev_tx_t;
111
112 /*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
dev_xmit_complete(int rc)116 static inline bool dev_xmit_complete(int rc)
117 {
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128 }
129
130 /*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
138 # else
139 # define LL_MAX_HEADER 96
140 # endif
141 #else
142 # define LL_MAX_HEADER 32
143 #endif
144
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
148 #else
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
150 #endif
151
152 /*
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
155 */
156
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181 };
182
183
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
186
187 #ifdef CONFIG_RPS
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
190 #endif
191
192 struct neighbour;
193 struct neigh_parms;
194 struct sk_buff;
195
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
205 bool global_use;
206 int sync_cnt;
207 int refcount;
208 int synced;
209 struct rcu_head rcu_head;
210 };
211
212 struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215 };
216
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
231
232 struct hh_cache {
233 u16 hh_len;
234 u16 __pad;
235 seqlock_t hh_lock;
236
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244 };
245
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258
259 struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268 bool (*validate)(const char *ll_header, unsigned int len);
269 };
270
271 /* These flag bits are private to the generic network queueing
272 * layer, they may not be explicitly referenced by any other
273 * code.
274 */
275
276 enum netdev_state_t {
277 __LINK_STATE_START,
278 __LINK_STATE_PRESENT,
279 __LINK_STATE_NOCARRIER,
280 __LINK_STATE_LINKWATCH_PENDING,
281 __LINK_STATE_DORMANT,
282 };
283
284
285 /*
286 * This structure holds at boot time configured netdevice settings. They
287 * are then used in the device probing.
288 */
289 struct netdev_boot_setup {
290 char name[IFNAMSIZ];
291 struct ifmap map;
292 };
293 #define NETDEV_BOOT_SETUP_MAX 8
294
295 int __init netdev_boot_setup(char *str);
296
297 /*
298 * Structure for NAPI scheduling similar to tasklet but with weighting
299 */
300 struct napi_struct {
301 /* The poll_list must only be managed by the entity which
302 * changes the state of the NAPI_STATE_SCHED bit. This means
303 * whoever atomically sets that bit can add this napi_struct
304 * to the per-cpu poll_list, and whoever clears that bit
305 * can remove from the list right before clearing the bit.
306 */
307 struct list_head poll_list;
308
309 unsigned long state;
310 int weight;
311 unsigned int gro_count;
312 int (*poll)(struct napi_struct *, int);
313 #ifdef CONFIG_NETPOLL
314 spinlock_t poll_lock;
315 int poll_owner;
316 #endif
317 struct net_device *dev;
318 struct sk_buff *gro_list;
319 struct sk_buff *skb;
320 struct hrtimer timer;
321 struct list_head dev_list;
322 struct hlist_node napi_hash_node;
323 unsigned int napi_id;
324 };
325
326 enum {
327 NAPI_STATE_SCHED, /* Poll is scheduled */
328 NAPI_STATE_DISABLE, /* Disable pending */
329 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
330 NAPI_STATE_HASHED, /* In NAPI hash */
331 };
332
333 enum gro_result {
334 GRO_MERGED,
335 GRO_MERGED_FREE,
336 GRO_HELD,
337 GRO_NORMAL,
338 GRO_DROP,
339 };
340 typedef enum gro_result gro_result_t;
341
342 /*
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * further.
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 *
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
353 *
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
356 *
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
360 *
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
362 * do with the skb.
363 *
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
367 *
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
371 *
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
374 * are registered on exact device (ptype->dev == skb->dev).
375 *
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
378 *
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
381 */
382
383 enum rx_handler_result {
384 RX_HANDLER_CONSUMED,
385 RX_HANDLER_ANOTHER,
386 RX_HANDLER_EXACT,
387 RX_HANDLER_PASS,
388 };
389 typedef enum rx_handler_result rx_handler_result_t;
390 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
391
392 void __napi_schedule(struct napi_struct *n);
393 void __napi_schedule_irqoff(struct napi_struct *n);
394
napi_disable_pending(struct napi_struct * n)395 static inline bool napi_disable_pending(struct napi_struct *n)
396 {
397 return test_bit(NAPI_STATE_DISABLE, &n->state);
398 }
399
400 /**
401 * napi_schedule_prep - check if napi can be scheduled
402 * @n: napi context
403 *
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
408 */
napi_schedule_prep(struct napi_struct * n)409 static inline bool napi_schedule_prep(struct napi_struct *n)
410 {
411 return !napi_disable_pending(n) &&
412 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
413 }
414
415 /**
416 * napi_schedule - schedule NAPI poll
417 * @n: napi context
418 *
419 * Schedule NAPI poll routine to be called if it is not already
420 * running.
421 */
napi_schedule(struct napi_struct * n)422 static inline void napi_schedule(struct napi_struct *n)
423 {
424 if (napi_schedule_prep(n))
425 __napi_schedule(n);
426 }
427
428 /**
429 * napi_schedule_irqoff - schedule NAPI poll
430 * @n: napi context
431 *
432 * Variant of napi_schedule(), assuming hard irqs are masked.
433 */
napi_schedule_irqoff(struct napi_struct * n)434 static inline void napi_schedule_irqoff(struct napi_struct *n)
435 {
436 if (napi_schedule_prep(n))
437 __napi_schedule_irqoff(n);
438 }
439
440 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)441 static inline bool napi_reschedule(struct napi_struct *napi)
442 {
443 if (napi_schedule_prep(napi)) {
444 __napi_schedule(napi);
445 return true;
446 }
447 return false;
448 }
449
450 void __napi_complete(struct napi_struct *n);
451 void napi_complete_done(struct napi_struct *n, int work_done);
452 /**
453 * napi_complete - NAPI processing complete
454 * @n: napi context
455 *
456 * Mark NAPI processing as complete.
457 * Consider using napi_complete_done() instead.
458 */
napi_complete(struct napi_struct * n)459 static inline void napi_complete(struct napi_struct *n)
460 {
461 return napi_complete_done(n, 0);
462 }
463
464 /**
465 * napi_by_id - lookup a NAPI by napi_id
466 * @napi_id: hashed napi_id
467 *
468 * lookup @napi_id in napi_hash table
469 * must be called under rcu_read_lock()
470 */
471 struct napi_struct *napi_by_id(unsigned int napi_id);
472
473 /**
474 * napi_hash_add - add a NAPI to global hashtable
475 * @napi: napi context
476 *
477 * generate a new napi_id and store a @napi under it in napi_hash
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
487 */
488 void napi_hash_del(struct napi_struct *napi);
489
490 /**
491 * napi_disable - prevent NAPI from scheduling
492 * @n: napi context
493 *
494 * Stop NAPI from being scheduled on this context.
495 * Waits till any outstanding processing completes.
496 */
497 void napi_disable(struct napi_struct *n);
498
499 /**
500 * napi_enable - enable NAPI scheduling
501 * @n: napi context
502 *
503 * Resume NAPI from being scheduled on this context.
504 * Must be paired with napi_disable.
505 */
napi_enable(struct napi_struct * n)506 static inline void napi_enable(struct napi_struct *n)
507 {
508 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
509 smp_mb__before_atomic();
510 clear_bit(NAPI_STATE_SCHED, &n->state);
511 clear_bit(NAPI_STATE_NPSVC, &n->state);
512 }
513
514 /**
515 * napi_synchronize - wait until NAPI is not running
516 * @n: napi context
517 *
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
521 */
napi_synchronize(const struct napi_struct * n)522 static inline void napi_synchronize(const struct napi_struct *n)
523 {
524 if (IS_ENABLED(CONFIG_SMP))
525 while (test_bit(NAPI_STATE_SCHED, &n->state))
526 msleep(1);
527 else
528 barrier();
529 }
530
531 enum netdev_queue_state_t {
532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
534 __QUEUE_STATE_FROZEN,
535 };
536
537 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540
541 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 QUEUE_STATE_FROZEN)
544 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
545 QUEUE_STATE_FROZEN)
546
547 /*
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
555 */
556
557 struct netdev_queue {
558 /*
559 * read mostly part
560 */
561 struct net_device *dev;
562 struct Qdisc __rcu *qdisc;
563 struct Qdisc *qdisc_sleeping;
564 #ifdef CONFIG_SYSFS
565 struct kobject kobj;
566 #endif
567 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
568 int numa_node;
569 #endif
570 /*
571 * write mostly part
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
575 /*
576 * please use this field instead of dev->trans_start
577 */
578 unsigned long trans_start;
579
580 /*
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
583 */
584 unsigned long trans_timeout;
585
586 unsigned long state;
587
588 #ifdef CONFIG_BQL
589 struct dql dql;
590 #endif
591 unsigned long tx_maxrate;
592 } ____cacheline_aligned_in_smp;
593
netdev_queue_numa_node_read(const struct netdev_queue * q)594 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
595 {
596 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 return q->numa_node;
598 #else
599 return NUMA_NO_NODE;
600 #endif
601 }
602
netdev_queue_numa_node_write(struct netdev_queue * q,int node)603 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
604 {
605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 q->numa_node = node;
607 #endif
608 }
609
610 #ifdef CONFIG_RPS
611 /*
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
614 */
615 struct rps_map {
616 unsigned int len;
617 struct rcu_head rcu;
618 u16 cpus[0];
619 };
620 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
621
622 /*
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
626 */
627 struct rps_dev_flow {
628 u16 cpu;
629 u16 filter;
630 unsigned int last_qtail;
631 };
632 #define RPS_NO_FILTER 0xffff
633
634 /*
635 * The rps_dev_flow_table structure contains a table of flow mappings.
636 */
637 struct rps_dev_flow_table {
638 unsigned int mask;
639 struct rcu_head rcu;
640 struct rps_dev_flow flows[0];
641 };
642 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
643 ((_num) * sizeof(struct rps_dev_flow)))
644
645 /*
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
648 * Each entry is a 32bit value. Upper part is the high order bits
649 * of flow hash, lower part is cpu number.
650 * rps_cpu_mask is used to partition the space, depending on number of
651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
653 * meaning we use 32-6=26 bits for the hash.
654 */
655 struct rps_sock_flow_table {
656 u32 mask;
657
658 u32 ents[0] ____cacheline_aligned_in_smp;
659 };
660 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
661
662 #define RPS_NO_CPU 0xffff
663
664 extern u32 rps_cpu_mask;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
666
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)667 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
668 u32 hash)
669 {
670 if (table && hash) {
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
673
674 /* We only give a hint, preemption can change cpu under us */
675 val |= raw_smp_processor_id();
676
677 if (table->ents[index] != val)
678 table->ents[index] = val;
679 }
680 }
681
682 #ifdef CONFIG_RFS_ACCEL
683 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
684 u16 filter_id);
685 #endif
686 #endif /* CONFIG_RPS */
687
688 /* This structure contains an instance of an RX queue. */
689 struct netdev_rx_queue {
690 #ifdef CONFIG_RPS
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
693 #endif
694 struct kobject kobj;
695 struct net_device *dev;
696 } ____cacheline_aligned_in_smp;
697
698 /*
699 * RX queue sysfs structures and functions.
700 */
701 struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
707 };
708
709 #ifdef CONFIG_XPS
710 /*
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
713 */
714 struct xps_map {
715 unsigned int len;
716 unsigned int alloc_len;
717 struct rcu_head rcu;
718 u16 queues[0];
719 };
720 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
721 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
722 - sizeof(struct xps_map)) / sizeof(u16))
723
724 /*
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 */
727 struct xps_dev_maps {
728 struct rcu_head rcu;
729 struct xps_map __rcu *cpu_map[0];
730 };
731 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733 #endif /* CONFIG_XPS */
734
735 #define TC_MAX_QUEUE 16
736 #define TC_BITMASK 15
737 /* HW offloaded queuing disciplines txq count and offset maps */
738 struct netdev_tc_txq {
739 u16 count;
740 u16 offset;
741 };
742
743 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744 /*
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
747 */
748 struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
755 char model[256];
756 char model_description[256];
757 };
758 #endif
759
760 #define MAX_PHYS_ITEM_ID_LEN 32
761
762 /* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
764 */
765 struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
767 unsigned char id_len;
768 };
769
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)770 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
771 struct netdev_phys_item_id *b)
772 {
773 return a->id_len == b->id_len &&
774 memcmp(a->id, b->id, a->id_len) == 0;
775 }
776
777 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
778 struct sk_buff *skb);
779
780 /*
781 * This structure defines the management hooks for network devices.
782 * The following hooks can be defined; unless noted otherwise, they are
783 * optional and can be filled with a null pointer.
784 *
785 * int (*ndo_init)(struct net_device *dev);
786 * This function is called once when network device is registered.
787 * The network device can use this to any late stage initializaton
788 * or semantic validattion. It can fail with an error code which will
789 * be propogated back to register_netdev
790 *
791 * void (*ndo_uninit)(struct net_device *dev);
792 * This function is called when device is unregistered or when registration
793 * fails. It is not called if init fails.
794 *
795 * int (*ndo_open)(struct net_device *dev);
796 * This function is called when network device transistions to the up
797 * state.
798 *
799 * int (*ndo_stop)(struct net_device *dev);
800 * This function is called when network device transistions to the down
801 * state.
802 *
803 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
804 * struct net_device *dev);
805 * Called when a packet needs to be transmitted.
806 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
807 * the queue before that can happen; it's for obsolete devices and weird
808 * corner cases, but the stack really does a non-trivial amount
809 * of useless work if you return NETDEV_TX_BUSY.
810 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
811 * Required can not be NULL.
812 *
813 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
814 * void *accel_priv, select_queue_fallback_t fallback);
815 * Called to decide which queue to when device supports multiple
816 * transmit queues.
817 *
818 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
819 * This function is called to allow device receiver to make
820 * changes to configuration when multicast or promiscious is enabled.
821 *
822 * void (*ndo_set_rx_mode)(struct net_device *dev);
823 * This function is called device changes address list filtering.
824 * If driver handles unicast address filtering, it should set
825 * IFF_UNICAST_FLT to its priv_flags.
826 *
827 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
828 * This function is called when the Media Access Control address
829 * needs to be changed. If this interface is not defined, the
830 * mac address can not be changed.
831 *
832 * int (*ndo_validate_addr)(struct net_device *dev);
833 * Test if Media Access Control address is valid for the device.
834 *
835 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
836 * Called when a user request an ioctl which can't be handled by
837 * the generic interface code. If not defined ioctl's return
838 * not supported error code.
839 *
840 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
841 * Used to set network devices bus interface parameters. This interface
842 * is retained for legacy reason, new devices should use the bus
843 * interface (PCI) for low level management.
844 *
845 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
846 * Called when a user wants to change the Maximum Transfer Unit
847 * of a device. If not defined, any request to change MTU will
848 * will return an error.
849 *
850 * void (*ndo_tx_timeout)(struct net_device *dev);
851 * Callback uses when the transmitter has not made any progress
852 * for dev->watchdog ticks.
853 *
854 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
855 * struct rtnl_link_stats64 *storage);
856 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
857 * Called when a user wants to get the network device usage
858 * statistics. Drivers must do one of the following:
859 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
860 * rtnl_link_stats64 structure passed by the caller.
861 * 2. Define @ndo_get_stats to update a net_device_stats structure
862 * (which should normally be dev->stats) and return a pointer to
863 * it. The structure may be changed asynchronously only if each
864 * field is written atomically.
865 * 3. Update dev->stats asynchronously and atomically, and define
866 * neither operation.
867 *
868 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
869 * If device support VLAN filtering this function is called when a
870 * VLAN id is registered.
871 *
872 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
873 * If device support VLAN filtering this function is called when a
874 * VLAN id is unregistered.
875 *
876 * void (*ndo_poll_controller)(struct net_device *dev);
877 *
878 * SR-IOV management functions.
879 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
880 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
881 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
882 * int max_tx_rate);
883 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
884 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
885 * int (*ndo_get_vf_config)(struct net_device *dev,
886 * int vf, struct ifla_vf_info *ivf);
887 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
888 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
889 * struct nlattr *port[]);
890 *
891 * Enable or disable the VF ability to query its RSS Redirection Table and
892 * Hash Key. This is needed since on some devices VF share this information
893 * with PF and querying it may adduce a theoretical security risk.
894 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
895 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
896 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
897 * Called to setup 'tc' number of traffic classes in the net device. This
898 * is always called from the stack with the rtnl lock held and netif tx
899 * queues stopped. This allows the netdevice to perform queue management
900 * safely.
901 *
902 * Fiber Channel over Ethernet (FCoE) offload functions.
903 * int (*ndo_fcoe_enable)(struct net_device *dev);
904 * Called when the FCoE protocol stack wants to start using LLD for FCoE
905 * so the underlying device can perform whatever needed configuration or
906 * initialization to support acceleration of FCoE traffic.
907 *
908 * int (*ndo_fcoe_disable)(struct net_device *dev);
909 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
910 * so the underlying device can perform whatever needed clean-ups to
911 * stop supporting acceleration of FCoE traffic.
912 *
913 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
914 * struct scatterlist *sgl, unsigned int sgc);
915 * Called when the FCoE Initiator wants to initialize an I/O that
916 * is a possible candidate for Direct Data Placement (DDP). The LLD can
917 * perform necessary setup and returns 1 to indicate the device is set up
918 * successfully to perform DDP on this I/O, otherwise this returns 0.
919 *
920 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
921 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
922 * indicated by the FC exchange id 'xid', so the underlying device can
923 * clean up and reuse resources for later DDP requests.
924 *
925 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
926 * struct scatterlist *sgl, unsigned int sgc);
927 * Called when the FCoE Target wants to initialize an I/O that
928 * is a possible candidate for Direct Data Placement (DDP). The LLD can
929 * perform necessary setup and returns 1 to indicate the device is set up
930 * successfully to perform DDP on this I/O, otherwise this returns 0.
931 *
932 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
933 * struct netdev_fcoe_hbainfo *hbainfo);
934 * Called when the FCoE Protocol stack wants information on the underlying
935 * device. This information is utilized by the FCoE protocol stack to
936 * register attributes with Fiber Channel management service as per the
937 * FC-GS Fabric Device Management Information(FDMI) specification.
938 *
939 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
940 * Called when the underlying device wants to override default World Wide
941 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
942 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
943 * protocol stack to use.
944 *
945 * RFS acceleration.
946 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
947 * u16 rxq_index, u32 flow_id);
948 * Set hardware filter for RFS. rxq_index is the target queue index;
949 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
950 * Return the filter ID on success, or a negative error code.
951 *
952 * Slave management functions (for bridge, bonding, etc).
953 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
954 * Called to make another netdev an underling.
955 *
956 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
957 * Called to release previously enslaved netdev.
958 *
959 * Feature/offload setting functions.
960 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
961 * netdev_features_t features);
962 * Adjusts the requested feature flags according to device-specific
963 * constraints, and returns the resulting flags. Must not modify
964 * the device state.
965 *
966 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
967 * Called to update device configuration to new features. Passed
968 * feature set might be less than what was returned by ndo_fix_features()).
969 * Must return >0 or -errno if it changed dev->features itself.
970 *
971 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
972 * struct net_device *dev,
973 * const unsigned char *addr, u16 vid, u16 flags)
974 * Adds an FDB entry to dev for addr.
975 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
976 * struct net_device *dev,
977 * const unsigned char *addr, u16 vid)
978 * Deletes the FDB entry from dev coresponding to addr.
979 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
980 * struct net_device *dev, struct net_device *filter_dev,
981 * int idx)
982 * Used to add FDB entries to dump requests. Implementers should add
983 * entries to skb and update idx with the number of entries.
984 *
985 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
986 * u16 flags)
987 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
988 * struct net_device *dev, u32 filter_mask,
989 * int nlflags)
990 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
991 * u16 flags);
992 *
993 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
994 * Called to change device carrier. Soft-devices (like dummy, team, etc)
995 * which do not represent real hardware may define this to allow their
996 * userspace components to manage their virtual carrier state. Devices
997 * that determine carrier state from physical hardware properties (eg
998 * network cables) or protocol-dependent mechanisms (eg
999 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1000 *
1001 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1002 * struct netdev_phys_item_id *ppid);
1003 * Called to get ID of physical port of this device. If driver does
1004 * not implement this, it is assumed that the hw is not able to have
1005 * multiple net devices on single physical port.
1006 *
1007 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1008 * sa_family_t sa_family, __be16 port);
1009 * Called by vxlan to notiy a driver about the UDP port and socket
1010 * address family that vxlan is listnening to. It is called only when
1011 * a new port starts listening. The operation is protected by the
1012 * vxlan_net->sock_lock.
1013 *
1014 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1015 * sa_family_t sa_family, __be16 port);
1016 * Called by vxlan to notify the driver about a UDP port and socket
1017 * address family that vxlan is not listening to anymore. The operation
1018 * is protected by the vxlan_net->sock_lock.
1019 *
1020 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1021 * struct net_device *dev)
1022 * Called by upper layer devices to accelerate switching or other
1023 * station functionality into hardware. 'pdev is the lowerdev
1024 * to use for the offload and 'dev' is the net device that will
1025 * back the offload. Returns a pointer to the private structure
1026 * the upper layer will maintain.
1027 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1028 * Called by upper layer device to delete the station created
1029 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1030 * the station and priv is the structure returned by the add
1031 * operation.
1032 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1033 * struct net_device *dev,
1034 * void *priv);
1035 * Callback to use for xmit over the accelerated station. This
1036 * is used in place of ndo_start_xmit on accelerated net
1037 * devices.
1038 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1039 * struct net_device *dev
1040 * netdev_features_t features);
1041 * Called by core transmit path to determine if device is capable of
1042 * performing offload operations on a given packet. This is to give
1043 * the device an opportunity to implement any restrictions that cannot
1044 * be otherwise expressed by feature flags. The check is called with
1045 * the set of features that the stack has calculated and it returns
1046 * those the driver believes to be appropriate.
1047 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1048 * int queue_index, u32 maxrate);
1049 * Called when a user wants to set a max-rate limitation of specific
1050 * TX queue.
1051 * int (*ndo_get_iflink)(const struct net_device *dev);
1052 * Called to get the iflink value of this device.
1053 * void (*ndo_change_proto_down)(struct net_device *dev,
1054 * bool proto_down);
1055 * This function is used to pass protocol port error state information
1056 * to the switch driver. The switch driver can react to the proto_down
1057 * by doing a phys down on the associated switch port.
1058 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1059 * This function is used to get egress tunnel information for given skb.
1060 * This is useful for retrieving outer tunnel header parameters while
1061 * sampling packet.
1062 *
1063 */
1064 struct net_device_ops {
1065 int (*ndo_init)(struct net_device *dev);
1066 void (*ndo_uninit)(struct net_device *dev);
1067 int (*ndo_open)(struct net_device *dev);
1068 int (*ndo_stop)(struct net_device *dev);
1069 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1070 struct net_device *dev);
1071 u16 (*ndo_select_queue)(struct net_device *dev,
1072 struct sk_buff *skb,
1073 void *accel_priv,
1074 select_queue_fallback_t fallback);
1075 void (*ndo_change_rx_flags)(struct net_device *dev,
1076 int flags);
1077 void (*ndo_set_rx_mode)(struct net_device *dev);
1078 int (*ndo_set_mac_address)(struct net_device *dev,
1079 void *addr);
1080 int (*ndo_validate_addr)(struct net_device *dev);
1081 int (*ndo_do_ioctl)(struct net_device *dev,
1082 struct ifreq *ifr, int cmd);
1083 int (*ndo_set_config)(struct net_device *dev,
1084 struct ifmap *map);
1085 int (*ndo_change_mtu)(struct net_device *dev,
1086 int new_mtu);
1087 int (*ndo_neigh_setup)(struct net_device *dev,
1088 struct neigh_parms *);
1089 void (*ndo_tx_timeout) (struct net_device *dev);
1090
1091 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1092 struct rtnl_link_stats64 *storage);
1093 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1094
1095 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1096 __be16 proto, u16 vid);
1097 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1098 __be16 proto, u16 vid);
1099 #ifdef CONFIG_NET_POLL_CONTROLLER
1100 void (*ndo_poll_controller)(struct net_device *dev);
1101 int (*ndo_netpoll_setup)(struct net_device *dev,
1102 struct netpoll_info *info);
1103 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1104 #endif
1105 #ifdef CONFIG_NET_RX_BUSY_POLL
1106 int (*ndo_busy_poll)(struct napi_struct *dev);
1107 #endif
1108 int (*ndo_set_vf_mac)(struct net_device *dev,
1109 int queue, u8 *mac);
1110 int (*ndo_set_vf_vlan)(struct net_device *dev,
1111 int queue, u16 vlan, u8 qos);
1112 int (*ndo_set_vf_rate)(struct net_device *dev,
1113 int vf, int min_tx_rate,
1114 int max_tx_rate);
1115 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1116 int vf, bool setting);
1117 int (*ndo_set_vf_trust)(struct net_device *dev,
1118 int vf, bool setting);
1119 int (*ndo_get_vf_config)(struct net_device *dev,
1120 int vf,
1121 struct ifla_vf_info *ivf);
1122 int (*ndo_set_vf_link_state)(struct net_device *dev,
1123 int vf, int link_state);
1124 int (*ndo_get_vf_stats)(struct net_device *dev,
1125 int vf,
1126 struct ifla_vf_stats
1127 *vf_stats);
1128 int (*ndo_set_vf_port)(struct net_device *dev,
1129 int vf,
1130 struct nlattr *port[]);
1131 int (*ndo_get_vf_port)(struct net_device *dev,
1132 int vf, struct sk_buff *skb);
1133 int (*ndo_set_vf_rss_query_en)(
1134 struct net_device *dev,
1135 int vf, bool setting);
1136 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1137 #if IS_ENABLED(CONFIG_FCOE)
1138 int (*ndo_fcoe_enable)(struct net_device *dev);
1139 int (*ndo_fcoe_disable)(struct net_device *dev);
1140 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1141 u16 xid,
1142 struct scatterlist *sgl,
1143 unsigned int sgc);
1144 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1145 u16 xid);
1146 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1147 u16 xid,
1148 struct scatterlist *sgl,
1149 unsigned int sgc);
1150 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1151 struct netdev_fcoe_hbainfo *hbainfo);
1152 #endif
1153
1154 #if IS_ENABLED(CONFIG_LIBFCOE)
1155 #define NETDEV_FCOE_WWNN 0
1156 #define NETDEV_FCOE_WWPN 1
1157 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1158 u64 *wwn, int type);
1159 #endif
1160
1161 #ifdef CONFIG_RFS_ACCEL
1162 int (*ndo_rx_flow_steer)(struct net_device *dev,
1163 const struct sk_buff *skb,
1164 u16 rxq_index,
1165 u32 flow_id);
1166 #endif
1167 int (*ndo_add_slave)(struct net_device *dev,
1168 struct net_device *slave_dev);
1169 int (*ndo_del_slave)(struct net_device *dev,
1170 struct net_device *slave_dev);
1171 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1172 netdev_features_t features);
1173 int (*ndo_set_features)(struct net_device *dev,
1174 netdev_features_t features);
1175 int (*ndo_neigh_construct)(struct neighbour *n);
1176 void (*ndo_neigh_destroy)(struct neighbour *n);
1177
1178 int (*ndo_fdb_add)(struct ndmsg *ndm,
1179 struct nlattr *tb[],
1180 struct net_device *dev,
1181 const unsigned char *addr,
1182 u16 vid,
1183 u16 flags);
1184 int (*ndo_fdb_del)(struct ndmsg *ndm,
1185 struct nlattr *tb[],
1186 struct net_device *dev,
1187 const unsigned char *addr,
1188 u16 vid);
1189 int (*ndo_fdb_dump)(struct sk_buff *skb,
1190 struct netlink_callback *cb,
1191 struct net_device *dev,
1192 struct net_device *filter_dev,
1193 int idx);
1194
1195 int (*ndo_bridge_setlink)(struct net_device *dev,
1196 struct nlmsghdr *nlh,
1197 u16 flags);
1198 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1199 u32 pid, u32 seq,
1200 struct net_device *dev,
1201 u32 filter_mask,
1202 int nlflags);
1203 int (*ndo_bridge_dellink)(struct net_device *dev,
1204 struct nlmsghdr *nlh,
1205 u16 flags);
1206 int (*ndo_change_carrier)(struct net_device *dev,
1207 bool new_carrier);
1208 int (*ndo_get_phys_port_id)(struct net_device *dev,
1209 struct netdev_phys_item_id *ppid);
1210 int (*ndo_get_phys_port_name)(struct net_device *dev,
1211 char *name, size_t len);
1212 void (*ndo_add_vxlan_port)(struct net_device *dev,
1213 sa_family_t sa_family,
1214 __be16 port);
1215 void (*ndo_del_vxlan_port)(struct net_device *dev,
1216 sa_family_t sa_family,
1217 __be16 port);
1218
1219 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1220 struct net_device *dev);
1221 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1222 void *priv);
1223
1224 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1225 struct net_device *dev,
1226 void *priv);
1227 int (*ndo_get_lock_subclass)(struct net_device *dev);
1228 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1229 struct net_device *dev,
1230 netdev_features_t features);
1231 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1232 int queue_index,
1233 u32 maxrate);
1234 int (*ndo_get_iflink)(const struct net_device *dev);
1235 int (*ndo_change_proto_down)(struct net_device *dev,
1236 bool proto_down);
1237 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1238 struct sk_buff *skb);
1239 };
1240
1241 /**
1242 * enum net_device_priv_flags - &struct net_device priv_flags
1243 *
1244 * These are the &struct net_device, they are only set internally
1245 * by drivers and used in the kernel. These flags are invisible to
1246 * userspace, this means that the order of these flags can change
1247 * during any kernel release.
1248 *
1249 * You should have a pretty good reason to be extending these flags.
1250 *
1251 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1252 * @IFF_EBRIDGE: Ethernet bridging device
1253 * @IFF_BONDING: bonding master or slave
1254 * @IFF_ISATAP: ISATAP interface (RFC4214)
1255 * @IFF_WAN_HDLC: WAN HDLC device
1256 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1257 * release skb->dst
1258 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1259 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1260 * @IFF_MACVLAN_PORT: device used as macvlan port
1261 * @IFF_BRIDGE_PORT: device used as bridge port
1262 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1263 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1264 * @IFF_UNICAST_FLT: Supports unicast filtering
1265 * @IFF_TEAM_PORT: device used as team port
1266 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1267 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1268 * change when it's running
1269 * @IFF_MACVLAN: Macvlan device
1270 * @IFF_L3MDEV_MASTER: device is an L3 master device
1271 * @IFF_NO_QUEUE: device can run without qdisc attached
1272 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1273 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1274 */
1275 enum netdev_priv_flags {
1276 IFF_802_1Q_VLAN = 1<<0,
1277 IFF_EBRIDGE = 1<<1,
1278 IFF_BONDING = 1<<2,
1279 IFF_ISATAP = 1<<3,
1280 IFF_WAN_HDLC = 1<<4,
1281 IFF_XMIT_DST_RELEASE = 1<<5,
1282 IFF_DONT_BRIDGE = 1<<6,
1283 IFF_DISABLE_NETPOLL = 1<<7,
1284 IFF_MACVLAN_PORT = 1<<8,
1285 IFF_BRIDGE_PORT = 1<<9,
1286 IFF_OVS_DATAPATH = 1<<10,
1287 IFF_TX_SKB_SHARING = 1<<11,
1288 IFF_UNICAST_FLT = 1<<12,
1289 IFF_TEAM_PORT = 1<<13,
1290 IFF_SUPP_NOFCS = 1<<14,
1291 IFF_LIVE_ADDR_CHANGE = 1<<15,
1292 IFF_MACVLAN = 1<<16,
1293 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1294 IFF_IPVLAN_MASTER = 1<<18,
1295 IFF_IPVLAN_SLAVE = 1<<19,
1296 IFF_L3MDEV_MASTER = 1<<20,
1297 IFF_NO_QUEUE = 1<<21,
1298 IFF_OPENVSWITCH = 1<<22,
1299 IFF_L3MDEV_SLAVE = 1<<23,
1300 };
1301
1302 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1303 #define IFF_EBRIDGE IFF_EBRIDGE
1304 #define IFF_BONDING IFF_BONDING
1305 #define IFF_ISATAP IFF_ISATAP
1306 #define IFF_WAN_HDLC IFF_WAN_HDLC
1307 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1308 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1309 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1310 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1311 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1312 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1313 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1314 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1315 #define IFF_TEAM_PORT IFF_TEAM_PORT
1316 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1317 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1318 #define IFF_MACVLAN IFF_MACVLAN
1319 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1320 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1321 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1322 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1323 #define IFF_NO_QUEUE IFF_NO_QUEUE
1324 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1325 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1326
1327 /**
1328 * struct net_device - The DEVICE structure.
1329 * Actually, this whole structure is a big mistake. It mixes I/O
1330 * data with strictly "high-level" data, and it has to know about
1331 * almost every data structure used in the INET module.
1332 *
1333 * @name: This is the first field of the "visible" part of this structure
1334 * (i.e. as seen by users in the "Space.c" file). It is the name
1335 * of the interface.
1336 *
1337 * @name_hlist: Device name hash chain, please keep it close to name[]
1338 * @ifalias: SNMP alias
1339 * @mem_end: Shared memory end
1340 * @mem_start: Shared memory start
1341 * @base_addr: Device I/O address
1342 * @irq: Device IRQ number
1343 *
1344 * @carrier_changes: Stats to monitor carrier on<->off transitions
1345 *
1346 * @state: Generic network queuing layer state, see netdev_state_t
1347 * @dev_list: The global list of network devices
1348 * @napi_list: List entry, that is used for polling napi devices
1349 * @unreg_list: List entry, that is used, when we are unregistering the
1350 * device, see the function unregister_netdev
1351 * @close_list: List entry, that is used, when we are closing the device
1352 *
1353 * @adj_list: Directly linked devices, like slaves for bonding
1354 * @all_adj_list: All linked devices, *including* neighbours
1355 * @features: Currently active device features
1356 * @hw_features: User-changeable features
1357 *
1358 * @wanted_features: User-requested features
1359 * @vlan_features: Mask of features inheritable by VLAN devices
1360 *
1361 * @hw_enc_features: Mask of features inherited by encapsulating devices
1362 * This field indicates what encapsulation
1363 * offloads the hardware is capable of doing,
1364 * and drivers will need to set them appropriately.
1365 *
1366 * @mpls_features: Mask of features inheritable by MPLS
1367 *
1368 * @ifindex: interface index
1369 * @group: The group, that the device belongs to
1370 *
1371 * @stats: Statistics struct, which was left as a legacy, use
1372 * rtnl_link_stats64 instead
1373 *
1374 * @rx_dropped: Dropped packets by core network,
1375 * do not use this in drivers
1376 * @tx_dropped: Dropped packets by core network,
1377 * do not use this in drivers
1378 *
1379 * @wireless_handlers: List of functions to handle Wireless Extensions,
1380 * instead of ioctl,
1381 * see <net/iw_handler.h> for details.
1382 * @wireless_data: Instance data managed by the core of wireless extensions
1383 *
1384 * @netdev_ops: Includes several pointers to callbacks,
1385 * if one wants to override the ndo_*() functions
1386 * @ethtool_ops: Management operations
1387 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1388 * of Layer 2 headers.
1389 *
1390 * @flags: Interface flags (a la BSD)
1391 * @priv_flags: Like 'flags' but invisible to userspace,
1392 * see if.h for the definitions
1393 * @gflags: Global flags ( kept as legacy )
1394 * @padded: How much padding added by alloc_netdev()
1395 * @operstate: RFC2863 operstate
1396 * @link_mode: Mapping policy to operstate
1397 * @if_port: Selectable AUI, TP, ...
1398 * @dma: DMA channel
1399 * @mtu: Interface MTU value
1400 * @type: Interface hardware type
1401 * @hard_header_len: Maximum hardware header length.
1402 * @min_header_len: Minimum hardware header length
1403 *
1404 * @needed_headroom: Extra headroom the hardware may need, but not in all
1405 * cases can this be guaranteed
1406 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1407 * cases can this be guaranteed. Some cases also use
1408 * LL_MAX_HEADER instead to allocate the skb
1409 *
1410 * interface address info:
1411 *
1412 * @perm_addr: Permanent hw address
1413 * @addr_assign_type: Hw address assignment type
1414 * @addr_len: Hardware address length
1415 * @neigh_priv_len; Used in neigh_alloc(),
1416 * initialized only in atm/clip.c
1417 * @dev_id: Used to differentiate devices that share
1418 * the same link layer address
1419 * @dev_port: Used to differentiate devices that share
1420 * the same function
1421 * @addr_list_lock: XXX: need comments on this one
1422 * @uc_promisc: Counter, that indicates, that promiscuous mode
1423 * has been enabled due to the need to listen to
1424 * additional unicast addresses in a device that
1425 * does not implement ndo_set_rx_mode()
1426 * @uc: unicast mac addresses
1427 * @mc: multicast mac addresses
1428 * @dev_addrs: list of device hw addresses
1429 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1430 * @promiscuity: Number of times, the NIC is told to work in
1431 * Promiscuous mode, if it becomes 0 the NIC will
1432 * exit from working in Promiscuous mode
1433 * @allmulti: Counter, enables or disables allmulticast mode
1434 *
1435 * @vlan_info: VLAN info
1436 * @dsa_ptr: dsa specific data
1437 * @tipc_ptr: TIPC specific data
1438 * @atalk_ptr: AppleTalk link
1439 * @ip_ptr: IPv4 specific data
1440 * @dn_ptr: DECnet specific data
1441 * @ip6_ptr: IPv6 specific data
1442 * @ax25_ptr: AX.25 specific data
1443 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1444 *
1445 * @last_rx: Time of last Rx
1446 * @dev_addr: Hw address (before bcast,
1447 * because most packets are unicast)
1448 *
1449 * @_rx: Array of RX queues
1450 * @num_rx_queues: Number of RX queues
1451 * allocated at register_netdev() time
1452 * @real_num_rx_queues: Number of RX queues currently active in device
1453 *
1454 * @rx_handler: handler for received packets
1455 * @rx_handler_data: XXX: need comments on this one
1456 * @ingress_queue: XXX: need comments on this one
1457 * @broadcast: hw bcast address
1458 *
1459 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1460 * indexed by RX queue number. Assigned by driver.
1461 * This must only be set if the ndo_rx_flow_steer
1462 * operation is defined
1463 * @index_hlist: Device index hash chain
1464 *
1465 * @_tx: Array of TX queues
1466 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1467 * @real_num_tx_queues: Number of TX queues currently active in device
1468 * @qdisc: Root qdisc from userspace point of view
1469 * @tx_queue_len: Max frames per queue allowed
1470 * @tx_global_lock: XXX: need comments on this one
1471 *
1472 * @xps_maps: XXX: need comments on this one
1473 *
1474 * @offload_fwd_mark: Offload device fwding mark
1475 *
1476 * @trans_start: Time (in jiffies) of last Tx
1477 * @watchdog_timeo: Represents the timeout that is used by
1478 * the watchdog ( see dev_watchdog() )
1479 * @watchdog_timer: List of timers
1480 *
1481 * @pcpu_refcnt: Number of references to this device
1482 * @todo_list: Delayed register/unregister
1483 * @link_watch_list: XXX: need comments on this one
1484 *
1485 * @reg_state: Register/unregister state machine
1486 * @dismantle: Device is going to be freed
1487 * @rtnl_link_state: This enum represents the phases of creating
1488 * a new link
1489 *
1490 * @destructor: Called from unregister,
1491 * can be used to call free_netdev
1492 * @npinfo: XXX: need comments on this one
1493 * @nd_net: Network namespace this network device is inside
1494 *
1495 * @ml_priv: Mid-layer private
1496 * @lstats: Loopback statistics
1497 * @tstats: Tunnel statistics
1498 * @dstats: Dummy statistics
1499 * @vstats: Virtual ethernet statistics
1500 *
1501 * @garp_port: GARP
1502 * @mrp_port: MRP
1503 *
1504 * @dev: Class/net/name entry
1505 * @sysfs_groups: Space for optional device, statistics and wireless
1506 * sysfs groups
1507 *
1508 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1509 * @rtnl_link_ops: Rtnl_link_ops
1510 *
1511 * @gso_max_size: Maximum size of generic segmentation offload
1512 * @gso_max_segs: Maximum number of segments that can be passed to the
1513 * NIC for GSO
1514 * @gso_min_segs: Minimum number of segments that can be passed to the
1515 * NIC for GSO
1516 *
1517 * @dcbnl_ops: Data Center Bridging netlink ops
1518 * @num_tc: Number of traffic classes in the net device
1519 * @tc_to_txq: XXX: need comments on this one
1520 * @prio_tc_map XXX: need comments on this one
1521 *
1522 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1523 *
1524 * @priomap: XXX: need comments on this one
1525 * @phydev: Physical device may attach itself
1526 * for hardware timestamping
1527 *
1528 * @qdisc_tx_busylock: XXX: need comments on this one
1529 *
1530 * @proto_down: protocol port state information can be sent to the
1531 * switch driver and used to set the phys state of the
1532 * switch port.
1533 *
1534 * FIXME: cleanup struct net_device such that network protocol info
1535 * moves out.
1536 */
1537
1538 struct net_device {
1539 char name[IFNAMSIZ];
1540 struct hlist_node name_hlist;
1541 char *ifalias;
1542 /*
1543 * I/O specific fields
1544 * FIXME: Merge these and struct ifmap into one
1545 */
1546 unsigned long mem_end;
1547 unsigned long mem_start;
1548 unsigned long base_addr;
1549 int irq;
1550
1551 atomic_t carrier_changes;
1552
1553 /*
1554 * Some hardware also needs these fields (state,dev_list,
1555 * napi_list,unreg_list,close_list) but they are not
1556 * part of the usual set specified in Space.c.
1557 */
1558
1559 unsigned long state;
1560
1561 struct list_head dev_list;
1562 struct list_head napi_list;
1563 struct list_head unreg_list;
1564 struct list_head close_list;
1565 struct list_head ptype_all;
1566 struct list_head ptype_specific;
1567
1568 struct {
1569 struct list_head upper;
1570 struct list_head lower;
1571 } adj_list;
1572
1573 struct {
1574 struct list_head upper;
1575 struct list_head lower;
1576 } all_adj_list;
1577
1578 netdev_features_t features;
1579 netdev_features_t hw_features;
1580 netdev_features_t wanted_features;
1581 netdev_features_t vlan_features;
1582 netdev_features_t hw_enc_features;
1583 netdev_features_t mpls_features;
1584
1585 int ifindex;
1586 int group;
1587
1588 struct net_device_stats stats;
1589
1590 atomic_long_t rx_dropped;
1591 atomic_long_t tx_dropped;
1592
1593 #ifdef CONFIG_WIRELESS_EXT
1594 const struct iw_handler_def * wireless_handlers;
1595 struct iw_public_data * wireless_data;
1596 #endif
1597 const struct net_device_ops *netdev_ops;
1598 const struct ethtool_ops *ethtool_ops;
1599 #ifdef CONFIG_NET_SWITCHDEV
1600 const struct switchdev_ops *switchdev_ops;
1601 #endif
1602 #ifdef CONFIG_NET_L3_MASTER_DEV
1603 const struct l3mdev_ops *l3mdev_ops;
1604 #endif
1605
1606 const struct header_ops *header_ops;
1607
1608 unsigned int flags;
1609 unsigned int priv_flags;
1610
1611 unsigned short gflags;
1612 unsigned short padded;
1613
1614 unsigned char operstate;
1615 unsigned char link_mode;
1616
1617 unsigned char if_port;
1618 unsigned char dma;
1619
1620 /* Note : dev->mtu is often read without holding a lock.
1621 * Writers usually hold RTNL.
1622 * It is recommended to use READ_ONCE() to annotate the reads,
1623 * and to use WRITE_ONCE() to annotate the writes.
1624 */
1625 unsigned int mtu;
1626 unsigned short type;
1627 unsigned short hard_header_len;
1628 unsigned short min_header_len;
1629
1630 unsigned short needed_headroom;
1631 unsigned short needed_tailroom;
1632
1633 /* Interface address info. */
1634 unsigned char perm_addr[MAX_ADDR_LEN];
1635 unsigned char addr_assign_type;
1636 unsigned char addr_len;
1637 unsigned short neigh_priv_len;
1638 unsigned short dev_id;
1639 unsigned short dev_port;
1640 spinlock_t addr_list_lock;
1641 unsigned char name_assign_type;
1642 bool uc_promisc;
1643 struct netdev_hw_addr_list uc;
1644 struct netdev_hw_addr_list mc;
1645 struct netdev_hw_addr_list dev_addrs;
1646
1647 #ifdef CONFIG_SYSFS
1648 struct kset *queues_kset;
1649 #endif
1650 unsigned int promiscuity;
1651 unsigned int allmulti;
1652
1653
1654 /* Protocol specific pointers */
1655
1656 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1657 struct vlan_info __rcu *vlan_info;
1658 #endif
1659 #if IS_ENABLED(CONFIG_NET_DSA)
1660 struct dsa_switch_tree *dsa_ptr;
1661 #endif
1662 #if IS_ENABLED(CONFIG_TIPC)
1663 struct tipc_bearer __rcu *tipc_ptr;
1664 #endif
1665 void *atalk_ptr;
1666 struct in_device __rcu *ip_ptr;
1667 struct dn_dev __rcu *dn_ptr;
1668 struct inet6_dev __rcu *ip6_ptr;
1669 void *ax25_ptr;
1670 struct wireless_dev *ieee80211_ptr;
1671 struct wpan_dev *ieee802154_ptr;
1672 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1673 struct mpls_dev __rcu *mpls_ptr;
1674 #endif
1675
1676 /*
1677 * Cache lines mostly used on receive path (including eth_type_trans())
1678 */
1679 unsigned long last_rx;
1680
1681 /* Interface address info used in eth_type_trans() */
1682 unsigned char *dev_addr;
1683
1684
1685 #ifdef CONFIG_SYSFS
1686 struct netdev_rx_queue *_rx;
1687
1688 unsigned int num_rx_queues;
1689 unsigned int real_num_rx_queues;
1690
1691 #endif
1692
1693 unsigned long gro_flush_timeout;
1694 rx_handler_func_t __rcu *rx_handler;
1695 void __rcu *rx_handler_data;
1696
1697 #ifdef CONFIG_NET_CLS_ACT
1698 struct tcf_proto __rcu *ingress_cl_list;
1699 #endif
1700 struct netdev_queue __rcu *ingress_queue;
1701 #ifdef CONFIG_NETFILTER_INGRESS
1702 struct list_head nf_hooks_ingress;
1703 #endif
1704
1705 unsigned char broadcast[MAX_ADDR_LEN];
1706 #ifdef CONFIG_RFS_ACCEL
1707 struct cpu_rmap *rx_cpu_rmap;
1708 #endif
1709 struct hlist_node index_hlist;
1710
1711 /*
1712 * Cache lines mostly used on transmit path
1713 */
1714 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1715 unsigned int num_tx_queues;
1716 unsigned int real_num_tx_queues;
1717 struct Qdisc *qdisc;
1718 unsigned long tx_queue_len;
1719 spinlock_t tx_global_lock;
1720 int watchdog_timeo;
1721
1722 #ifdef CONFIG_XPS
1723 struct xps_dev_maps __rcu *xps_maps;
1724 #endif
1725
1726 #ifdef CONFIG_NET_SWITCHDEV
1727 u32 offload_fwd_mark;
1728 #endif
1729
1730 /* These may be needed for future network-power-down code. */
1731
1732 /*
1733 * trans_start here is expensive for high speed devices on SMP,
1734 * please use netdev_queue->trans_start instead.
1735 */
1736 unsigned long trans_start;
1737
1738 struct timer_list watchdog_timer;
1739
1740 int __percpu *pcpu_refcnt;
1741 struct list_head todo_list;
1742
1743 struct list_head link_watch_list;
1744
1745 enum { NETREG_UNINITIALIZED=0,
1746 NETREG_REGISTERED, /* completed register_netdevice */
1747 NETREG_UNREGISTERING, /* called unregister_netdevice */
1748 NETREG_UNREGISTERED, /* completed unregister todo */
1749 NETREG_RELEASED, /* called free_netdev */
1750 NETREG_DUMMY, /* dummy device for NAPI poll */
1751 } reg_state:8;
1752
1753 bool dismantle;
1754
1755 enum {
1756 RTNL_LINK_INITIALIZED,
1757 RTNL_LINK_INITIALIZING,
1758 } rtnl_link_state:16;
1759
1760 void (*destructor)(struct net_device *dev);
1761
1762 #ifdef CONFIG_NETPOLL
1763 struct netpoll_info __rcu *npinfo;
1764 #endif
1765
1766 possible_net_t nd_net;
1767
1768 /* mid-layer private */
1769 union {
1770 void *ml_priv;
1771 struct pcpu_lstats __percpu *lstats;
1772 struct pcpu_sw_netstats __percpu *tstats;
1773 struct pcpu_dstats __percpu *dstats;
1774 struct pcpu_vstats __percpu *vstats;
1775 };
1776
1777 struct garp_port __rcu *garp_port;
1778 struct mrp_port __rcu *mrp_port;
1779
1780 struct device dev;
1781 const struct attribute_group *sysfs_groups[4];
1782 const struct attribute_group *sysfs_rx_queue_group;
1783
1784 const struct rtnl_link_ops *rtnl_link_ops;
1785
1786 /* for setting kernel sock attribute on TCP connection setup */
1787 #define GSO_MAX_SIZE 65536
1788 unsigned int gso_max_size;
1789 #define GSO_MAX_SEGS 65535
1790 u16 gso_max_segs;
1791 u16 gso_min_segs;
1792 #ifdef CONFIG_DCB
1793 const struct dcbnl_rtnl_ops *dcbnl_ops;
1794 #endif
1795 u8 num_tc;
1796 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1797 u8 prio_tc_map[TC_BITMASK + 1];
1798
1799 #if IS_ENABLED(CONFIG_FCOE)
1800 unsigned int fcoe_ddp_xid;
1801 #endif
1802 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1803 struct netprio_map __rcu *priomap;
1804 #endif
1805 struct phy_device *phydev;
1806 struct lock_class_key *qdisc_tx_busylock;
1807 bool proto_down;
1808 };
1809 #define to_net_dev(d) container_of(d, struct net_device, dev)
1810
1811 #define NETDEV_ALIGN 32
1812
1813 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)1814 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1815 {
1816 return dev->prio_tc_map[prio & TC_BITMASK];
1817 }
1818
1819 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)1820 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1821 {
1822 if (tc >= dev->num_tc)
1823 return -EINVAL;
1824
1825 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1826 return 0;
1827 }
1828
1829 static inline
netdev_reset_tc(struct net_device * dev)1830 void netdev_reset_tc(struct net_device *dev)
1831 {
1832 dev->num_tc = 0;
1833 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1834 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1835 }
1836
1837 static inline
netdev_set_tc_queue(struct net_device * dev,u8 tc,u16 count,u16 offset)1838 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1839 {
1840 if (tc >= dev->num_tc)
1841 return -EINVAL;
1842
1843 dev->tc_to_txq[tc].count = count;
1844 dev->tc_to_txq[tc].offset = offset;
1845 return 0;
1846 }
1847
1848 static inline
netdev_set_num_tc(struct net_device * dev,u8 num_tc)1849 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1850 {
1851 if (num_tc > TC_MAX_QUEUE)
1852 return -EINVAL;
1853
1854 dev->num_tc = num_tc;
1855 return 0;
1856 }
1857
1858 static inline
netdev_get_num_tc(struct net_device * dev)1859 int netdev_get_num_tc(struct net_device *dev)
1860 {
1861 return dev->num_tc;
1862 }
1863
1864 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)1865 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1866 unsigned int index)
1867 {
1868 return &dev->_tx[index];
1869 }
1870
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)1871 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1872 const struct sk_buff *skb)
1873 {
1874 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1875 }
1876
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)1877 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1878 void (*f)(struct net_device *,
1879 struct netdev_queue *,
1880 void *),
1881 void *arg)
1882 {
1883 unsigned int i;
1884
1885 for (i = 0; i < dev->num_tx_queues; i++)
1886 f(dev, &dev->_tx[i], arg);
1887 }
1888
1889 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1890 struct sk_buff *skb,
1891 void *accel_priv);
1892
1893 /*
1894 * Net namespace inlines
1895 */
1896 static inline
dev_net(const struct net_device * dev)1897 struct net *dev_net(const struct net_device *dev)
1898 {
1899 return read_pnet(&dev->nd_net);
1900 }
1901
1902 static inline
dev_net_set(struct net_device * dev,struct net * net)1903 void dev_net_set(struct net_device *dev, struct net *net)
1904 {
1905 write_pnet(&dev->nd_net, net);
1906 }
1907
netdev_uses_dsa(struct net_device * dev)1908 static inline bool netdev_uses_dsa(struct net_device *dev)
1909 {
1910 #if IS_ENABLED(CONFIG_NET_DSA)
1911 if (dev->dsa_ptr != NULL)
1912 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1913 #endif
1914 return false;
1915 }
1916
1917 /**
1918 * netdev_priv - access network device private data
1919 * @dev: network device
1920 *
1921 * Get network device private data
1922 */
netdev_priv(const struct net_device * dev)1923 static inline void *netdev_priv(const struct net_device *dev)
1924 {
1925 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1926 }
1927
1928 /* Set the sysfs physical device reference for the network logical device
1929 * if set prior to registration will cause a symlink during initialization.
1930 */
1931 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1932
1933 /* Set the sysfs device type for the network logical device to allow
1934 * fine-grained identification of different network device types. For
1935 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1936 */
1937 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1938
1939 /* Default NAPI poll() weight
1940 * Device drivers are strongly advised to not use bigger value
1941 */
1942 #define NAPI_POLL_WEIGHT 64
1943
1944 /**
1945 * netif_napi_add - initialize a napi context
1946 * @dev: network device
1947 * @napi: napi context
1948 * @poll: polling function
1949 * @weight: default weight
1950 *
1951 * netif_napi_add() must be used to initialize a napi context prior to calling
1952 * *any* of the other napi related functions.
1953 */
1954 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1955 int (*poll)(struct napi_struct *, int), int weight);
1956
1957 /**
1958 * netif_napi_del - remove a napi context
1959 * @napi: napi context
1960 *
1961 * netif_napi_del() removes a napi context from the network device napi list
1962 */
1963 void netif_napi_del(struct napi_struct *napi);
1964
1965 struct napi_gro_cb {
1966 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1967 void *frag0;
1968
1969 /* Length of frag0. */
1970 unsigned int frag0_len;
1971
1972 /* This indicates where we are processing relative to skb->data. */
1973 int data_offset;
1974
1975 /* This is non-zero if the packet cannot be merged with the new skb. */
1976 u16 flush;
1977
1978 /* Save the IP ID here and check when we get to the transport layer */
1979 u16 flush_id;
1980
1981 /* Number of segments aggregated. */
1982 u16 count;
1983
1984 /* Start offset for remote checksum offload */
1985 u16 gro_remcsum_start;
1986
1987 /* jiffies when first packet was created/queued */
1988 unsigned long age;
1989
1990 /* Used in ipv6_gro_receive() and foo-over-udp */
1991 u16 proto;
1992
1993 /* This is non-zero if the packet may be of the same flow. */
1994 u8 same_flow:1;
1995
1996 /* Used in tunnel GRO receive */
1997 u8 encap_mark:1;
1998
1999 /* GRO checksum is valid */
2000 u8 csum_valid:1;
2001
2002 /* Number of checksums via CHECKSUM_UNNECESSARY */
2003 u8 csum_cnt:3;
2004
2005 /* Free the skb? */
2006 u8 free:2;
2007 #define NAPI_GRO_FREE 1
2008 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2009
2010 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2011 u8 is_ipv6:1;
2012
2013 /* Number of gro_receive callbacks this packet already went through */
2014 u8 recursion_counter:4;
2015
2016 /* Used in GRE, set in fou/gue_gro_receive */
2017 u8 is_fou:1;
2018
2019 /* 2 bit hole */
2020
2021 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2022 __wsum csum;
2023
2024 /* used in skb_gro_receive() slow path */
2025 struct sk_buff *last;
2026 };
2027
2028 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2029
2030 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2031 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2032 {
2033 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2034 }
2035
2036 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)2037 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2038 struct sk_buff **head,
2039 struct sk_buff *skb)
2040 {
2041 if (unlikely(gro_recursion_inc_test(skb))) {
2042 NAPI_GRO_CB(skb)->flush |= 1;
2043 return NULL;
2044 }
2045
2046 return cb(head, skb);
2047 }
2048
2049 struct packet_type {
2050 __be16 type; /* This is really htons(ether_type). */
2051 struct net_device *dev; /* NULL is wildcarded here */
2052 int (*func) (struct sk_buff *,
2053 struct net_device *,
2054 struct packet_type *,
2055 struct net_device *);
2056 bool (*id_match)(struct packet_type *ptype,
2057 struct sock *sk);
2058 struct net *af_packet_net;
2059 void *af_packet_priv;
2060 struct list_head list;
2061 };
2062
2063 struct offload_callbacks {
2064 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2065 netdev_features_t features);
2066 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2067 struct sk_buff *skb);
2068 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2069 };
2070
2071 struct packet_offload {
2072 __be16 type; /* This is really htons(ether_type). */
2073 u16 priority;
2074 struct offload_callbacks callbacks;
2075 struct list_head list;
2076 };
2077
2078 struct udp_offload;
2079
2080 struct udp_offload_callbacks {
2081 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2082 struct sk_buff *skb,
2083 struct udp_offload *uoff);
2084 int (*gro_complete)(struct sk_buff *skb,
2085 int nhoff,
2086 struct udp_offload *uoff);
2087 };
2088
2089 struct udp_offload {
2090 __be16 port;
2091 u8 ipproto;
2092 struct udp_offload_callbacks callbacks;
2093 };
2094
2095 typedef struct sk_buff **(*gro_receive_udp_t)(struct sk_buff **,
2096 struct sk_buff *,
2097 struct udp_offload *);
call_gro_receive_udp(gro_receive_udp_t cb,struct sk_buff ** head,struct sk_buff * skb,struct udp_offload * uoff)2098 static inline struct sk_buff **call_gro_receive_udp(gro_receive_udp_t cb,
2099 struct sk_buff **head,
2100 struct sk_buff *skb,
2101 struct udp_offload *uoff)
2102 {
2103 if (unlikely(gro_recursion_inc_test(skb))) {
2104 NAPI_GRO_CB(skb)->flush |= 1;
2105 return NULL;
2106 }
2107
2108 return cb(head, skb, uoff);
2109 }
2110
2111 /* often modified stats are per cpu, other are shared (netdev->stats) */
2112 struct pcpu_sw_netstats {
2113 u64 rx_packets;
2114 u64 rx_bytes;
2115 u64 tx_packets;
2116 u64 tx_bytes;
2117 struct u64_stats_sync syncp;
2118 };
2119
2120 #define __netdev_alloc_pcpu_stats(type, gfp) \
2121 ({ \
2122 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2123 if (pcpu_stats) { \
2124 int __cpu; \
2125 for_each_possible_cpu(__cpu) { \
2126 typeof(type) *stat; \
2127 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2128 u64_stats_init(&stat->syncp); \
2129 } \
2130 } \
2131 pcpu_stats; \
2132 })
2133
2134 #define netdev_alloc_pcpu_stats(type) \
2135 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2136
2137 #include <linux/notifier.h>
2138
2139 /* netdevice notifier chain. Please remember to update the rtnetlink
2140 * notification exclusion list in rtnetlink_event() when adding new
2141 * types.
2142 */
2143 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2144 #define NETDEV_DOWN 0x0002
2145 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2146 detected a hardware crash and restarted
2147 - we can use this eg to kick tcp sessions
2148 once done */
2149 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2150 #define NETDEV_REGISTER 0x0005
2151 #define NETDEV_UNREGISTER 0x0006
2152 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2153 #define NETDEV_CHANGEADDR 0x0008
2154 #define NETDEV_GOING_DOWN 0x0009
2155 #define NETDEV_CHANGENAME 0x000A
2156 #define NETDEV_FEAT_CHANGE 0x000B
2157 #define NETDEV_BONDING_FAILOVER 0x000C
2158 #define NETDEV_PRE_UP 0x000D
2159 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2160 #define NETDEV_POST_TYPE_CHANGE 0x000F
2161 #define NETDEV_POST_INIT 0x0010
2162 #define NETDEV_UNREGISTER_FINAL 0x0011
2163 #define NETDEV_RELEASE 0x0012
2164 #define NETDEV_NOTIFY_PEERS 0x0013
2165 #define NETDEV_JOIN 0x0014
2166 #define NETDEV_CHANGEUPPER 0x0015
2167 #define NETDEV_RESEND_IGMP 0x0016
2168 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2169 #define NETDEV_CHANGEINFODATA 0x0018
2170 #define NETDEV_BONDING_INFO 0x0019
2171 #define NETDEV_PRECHANGEUPPER 0x001A
2172
2173 int register_netdevice_notifier(struct notifier_block *nb);
2174 int unregister_netdevice_notifier(struct notifier_block *nb);
2175
2176 struct netdev_notifier_info {
2177 struct net_device *dev;
2178 };
2179
2180 struct netdev_notifier_info_ext {
2181 struct netdev_notifier_info info; /* must be first */
2182 union {
2183 u32 mtu;
2184 } ext;
2185 };
2186
2187 struct netdev_notifier_change_info {
2188 struct netdev_notifier_info info; /* must be first */
2189 unsigned int flags_changed;
2190 };
2191
2192 struct netdev_notifier_changeupper_info {
2193 struct netdev_notifier_info info; /* must be first */
2194 struct net_device *upper_dev; /* new upper dev */
2195 bool master; /* is upper dev master */
2196 bool linking; /* is the nofication for link or unlink */
2197 };
2198
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2199 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2200 struct net_device *dev)
2201 {
2202 info->dev = dev;
2203 }
2204
2205 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2206 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2207 {
2208 return info->dev;
2209 }
2210
2211 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2212
2213
2214 extern rwlock_t dev_base_lock; /* Device list lock */
2215
2216 #define for_each_netdev(net, d) \
2217 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2218 #define for_each_netdev_reverse(net, d) \
2219 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2220 #define for_each_netdev_rcu(net, d) \
2221 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2222 #define for_each_netdev_safe(net, d, n) \
2223 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2224 #define for_each_netdev_continue(net, d) \
2225 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2226 #define for_each_netdev_continue_rcu(net, d) \
2227 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2228 #define for_each_netdev_in_bond_rcu(bond, slave) \
2229 for_each_netdev_rcu(&init_net, slave) \
2230 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2231 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2232
next_net_device(struct net_device * dev)2233 static inline struct net_device *next_net_device(struct net_device *dev)
2234 {
2235 struct list_head *lh;
2236 struct net *net;
2237
2238 net = dev_net(dev);
2239 lh = dev->dev_list.next;
2240 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2241 }
2242
next_net_device_rcu(struct net_device * dev)2243 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2244 {
2245 struct list_head *lh;
2246 struct net *net;
2247
2248 net = dev_net(dev);
2249 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2250 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2251 }
2252
first_net_device(struct net * net)2253 static inline struct net_device *first_net_device(struct net *net)
2254 {
2255 return list_empty(&net->dev_base_head) ? NULL :
2256 net_device_entry(net->dev_base_head.next);
2257 }
2258
first_net_device_rcu(struct net * net)2259 static inline struct net_device *first_net_device_rcu(struct net *net)
2260 {
2261 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2262
2263 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2264 }
2265
2266 int netdev_boot_setup_check(struct net_device *dev);
2267 unsigned long netdev_boot_base(const char *prefix, int unit);
2268 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2269 const char *hwaddr);
2270 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2271 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2272 void dev_add_pack(struct packet_type *pt);
2273 void dev_remove_pack(struct packet_type *pt);
2274 void __dev_remove_pack(struct packet_type *pt);
2275 void dev_add_offload(struct packet_offload *po);
2276 void dev_remove_offload(struct packet_offload *po);
2277
2278 int dev_get_iflink(const struct net_device *dev);
2279 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2280 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2281 unsigned short mask);
2282 struct net_device *dev_get_by_name(struct net *net, const char *name);
2283 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2284 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2285 int dev_alloc_name(struct net_device *dev, const char *name);
2286 int dev_open(struct net_device *dev);
2287 int dev_close(struct net_device *dev);
2288 int dev_close_many(struct list_head *head, bool unlink);
2289 void dev_disable_lro(struct net_device *dev);
2290 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2291 int dev_queue_xmit(struct sk_buff *skb);
2292 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2293 int register_netdevice(struct net_device *dev);
2294 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2295 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2296 static inline void unregister_netdevice(struct net_device *dev)
2297 {
2298 unregister_netdevice_queue(dev, NULL);
2299 }
2300
2301 int netdev_refcnt_read(const struct net_device *dev);
2302 void free_netdev(struct net_device *dev);
2303 void netdev_freemem(struct net_device *dev);
2304 void synchronize_net(void);
2305 int init_dummy_netdev(struct net_device *dev);
2306
2307 DECLARE_PER_CPU(int, xmit_recursion);
dev_recursion_level(void)2308 static inline int dev_recursion_level(void)
2309 {
2310 return this_cpu_read(xmit_recursion);
2311 }
2312
2313 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2314 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2315 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2316 int netdev_get_name(struct net *net, char *name, int ifindex);
2317 int dev_restart(struct net_device *dev);
2318 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2319
skb_gro_offset(const struct sk_buff * skb)2320 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2321 {
2322 return NAPI_GRO_CB(skb)->data_offset;
2323 }
2324
skb_gro_len(const struct sk_buff * skb)2325 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2326 {
2327 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2328 }
2329
skb_gro_pull(struct sk_buff * skb,unsigned int len)2330 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2331 {
2332 NAPI_GRO_CB(skb)->data_offset += len;
2333 }
2334
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2335 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2336 unsigned int offset)
2337 {
2338 return NAPI_GRO_CB(skb)->frag0 + offset;
2339 }
2340
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2341 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2342 {
2343 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2344 }
2345
skb_gro_frag0_invalidate(struct sk_buff * skb)2346 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2347 {
2348 NAPI_GRO_CB(skb)->frag0 = NULL;
2349 NAPI_GRO_CB(skb)->frag0_len = 0;
2350 }
2351
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2352 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2353 unsigned int offset)
2354 {
2355 if (!pskb_may_pull(skb, hlen))
2356 return NULL;
2357
2358 skb_gro_frag0_invalidate(skb);
2359 return skb->data + offset;
2360 }
2361
skb_gro_network_header(struct sk_buff * skb)2362 static inline void *skb_gro_network_header(struct sk_buff *skb)
2363 {
2364 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2365 skb_network_offset(skb);
2366 }
2367
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2368 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2369 const void *start, unsigned int len)
2370 {
2371 if (NAPI_GRO_CB(skb)->csum_valid)
2372 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2373 csum_partial(start, len, 0));
2374 }
2375
2376 /* GRO checksum functions. These are logical equivalents of the normal
2377 * checksum functions (in skbuff.h) except that they operate on the GRO
2378 * offsets and fields in sk_buff.
2379 */
2380
2381 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2382
skb_at_gro_remcsum_start(struct sk_buff * skb)2383 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2384 {
2385 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2386 }
2387
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2388 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2389 bool zero_okay,
2390 __sum16 check)
2391 {
2392 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2393 skb_checksum_start_offset(skb) <
2394 skb_gro_offset(skb)) &&
2395 !skb_at_gro_remcsum_start(skb) &&
2396 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2397 (!zero_okay || check));
2398 }
2399
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2400 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2401 __wsum psum)
2402 {
2403 if (NAPI_GRO_CB(skb)->csum_valid &&
2404 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2405 return 0;
2406
2407 NAPI_GRO_CB(skb)->csum = psum;
2408
2409 return __skb_gro_checksum_complete(skb);
2410 }
2411
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)2412 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2413 {
2414 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2415 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2416 NAPI_GRO_CB(skb)->csum_cnt--;
2417 } else {
2418 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2419 * verified a new top level checksum or an encapsulated one
2420 * during GRO. This saves work if we fallback to normal path.
2421 */
2422 __skb_incr_checksum_unnecessary(skb);
2423 }
2424 }
2425
2426 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2427 compute_pseudo) \
2428 ({ \
2429 __sum16 __ret = 0; \
2430 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2431 __ret = __skb_gro_checksum_validate_complete(skb, \
2432 compute_pseudo(skb, proto)); \
2433 if (__ret) \
2434 __skb_mark_checksum_bad(skb); \
2435 else \
2436 skb_gro_incr_csum_unnecessary(skb); \
2437 __ret; \
2438 })
2439
2440 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2441 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2442
2443 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2444 compute_pseudo) \
2445 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2446
2447 #define skb_gro_checksum_simple_validate(skb) \
2448 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2449
__skb_gro_checksum_convert_check(struct sk_buff * skb)2450 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2451 {
2452 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2453 !NAPI_GRO_CB(skb)->csum_valid);
2454 }
2455
__skb_gro_checksum_convert(struct sk_buff * skb,__sum16 check,__wsum pseudo)2456 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2457 __sum16 check, __wsum pseudo)
2458 {
2459 NAPI_GRO_CB(skb)->csum = ~pseudo;
2460 NAPI_GRO_CB(skb)->csum_valid = 1;
2461 }
2462
2463 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2464 do { \
2465 if (__skb_gro_checksum_convert_check(skb)) \
2466 __skb_gro_checksum_convert(skb, check, \
2467 compute_pseudo(skb, proto)); \
2468 } while (0)
2469
2470 struct gro_remcsum {
2471 int offset;
2472 __wsum delta;
2473 };
2474
skb_gro_remcsum_init(struct gro_remcsum * grc)2475 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2476 {
2477 grc->offset = 0;
2478 grc->delta = 0;
2479 }
2480
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)2481 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2482 unsigned int off, size_t hdrlen,
2483 int start, int offset,
2484 struct gro_remcsum *grc,
2485 bool nopartial)
2486 {
2487 __wsum delta;
2488 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2489
2490 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2491
2492 if (!nopartial) {
2493 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2494 return ptr;
2495 }
2496
2497 ptr = skb_gro_header_fast(skb, off);
2498 if (skb_gro_header_hard(skb, off + plen)) {
2499 ptr = skb_gro_header_slow(skb, off + plen, off);
2500 if (!ptr)
2501 return NULL;
2502 }
2503
2504 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2505 start, offset);
2506
2507 /* Adjust skb->csum since we changed the packet */
2508 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2509
2510 grc->offset = off + hdrlen + offset;
2511 grc->delta = delta;
2512
2513 return ptr;
2514 }
2515
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)2516 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2517 struct gro_remcsum *grc)
2518 {
2519 void *ptr;
2520 size_t plen = grc->offset + sizeof(u16);
2521
2522 if (!grc->delta)
2523 return;
2524
2525 ptr = skb_gro_header_fast(skb, grc->offset);
2526 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2527 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2528 if (!ptr)
2529 return;
2530 }
2531
2532 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2533 }
2534
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)2535 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2536 unsigned short type,
2537 const void *daddr, const void *saddr,
2538 unsigned int len)
2539 {
2540 if (!dev->header_ops || !dev->header_ops->create)
2541 return 0;
2542
2543 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2544 }
2545
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)2546 static inline int dev_parse_header(const struct sk_buff *skb,
2547 unsigned char *haddr)
2548 {
2549 const struct net_device *dev = skb->dev;
2550
2551 if (!dev->header_ops || !dev->header_ops->parse)
2552 return 0;
2553 return dev->header_ops->parse(skb, haddr);
2554 }
2555
2556 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)2557 static inline bool dev_validate_header(const struct net_device *dev,
2558 char *ll_header, int len)
2559 {
2560 if (likely(len >= dev->hard_header_len))
2561 return true;
2562 if (len < dev->min_header_len)
2563 return false;
2564
2565 if (capable(CAP_SYS_RAWIO)) {
2566 memset(ll_header + len, 0, dev->hard_header_len - len);
2567 return true;
2568 }
2569
2570 if (dev->header_ops && dev->header_ops->validate)
2571 return dev->header_ops->validate(ll_header, len);
2572
2573 return false;
2574 }
2575
2576 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2577 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
unregister_gifconf(unsigned int family)2578 static inline int unregister_gifconf(unsigned int family)
2579 {
2580 return register_gifconf(family, NULL);
2581 }
2582
2583 #ifdef CONFIG_NET_FLOW_LIMIT
2584 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2585 struct sd_flow_limit {
2586 u64 count;
2587 unsigned int num_buckets;
2588 unsigned int history_head;
2589 u16 history[FLOW_LIMIT_HISTORY];
2590 u8 buckets[];
2591 };
2592
2593 extern int netdev_flow_limit_table_len;
2594 #endif /* CONFIG_NET_FLOW_LIMIT */
2595
2596 /*
2597 * Incoming packets are placed on per-cpu queues
2598 */
2599 struct softnet_data {
2600 struct list_head poll_list;
2601 struct sk_buff_head process_queue;
2602
2603 /* stats */
2604 unsigned int processed;
2605 unsigned int time_squeeze;
2606 unsigned int cpu_collision;
2607 unsigned int received_rps;
2608 #ifdef CONFIG_RPS
2609 struct softnet_data *rps_ipi_list;
2610 #endif
2611 #ifdef CONFIG_NET_FLOW_LIMIT
2612 struct sd_flow_limit __rcu *flow_limit;
2613 #endif
2614 struct Qdisc *output_queue;
2615 struct Qdisc **output_queue_tailp;
2616 struct sk_buff *completion_queue;
2617
2618 #ifdef CONFIG_RPS
2619 /* Elements below can be accessed between CPUs for RPS */
2620 struct call_single_data csd ____cacheline_aligned_in_smp;
2621 struct softnet_data *rps_ipi_next;
2622 unsigned int cpu;
2623 unsigned int input_queue_head;
2624 unsigned int input_queue_tail;
2625 #endif
2626 unsigned int dropped;
2627 struct sk_buff_head input_pkt_queue;
2628 struct napi_struct backlog;
2629
2630 };
2631
input_queue_head_incr(struct softnet_data * sd)2632 static inline void input_queue_head_incr(struct softnet_data *sd)
2633 {
2634 #ifdef CONFIG_RPS
2635 sd->input_queue_head++;
2636 #endif
2637 }
2638
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)2639 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2640 unsigned int *qtail)
2641 {
2642 #ifdef CONFIG_RPS
2643 *qtail = ++sd->input_queue_tail;
2644 #endif
2645 }
2646
2647 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2648
2649 void __netif_schedule(struct Qdisc *q);
2650 void netif_schedule_queue(struct netdev_queue *txq);
2651
netif_tx_schedule_all(struct net_device * dev)2652 static inline void netif_tx_schedule_all(struct net_device *dev)
2653 {
2654 unsigned int i;
2655
2656 for (i = 0; i < dev->num_tx_queues; i++)
2657 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2658 }
2659
netif_tx_start_queue(struct netdev_queue * dev_queue)2660 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2661 {
2662 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2663 }
2664
2665 /**
2666 * netif_start_queue - allow transmit
2667 * @dev: network device
2668 *
2669 * Allow upper layers to call the device hard_start_xmit routine.
2670 */
netif_start_queue(struct net_device * dev)2671 static inline void netif_start_queue(struct net_device *dev)
2672 {
2673 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2674 }
2675
netif_tx_start_all_queues(struct net_device * dev)2676 static inline void netif_tx_start_all_queues(struct net_device *dev)
2677 {
2678 unsigned int i;
2679
2680 for (i = 0; i < dev->num_tx_queues; i++) {
2681 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2682 netif_tx_start_queue(txq);
2683 }
2684 }
2685
2686 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2687
2688 /**
2689 * netif_wake_queue - restart transmit
2690 * @dev: network device
2691 *
2692 * Allow upper layers to call the device hard_start_xmit routine.
2693 * Used for flow control when transmit resources are available.
2694 */
netif_wake_queue(struct net_device * dev)2695 static inline void netif_wake_queue(struct net_device *dev)
2696 {
2697 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2698 }
2699
netif_tx_wake_all_queues(struct net_device * dev)2700 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2701 {
2702 unsigned int i;
2703
2704 for (i = 0; i < dev->num_tx_queues; i++) {
2705 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2706 netif_tx_wake_queue(txq);
2707 }
2708 }
2709
netif_tx_stop_queue(struct netdev_queue * dev_queue)2710 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2711 {
2712 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2713 }
2714
2715 /**
2716 * netif_stop_queue - stop transmitted packets
2717 * @dev: network device
2718 *
2719 * Stop upper layers calling the device hard_start_xmit routine.
2720 * Used for flow control when transmit resources are unavailable.
2721 */
netif_stop_queue(struct net_device * dev)2722 static inline void netif_stop_queue(struct net_device *dev)
2723 {
2724 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2725 }
2726
2727 void netif_tx_stop_all_queues(struct net_device *dev);
2728
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)2729 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2730 {
2731 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2732 }
2733
2734 /**
2735 * netif_queue_stopped - test if transmit queue is flowblocked
2736 * @dev: network device
2737 *
2738 * Test if transmit queue on device is currently unable to send.
2739 */
netif_queue_stopped(const struct net_device * dev)2740 static inline bool netif_queue_stopped(const struct net_device *dev)
2741 {
2742 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2743 }
2744
netif_xmit_stopped(const struct netdev_queue * dev_queue)2745 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2746 {
2747 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2748 }
2749
2750 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)2751 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2752 {
2753 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2754 }
2755
2756 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)2757 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2758 {
2759 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2760 }
2761
2762 /**
2763 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2764 * @dev_queue: pointer to transmit queue
2765 *
2766 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2767 * to give appropriate hint to the cpu.
2768 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)2769 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2770 {
2771 #ifdef CONFIG_BQL
2772 prefetchw(&dev_queue->dql.num_queued);
2773 #endif
2774 }
2775
2776 /**
2777 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2778 * @dev_queue: pointer to transmit queue
2779 *
2780 * BQL enabled drivers might use this helper in their TX completion path,
2781 * to give appropriate hint to the cpu.
2782 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)2783 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2784 {
2785 #ifdef CONFIG_BQL
2786 prefetchw(&dev_queue->dql.limit);
2787 #endif
2788 }
2789
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)2790 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2791 unsigned int bytes)
2792 {
2793 #ifdef CONFIG_BQL
2794 dql_queued(&dev_queue->dql, bytes);
2795
2796 if (likely(dql_avail(&dev_queue->dql) >= 0))
2797 return;
2798
2799 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2800
2801 /*
2802 * The XOFF flag must be set before checking the dql_avail below,
2803 * because in netdev_tx_completed_queue we update the dql_completed
2804 * before checking the XOFF flag.
2805 */
2806 smp_mb();
2807
2808 /* check again in case another CPU has just made room avail */
2809 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2810 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2811 #endif
2812 }
2813
2814 /**
2815 * netdev_sent_queue - report the number of bytes queued to hardware
2816 * @dev: network device
2817 * @bytes: number of bytes queued to the hardware device queue
2818 *
2819 * Report the number of bytes queued for sending/completion to the network
2820 * device hardware queue. @bytes should be a good approximation and should
2821 * exactly match netdev_completed_queue() @bytes
2822 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)2823 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2824 {
2825 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2826 }
2827
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)2828 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2829 unsigned int pkts, unsigned int bytes)
2830 {
2831 #ifdef CONFIG_BQL
2832 if (unlikely(!bytes))
2833 return;
2834
2835 dql_completed(&dev_queue->dql, bytes);
2836
2837 /*
2838 * Without the memory barrier there is a small possiblity that
2839 * netdev_tx_sent_queue will miss the update and cause the queue to
2840 * be stopped forever
2841 */
2842 smp_mb();
2843
2844 if (dql_avail(&dev_queue->dql) < 0)
2845 return;
2846
2847 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2848 netif_schedule_queue(dev_queue);
2849 #endif
2850 }
2851
2852 /**
2853 * netdev_completed_queue - report bytes and packets completed by device
2854 * @dev: network device
2855 * @pkts: actual number of packets sent over the medium
2856 * @bytes: actual number of bytes sent over the medium
2857 *
2858 * Report the number of bytes and packets transmitted by the network device
2859 * hardware queue over the physical medium, @bytes must exactly match the
2860 * @bytes amount passed to netdev_sent_queue()
2861 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)2862 static inline void netdev_completed_queue(struct net_device *dev,
2863 unsigned int pkts, unsigned int bytes)
2864 {
2865 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2866 }
2867
netdev_tx_reset_queue(struct netdev_queue * q)2868 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2869 {
2870 #ifdef CONFIG_BQL
2871 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2872 dql_reset(&q->dql);
2873 #endif
2874 }
2875
2876 /**
2877 * netdev_reset_queue - reset the packets and bytes count of a network device
2878 * @dev_queue: network device
2879 *
2880 * Reset the bytes and packet count of a network device and clear the
2881 * software flow control OFF bit for this network device
2882 */
netdev_reset_queue(struct net_device * dev_queue)2883 static inline void netdev_reset_queue(struct net_device *dev_queue)
2884 {
2885 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2886 }
2887
2888 /**
2889 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2890 * @dev: network device
2891 * @queue_index: given tx queue index
2892 *
2893 * Returns 0 if given tx queue index >= number of device tx queues,
2894 * otherwise returns the originally passed tx queue index.
2895 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)2896 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2897 {
2898 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2899 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2900 dev->name, queue_index,
2901 dev->real_num_tx_queues);
2902 return 0;
2903 }
2904
2905 return queue_index;
2906 }
2907
2908 /**
2909 * netif_running - test if up
2910 * @dev: network device
2911 *
2912 * Test if the device has been brought up.
2913 */
netif_running(const struct net_device * dev)2914 static inline bool netif_running(const struct net_device *dev)
2915 {
2916 return test_bit(__LINK_STATE_START, &dev->state);
2917 }
2918
2919 /*
2920 * Routines to manage the subqueues on a device. We only need start
2921 * stop, and a check if it's stopped. All other device management is
2922 * done at the overall netdevice level.
2923 * Also test the device if we're multiqueue.
2924 */
2925
2926 /**
2927 * netif_start_subqueue - allow sending packets on subqueue
2928 * @dev: network device
2929 * @queue_index: sub queue index
2930 *
2931 * Start individual transmit queue of a device with multiple transmit queues.
2932 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)2933 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2934 {
2935 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2936
2937 netif_tx_start_queue(txq);
2938 }
2939
2940 /**
2941 * netif_stop_subqueue - stop sending packets on subqueue
2942 * @dev: network device
2943 * @queue_index: sub queue index
2944 *
2945 * Stop individual transmit queue of a device with multiple transmit queues.
2946 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)2947 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2948 {
2949 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2950 netif_tx_stop_queue(txq);
2951 }
2952
2953 /**
2954 * netif_subqueue_stopped - test status of subqueue
2955 * @dev: network device
2956 * @queue_index: sub queue index
2957 *
2958 * Check individual transmit queue of a device with multiple transmit queues.
2959 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)2960 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2961 u16 queue_index)
2962 {
2963 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2964
2965 return netif_tx_queue_stopped(txq);
2966 }
2967
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)2968 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2969 struct sk_buff *skb)
2970 {
2971 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2972 }
2973
2974 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2975
2976 #ifdef CONFIG_XPS
2977 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2978 u16 index);
2979 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)2980 static inline int netif_set_xps_queue(struct net_device *dev,
2981 const struct cpumask *mask,
2982 u16 index)
2983 {
2984 return 0;
2985 }
2986 #endif
2987
2988 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2989 unsigned int num_tx_queues);
2990
2991 /*
2992 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2993 * as a distribution range limit for the returned value.
2994 */
skb_tx_hash(const struct net_device * dev,struct sk_buff * skb)2995 static inline u16 skb_tx_hash(const struct net_device *dev,
2996 struct sk_buff *skb)
2997 {
2998 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2999 }
3000
3001 /**
3002 * netif_is_multiqueue - test if device has multiple transmit queues
3003 * @dev: network device
3004 *
3005 * Check if device has multiple transmit queues
3006 */
netif_is_multiqueue(const struct net_device * dev)3007 static inline bool netif_is_multiqueue(const struct net_device *dev)
3008 {
3009 return dev->num_tx_queues > 1;
3010 }
3011
3012 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3013
3014 #ifdef CONFIG_SYSFS
3015 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3016 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxq)3017 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3018 unsigned int rxq)
3019 {
3020 return 0;
3021 }
3022 #endif
3023
3024 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3025 static inline unsigned int get_netdev_rx_queue_index(
3026 struct netdev_rx_queue *queue)
3027 {
3028 struct net_device *dev = queue->dev;
3029 int index = queue - dev->_rx;
3030
3031 BUG_ON(index >= dev->num_rx_queues);
3032 return index;
3033 }
3034 #endif
3035
3036 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3037 int netif_get_num_default_rss_queues(void);
3038
3039 enum skb_free_reason {
3040 SKB_REASON_CONSUMED,
3041 SKB_REASON_DROPPED,
3042 };
3043
3044 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3045 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3046
3047 /*
3048 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3049 * interrupt context or with hardware interrupts being disabled.
3050 * (in_irq() || irqs_disabled())
3051 *
3052 * We provide four helpers that can be used in following contexts :
3053 *
3054 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3055 * replacing kfree_skb(skb)
3056 *
3057 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3058 * Typically used in place of consume_skb(skb) in TX completion path
3059 *
3060 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3061 * replacing kfree_skb(skb)
3062 *
3063 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3064 * and consumed a packet. Used in place of consume_skb(skb)
3065 */
dev_kfree_skb_irq(struct sk_buff * skb)3066 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3067 {
3068 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3069 }
3070
dev_consume_skb_irq(struct sk_buff * skb)3071 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3072 {
3073 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3074 }
3075
dev_kfree_skb_any(struct sk_buff * skb)3076 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3077 {
3078 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3079 }
3080
dev_consume_skb_any(struct sk_buff * skb)3081 static inline void dev_consume_skb_any(struct sk_buff *skb)
3082 {
3083 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3084 }
3085
3086 int netif_rx(struct sk_buff *skb);
3087 int netif_rx_ni(struct sk_buff *skb);
3088 int netif_receive_skb(struct sk_buff *skb);
3089 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3090 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3091 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3092 gro_result_t napi_gro_frags(struct napi_struct *napi);
3093 struct packet_offload *gro_find_receive_by_type(__be16 type);
3094 struct packet_offload *gro_find_complete_by_type(__be16 type);
3095
napi_free_frags(struct napi_struct * napi)3096 static inline void napi_free_frags(struct napi_struct *napi)
3097 {
3098 kfree_skb(napi->skb);
3099 napi->skb = NULL;
3100 }
3101
3102 bool netdev_is_rx_handler_busy(struct net_device *dev);
3103 int netdev_rx_handler_register(struct net_device *dev,
3104 rx_handler_func_t *rx_handler,
3105 void *rx_handler_data);
3106 void netdev_rx_handler_unregister(struct net_device *dev);
3107
3108 bool dev_valid_name(const char *name);
3109 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3110 int dev_ethtool(struct net *net, struct ifreq *);
3111 unsigned int dev_get_flags(const struct net_device *);
3112 int __dev_change_flags(struct net_device *, unsigned int flags);
3113 int dev_change_flags(struct net_device *, unsigned int);
3114 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3115 unsigned int gchanges);
3116 int dev_change_name(struct net_device *, const char *);
3117 int dev_set_alias(struct net_device *, const char *, size_t);
3118 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3119 int dev_set_mtu(struct net_device *, int);
3120 void dev_set_group(struct net_device *, int);
3121 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3122 int dev_change_carrier(struct net_device *, bool new_carrier);
3123 int dev_get_phys_port_id(struct net_device *dev,
3124 struct netdev_phys_item_id *ppid);
3125 int dev_get_phys_port_name(struct net_device *dev,
3126 char *name, size_t len);
3127 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3128 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3129 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3130 struct netdev_queue *txq, int *ret);
3131 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3132 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3133 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3134
3135 extern int netdev_budget;
3136
3137 /* Called by rtnetlink.c:rtnl_unlock() */
3138 void netdev_run_todo(void);
3139
3140 /**
3141 * dev_put - release reference to device
3142 * @dev: network device
3143 *
3144 * Release reference to device to allow it to be freed.
3145 */
dev_put(struct net_device * dev)3146 static inline void dev_put(struct net_device *dev)
3147 {
3148 this_cpu_dec(*dev->pcpu_refcnt);
3149 }
3150
3151 /**
3152 * dev_hold - get reference to device
3153 * @dev: network device
3154 *
3155 * Hold reference to device to keep it from being freed.
3156 */
dev_hold(struct net_device * dev)3157 static inline void dev_hold(struct net_device *dev)
3158 {
3159 this_cpu_inc(*dev->pcpu_refcnt);
3160 }
3161
3162 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3163 * and _off may be called from IRQ context, but it is caller
3164 * who is responsible for serialization of these calls.
3165 *
3166 * The name carrier is inappropriate, these functions should really be
3167 * called netif_lowerlayer_*() because they represent the state of any
3168 * kind of lower layer not just hardware media.
3169 */
3170
3171 void linkwatch_init_dev(struct net_device *dev);
3172 void linkwatch_fire_event(struct net_device *dev);
3173 void linkwatch_forget_dev(struct net_device *dev);
3174
3175 /**
3176 * netif_carrier_ok - test if carrier present
3177 * @dev: network device
3178 *
3179 * Check if carrier is present on device
3180 */
netif_carrier_ok(const struct net_device * dev)3181 static inline bool netif_carrier_ok(const struct net_device *dev)
3182 {
3183 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3184 }
3185
3186 unsigned long dev_trans_start(struct net_device *dev);
3187
3188 void __netdev_watchdog_up(struct net_device *dev);
3189
3190 void netif_carrier_on(struct net_device *dev);
3191
3192 void netif_carrier_off(struct net_device *dev);
3193
3194 /**
3195 * netif_dormant_on - mark device as dormant.
3196 * @dev: network device
3197 *
3198 * Mark device as dormant (as per RFC2863).
3199 *
3200 * The dormant state indicates that the relevant interface is not
3201 * actually in a condition to pass packets (i.e., it is not 'up') but is
3202 * in a "pending" state, waiting for some external event. For "on-
3203 * demand" interfaces, this new state identifies the situation where the
3204 * interface is waiting for events to place it in the up state.
3205 *
3206 */
netif_dormant_on(struct net_device * dev)3207 static inline void netif_dormant_on(struct net_device *dev)
3208 {
3209 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3210 linkwatch_fire_event(dev);
3211 }
3212
3213 /**
3214 * netif_dormant_off - set device as not dormant.
3215 * @dev: network device
3216 *
3217 * Device is not in dormant state.
3218 */
netif_dormant_off(struct net_device * dev)3219 static inline void netif_dormant_off(struct net_device *dev)
3220 {
3221 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3222 linkwatch_fire_event(dev);
3223 }
3224
3225 /**
3226 * netif_dormant - test if carrier present
3227 * @dev: network device
3228 *
3229 * Check if carrier is present on device
3230 */
netif_dormant(const struct net_device * dev)3231 static inline bool netif_dormant(const struct net_device *dev)
3232 {
3233 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3234 }
3235
3236
3237 /**
3238 * netif_oper_up - test if device is operational
3239 * @dev: network device
3240 *
3241 * Check if carrier is operational
3242 */
netif_oper_up(const struct net_device * dev)3243 static inline bool netif_oper_up(const struct net_device *dev)
3244 {
3245 return (dev->operstate == IF_OPER_UP ||
3246 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3247 }
3248
3249 /**
3250 * netif_device_present - is device available or removed
3251 * @dev: network device
3252 *
3253 * Check if device has not been removed from system.
3254 */
netif_device_present(struct net_device * dev)3255 static inline bool netif_device_present(struct net_device *dev)
3256 {
3257 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3258 }
3259
3260 void netif_device_detach(struct net_device *dev);
3261
3262 void netif_device_attach(struct net_device *dev);
3263
3264 /*
3265 * Network interface message level settings
3266 */
3267
3268 enum {
3269 NETIF_MSG_DRV = 0x0001,
3270 NETIF_MSG_PROBE = 0x0002,
3271 NETIF_MSG_LINK = 0x0004,
3272 NETIF_MSG_TIMER = 0x0008,
3273 NETIF_MSG_IFDOWN = 0x0010,
3274 NETIF_MSG_IFUP = 0x0020,
3275 NETIF_MSG_RX_ERR = 0x0040,
3276 NETIF_MSG_TX_ERR = 0x0080,
3277 NETIF_MSG_TX_QUEUED = 0x0100,
3278 NETIF_MSG_INTR = 0x0200,
3279 NETIF_MSG_TX_DONE = 0x0400,
3280 NETIF_MSG_RX_STATUS = 0x0800,
3281 NETIF_MSG_PKTDATA = 0x1000,
3282 NETIF_MSG_HW = 0x2000,
3283 NETIF_MSG_WOL = 0x4000,
3284 };
3285
3286 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3287 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3288 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3289 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3290 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3291 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3292 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3293 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3294 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3295 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3296 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3297 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3298 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3299 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3300 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3301
netif_msg_init(int debug_value,int default_msg_enable_bits)3302 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3303 {
3304 /* use default */
3305 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3306 return default_msg_enable_bits;
3307 if (debug_value == 0) /* no output */
3308 return 0;
3309 /* set low N bits */
3310 return (1U << debug_value) - 1;
3311 }
3312
__netif_tx_lock(struct netdev_queue * txq,int cpu)3313 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3314 {
3315 spin_lock(&txq->_xmit_lock);
3316 txq->xmit_lock_owner = cpu;
3317 }
3318
__netif_tx_lock_bh(struct netdev_queue * txq)3319 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3320 {
3321 spin_lock_bh(&txq->_xmit_lock);
3322 txq->xmit_lock_owner = smp_processor_id();
3323 }
3324
__netif_tx_trylock(struct netdev_queue * txq)3325 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3326 {
3327 bool ok = spin_trylock(&txq->_xmit_lock);
3328 if (likely(ok))
3329 txq->xmit_lock_owner = smp_processor_id();
3330 return ok;
3331 }
3332
__netif_tx_unlock(struct netdev_queue * txq)3333 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3334 {
3335 txq->xmit_lock_owner = -1;
3336 spin_unlock(&txq->_xmit_lock);
3337 }
3338
__netif_tx_unlock_bh(struct netdev_queue * txq)3339 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3340 {
3341 txq->xmit_lock_owner = -1;
3342 spin_unlock_bh(&txq->_xmit_lock);
3343 }
3344
txq_trans_update(struct netdev_queue * txq)3345 static inline void txq_trans_update(struct netdev_queue *txq)
3346 {
3347 if (txq->xmit_lock_owner != -1)
3348 txq->trans_start = jiffies;
3349 }
3350
3351 /**
3352 * netif_tx_lock - grab network device transmit lock
3353 * @dev: network device
3354 *
3355 * Get network device transmit lock
3356 */
netif_tx_lock(struct net_device * dev)3357 static inline void netif_tx_lock(struct net_device *dev)
3358 {
3359 unsigned int i;
3360 int cpu;
3361
3362 spin_lock(&dev->tx_global_lock);
3363 cpu = smp_processor_id();
3364 for (i = 0; i < dev->num_tx_queues; i++) {
3365 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3366
3367 /* We are the only thread of execution doing a
3368 * freeze, but we have to grab the _xmit_lock in
3369 * order to synchronize with threads which are in
3370 * the ->hard_start_xmit() handler and already
3371 * checked the frozen bit.
3372 */
3373 __netif_tx_lock(txq, cpu);
3374 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3375 __netif_tx_unlock(txq);
3376 }
3377 }
3378
netif_tx_lock_bh(struct net_device * dev)3379 static inline void netif_tx_lock_bh(struct net_device *dev)
3380 {
3381 local_bh_disable();
3382 netif_tx_lock(dev);
3383 }
3384
netif_tx_unlock(struct net_device * dev)3385 static inline void netif_tx_unlock(struct net_device *dev)
3386 {
3387 unsigned int i;
3388
3389 for (i = 0; i < dev->num_tx_queues; i++) {
3390 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3391
3392 /* No need to grab the _xmit_lock here. If the
3393 * queue is not stopped for another reason, we
3394 * force a schedule.
3395 */
3396 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3397 netif_schedule_queue(txq);
3398 }
3399 spin_unlock(&dev->tx_global_lock);
3400 }
3401
netif_tx_unlock_bh(struct net_device * dev)3402 static inline void netif_tx_unlock_bh(struct net_device *dev)
3403 {
3404 netif_tx_unlock(dev);
3405 local_bh_enable();
3406 }
3407
3408 #define HARD_TX_LOCK(dev, txq, cpu) { \
3409 if ((dev->features & NETIF_F_LLTX) == 0) { \
3410 __netif_tx_lock(txq, cpu); \
3411 } \
3412 }
3413
3414 #define HARD_TX_TRYLOCK(dev, txq) \
3415 (((dev->features & NETIF_F_LLTX) == 0) ? \
3416 __netif_tx_trylock(txq) : \
3417 true )
3418
3419 #define HARD_TX_UNLOCK(dev, txq) { \
3420 if ((dev->features & NETIF_F_LLTX) == 0) { \
3421 __netif_tx_unlock(txq); \
3422 } \
3423 }
3424
netif_tx_disable(struct net_device * dev)3425 static inline void netif_tx_disable(struct net_device *dev)
3426 {
3427 unsigned int i;
3428 int cpu;
3429
3430 local_bh_disable();
3431 cpu = smp_processor_id();
3432 spin_lock(&dev->tx_global_lock);
3433 for (i = 0; i < dev->num_tx_queues; i++) {
3434 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3435
3436 __netif_tx_lock(txq, cpu);
3437 netif_tx_stop_queue(txq);
3438 __netif_tx_unlock(txq);
3439 }
3440 spin_unlock(&dev->tx_global_lock);
3441 local_bh_enable();
3442 }
3443
netif_addr_lock(struct net_device * dev)3444 static inline void netif_addr_lock(struct net_device *dev)
3445 {
3446 spin_lock(&dev->addr_list_lock);
3447 }
3448
netif_addr_lock_nested(struct net_device * dev)3449 static inline void netif_addr_lock_nested(struct net_device *dev)
3450 {
3451 int subclass = SINGLE_DEPTH_NESTING;
3452
3453 if (dev->netdev_ops->ndo_get_lock_subclass)
3454 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3455
3456 spin_lock_nested(&dev->addr_list_lock, subclass);
3457 }
3458
netif_addr_lock_bh(struct net_device * dev)3459 static inline void netif_addr_lock_bh(struct net_device *dev)
3460 {
3461 spin_lock_bh(&dev->addr_list_lock);
3462 }
3463
netif_addr_unlock(struct net_device * dev)3464 static inline void netif_addr_unlock(struct net_device *dev)
3465 {
3466 spin_unlock(&dev->addr_list_lock);
3467 }
3468
netif_addr_unlock_bh(struct net_device * dev)3469 static inline void netif_addr_unlock_bh(struct net_device *dev)
3470 {
3471 spin_unlock_bh(&dev->addr_list_lock);
3472 }
3473
3474 /*
3475 * dev_addrs walker. Should be used only for read access. Call with
3476 * rcu_read_lock held.
3477 */
3478 #define for_each_dev_addr(dev, ha) \
3479 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3480
3481 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3482
3483 void ether_setup(struct net_device *dev);
3484
3485 /* Support for loadable net-drivers */
3486 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3487 unsigned char name_assign_type,
3488 void (*setup)(struct net_device *),
3489 unsigned int txqs, unsigned int rxqs);
3490 int dev_get_valid_name(struct net *net, struct net_device *dev,
3491 const char *name);
3492
3493 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3494 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3495
3496 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3497 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3498 count)
3499
3500 int register_netdev(struct net_device *dev);
3501 void unregister_netdev(struct net_device *dev);
3502
3503 /* General hardware address lists handling functions */
3504 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3505 struct netdev_hw_addr_list *from_list, int addr_len);
3506 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3507 struct netdev_hw_addr_list *from_list, int addr_len);
3508 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3509 struct net_device *dev,
3510 int (*sync)(struct net_device *, const unsigned char *),
3511 int (*unsync)(struct net_device *,
3512 const unsigned char *));
3513 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3514 struct net_device *dev,
3515 int (*unsync)(struct net_device *,
3516 const unsigned char *));
3517 void __hw_addr_init(struct netdev_hw_addr_list *list);
3518
3519 /* Functions used for device addresses handling */
3520 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3521 unsigned char addr_type);
3522 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3523 unsigned char addr_type);
3524 void dev_addr_flush(struct net_device *dev);
3525 int dev_addr_init(struct net_device *dev);
3526
3527 /* Functions used for unicast addresses handling */
3528 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3529 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3530 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3531 int dev_uc_sync(struct net_device *to, struct net_device *from);
3532 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3533 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3534 void dev_uc_flush(struct net_device *dev);
3535 void dev_uc_init(struct net_device *dev);
3536
3537 /**
3538 * __dev_uc_sync - Synchonize device's unicast list
3539 * @dev: device to sync
3540 * @sync: function to call if address should be added
3541 * @unsync: function to call if address should be removed
3542 *
3543 * Add newly added addresses to the interface, and release
3544 * addresses that have been deleted.
3545 **/
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))3546 static inline int __dev_uc_sync(struct net_device *dev,
3547 int (*sync)(struct net_device *,
3548 const unsigned char *),
3549 int (*unsync)(struct net_device *,
3550 const unsigned char *))
3551 {
3552 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3553 }
3554
3555 /**
3556 * __dev_uc_unsync - Remove synchronized addresses from device
3557 * @dev: device to sync
3558 * @unsync: function to call if address should be removed
3559 *
3560 * Remove all addresses that were added to the device by dev_uc_sync().
3561 **/
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))3562 static inline void __dev_uc_unsync(struct net_device *dev,
3563 int (*unsync)(struct net_device *,
3564 const unsigned char *))
3565 {
3566 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3567 }
3568
3569 /* Functions used for multicast addresses handling */
3570 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3571 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3572 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3573 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3574 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3575 int dev_mc_sync(struct net_device *to, struct net_device *from);
3576 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3577 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3578 void dev_mc_flush(struct net_device *dev);
3579 void dev_mc_init(struct net_device *dev);
3580
3581 /**
3582 * __dev_mc_sync - Synchonize device's multicast list
3583 * @dev: device to sync
3584 * @sync: function to call if address should be added
3585 * @unsync: function to call if address should be removed
3586 *
3587 * Add newly added addresses to the interface, and release
3588 * addresses that have been deleted.
3589 **/
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))3590 static inline int __dev_mc_sync(struct net_device *dev,
3591 int (*sync)(struct net_device *,
3592 const unsigned char *),
3593 int (*unsync)(struct net_device *,
3594 const unsigned char *))
3595 {
3596 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3597 }
3598
3599 /**
3600 * __dev_mc_unsync - Remove synchronized addresses from device
3601 * @dev: device to sync
3602 * @unsync: function to call if address should be removed
3603 *
3604 * Remove all addresses that were added to the device by dev_mc_sync().
3605 **/
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))3606 static inline void __dev_mc_unsync(struct net_device *dev,
3607 int (*unsync)(struct net_device *,
3608 const unsigned char *))
3609 {
3610 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3611 }
3612
3613 /* Functions used for secondary unicast and multicast support */
3614 void dev_set_rx_mode(struct net_device *dev);
3615 void __dev_set_rx_mode(struct net_device *dev);
3616 int dev_set_promiscuity(struct net_device *dev, int inc);
3617 int dev_set_allmulti(struct net_device *dev, int inc);
3618 void netdev_state_change(struct net_device *dev);
3619 void netdev_notify_peers(struct net_device *dev);
3620 void netdev_features_change(struct net_device *dev);
3621 /* Load a device via the kmod */
3622 void dev_load(struct net *net, const char *name);
3623 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3624 struct rtnl_link_stats64 *storage);
3625 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3626 const struct net_device_stats *netdev_stats);
3627
3628 extern int netdev_max_backlog;
3629 extern int netdev_tstamp_prequeue;
3630 extern int weight_p;
3631 extern int bpf_jit_enable;
3632
3633 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3634 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3635 struct list_head **iter);
3636 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3637 struct list_head **iter);
3638
3639 /* iterate through upper list, must be called under RCU read lock */
3640 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3641 for (iter = &(dev)->adj_list.upper, \
3642 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3643 updev; \
3644 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3645
3646 /* iterate through upper list, must be called under RCU read lock */
3647 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3648 for (iter = &(dev)->all_adj_list.upper, \
3649 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3650 updev; \
3651 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3652
3653 void *netdev_lower_get_next_private(struct net_device *dev,
3654 struct list_head **iter);
3655 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3656 struct list_head **iter);
3657
3658 #define netdev_for_each_lower_private(dev, priv, iter) \
3659 for (iter = (dev)->adj_list.lower.next, \
3660 priv = netdev_lower_get_next_private(dev, &(iter)); \
3661 priv; \
3662 priv = netdev_lower_get_next_private(dev, &(iter)))
3663
3664 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3665 for (iter = &(dev)->adj_list.lower, \
3666 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3667 priv; \
3668 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3669
3670 void *netdev_lower_get_next(struct net_device *dev,
3671 struct list_head **iter);
3672 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3673 for (iter = &(dev)->adj_list.lower, \
3674 ldev = netdev_lower_get_next(dev, &(iter)); \
3675 ldev; \
3676 ldev = netdev_lower_get_next(dev, &(iter)))
3677
3678 void *netdev_adjacent_get_private(struct list_head *adj_list);
3679 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3680 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3681 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3682 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3683 int netdev_master_upper_dev_link(struct net_device *dev,
3684 struct net_device *upper_dev);
3685 int netdev_master_upper_dev_link_private(struct net_device *dev,
3686 struct net_device *upper_dev,
3687 void *private);
3688 void netdev_upper_dev_unlink(struct net_device *dev,
3689 struct net_device *upper_dev);
3690 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3691 void *netdev_lower_dev_get_private(struct net_device *dev,
3692 struct net_device *lower_dev);
3693
3694 /* RSS keys are 40 or 52 bytes long */
3695 #define NETDEV_RSS_KEY_LEN 52
3696 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3697 void netdev_rss_key_fill(void *buffer, size_t len);
3698
3699 int dev_get_nest_level(struct net_device *dev,
3700 bool (*type_check)(struct net_device *dev));
3701 int skb_checksum_help(struct sk_buff *skb);
3702 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3703 netdev_features_t features, bool tx_path);
3704 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3705 netdev_features_t features);
3706
3707 struct netdev_bonding_info {
3708 ifslave slave;
3709 ifbond master;
3710 };
3711
3712 struct netdev_notifier_bonding_info {
3713 struct netdev_notifier_info info; /* must be first */
3714 struct netdev_bonding_info bonding_info;
3715 };
3716
3717 void netdev_bonding_info_change(struct net_device *dev,
3718 struct netdev_bonding_info *bonding_info);
3719
3720 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)3721 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3722 {
3723 return __skb_gso_segment(skb, features, true);
3724 }
3725 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3726
can_checksum_protocol(netdev_features_t features,__be16 protocol)3727 static inline bool can_checksum_protocol(netdev_features_t features,
3728 __be16 protocol)
3729 {
3730 return ((features & NETIF_F_GEN_CSUM) ||
3731 ((features & NETIF_F_V4_CSUM) &&
3732 protocol == htons(ETH_P_IP)) ||
3733 ((features & NETIF_F_V6_CSUM) &&
3734 protocol == htons(ETH_P_IPV6)) ||
3735 ((features & NETIF_F_FCOE_CRC) &&
3736 protocol == htons(ETH_P_FCOE)));
3737 }
3738
3739 #ifdef CONFIG_BUG
3740 void netdev_rx_csum_fault(struct net_device *dev);
3741 #else
netdev_rx_csum_fault(struct net_device * dev)3742 static inline void netdev_rx_csum_fault(struct net_device *dev)
3743 {
3744 }
3745 #endif
3746 /* rx skb timestamps */
3747 void net_enable_timestamp(void);
3748 void net_disable_timestamp(void);
3749
3750 #ifdef CONFIG_PROC_FS
3751 int __init dev_proc_init(void);
3752 #else
3753 #define dev_proc_init() 0
3754 #endif
3755
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)3756 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3757 struct sk_buff *skb, struct net_device *dev,
3758 bool more)
3759 {
3760 skb->xmit_more = more ? 1 : 0;
3761 return ops->ndo_start_xmit(skb, dev);
3762 }
3763
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)3764 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3765 struct netdev_queue *txq, bool more)
3766 {
3767 const struct net_device_ops *ops = dev->netdev_ops;
3768 int rc;
3769
3770 rc = __netdev_start_xmit(ops, skb, dev, more);
3771 if (rc == NETDEV_TX_OK)
3772 txq_trans_update(txq);
3773
3774 return rc;
3775 }
3776
3777 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3778 const void *ns);
3779 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3780 const void *ns);
3781
netdev_class_create_file(struct class_attribute * class_attr)3782 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3783 {
3784 return netdev_class_create_file_ns(class_attr, NULL);
3785 }
3786
netdev_class_remove_file(struct class_attribute * class_attr)3787 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3788 {
3789 netdev_class_remove_file_ns(class_attr, NULL);
3790 }
3791
3792 extern struct kobj_ns_type_operations net_ns_type_operations;
3793
3794 const char *netdev_drivername(const struct net_device *dev);
3795
3796 void linkwatch_run_queue(void);
3797
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)3798 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3799 netdev_features_t f2)
3800 {
3801 if (f1 & NETIF_F_GEN_CSUM)
3802 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3803 if (f2 & NETIF_F_GEN_CSUM)
3804 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3805 f1 &= f2;
3806 if (f1 & NETIF_F_GEN_CSUM)
3807 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3808
3809 return f1;
3810 }
3811
netdev_get_wanted_features(struct net_device * dev)3812 static inline netdev_features_t netdev_get_wanted_features(
3813 struct net_device *dev)
3814 {
3815 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3816 }
3817 netdev_features_t netdev_increment_features(netdev_features_t all,
3818 netdev_features_t one, netdev_features_t mask);
3819
3820 /* Allow TSO being used on stacked device :
3821 * Performing the GSO segmentation before last device
3822 * is a performance improvement.
3823 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)3824 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3825 netdev_features_t mask)
3826 {
3827 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3828 }
3829
3830 int __netdev_update_features(struct net_device *dev);
3831 void netdev_update_features(struct net_device *dev);
3832 void netdev_change_features(struct net_device *dev);
3833
3834 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3835 struct net_device *dev);
3836
3837 netdev_features_t passthru_features_check(struct sk_buff *skb,
3838 struct net_device *dev,
3839 netdev_features_t features);
3840 netdev_features_t netif_skb_features(struct sk_buff *skb);
3841
net_gso_ok(netdev_features_t features,int gso_type)3842 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3843 {
3844 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3845
3846 /* check flags correspondence */
3847 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3848 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3849 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3850 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3851 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3852 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3853 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3854 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3855 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3856 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3857 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3858 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3859 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3860
3861 return (features & feature) == feature;
3862 }
3863
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)3864 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3865 {
3866 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3867 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3868 }
3869
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)3870 static inline bool netif_needs_gso(struct sk_buff *skb,
3871 netdev_features_t features)
3872 {
3873 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3874 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3875 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3876 }
3877
netif_set_gso_max_size(struct net_device * dev,unsigned int size)3878 static inline void netif_set_gso_max_size(struct net_device *dev,
3879 unsigned int size)
3880 {
3881 dev->gso_max_size = size;
3882 }
3883
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)3884 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3885 int pulled_hlen, u16 mac_offset,
3886 int mac_len)
3887 {
3888 skb->protocol = protocol;
3889 skb->encapsulation = 1;
3890 skb_push(skb, pulled_hlen);
3891 skb_reset_transport_header(skb);
3892 skb->mac_header = mac_offset;
3893 skb->network_header = skb->mac_header + mac_len;
3894 skb->mac_len = mac_len;
3895 }
3896
netif_is_macvlan(struct net_device * dev)3897 static inline bool netif_is_macvlan(struct net_device *dev)
3898 {
3899 return dev->priv_flags & IFF_MACVLAN;
3900 }
3901
netif_is_macvlan_port(struct net_device * dev)3902 static inline bool netif_is_macvlan_port(struct net_device *dev)
3903 {
3904 return dev->priv_flags & IFF_MACVLAN_PORT;
3905 }
3906
netif_is_ipvlan(struct net_device * dev)3907 static inline bool netif_is_ipvlan(struct net_device *dev)
3908 {
3909 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3910 }
3911
netif_is_ipvlan_port(struct net_device * dev)3912 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3913 {
3914 return dev->priv_flags & IFF_IPVLAN_MASTER;
3915 }
3916
netif_is_bond_master(struct net_device * dev)3917 static inline bool netif_is_bond_master(struct net_device *dev)
3918 {
3919 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3920 }
3921
netif_is_bond_slave(struct net_device * dev)3922 static inline bool netif_is_bond_slave(struct net_device *dev)
3923 {
3924 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3925 }
3926
netif_supports_nofcs(struct net_device * dev)3927 static inline bool netif_supports_nofcs(struct net_device *dev)
3928 {
3929 return dev->priv_flags & IFF_SUPP_NOFCS;
3930 }
3931
netif_is_l3_master(const struct net_device * dev)3932 static inline bool netif_is_l3_master(const struct net_device *dev)
3933 {
3934 return dev->priv_flags & IFF_L3MDEV_MASTER;
3935 }
3936
netif_is_l3_slave(const struct net_device * dev)3937 static inline bool netif_is_l3_slave(const struct net_device *dev)
3938 {
3939 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3940 }
3941
netif_is_bridge_master(const struct net_device * dev)3942 static inline bool netif_is_bridge_master(const struct net_device *dev)
3943 {
3944 return dev->priv_flags & IFF_EBRIDGE;
3945 }
3946
netif_is_bridge_port(const struct net_device * dev)3947 static inline bool netif_is_bridge_port(const struct net_device *dev)
3948 {
3949 return dev->priv_flags & IFF_BRIDGE_PORT;
3950 }
3951
netif_is_ovs_master(const struct net_device * dev)3952 static inline bool netif_is_ovs_master(const struct net_device *dev)
3953 {
3954 return dev->priv_flags & IFF_OPENVSWITCH;
3955 }
3956
3957 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)3958 static inline void netif_keep_dst(struct net_device *dev)
3959 {
3960 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3961 }
3962
3963 extern struct pernet_operations __net_initdata loopback_net_ops;
3964
3965 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3966
3967 /* netdev_printk helpers, similar to dev_printk */
3968
netdev_name(const struct net_device * dev)3969 static inline const char *netdev_name(const struct net_device *dev)
3970 {
3971 if (!dev->name[0] || strchr(dev->name, '%'))
3972 return "(unnamed net_device)";
3973 return dev->name;
3974 }
3975
netdev_reg_state(const struct net_device * dev)3976 static inline const char *netdev_reg_state(const struct net_device *dev)
3977 {
3978 switch (dev->reg_state) {
3979 case NETREG_UNINITIALIZED: return " (uninitialized)";
3980 case NETREG_REGISTERED: return "";
3981 case NETREG_UNREGISTERING: return " (unregistering)";
3982 case NETREG_UNREGISTERED: return " (unregistered)";
3983 case NETREG_RELEASED: return " (released)";
3984 case NETREG_DUMMY: return " (dummy)";
3985 }
3986
3987 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3988 return " (unknown)";
3989 }
3990
3991 __printf(3, 4)
3992 void netdev_printk(const char *level, const struct net_device *dev,
3993 const char *format, ...);
3994 __printf(2, 3)
3995 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3996 __printf(2, 3)
3997 void netdev_alert(const struct net_device *dev, const char *format, ...);
3998 __printf(2, 3)
3999 void netdev_crit(const struct net_device *dev, const char *format, ...);
4000 __printf(2, 3)
4001 void netdev_err(const struct net_device *dev, const char *format, ...);
4002 __printf(2, 3)
4003 void netdev_warn(const struct net_device *dev, const char *format, ...);
4004 __printf(2, 3)
4005 void netdev_notice(const struct net_device *dev, const char *format, ...);
4006 __printf(2, 3)
4007 void netdev_info(const struct net_device *dev, const char *format, ...);
4008
4009 #define MODULE_ALIAS_NETDEV(device) \
4010 MODULE_ALIAS("netdev-" device)
4011
4012 #if defined(CONFIG_DYNAMIC_DEBUG)
4013 #define netdev_dbg(__dev, format, args...) \
4014 do { \
4015 dynamic_netdev_dbg(__dev, format, ##args); \
4016 } while (0)
4017 #elif defined(DEBUG)
4018 #define netdev_dbg(__dev, format, args...) \
4019 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4020 #else
4021 #define netdev_dbg(__dev, format, args...) \
4022 ({ \
4023 if (0) \
4024 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4025 })
4026 #endif
4027
4028 #if defined(VERBOSE_DEBUG)
4029 #define netdev_vdbg netdev_dbg
4030 #else
4031
4032 #define netdev_vdbg(dev, format, args...) \
4033 ({ \
4034 if (0) \
4035 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4036 0; \
4037 })
4038 #endif
4039
4040 /*
4041 * netdev_WARN() acts like dev_printk(), but with the key difference
4042 * of using a WARN/WARN_ON to get the message out, including the
4043 * file/line information and a backtrace.
4044 */
4045 #define netdev_WARN(dev, format, args...) \
4046 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4047 netdev_reg_state(dev), ##args)
4048
4049 /* netif printk helpers, similar to netdev_printk */
4050
4051 #define netif_printk(priv, type, level, dev, fmt, args...) \
4052 do { \
4053 if (netif_msg_##type(priv)) \
4054 netdev_printk(level, (dev), fmt, ##args); \
4055 } while (0)
4056
4057 #define netif_level(level, priv, type, dev, fmt, args...) \
4058 do { \
4059 if (netif_msg_##type(priv)) \
4060 netdev_##level(dev, fmt, ##args); \
4061 } while (0)
4062
4063 #define netif_emerg(priv, type, dev, fmt, args...) \
4064 netif_level(emerg, priv, type, dev, fmt, ##args)
4065 #define netif_alert(priv, type, dev, fmt, args...) \
4066 netif_level(alert, priv, type, dev, fmt, ##args)
4067 #define netif_crit(priv, type, dev, fmt, args...) \
4068 netif_level(crit, priv, type, dev, fmt, ##args)
4069 #define netif_err(priv, type, dev, fmt, args...) \
4070 netif_level(err, priv, type, dev, fmt, ##args)
4071 #define netif_warn(priv, type, dev, fmt, args...) \
4072 netif_level(warn, priv, type, dev, fmt, ##args)
4073 #define netif_notice(priv, type, dev, fmt, args...) \
4074 netif_level(notice, priv, type, dev, fmt, ##args)
4075 #define netif_info(priv, type, dev, fmt, args...) \
4076 netif_level(info, priv, type, dev, fmt, ##args)
4077
4078 #if defined(CONFIG_DYNAMIC_DEBUG)
4079 #define netif_dbg(priv, type, netdev, format, args...) \
4080 do { \
4081 if (netif_msg_##type(priv)) \
4082 dynamic_netdev_dbg(netdev, format, ##args); \
4083 } while (0)
4084 #elif defined(DEBUG)
4085 #define netif_dbg(priv, type, dev, format, args...) \
4086 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4087 #else
4088 #define netif_dbg(priv, type, dev, format, args...) \
4089 ({ \
4090 if (0) \
4091 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4092 0; \
4093 })
4094 #endif
4095
4096 #if defined(VERBOSE_DEBUG)
4097 #define netif_vdbg netif_dbg
4098 #else
4099 #define netif_vdbg(priv, type, dev, format, args...) \
4100 ({ \
4101 if (0) \
4102 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4103 0; \
4104 })
4105 #endif
4106
4107 /*
4108 * The list of packet types we will receive (as opposed to discard)
4109 * and the routines to invoke.
4110 *
4111 * Why 16. Because with 16 the only overlap we get on a hash of the
4112 * low nibble of the protocol value is RARP/SNAP/X.25.
4113 *
4114 * NOTE: That is no longer true with the addition of VLAN tags. Not
4115 * sure which should go first, but I bet it won't make much
4116 * difference if we are running VLANs. The good news is that
4117 * this protocol won't be in the list unless compiled in, so
4118 * the average user (w/out VLANs) will not be adversely affected.
4119 * --BLG
4120 *
4121 * 0800 IP
4122 * 8100 802.1Q VLAN
4123 * 0001 802.3
4124 * 0002 AX.25
4125 * 0004 802.2
4126 * 8035 RARP
4127 * 0005 SNAP
4128 * 0805 X.25
4129 * 0806 ARP
4130 * 8137 IPX
4131 * 0009 Localtalk
4132 * 86DD IPv6
4133 */
4134 #define PTYPE_HASH_SIZE (16)
4135 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4136
4137 #endif /* _LINUX_NETDEVICE_H */
4138