]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - include/linux/netdevice.h
Merge tag 'master-2014-11-20' of git://git.kernel.org/pub/scm/linux/kernel/git/linvil...
[thirdparty/kernel/stable.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
36
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
42
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
45 #include <net/dsa.h>
46 #ifdef CONFIG_DCB
47 #include <net/dcbnl.h>
48 #endif
49 #include <net/netprio_cgroup.h>
50
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54
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
63 void netdev_set_default_ethtool_ops(struct net_device *dev,
64 const struct ethtool_ops *ops);
65
66 /* Backlog congestion levels */
67 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
68 #define NET_RX_DROP 1 /* packet dropped */
69
70 /*
71 * Transmit return codes: transmit return codes originate from three different
72 * namespaces:
73 *
74 * - qdisc return codes
75 * - driver transmit return codes
76 * - errno values
77 *
78 * Drivers are allowed to return any one of those in their hard_start_xmit()
79 * function. Real network devices commonly used with qdiscs should only return
80 * the driver transmit return codes though - when qdiscs are used, the actual
81 * transmission happens asynchronously, so the value is not propagated to
82 * higher layers. Virtual network devices transmit synchronously, in this case
83 * the driver transmit return codes are consumed by dev_queue_xmit(), all
84 * others are propagated to higher layers.
85 */
86
87 /* qdisc ->enqueue() return codes. */
88 #define NET_XMIT_SUCCESS 0x00
89 #define NET_XMIT_DROP 0x01 /* skb dropped */
90 #define NET_XMIT_CN 0x02 /* congestion notification */
91 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
92 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
93
94 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
95 * indicates that the device will soon be dropping packets, or already drops
96 * some packets of the same priority; prompting us to send less aggressively. */
97 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
98 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
99
100 /* Driver transmit return codes */
101 #define NETDEV_TX_MASK 0xf0
102
103 enum netdev_tx {
104 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
105 NETDEV_TX_OK = 0x00, /* driver took care of packet */
106 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
107 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
108 };
109 typedef enum netdev_tx netdev_tx_t;
110
111 /*
112 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
113 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
114 */
115 static inline bool dev_xmit_complete(int rc)
116 {
117 /*
118 * Positive cases with an skb consumed by a driver:
119 * - successful transmission (rc == NETDEV_TX_OK)
120 * - error while transmitting (rc < 0)
121 * - error while queueing to a different device (rc & NET_XMIT_MASK)
122 */
123 if (likely(rc < NET_XMIT_MASK))
124 return true;
125
126 return false;
127 }
128
129 /*
130 * Compute the worst case header length according to the protocols
131 * used.
132 */
133
134 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
135 # if defined(CONFIG_MAC80211_MESH)
136 # define LL_MAX_HEADER 128
137 # else
138 # define LL_MAX_HEADER 96
139 # endif
140 #else
141 # define LL_MAX_HEADER 32
142 #endif
143
144 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
145 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
146 #define MAX_HEADER LL_MAX_HEADER
147 #else
148 #define MAX_HEADER (LL_MAX_HEADER + 48)
149 #endif
150
151 /*
152 * Old network device statistics. Fields are native words
153 * (unsigned long) so they can be read and written atomically.
154 */
155
156 struct net_device_stats {
157 unsigned long rx_packets;
158 unsigned long tx_packets;
159 unsigned long rx_bytes;
160 unsigned long tx_bytes;
161 unsigned long rx_errors;
162 unsigned long tx_errors;
163 unsigned long rx_dropped;
164 unsigned long tx_dropped;
165 unsigned long multicast;
166 unsigned long collisions;
167 unsigned long rx_length_errors;
168 unsigned long rx_over_errors;
169 unsigned long rx_crc_errors;
170 unsigned long rx_frame_errors;
171 unsigned long rx_fifo_errors;
172 unsigned long rx_missed_errors;
173 unsigned long tx_aborted_errors;
174 unsigned long tx_carrier_errors;
175 unsigned long tx_fifo_errors;
176 unsigned long tx_heartbeat_errors;
177 unsigned long tx_window_errors;
178 unsigned long rx_compressed;
179 unsigned long tx_compressed;
180 };
181
182
183 #include <linux/cache.h>
184 #include <linux/skbuff.h>
185
186 #ifdef CONFIG_RPS
187 #include <linux/static_key.h>
188 extern struct static_key rps_needed;
189 #endif
190
191 struct neighbour;
192 struct neigh_parms;
193 struct sk_buff;
194
195 struct netdev_hw_addr {
196 struct list_head list;
197 unsigned char addr[MAX_ADDR_LEN];
198 unsigned char type;
199 #define NETDEV_HW_ADDR_T_LAN 1
200 #define NETDEV_HW_ADDR_T_SAN 2
201 #define NETDEV_HW_ADDR_T_SLAVE 3
202 #define NETDEV_HW_ADDR_T_UNICAST 4
203 #define NETDEV_HW_ADDR_T_MULTICAST 5
204 bool global_use;
205 int sync_cnt;
206 int refcount;
207 int synced;
208 struct rcu_head rcu_head;
209 };
210
211 struct netdev_hw_addr_list {
212 struct list_head list;
213 int count;
214 };
215
216 #define netdev_hw_addr_list_count(l) ((l)->count)
217 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
218 #define netdev_hw_addr_list_for_each(ha, l) \
219 list_for_each_entry(ha, &(l)->list, list)
220
221 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
222 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
223 #define netdev_for_each_uc_addr(ha, dev) \
224 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
225
226 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
227 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
228 #define netdev_for_each_mc_addr(ha, dev) \
229 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
230
231 struct hh_cache {
232 u16 hh_len;
233 u16 __pad;
234 seqlock_t hh_lock;
235
236 /* cached hardware header; allow for machine alignment needs. */
237 #define HH_DATA_MOD 16
238 #define HH_DATA_OFF(__len) \
239 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
240 #define HH_DATA_ALIGN(__len) \
241 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
242 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
243 };
244
245 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
246 * Alternative is:
247 * dev->hard_header_len ? (dev->hard_header_len +
248 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
249 *
250 * We could use other alignment values, but we must maintain the
251 * relationship HH alignment <= LL alignment.
252 */
253 #define LL_RESERVED_SPACE(dev) \
254 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
255 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
256 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
257
258 struct header_ops {
259 int (*create) (struct sk_buff *skb, struct net_device *dev,
260 unsigned short type, const void *daddr,
261 const void *saddr, unsigned int len);
262 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
263 int (*rebuild)(struct sk_buff *skb);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268 };
269
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
275 enum netdev_state_t {
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
281 };
282
283
284 /*
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
287 */
288 struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291 };
292 #define NETDEV_BOOT_SETUP_MAX 8
293
294 int __init netdev_boot_setup(char *str);
295
296 /*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299 struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
315 #endif
316 struct net_device *dev;
317 struct sk_buff *gro_list;
318 struct sk_buff *skb;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
323 };
324
325 enum {
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
330 };
331
332 enum gro_result {
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338 };
339 typedef enum gro_result gro_result_t;
340
341 /*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382 enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387 };
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
393
394 static inline bool napi_disable_pending(struct napi_struct *n)
395 {
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397 }
398
399 /**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
407 */
408 static inline bool napi_schedule_prep(struct napi_struct *n)
409 {
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
412 }
413
414 /**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421 static inline void napi_schedule(struct napi_struct *n)
422 {
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425 }
426
427 /**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
434 {
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437 }
438
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
441 {
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
444 return true;
445 }
446 return false;
447 }
448
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
451 /**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
457 */
458 static inline void napi_complete(struct napi_struct *n)
459 {
460 return napi_complete_done(n, 0);
461 }
462
463 /**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
470 struct napi_struct *napi_by_id(unsigned int napi_id);
471
472 /**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
478 void napi_hash_add(struct napi_struct *napi);
479
480 /**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
487 void napi_hash_del(struct napi_struct *napi);
488
489 /**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
496 void napi_disable(struct napi_struct *n);
497
498 /**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505 static inline void napi_enable(struct napi_struct *n)
506 {
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510 }
511
512 #ifdef CONFIG_SMP
513 /**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521 static inline void napi_synchronize(const struct napi_struct *n)
522 {
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525 }
526 #else
527 # define napi_synchronize(n) barrier()
528 #endif
529
530 enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
534 };
535
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
546 /*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
555
556 struct netdev_queue {
557 /*
558 * read mostly part
559 */
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
563 #ifdef CONFIG_SYSFS
564 struct kobject kobj;
565 #endif
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568 #endif
569 /*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
584
585 unsigned long state;
586
587 #ifdef CONFIG_BQL
588 struct dql dql;
589 #endif
590 } ____cacheline_aligned_in_smp;
591
592 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
593 {
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
595 return q->numa_node;
596 #else
597 return NUMA_NO_NODE;
598 #endif
599 }
600
601 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
602 {
603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
604 q->numa_node = node;
605 #endif
606 }
607
608 #ifdef CONFIG_RPS
609 /*
610 * This structure holds an RPS map which can be of variable length. The
611 * map is an array of CPUs.
612 */
613 struct rps_map {
614 unsigned int len;
615 struct rcu_head rcu;
616 u16 cpus[0];
617 };
618 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
619
620 /*
621 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
622 * tail pointer for that CPU's input queue at the time of last enqueue, and
623 * a hardware filter index.
624 */
625 struct rps_dev_flow {
626 u16 cpu;
627 u16 filter;
628 unsigned int last_qtail;
629 };
630 #define RPS_NO_FILTER 0xffff
631
632 /*
633 * The rps_dev_flow_table structure contains a table of flow mappings.
634 */
635 struct rps_dev_flow_table {
636 unsigned int mask;
637 struct rcu_head rcu;
638 struct rps_dev_flow flows[0];
639 };
640 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
641 ((_num) * sizeof(struct rps_dev_flow)))
642
643 /*
644 * The rps_sock_flow_table contains mappings of flows to the last CPU
645 * on which they were processed by the application (set in recvmsg).
646 */
647 struct rps_sock_flow_table {
648 unsigned int mask;
649 u16 ents[0];
650 };
651 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
652 ((_num) * sizeof(u16)))
653
654 #define RPS_NO_CPU 0xffff
655
656 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
657 u32 hash)
658 {
659 if (table && hash) {
660 unsigned int cpu, index = hash & table->mask;
661
662 /* We only give a hint, preemption can change cpu under us */
663 cpu = raw_smp_processor_id();
664
665 if (table->ents[index] != cpu)
666 table->ents[index] = cpu;
667 }
668 }
669
670 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672 {
673 if (table && hash)
674 table->ents[hash & table->mask] = RPS_NO_CPU;
675 }
676
677 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
678
679 #ifdef CONFIG_RFS_ACCEL
680 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
681 u16 filter_id);
682 #endif
683 #endif /* CONFIG_RPS */
684
685 /* This structure contains an instance of an RX queue. */
686 struct netdev_rx_queue {
687 #ifdef CONFIG_RPS
688 struct rps_map __rcu *rps_map;
689 struct rps_dev_flow_table __rcu *rps_flow_table;
690 #endif
691 struct kobject kobj;
692 struct net_device *dev;
693 } ____cacheline_aligned_in_smp;
694
695 /*
696 * RX queue sysfs structures and functions.
697 */
698 struct rx_queue_attribute {
699 struct attribute attr;
700 ssize_t (*show)(struct netdev_rx_queue *queue,
701 struct rx_queue_attribute *attr, char *buf);
702 ssize_t (*store)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, const char *buf, size_t len);
704 };
705
706 #ifdef CONFIG_XPS
707 /*
708 * This structure holds an XPS map which can be of variable length. The
709 * map is an array of queues.
710 */
711 struct xps_map {
712 unsigned int len;
713 unsigned int alloc_len;
714 struct rcu_head rcu;
715 u16 queues[0];
716 };
717 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
718 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
719 / sizeof(u16))
720
721 /*
722 * This structure holds all XPS maps for device. Maps are indexed by CPU.
723 */
724 struct xps_dev_maps {
725 struct rcu_head rcu;
726 struct xps_map __rcu *cpu_map[0];
727 };
728 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
729 (nr_cpu_ids * sizeof(struct xps_map *)))
730 #endif /* CONFIG_XPS */
731
732 #define TC_MAX_QUEUE 16
733 #define TC_BITMASK 15
734 /* HW offloaded queuing disciplines txq count and offset maps */
735 struct netdev_tc_txq {
736 u16 count;
737 u16 offset;
738 };
739
740 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
741 /*
742 * This structure is to hold information about the device
743 * configured to run FCoE protocol stack.
744 */
745 struct netdev_fcoe_hbainfo {
746 char manufacturer[64];
747 char serial_number[64];
748 char hardware_version[64];
749 char driver_version[64];
750 char optionrom_version[64];
751 char firmware_version[64];
752 char model[256];
753 char model_description[256];
754 };
755 #endif
756
757 #define MAX_PHYS_PORT_ID_LEN 32
758
759 /* This structure holds a unique identifier to identify the
760 * physical port used by a netdevice.
761 */
762 struct netdev_phys_port_id {
763 unsigned char id[MAX_PHYS_PORT_ID_LEN];
764 unsigned char id_len;
765 };
766
767 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
768 struct sk_buff *skb);
769
770 /*
771 * This structure defines the management hooks for network devices.
772 * The following hooks can be defined; unless noted otherwise, they are
773 * optional and can be filled with a null pointer.
774 *
775 * int (*ndo_init)(struct net_device *dev);
776 * This function is called once when network device is registered.
777 * The network device can use this to any late stage initializaton
778 * or semantic validattion. It can fail with an error code which will
779 * be propogated back to register_netdev
780 *
781 * void (*ndo_uninit)(struct net_device *dev);
782 * This function is called when device is unregistered or when registration
783 * fails. It is not called if init fails.
784 *
785 * int (*ndo_open)(struct net_device *dev);
786 * This function is called when network device transistions to the up
787 * state.
788 *
789 * int (*ndo_stop)(struct net_device *dev);
790 * This function is called when network device transistions to the down
791 * state.
792 *
793 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
794 * struct net_device *dev);
795 * Called when a packet needs to be transmitted.
796 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
797 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
798 * Required can not be NULL.
799 *
800 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
801 * void *accel_priv, select_queue_fallback_t fallback);
802 * Called to decide which queue to when device supports multiple
803 * transmit queues.
804 *
805 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
806 * This function is called to allow device receiver to make
807 * changes to configuration when multicast or promiscious is enabled.
808 *
809 * void (*ndo_set_rx_mode)(struct net_device *dev);
810 * This function is called device changes address list filtering.
811 * If driver handles unicast address filtering, it should set
812 * IFF_UNICAST_FLT to its priv_flags.
813 *
814 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
815 * This function is called when the Media Access Control address
816 * needs to be changed. If this interface is not defined, the
817 * mac address can not be changed.
818 *
819 * int (*ndo_validate_addr)(struct net_device *dev);
820 * Test if Media Access Control address is valid for the device.
821 *
822 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
823 * Called when a user request an ioctl which can't be handled by
824 * the generic interface code. If not defined ioctl's return
825 * not supported error code.
826 *
827 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
828 * Used to set network devices bus interface parameters. This interface
829 * is retained for legacy reason, new devices should use the bus
830 * interface (PCI) for low level management.
831 *
832 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
833 * Called when a user wants to change the Maximum Transfer Unit
834 * of a device. If not defined, any request to change MTU will
835 * will return an error.
836 *
837 * void (*ndo_tx_timeout)(struct net_device *dev);
838 * Callback uses when the transmitter has not made any progress
839 * for dev->watchdog ticks.
840 *
841 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
842 * struct rtnl_link_stats64 *storage);
843 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
844 * Called when a user wants to get the network device usage
845 * statistics. Drivers must do one of the following:
846 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
847 * rtnl_link_stats64 structure passed by the caller.
848 * 2. Define @ndo_get_stats to update a net_device_stats structure
849 * (which should normally be dev->stats) and return a pointer to
850 * it. The structure may be changed asynchronously only if each
851 * field is written atomically.
852 * 3. Update dev->stats asynchronously and atomically, and define
853 * neither operation.
854 *
855 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
856 * If device support VLAN filtering this function is called when a
857 * VLAN id is registered.
858 *
859 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
860 * If device support VLAN filtering this function is called when a
861 * VLAN id is unregistered.
862 *
863 * void (*ndo_poll_controller)(struct net_device *dev);
864 *
865 * SR-IOV management functions.
866 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
867 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
868 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
869 * int max_tx_rate);
870 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
871 * int (*ndo_get_vf_config)(struct net_device *dev,
872 * int vf, struct ifla_vf_info *ivf);
873 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
874 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
875 * struct nlattr *port[]);
876 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
877 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
878 * Called to setup 'tc' number of traffic classes in the net device. This
879 * is always called from the stack with the rtnl lock held and netif tx
880 * queues stopped. This allows the netdevice to perform queue management
881 * safely.
882 *
883 * Fiber Channel over Ethernet (FCoE) offload functions.
884 * int (*ndo_fcoe_enable)(struct net_device *dev);
885 * Called when the FCoE protocol stack wants to start using LLD for FCoE
886 * so the underlying device can perform whatever needed configuration or
887 * initialization to support acceleration of FCoE traffic.
888 *
889 * int (*ndo_fcoe_disable)(struct net_device *dev);
890 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
891 * so the underlying device can perform whatever needed clean-ups to
892 * stop supporting acceleration of FCoE traffic.
893 *
894 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
895 * struct scatterlist *sgl, unsigned int sgc);
896 * Called when the FCoE Initiator wants to initialize an I/O that
897 * is a possible candidate for Direct Data Placement (DDP). The LLD can
898 * perform necessary setup and returns 1 to indicate the device is set up
899 * successfully to perform DDP on this I/O, otherwise this returns 0.
900 *
901 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
902 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
903 * indicated by the FC exchange id 'xid', so the underlying device can
904 * clean up and reuse resources for later DDP requests.
905 *
906 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
907 * struct scatterlist *sgl, unsigned int sgc);
908 * Called when the FCoE Target wants to initialize an I/O that
909 * is a possible candidate for Direct Data Placement (DDP). The LLD can
910 * perform necessary setup and returns 1 to indicate the device is set up
911 * successfully to perform DDP on this I/O, otherwise this returns 0.
912 *
913 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
914 * struct netdev_fcoe_hbainfo *hbainfo);
915 * Called when the FCoE Protocol stack wants information on the underlying
916 * device. This information is utilized by the FCoE protocol stack to
917 * register attributes with Fiber Channel management service as per the
918 * FC-GS Fabric Device Management Information(FDMI) specification.
919 *
920 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
921 * Called when the underlying device wants to override default World Wide
922 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
923 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
924 * protocol stack to use.
925 *
926 * RFS acceleration.
927 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
928 * u16 rxq_index, u32 flow_id);
929 * Set hardware filter for RFS. rxq_index is the target queue index;
930 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
931 * Return the filter ID on success, or a negative error code.
932 *
933 * Slave management functions (for bridge, bonding, etc).
934 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
935 * Called to make another netdev an underling.
936 *
937 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
938 * Called to release previously enslaved netdev.
939 *
940 * Feature/offload setting functions.
941 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
942 * netdev_features_t features);
943 * Adjusts the requested feature flags according to device-specific
944 * constraints, and returns the resulting flags. Must not modify
945 * the device state.
946 *
947 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
948 * Called to update device configuration to new features. Passed
949 * feature set might be less than what was returned by ndo_fix_features()).
950 * Must return >0 or -errno if it changed dev->features itself.
951 *
952 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
953 * struct net_device *dev,
954 * const unsigned char *addr, u16 flags)
955 * Adds an FDB entry to dev for addr.
956 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
957 * struct net_device *dev,
958 * const unsigned char *addr)
959 * Deletes the FDB entry from dev coresponding to addr.
960 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
961 * struct net_device *dev, struct net_device *filter_dev,
962 * int idx)
963 * Used to add FDB entries to dump requests. Implementers should add
964 * entries to skb and update idx with the number of entries.
965 *
966 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
967 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
968 * struct net_device *dev, u32 filter_mask)
969 *
970 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
971 * Called to change device carrier. Soft-devices (like dummy, team, etc)
972 * which do not represent real hardware may define this to allow their
973 * userspace components to manage their virtual carrier state. Devices
974 * that determine carrier state from physical hardware properties (eg
975 * network cables) or protocol-dependent mechanisms (eg
976 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
977 *
978 * int (*ndo_get_phys_port_id)(struct net_device *dev,
979 * struct netdev_phys_port_id *ppid);
980 * Called to get ID of physical port of this device. If driver does
981 * not implement this, it is assumed that the hw is not able to have
982 * multiple net devices on single physical port.
983 *
984 * void (*ndo_add_vxlan_port)(struct net_device *dev,
985 * sa_family_t sa_family, __be16 port);
986 * Called by vxlan to notiy a driver about the UDP port and socket
987 * address family that vxlan is listnening to. It is called only when
988 * a new port starts listening. The operation is protected by the
989 * vxlan_net->sock_lock.
990 *
991 * void (*ndo_del_vxlan_port)(struct net_device *dev,
992 * sa_family_t sa_family, __be16 port);
993 * Called by vxlan to notify the driver about a UDP port and socket
994 * address family that vxlan is not listening to anymore. The operation
995 * is protected by the vxlan_net->sock_lock.
996 *
997 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
998 * struct net_device *dev)
999 * Called by upper layer devices to accelerate switching or other
1000 * station functionality into hardware. 'pdev is the lowerdev
1001 * to use for the offload and 'dev' is the net device that will
1002 * back the offload. Returns a pointer to the private structure
1003 * the upper layer will maintain.
1004 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1005 * Called by upper layer device to delete the station created
1006 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1007 * the station and priv is the structure returned by the add
1008 * operation.
1009 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1010 * struct net_device *dev,
1011 * void *priv);
1012 * Callback to use for xmit over the accelerated station. This
1013 * is used in place of ndo_start_xmit on accelerated net
1014 * devices.
1015 * bool (*ndo_gso_check) (struct sk_buff *skb,
1016 * struct net_device *dev);
1017 * Called by core transmit path to determine if device is capable of
1018 * performing GSO on a packet. The device returns true if it is
1019 * able to GSO the packet, false otherwise. If the return value is
1020 * false the stack will do software GSO.
1021 */
1022 struct net_device_ops {
1023 int (*ndo_init)(struct net_device *dev);
1024 void (*ndo_uninit)(struct net_device *dev);
1025 int (*ndo_open)(struct net_device *dev);
1026 int (*ndo_stop)(struct net_device *dev);
1027 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1028 struct net_device *dev);
1029 u16 (*ndo_select_queue)(struct net_device *dev,
1030 struct sk_buff *skb,
1031 void *accel_priv,
1032 select_queue_fallback_t fallback);
1033 void (*ndo_change_rx_flags)(struct net_device *dev,
1034 int flags);
1035 void (*ndo_set_rx_mode)(struct net_device *dev);
1036 int (*ndo_set_mac_address)(struct net_device *dev,
1037 void *addr);
1038 int (*ndo_validate_addr)(struct net_device *dev);
1039 int (*ndo_do_ioctl)(struct net_device *dev,
1040 struct ifreq *ifr, int cmd);
1041 int (*ndo_set_config)(struct net_device *dev,
1042 struct ifmap *map);
1043 int (*ndo_change_mtu)(struct net_device *dev,
1044 int new_mtu);
1045 int (*ndo_neigh_setup)(struct net_device *dev,
1046 struct neigh_parms *);
1047 void (*ndo_tx_timeout) (struct net_device *dev);
1048
1049 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1050 struct rtnl_link_stats64 *storage);
1051 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1052
1053 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1054 __be16 proto, u16 vid);
1055 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1056 __be16 proto, u16 vid);
1057 #ifdef CONFIG_NET_POLL_CONTROLLER
1058 void (*ndo_poll_controller)(struct net_device *dev);
1059 int (*ndo_netpoll_setup)(struct net_device *dev,
1060 struct netpoll_info *info);
1061 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1062 #endif
1063 #ifdef CONFIG_NET_RX_BUSY_POLL
1064 int (*ndo_busy_poll)(struct napi_struct *dev);
1065 #endif
1066 int (*ndo_set_vf_mac)(struct net_device *dev,
1067 int queue, u8 *mac);
1068 int (*ndo_set_vf_vlan)(struct net_device *dev,
1069 int queue, u16 vlan, u8 qos);
1070 int (*ndo_set_vf_rate)(struct net_device *dev,
1071 int vf, int min_tx_rate,
1072 int max_tx_rate);
1073 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1074 int vf, bool setting);
1075 int (*ndo_get_vf_config)(struct net_device *dev,
1076 int vf,
1077 struct ifla_vf_info *ivf);
1078 int (*ndo_set_vf_link_state)(struct net_device *dev,
1079 int vf, int link_state);
1080 int (*ndo_set_vf_port)(struct net_device *dev,
1081 int vf,
1082 struct nlattr *port[]);
1083 int (*ndo_get_vf_port)(struct net_device *dev,
1084 int vf, struct sk_buff *skb);
1085 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1086 #if IS_ENABLED(CONFIG_FCOE)
1087 int (*ndo_fcoe_enable)(struct net_device *dev);
1088 int (*ndo_fcoe_disable)(struct net_device *dev);
1089 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1090 u16 xid,
1091 struct scatterlist *sgl,
1092 unsigned int sgc);
1093 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1094 u16 xid);
1095 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1096 u16 xid,
1097 struct scatterlist *sgl,
1098 unsigned int sgc);
1099 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1100 struct netdev_fcoe_hbainfo *hbainfo);
1101 #endif
1102
1103 #if IS_ENABLED(CONFIG_LIBFCOE)
1104 #define NETDEV_FCOE_WWNN 0
1105 #define NETDEV_FCOE_WWPN 1
1106 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1107 u64 *wwn, int type);
1108 #endif
1109
1110 #ifdef CONFIG_RFS_ACCEL
1111 int (*ndo_rx_flow_steer)(struct net_device *dev,
1112 const struct sk_buff *skb,
1113 u16 rxq_index,
1114 u32 flow_id);
1115 #endif
1116 int (*ndo_add_slave)(struct net_device *dev,
1117 struct net_device *slave_dev);
1118 int (*ndo_del_slave)(struct net_device *dev,
1119 struct net_device *slave_dev);
1120 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1121 netdev_features_t features);
1122 int (*ndo_set_features)(struct net_device *dev,
1123 netdev_features_t features);
1124 int (*ndo_neigh_construct)(struct neighbour *n);
1125 void (*ndo_neigh_destroy)(struct neighbour *n);
1126
1127 int (*ndo_fdb_add)(struct ndmsg *ndm,
1128 struct nlattr *tb[],
1129 struct net_device *dev,
1130 const unsigned char *addr,
1131 u16 flags);
1132 int (*ndo_fdb_del)(struct ndmsg *ndm,
1133 struct nlattr *tb[],
1134 struct net_device *dev,
1135 const unsigned char *addr);
1136 int (*ndo_fdb_dump)(struct sk_buff *skb,
1137 struct netlink_callback *cb,
1138 struct net_device *dev,
1139 struct net_device *filter_dev,
1140 int idx);
1141
1142 int (*ndo_bridge_setlink)(struct net_device *dev,
1143 struct nlmsghdr *nlh);
1144 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1145 u32 pid, u32 seq,
1146 struct net_device *dev,
1147 u32 filter_mask);
1148 int (*ndo_bridge_dellink)(struct net_device *dev,
1149 struct nlmsghdr *nlh);
1150 int (*ndo_change_carrier)(struct net_device *dev,
1151 bool new_carrier);
1152 int (*ndo_get_phys_port_id)(struct net_device *dev,
1153 struct netdev_phys_port_id *ppid);
1154 void (*ndo_add_vxlan_port)(struct net_device *dev,
1155 sa_family_t sa_family,
1156 __be16 port);
1157 void (*ndo_del_vxlan_port)(struct net_device *dev,
1158 sa_family_t sa_family,
1159 __be16 port);
1160
1161 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1162 struct net_device *dev);
1163 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1164 void *priv);
1165
1166 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1167 struct net_device *dev,
1168 void *priv);
1169 int (*ndo_get_lock_subclass)(struct net_device *dev);
1170 bool (*ndo_gso_check) (struct sk_buff *skb,
1171 struct net_device *dev);
1172 };
1173
1174 /**
1175 * enum net_device_priv_flags - &struct net_device priv_flags
1176 *
1177 * These are the &struct net_device, they are only set internally
1178 * by drivers and used in the kernel. These flags are invisible to
1179 * userspace, this means that the order of these flags can change
1180 * during any kernel release.
1181 *
1182 * You should have a pretty good reason to be extending these flags.
1183 *
1184 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1185 * @IFF_EBRIDGE: Ethernet bridging device
1186 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1187 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1188 * @IFF_MASTER_ALB: bonding master, balance-alb
1189 * @IFF_BONDING: bonding master or slave
1190 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1191 * @IFF_ISATAP: ISATAP interface (RFC4214)
1192 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1193 * @IFF_WAN_HDLC: WAN HDLC device
1194 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1195 * release skb->dst
1196 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1197 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1198 * @IFF_MACVLAN_PORT: device used as macvlan port
1199 * @IFF_BRIDGE_PORT: device used as bridge port
1200 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1201 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1202 * @IFF_UNICAST_FLT: Supports unicast filtering
1203 * @IFF_TEAM_PORT: device used as team port
1204 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1205 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1206 * change when it's running
1207 * @IFF_MACVLAN: Macvlan device
1208 */
1209 enum netdev_priv_flags {
1210 IFF_802_1Q_VLAN = 1<<0,
1211 IFF_EBRIDGE = 1<<1,
1212 IFF_SLAVE_INACTIVE = 1<<2,
1213 IFF_MASTER_8023AD = 1<<3,
1214 IFF_MASTER_ALB = 1<<4,
1215 IFF_BONDING = 1<<5,
1216 IFF_SLAVE_NEEDARP = 1<<6,
1217 IFF_ISATAP = 1<<7,
1218 IFF_MASTER_ARPMON = 1<<8,
1219 IFF_WAN_HDLC = 1<<9,
1220 IFF_XMIT_DST_RELEASE = 1<<10,
1221 IFF_DONT_BRIDGE = 1<<11,
1222 IFF_DISABLE_NETPOLL = 1<<12,
1223 IFF_MACVLAN_PORT = 1<<13,
1224 IFF_BRIDGE_PORT = 1<<14,
1225 IFF_OVS_DATAPATH = 1<<15,
1226 IFF_TX_SKB_SHARING = 1<<16,
1227 IFF_UNICAST_FLT = 1<<17,
1228 IFF_TEAM_PORT = 1<<18,
1229 IFF_SUPP_NOFCS = 1<<19,
1230 IFF_LIVE_ADDR_CHANGE = 1<<20,
1231 IFF_MACVLAN = 1<<21,
1232 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1233 };
1234
1235 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1236 #define IFF_EBRIDGE IFF_EBRIDGE
1237 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1238 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1239 #define IFF_MASTER_ALB IFF_MASTER_ALB
1240 #define IFF_BONDING IFF_BONDING
1241 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1242 #define IFF_ISATAP IFF_ISATAP
1243 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1244 #define IFF_WAN_HDLC IFF_WAN_HDLC
1245 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1246 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1247 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1248 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1249 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1250 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1251 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1252 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1253 #define IFF_TEAM_PORT IFF_TEAM_PORT
1254 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1255 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1256 #define IFF_MACVLAN IFF_MACVLAN
1257 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1258
1259 /**
1260 * struct net_device - The DEVICE structure.
1261 * Actually, this whole structure is a big mistake. It mixes I/O
1262 * data with strictly "high-level" data, and it has to know about
1263 * almost every data structure used in the INET module.
1264 *
1265 * @name: This is the first field of the "visible" part of this structure
1266 * (i.e. as seen by users in the "Space.c" file). It is the name
1267 * of the interface.
1268 *
1269 * @name_hlist: Device name hash chain, please keep it close to name[]
1270 * @ifalias: SNMP alias
1271 * @mem_end: Shared memory end
1272 * @mem_start: Shared memory start
1273 * @base_addr: Device I/O address
1274 * @irq: Device IRQ number
1275 *
1276 * @state: Generic network queuing layer state, see netdev_state_t
1277 * @dev_list: The global list of network devices
1278 * @napi_list: List entry, that is used for polling napi devices
1279 * @unreg_list: List entry, that is used, when we are unregistering the
1280 * device, see the function unregister_netdev
1281 * @close_list: List entry, that is used, when we are closing the device
1282 *
1283 * @adj_list: Directly linked devices, like slaves for bonding
1284 * @all_adj_list: All linked devices, *including* neighbours
1285 * @features: Currently active device features
1286 * @hw_features: User-changeable features
1287 *
1288 * @wanted_features: User-requested features
1289 * @vlan_features: Mask of features inheritable by VLAN devices
1290 *
1291 * @hw_enc_features: Mask of features inherited by encapsulating devices
1292 * This field indicates what encapsulation
1293 * offloads the hardware is capable of doing,
1294 * and drivers will need to set them appropriately.
1295 *
1296 * @mpls_features: Mask of features inheritable by MPLS
1297 *
1298 * @ifindex: interface index
1299 * @iflink: unique device identifier
1300 *
1301 * @stats: Statistics struct, which was left as a legacy, use
1302 * rtnl_link_stats64 instead
1303 *
1304 * @rx_dropped: Dropped packets by core network,
1305 * do not use this in drivers
1306 * @tx_dropped: Dropped packets by core network,
1307 * do not use this in drivers
1308 *
1309 * @carrier_changes: Stats to monitor carrier on<->off transitions
1310 *
1311 * @wireless_handlers: List of functions to handle Wireless Extensions,
1312 * instead of ioctl,
1313 * see <net/iw_handler.h> for details.
1314 * @wireless_data: Instance data managed by the core of wireless extensions
1315 *
1316 * @netdev_ops: Includes several pointers to callbacks,
1317 * if one wants to override the ndo_*() functions
1318 * @ethtool_ops: Management operations
1319 * @fwd_ops: Management operations
1320 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1321 * of Layer 2 headers.
1322 *
1323 * @flags: Interface flags (a la BSD)
1324 * @priv_flags: Like 'flags' but invisible to userspace,
1325 * see if.h for the definitions
1326 * @gflags: Global flags ( kept as legacy )
1327 * @padded: How much padding added by alloc_netdev()
1328 * @operstate: RFC2863 operstate
1329 * @link_mode: Mapping policy to operstate
1330 * @if_port: Selectable AUI, TP, ...
1331 * @dma: DMA channel
1332 * @mtu: Interface MTU value
1333 * @type: Interface hardware type
1334 * @hard_header_len: Hardware header length
1335 *
1336 * @needed_headroom: Extra headroom the hardware may need, but not in all
1337 * cases can this be guaranteed
1338 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1339 * cases can this be guaranteed. Some cases also use
1340 * LL_MAX_HEADER instead to allocate the skb
1341 *
1342 * interface address info:
1343 *
1344 * @perm_addr: Permanent hw address
1345 * @addr_assign_type: Hw address assignment type
1346 * @addr_len: Hardware address length
1347 * @neigh_priv_len; Used in neigh_alloc(),
1348 * initialized only in atm/clip.c
1349 * @dev_id: Used to differentiate devices that share
1350 * the same link layer address
1351 * @dev_port: Used to differentiate devices that share
1352 * the same function
1353 * @addr_list_lock: XXX: need comments on this one
1354 * @uc: unicast mac addresses
1355 * @mc: multicast mac addresses
1356 * @dev_addrs: list of device hw addresses
1357 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1358 * @uc_promisc: Counter, that indicates, that promiscuous mode
1359 * has been enabled due to the need to listen to
1360 * additional unicast addresses in a device that
1361 * does not implement ndo_set_rx_mode()
1362 * @promiscuity: Number of times, the NIC is told to work in
1363 * Promiscuous mode, if it becomes 0 the NIC will
1364 * exit from working in Promiscuous mode
1365 * @allmulti: Counter, enables or disables allmulticast mode
1366 *
1367 * @vlan_info: VLAN info
1368 * @dsa_ptr: dsa specific data
1369 * @tipc_ptr: TIPC specific data
1370 * @atalk_ptr: AppleTalk link
1371 * @ip_ptr: IPv4 specific data
1372 * @dn_ptr: DECnet specific data
1373 * @ip6_ptr: IPv6 specific data
1374 * @ax25_ptr: AX.25 specific data
1375 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1376 *
1377 * @last_rx: Time of last Rx
1378 * @dev_addr: Hw address (before bcast,
1379 * because most packets are unicast)
1380 *
1381 * @_rx: Array of RX queues
1382 * @num_rx_queues: Number of RX queues
1383 * allocated at register_netdev() time
1384 * @real_num_rx_queues: Number of RX queues currently active in device
1385 *
1386 * @rx_handler: handler for received packets
1387 * @rx_handler_data: XXX: need comments on this one
1388 * @ingress_queue: XXX: need comments on this one
1389 * @broadcast: hw bcast address
1390 *
1391 * @_tx: Array of TX queues
1392 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1393 * @real_num_tx_queues: Number of TX queues currently active in device
1394 * @qdisc: Root qdisc from userspace point of view
1395 * @tx_queue_len: Max frames per queue allowed
1396 * @tx_global_lock: XXX: need comments on this one
1397 *
1398 * @xps_maps: XXX: need comments on this one
1399 *
1400 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1401 * indexed by RX queue number. Assigned by driver.
1402 * This must only be set if the ndo_rx_flow_steer
1403 * operation is defined
1404 *
1405 * @trans_start: Time (in jiffies) of last Tx
1406 * @watchdog_timeo: Represents the timeout that is used by
1407 * the watchdog ( see dev_watchdog() )
1408 * @watchdog_timer: List of timers
1409 *
1410 * @pcpu_refcnt: Number of references to this device
1411 * @todo_list: Delayed register/unregister
1412 * @index_hlist: Device index hash chain
1413 * @link_watch_list: XXX: need comments on this one
1414 *
1415 * @reg_state: Register/unregister state machine
1416 * @dismantle: Device is going to be freed
1417 * @rtnl_link_state: This enum represents the phases of creating
1418 * a new link
1419 *
1420 * @destructor: Called from unregister,
1421 * can be used to call free_netdev
1422 * @npinfo: XXX: need comments on this one
1423 * @nd_net: Network namespace this network device is inside
1424 *
1425 * @ml_priv: Mid-layer private
1426 * @lstats: Loopback statistics
1427 * @tstats: Tunnel statistics
1428 * @dstats: Dummy statistics
1429 * @vstats: Virtual ethernet statistics
1430 *
1431 * @garp_port: GARP
1432 * @mrp_port: MRP
1433 *
1434 * @dev: Class/net/name entry
1435 * @sysfs_groups: Space for optional device, statistics and wireless
1436 * sysfs groups
1437 *
1438 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1439 * @rtnl_link_ops: Rtnl_link_ops
1440 *
1441 * @gso_max_size: Maximum size of generic segmentation offload
1442 * @gso_max_segs: Maximum number of segments that can be passed to the
1443 * NIC for GSO
1444 * @gso_min_segs: Minimum number of segments that can be passed to the
1445 * NIC for GSO
1446 *
1447 * @dcbnl_ops: Data Center Bridging netlink ops
1448 * @num_tc: Number of traffic classes in the net device
1449 * @tc_to_txq: XXX: need comments on this one
1450 * @prio_tc_map XXX: need comments on this one
1451 *
1452 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1453 *
1454 * @priomap: XXX: need comments on this one
1455 * @phydev: Physical device may attach itself
1456 * for hardware timestamping
1457 *
1458 * @qdisc_tx_busylock: XXX: need comments on this one
1459 *
1460 * @group: The group, that the device belongs to
1461 * @pm_qos_req: Power Management QoS object
1462 *
1463 * FIXME: cleanup struct net_device such that network protocol info
1464 * moves out.
1465 */
1466
1467 struct net_device {
1468 char name[IFNAMSIZ];
1469 struct hlist_node name_hlist;
1470 char *ifalias;
1471 /*
1472 * I/O specific fields
1473 * FIXME: Merge these and struct ifmap into one
1474 */
1475 unsigned long mem_end;
1476 unsigned long mem_start;
1477 unsigned long base_addr;
1478 int irq;
1479
1480 /*
1481 * Some hardware also needs these fields (state,dev_list,
1482 * napi_list,unreg_list,close_list) but they are not
1483 * part of the usual set specified in Space.c.
1484 */
1485
1486 unsigned long state;
1487
1488 struct list_head dev_list;
1489 struct list_head napi_list;
1490 struct list_head unreg_list;
1491 struct list_head close_list;
1492
1493 struct {
1494 struct list_head upper;
1495 struct list_head lower;
1496 } adj_list;
1497
1498 struct {
1499 struct list_head upper;
1500 struct list_head lower;
1501 } all_adj_list;
1502
1503 netdev_features_t features;
1504 netdev_features_t hw_features;
1505 netdev_features_t wanted_features;
1506 netdev_features_t vlan_features;
1507 netdev_features_t hw_enc_features;
1508 netdev_features_t mpls_features;
1509
1510 int ifindex;
1511 int iflink;
1512
1513 struct net_device_stats stats;
1514
1515 atomic_long_t rx_dropped;
1516 atomic_long_t tx_dropped;
1517
1518 atomic_t carrier_changes;
1519
1520 #ifdef CONFIG_WIRELESS_EXT
1521 const struct iw_handler_def * wireless_handlers;
1522 struct iw_public_data * wireless_data;
1523 #endif
1524 const struct net_device_ops *netdev_ops;
1525 const struct ethtool_ops *ethtool_ops;
1526 const struct forwarding_accel_ops *fwd_ops;
1527
1528 const struct header_ops *header_ops;
1529
1530 unsigned int flags;
1531 unsigned int priv_flags;
1532
1533 unsigned short gflags;
1534 unsigned short padded;
1535
1536 unsigned char operstate;
1537 unsigned char link_mode;
1538
1539 unsigned char if_port;
1540 unsigned char dma;
1541
1542 unsigned int mtu;
1543 unsigned short type;
1544 unsigned short hard_header_len;
1545
1546 unsigned short needed_headroom;
1547 unsigned short needed_tailroom;
1548
1549 /* Interface address info. */
1550 unsigned char perm_addr[MAX_ADDR_LEN];
1551 unsigned char addr_assign_type;
1552 unsigned char addr_len;
1553 unsigned short neigh_priv_len;
1554 unsigned short dev_id;
1555 unsigned short dev_port;
1556 spinlock_t addr_list_lock;
1557 struct netdev_hw_addr_list uc;
1558 struct netdev_hw_addr_list mc;
1559 struct netdev_hw_addr_list dev_addrs;
1560
1561 #ifdef CONFIG_SYSFS
1562 struct kset *queues_kset;
1563 #endif
1564
1565 unsigned char name_assign_type;
1566
1567 bool uc_promisc;
1568 unsigned int promiscuity;
1569 unsigned int allmulti;
1570
1571
1572 /* Protocol specific pointers */
1573
1574 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1575 struct vlan_info __rcu *vlan_info;
1576 #endif
1577 #if IS_ENABLED(CONFIG_NET_DSA)
1578 struct dsa_switch_tree *dsa_ptr;
1579 #endif
1580 #if IS_ENABLED(CONFIG_TIPC)
1581 struct tipc_bearer __rcu *tipc_ptr;
1582 #endif
1583 void *atalk_ptr;
1584 struct in_device __rcu *ip_ptr;
1585 struct dn_dev __rcu *dn_ptr;
1586 struct inet6_dev __rcu *ip6_ptr;
1587 void *ax25_ptr;
1588 struct wireless_dev *ieee80211_ptr;
1589 struct wpan_dev *ieee802154_ptr;
1590
1591 /*
1592 * Cache lines mostly used on receive path (including eth_type_trans())
1593 */
1594 unsigned long last_rx;
1595
1596 /* Interface address info used in eth_type_trans() */
1597 unsigned char *dev_addr;
1598
1599
1600 #ifdef CONFIG_SYSFS
1601 struct netdev_rx_queue *_rx;
1602
1603 unsigned int num_rx_queues;
1604 unsigned int real_num_rx_queues;
1605
1606 #endif
1607
1608 unsigned long gro_flush_timeout;
1609 rx_handler_func_t __rcu *rx_handler;
1610 void __rcu *rx_handler_data;
1611
1612 struct netdev_queue __rcu *ingress_queue;
1613 unsigned char broadcast[MAX_ADDR_LEN];
1614
1615
1616 /*
1617 * Cache lines mostly used on transmit path
1618 */
1619 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1620 unsigned int num_tx_queues;
1621 unsigned int real_num_tx_queues;
1622 struct Qdisc *qdisc;
1623 unsigned long tx_queue_len;
1624 spinlock_t tx_global_lock;
1625
1626 #ifdef CONFIG_XPS
1627 struct xps_dev_maps __rcu *xps_maps;
1628 #endif
1629 #ifdef CONFIG_RFS_ACCEL
1630 struct cpu_rmap *rx_cpu_rmap;
1631 #endif
1632
1633 /* These may be needed for future network-power-down code. */
1634
1635 /*
1636 * trans_start here is expensive for high speed devices on SMP,
1637 * please use netdev_queue->trans_start instead.
1638 */
1639 unsigned long trans_start;
1640
1641 int watchdog_timeo;
1642 struct timer_list watchdog_timer;
1643
1644 int __percpu *pcpu_refcnt;
1645 struct list_head todo_list;
1646
1647 struct hlist_node index_hlist;
1648 struct list_head link_watch_list;
1649
1650 enum { NETREG_UNINITIALIZED=0,
1651 NETREG_REGISTERED, /* completed register_netdevice */
1652 NETREG_UNREGISTERING, /* called unregister_netdevice */
1653 NETREG_UNREGISTERED, /* completed unregister todo */
1654 NETREG_RELEASED, /* called free_netdev */
1655 NETREG_DUMMY, /* dummy device for NAPI poll */
1656 } reg_state:8;
1657
1658 bool dismantle;
1659
1660 enum {
1661 RTNL_LINK_INITIALIZED,
1662 RTNL_LINK_INITIALIZING,
1663 } rtnl_link_state:16;
1664
1665 void (*destructor)(struct net_device *dev);
1666
1667 #ifdef CONFIG_NETPOLL
1668 struct netpoll_info __rcu *npinfo;
1669 #endif
1670
1671 #ifdef CONFIG_NET_NS
1672 struct net *nd_net;
1673 #endif
1674
1675 /* mid-layer private */
1676 union {
1677 void *ml_priv;
1678 struct pcpu_lstats __percpu *lstats;
1679 struct pcpu_sw_netstats __percpu *tstats;
1680 struct pcpu_dstats __percpu *dstats;
1681 struct pcpu_vstats __percpu *vstats;
1682 };
1683
1684 struct garp_port __rcu *garp_port;
1685 struct mrp_port __rcu *mrp_port;
1686
1687 struct device dev;
1688 const struct attribute_group *sysfs_groups[4];
1689 const struct attribute_group *sysfs_rx_queue_group;
1690
1691 const struct rtnl_link_ops *rtnl_link_ops;
1692
1693 /* for setting kernel sock attribute on TCP connection setup */
1694 #define GSO_MAX_SIZE 65536
1695 unsigned int gso_max_size;
1696 #define GSO_MAX_SEGS 65535
1697 u16 gso_max_segs;
1698 u16 gso_min_segs;
1699 #ifdef CONFIG_DCB
1700 const struct dcbnl_rtnl_ops *dcbnl_ops;
1701 #endif
1702 u8 num_tc;
1703 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1704 u8 prio_tc_map[TC_BITMASK + 1];
1705
1706 #if IS_ENABLED(CONFIG_FCOE)
1707 unsigned int fcoe_ddp_xid;
1708 #endif
1709 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1710 struct netprio_map __rcu *priomap;
1711 #endif
1712 struct phy_device *phydev;
1713 struct lock_class_key *qdisc_tx_busylock;
1714 int group;
1715 struct pm_qos_request pm_qos_req;
1716 };
1717 #define to_net_dev(d) container_of(d, struct net_device, dev)
1718
1719 #define NETDEV_ALIGN 32
1720
1721 static inline
1722 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1723 {
1724 return dev->prio_tc_map[prio & TC_BITMASK];
1725 }
1726
1727 static inline
1728 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1729 {
1730 if (tc >= dev->num_tc)
1731 return -EINVAL;
1732
1733 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1734 return 0;
1735 }
1736
1737 static inline
1738 void netdev_reset_tc(struct net_device *dev)
1739 {
1740 dev->num_tc = 0;
1741 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1742 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1743 }
1744
1745 static inline
1746 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1747 {
1748 if (tc >= dev->num_tc)
1749 return -EINVAL;
1750
1751 dev->tc_to_txq[tc].count = count;
1752 dev->tc_to_txq[tc].offset = offset;
1753 return 0;
1754 }
1755
1756 static inline
1757 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1758 {
1759 if (num_tc > TC_MAX_QUEUE)
1760 return -EINVAL;
1761
1762 dev->num_tc = num_tc;
1763 return 0;
1764 }
1765
1766 static inline
1767 int netdev_get_num_tc(struct net_device *dev)
1768 {
1769 return dev->num_tc;
1770 }
1771
1772 static inline
1773 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1774 unsigned int index)
1775 {
1776 return &dev->_tx[index];
1777 }
1778
1779 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1780 const struct sk_buff *skb)
1781 {
1782 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1783 }
1784
1785 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1786 void (*f)(struct net_device *,
1787 struct netdev_queue *,
1788 void *),
1789 void *arg)
1790 {
1791 unsigned int i;
1792
1793 for (i = 0; i < dev->num_tx_queues; i++)
1794 f(dev, &dev->_tx[i], arg);
1795 }
1796
1797 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1798 struct sk_buff *skb,
1799 void *accel_priv);
1800
1801 /*
1802 * Net namespace inlines
1803 */
1804 static inline
1805 struct net *dev_net(const struct net_device *dev)
1806 {
1807 return read_pnet(&dev->nd_net);
1808 }
1809
1810 static inline
1811 void dev_net_set(struct net_device *dev, struct net *net)
1812 {
1813 #ifdef CONFIG_NET_NS
1814 release_net(dev->nd_net);
1815 dev->nd_net = hold_net(net);
1816 #endif
1817 }
1818
1819 static inline bool netdev_uses_dsa(struct net_device *dev)
1820 {
1821 #if IS_ENABLED(CONFIG_NET_DSA)
1822 if (dev->dsa_ptr != NULL)
1823 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1824 #endif
1825 return false;
1826 }
1827
1828 /**
1829 * netdev_priv - access network device private data
1830 * @dev: network device
1831 *
1832 * Get network device private data
1833 */
1834 static inline void *netdev_priv(const struct net_device *dev)
1835 {
1836 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1837 }
1838
1839 /* Set the sysfs physical device reference for the network logical device
1840 * if set prior to registration will cause a symlink during initialization.
1841 */
1842 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1843
1844 /* Set the sysfs device type for the network logical device to allow
1845 * fine-grained identification of different network device types. For
1846 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1847 */
1848 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1849
1850 /* Default NAPI poll() weight
1851 * Device drivers are strongly advised to not use bigger value
1852 */
1853 #define NAPI_POLL_WEIGHT 64
1854
1855 /**
1856 * netif_napi_add - initialize a napi context
1857 * @dev: network device
1858 * @napi: napi context
1859 * @poll: polling function
1860 * @weight: default weight
1861 *
1862 * netif_napi_add() must be used to initialize a napi context prior to calling
1863 * *any* of the other napi related functions.
1864 */
1865 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1866 int (*poll)(struct napi_struct *, int), int weight);
1867
1868 /**
1869 * netif_napi_del - remove a napi context
1870 * @napi: napi context
1871 *
1872 * netif_napi_del() removes a napi context from the network device napi list
1873 */
1874 void netif_napi_del(struct napi_struct *napi);
1875
1876 struct napi_gro_cb {
1877 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1878 void *frag0;
1879
1880 /* Length of frag0. */
1881 unsigned int frag0_len;
1882
1883 /* This indicates where we are processing relative to skb->data. */
1884 int data_offset;
1885
1886 /* This is non-zero if the packet cannot be merged with the new skb. */
1887 u16 flush;
1888
1889 /* Save the IP ID here and check when we get to the transport layer */
1890 u16 flush_id;
1891
1892 /* Number of segments aggregated. */
1893 u16 count;
1894
1895 /* This is non-zero if the packet may be of the same flow. */
1896 u8 same_flow;
1897
1898 /* Free the skb? */
1899 u8 free;
1900 #define NAPI_GRO_FREE 1
1901 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1902
1903 /* jiffies when first packet was created/queued */
1904 unsigned long age;
1905
1906 /* Used in ipv6_gro_receive() and foo-over-udp */
1907 u16 proto;
1908
1909 /* Used in udp_gro_receive */
1910 u8 udp_mark:1;
1911
1912 /* GRO checksum is valid */
1913 u8 csum_valid:1;
1914
1915 /* Number of checksums via CHECKSUM_UNNECESSARY */
1916 u8 csum_cnt:3;
1917
1918 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1919 u8 is_ipv6:1;
1920
1921 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1922 __wsum csum;
1923
1924 /* used in skb_gro_receive() slow path */
1925 struct sk_buff *last;
1926 };
1927
1928 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1929
1930 struct packet_type {
1931 __be16 type; /* This is really htons(ether_type). */
1932 struct net_device *dev; /* NULL is wildcarded here */
1933 int (*func) (struct sk_buff *,
1934 struct net_device *,
1935 struct packet_type *,
1936 struct net_device *);
1937 bool (*id_match)(struct packet_type *ptype,
1938 struct sock *sk);
1939 void *af_packet_priv;
1940 struct list_head list;
1941 };
1942
1943 struct offload_callbacks {
1944 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1945 netdev_features_t features);
1946 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1947 struct sk_buff *skb);
1948 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1949 };
1950
1951 struct packet_offload {
1952 __be16 type; /* This is really htons(ether_type). */
1953 struct offload_callbacks callbacks;
1954 struct list_head list;
1955 };
1956
1957 struct udp_offload {
1958 __be16 port;
1959 u8 ipproto;
1960 struct offload_callbacks callbacks;
1961 };
1962
1963 /* often modified stats are per cpu, other are shared (netdev->stats) */
1964 struct pcpu_sw_netstats {
1965 u64 rx_packets;
1966 u64 rx_bytes;
1967 u64 tx_packets;
1968 u64 tx_bytes;
1969 struct u64_stats_sync syncp;
1970 };
1971
1972 #define netdev_alloc_pcpu_stats(type) \
1973 ({ \
1974 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1975 if (pcpu_stats) { \
1976 int i; \
1977 for_each_possible_cpu(i) { \
1978 typeof(type) *stat; \
1979 stat = per_cpu_ptr(pcpu_stats, i); \
1980 u64_stats_init(&stat->syncp); \
1981 } \
1982 } \
1983 pcpu_stats; \
1984 })
1985
1986 #include <linux/notifier.h>
1987
1988 /* netdevice notifier chain. Please remember to update the rtnetlink
1989 * notification exclusion list in rtnetlink_event() when adding new
1990 * types.
1991 */
1992 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1993 #define NETDEV_DOWN 0x0002
1994 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1995 detected a hardware crash and restarted
1996 - we can use this eg to kick tcp sessions
1997 once done */
1998 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1999 #define NETDEV_REGISTER 0x0005
2000 #define NETDEV_UNREGISTER 0x0006
2001 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2002 #define NETDEV_CHANGEADDR 0x0008
2003 #define NETDEV_GOING_DOWN 0x0009
2004 #define NETDEV_CHANGENAME 0x000A
2005 #define NETDEV_FEAT_CHANGE 0x000B
2006 #define NETDEV_BONDING_FAILOVER 0x000C
2007 #define NETDEV_PRE_UP 0x000D
2008 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2009 #define NETDEV_POST_TYPE_CHANGE 0x000F
2010 #define NETDEV_POST_INIT 0x0010
2011 #define NETDEV_UNREGISTER_FINAL 0x0011
2012 #define NETDEV_RELEASE 0x0012
2013 #define NETDEV_NOTIFY_PEERS 0x0013
2014 #define NETDEV_JOIN 0x0014
2015 #define NETDEV_CHANGEUPPER 0x0015
2016 #define NETDEV_RESEND_IGMP 0x0016
2017 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2018 #define NETDEV_CHANGEINFODATA 0x0018
2019
2020 int register_netdevice_notifier(struct notifier_block *nb);
2021 int unregister_netdevice_notifier(struct notifier_block *nb);
2022
2023 struct netdev_notifier_info {
2024 struct net_device *dev;
2025 };
2026
2027 struct netdev_notifier_change_info {
2028 struct netdev_notifier_info info; /* must be first */
2029 unsigned int flags_changed;
2030 };
2031
2032 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2033 struct net_device *dev)
2034 {
2035 info->dev = dev;
2036 }
2037
2038 static inline struct net_device *
2039 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2040 {
2041 return info->dev;
2042 }
2043
2044 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2045
2046
2047 extern rwlock_t dev_base_lock; /* Device list lock */
2048
2049 #define for_each_netdev(net, d) \
2050 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2051 #define for_each_netdev_reverse(net, d) \
2052 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2053 #define for_each_netdev_rcu(net, d) \
2054 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2055 #define for_each_netdev_safe(net, d, n) \
2056 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2057 #define for_each_netdev_continue(net, d) \
2058 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2059 #define for_each_netdev_continue_rcu(net, d) \
2060 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2061 #define for_each_netdev_in_bond_rcu(bond, slave) \
2062 for_each_netdev_rcu(&init_net, slave) \
2063 if (netdev_master_upper_dev_get_rcu(slave) == bond)
2064 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2065
2066 static inline struct net_device *next_net_device(struct net_device *dev)
2067 {
2068 struct list_head *lh;
2069 struct net *net;
2070
2071 net = dev_net(dev);
2072 lh = dev->dev_list.next;
2073 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2074 }
2075
2076 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2077 {
2078 struct list_head *lh;
2079 struct net *net;
2080
2081 net = dev_net(dev);
2082 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2083 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2084 }
2085
2086 static inline struct net_device *first_net_device(struct net *net)
2087 {
2088 return list_empty(&net->dev_base_head) ? NULL :
2089 net_device_entry(net->dev_base_head.next);
2090 }
2091
2092 static inline struct net_device *first_net_device_rcu(struct net *net)
2093 {
2094 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2095
2096 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2097 }
2098
2099 int netdev_boot_setup_check(struct net_device *dev);
2100 unsigned long netdev_boot_base(const char *prefix, int unit);
2101 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2102 const char *hwaddr);
2103 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2104 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2105 void dev_add_pack(struct packet_type *pt);
2106 void dev_remove_pack(struct packet_type *pt);
2107 void __dev_remove_pack(struct packet_type *pt);
2108 void dev_add_offload(struct packet_offload *po);
2109 void dev_remove_offload(struct packet_offload *po);
2110
2111 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2112 unsigned short mask);
2113 struct net_device *dev_get_by_name(struct net *net, const char *name);
2114 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2115 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2116 int dev_alloc_name(struct net_device *dev, const char *name);
2117 int dev_open(struct net_device *dev);
2118 int dev_close(struct net_device *dev);
2119 void dev_disable_lro(struct net_device *dev);
2120 int dev_loopback_xmit(struct sk_buff *newskb);
2121 int dev_queue_xmit(struct sk_buff *skb);
2122 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2123 int register_netdevice(struct net_device *dev);
2124 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2125 void unregister_netdevice_many(struct list_head *head);
2126 static inline void unregister_netdevice(struct net_device *dev)
2127 {
2128 unregister_netdevice_queue(dev, NULL);
2129 }
2130
2131 int netdev_refcnt_read(const struct net_device *dev);
2132 void free_netdev(struct net_device *dev);
2133 void netdev_freemem(struct net_device *dev);
2134 void synchronize_net(void);
2135 int init_dummy_netdev(struct net_device *dev);
2136
2137 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2138 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2139 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2140 int netdev_get_name(struct net *net, char *name, int ifindex);
2141 int dev_restart(struct net_device *dev);
2142 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2143
2144 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2145 {
2146 return NAPI_GRO_CB(skb)->data_offset;
2147 }
2148
2149 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2150 {
2151 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2152 }
2153
2154 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2155 {
2156 NAPI_GRO_CB(skb)->data_offset += len;
2157 }
2158
2159 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2160 unsigned int offset)
2161 {
2162 return NAPI_GRO_CB(skb)->frag0 + offset;
2163 }
2164
2165 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2166 {
2167 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2168 }
2169
2170 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2171 unsigned int offset)
2172 {
2173 if (!pskb_may_pull(skb, hlen))
2174 return NULL;
2175
2176 NAPI_GRO_CB(skb)->frag0 = NULL;
2177 NAPI_GRO_CB(skb)->frag0_len = 0;
2178 return skb->data + offset;
2179 }
2180
2181 static inline void *skb_gro_network_header(struct sk_buff *skb)
2182 {
2183 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2184 skb_network_offset(skb);
2185 }
2186
2187 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2188 const void *start, unsigned int len)
2189 {
2190 if (NAPI_GRO_CB(skb)->csum_valid)
2191 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2192 csum_partial(start, len, 0));
2193 }
2194
2195 /* GRO checksum functions. These are logical equivalents of the normal
2196 * checksum functions (in skbuff.h) except that they operate on the GRO
2197 * offsets and fields in sk_buff.
2198 */
2199
2200 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2201
2202 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2203 bool zero_okay,
2204 __sum16 check)
2205 {
2206 return (skb->ip_summed != CHECKSUM_PARTIAL &&
2207 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2208 (!zero_okay || check));
2209 }
2210
2211 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2212 __wsum psum)
2213 {
2214 if (NAPI_GRO_CB(skb)->csum_valid &&
2215 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2216 return 0;
2217
2218 NAPI_GRO_CB(skb)->csum = psum;
2219
2220 return __skb_gro_checksum_complete(skb);
2221 }
2222
2223 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2224 {
2225 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2226 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2227 NAPI_GRO_CB(skb)->csum_cnt--;
2228 } else {
2229 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2230 * verified a new top level checksum or an encapsulated one
2231 * during GRO. This saves work if we fallback to normal path.
2232 */
2233 __skb_incr_checksum_unnecessary(skb);
2234 }
2235 }
2236
2237 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2238 compute_pseudo) \
2239 ({ \
2240 __sum16 __ret = 0; \
2241 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2242 __ret = __skb_gro_checksum_validate_complete(skb, \
2243 compute_pseudo(skb, proto)); \
2244 if (__ret) \
2245 __skb_mark_checksum_bad(skb); \
2246 else \
2247 skb_gro_incr_csum_unnecessary(skb); \
2248 __ret; \
2249 })
2250
2251 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2252 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2253
2254 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2255 compute_pseudo) \
2256 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2257
2258 #define skb_gro_checksum_simple_validate(skb) \
2259 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2260
2261 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2262 {
2263 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2264 !NAPI_GRO_CB(skb)->csum_valid);
2265 }
2266
2267 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2268 __sum16 check, __wsum pseudo)
2269 {
2270 NAPI_GRO_CB(skb)->csum = ~pseudo;
2271 NAPI_GRO_CB(skb)->csum_valid = 1;
2272 }
2273
2274 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2275 do { \
2276 if (__skb_gro_checksum_convert_check(skb)) \
2277 __skb_gro_checksum_convert(skb, check, \
2278 compute_pseudo(skb, proto)); \
2279 } while (0)
2280
2281 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2282 unsigned short type,
2283 const void *daddr, const void *saddr,
2284 unsigned int len)
2285 {
2286 if (!dev->header_ops || !dev->header_ops->create)
2287 return 0;
2288
2289 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2290 }
2291
2292 static inline int dev_parse_header(const struct sk_buff *skb,
2293 unsigned char *haddr)
2294 {
2295 const struct net_device *dev = skb->dev;
2296
2297 if (!dev->header_ops || !dev->header_ops->parse)
2298 return 0;
2299 return dev->header_ops->parse(skb, haddr);
2300 }
2301
2302 static inline int dev_rebuild_header(struct sk_buff *skb)
2303 {
2304 const struct net_device *dev = skb->dev;
2305
2306 if (!dev->header_ops || !dev->header_ops->rebuild)
2307 return 0;
2308 return dev->header_ops->rebuild(skb);
2309 }
2310
2311 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2312 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2313 static inline int unregister_gifconf(unsigned int family)
2314 {
2315 return register_gifconf(family, NULL);
2316 }
2317
2318 #ifdef CONFIG_NET_FLOW_LIMIT
2319 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2320 struct sd_flow_limit {
2321 u64 count;
2322 unsigned int num_buckets;
2323 unsigned int history_head;
2324 u16 history[FLOW_LIMIT_HISTORY];
2325 u8 buckets[];
2326 };
2327
2328 extern int netdev_flow_limit_table_len;
2329 #endif /* CONFIG_NET_FLOW_LIMIT */
2330
2331 /*
2332 * Incoming packets are placed on per-cpu queues
2333 */
2334 struct softnet_data {
2335 struct list_head poll_list;
2336 struct sk_buff_head process_queue;
2337
2338 /* stats */
2339 unsigned int processed;
2340 unsigned int time_squeeze;
2341 unsigned int cpu_collision;
2342 unsigned int received_rps;
2343 #ifdef CONFIG_RPS
2344 struct softnet_data *rps_ipi_list;
2345 #endif
2346 #ifdef CONFIG_NET_FLOW_LIMIT
2347 struct sd_flow_limit __rcu *flow_limit;
2348 #endif
2349 struct Qdisc *output_queue;
2350 struct Qdisc **output_queue_tailp;
2351 struct sk_buff *completion_queue;
2352
2353 #ifdef CONFIG_RPS
2354 /* Elements below can be accessed between CPUs for RPS */
2355 struct call_single_data csd ____cacheline_aligned_in_smp;
2356 struct softnet_data *rps_ipi_next;
2357 unsigned int cpu;
2358 unsigned int input_queue_head;
2359 unsigned int input_queue_tail;
2360 #endif
2361 unsigned int dropped;
2362 struct sk_buff_head input_pkt_queue;
2363 struct napi_struct backlog;
2364
2365 };
2366
2367 static inline void input_queue_head_incr(struct softnet_data *sd)
2368 {
2369 #ifdef CONFIG_RPS
2370 sd->input_queue_head++;
2371 #endif
2372 }
2373
2374 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2375 unsigned int *qtail)
2376 {
2377 #ifdef CONFIG_RPS
2378 *qtail = ++sd->input_queue_tail;
2379 #endif
2380 }
2381
2382 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2383
2384 void __netif_schedule(struct Qdisc *q);
2385 void netif_schedule_queue(struct netdev_queue *txq);
2386
2387 static inline void netif_tx_schedule_all(struct net_device *dev)
2388 {
2389 unsigned int i;
2390
2391 for (i = 0; i < dev->num_tx_queues; i++)
2392 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2393 }
2394
2395 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2396 {
2397 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2398 }
2399
2400 /**
2401 * netif_start_queue - allow transmit
2402 * @dev: network device
2403 *
2404 * Allow upper layers to call the device hard_start_xmit routine.
2405 */
2406 static inline void netif_start_queue(struct net_device *dev)
2407 {
2408 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2409 }
2410
2411 static inline void netif_tx_start_all_queues(struct net_device *dev)
2412 {
2413 unsigned int i;
2414
2415 for (i = 0; i < dev->num_tx_queues; i++) {
2416 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2417 netif_tx_start_queue(txq);
2418 }
2419 }
2420
2421 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2422
2423 /**
2424 * netif_wake_queue - restart transmit
2425 * @dev: network device
2426 *
2427 * Allow upper layers to call the device hard_start_xmit routine.
2428 * Used for flow control when transmit resources are available.
2429 */
2430 static inline void netif_wake_queue(struct net_device *dev)
2431 {
2432 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2433 }
2434
2435 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2436 {
2437 unsigned int i;
2438
2439 for (i = 0; i < dev->num_tx_queues; i++) {
2440 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2441 netif_tx_wake_queue(txq);
2442 }
2443 }
2444
2445 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2446 {
2447 if (WARN_ON(!dev_queue)) {
2448 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2449 return;
2450 }
2451 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2452 }
2453
2454 /**
2455 * netif_stop_queue - stop transmitted packets
2456 * @dev: network device
2457 *
2458 * Stop upper layers calling the device hard_start_xmit routine.
2459 * Used for flow control when transmit resources are unavailable.
2460 */
2461 static inline void netif_stop_queue(struct net_device *dev)
2462 {
2463 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2464 }
2465
2466 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2467 {
2468 unsigned int i;
2469
2470 for (i = 0; i < dev->num_tx_queues; i++) {
2471 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2472 netif_tx_stop_queue(txq);
2473 }
2474 }
2475
2476 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2477 {
2478 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2479 }
2480
2481 /**
2482 * netif_queue_stopped - test if transmit queue is flowblocked
2483 * @dev: network device
2484 *
2485 * Test if transmit queue on device is currently unable to send.
2486 */
2487 static inline bool netif_queue_stopped(const struct net_device *dev)
2488 {
2489 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2490 }
2491
2492 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2493 {
2494 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2495 }
2496
2497 static inline bool
2498 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2499 {
2500 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2501 }
2502
2503 static inline bool
2504 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2505 {
2506 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2507 }
2508
2509 /**
2510 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2511 * @dev_queue: pointer to transmit queue
2512 *
2513 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2514 * to give appropriate hint to the cpu.
2515 */
2516 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2517 {
2518 #ifdef CONFIG_BQL
2519 prefetchw(&dev_queue->dql.num_queued);
2520 #endif
2521 }
2522
2523 /**
2524 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2525 * @dev_queue: pointer to transmit queue
2526 *
2527 * BQL enabled drivers might use this helper in their TX completion path,
2528 * to give appropriate hint to the cpu.
2529 */
2530 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2531 {
2532 #ifdef CONFIG_BQL
2533 prefetchw(&dev_queue->dql.limit);
2534 #endif
2535 }
2536
2537 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2538 unsigned int bytes)
2539 {
2540 #ifdef CONFIG_BQL
2541 dql_queued(&dev_queue->dql, bytes);
2542
2543 if (likely(dql_avail(&dev_queue->dql) >= 0))
2544 return;
2545
2546 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2547
2548 /*
2549 * The XOFF flag must be set before checking the dql_avail below,
2550 * because in netdev_tx_completed_queue we update the dql_completed
2551 * before checking the XOFF flag.
2552 */
2553 smp_mb();
2554
2555 /* check again in case another CPU has just made room avail */
2556 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2557 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2558 #endif
2559 }
2560
2561 /**
2562 * netdev_sent_queue - report the number of bytes queued to hardware
2563 * @dev: network device
2564 * @bytes: number of bytes queued to the hardware device queue
2565 *
2566 * Report the number of bytes queued for sending/completion to the network
2567 * device hardware queue. @bytes should be a good approximation and should
2568 * exactly match netdev_completed_queue() @bytes
2569 */
2570 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2571 {
2572 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2573 }
2574
2575 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2576 unsigned int pkts, unsigned int bytes)
2577 {
2578 #ifdef CONFIG_BQL
2579 if (unlikely(!bytes))
2580 return;
2581
2582 dql_completed(&dev_queue->dql, bytes);
2583
2584 /*
2585 * Without the memory barrier there is a small possiblity that
2586 * netdev_tx_sent_queue will miss the update and cause the queue to
2587 * be stopped forever
2588 */
2589 smp_mb();
2590
2591 if (dql_avail(&dev_queue->dql) < 0)
2592 return;
2593
2594 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2595 netif_schedule_queue(dev_queue);
2596 #endif
2597 }
2598
2599 /**
2600 * netdev_completed_queue - report bytes and packets completed by device
2601 * @dev: network device
2602 * @pkts: actual number of packets sent over the medium
2603 * @bytes: actual number of bytes sent over the medium
2604 *
2605 * Report the number of bytes and packets transmitted by the network device
2606 * hardware queue over the physical medium, @bytes must exactly match the
2607 * @bytes amount passed to netdev_sent_queue()
2608 */
2609 static inline void netdev_completed_queue(struct net_device *dev,
2610 unsigned int pkts, unsigned int bytes)
2611 {
2612 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2613 }
2614
2615 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2616 {
2617 #ifdef CONFIG_BQL
2618 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2619 dql_reset(&q->dql);
2620 #endif
2621 }
2622
2623 /**
2624 * netdev_reset_queue - reset the packets and bytes count of a network device
2625 * @dev_queue: network device
2626 *
2627 * Reset the bytes and packet count of a network device and clear the
2628 * software flow control OFF bit for this network device
2629 */
2630 static inline void netdev_reset_queue(struct net_device *dev_queue)
2631 {
2632 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2633 }
2634
2635 /**
2636 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2637 * @dev: network device
2638 * @queue_index: given tx queue index
2639 *
2640 * Returns 0 if given tx queue index >= number of device tx queues,
2641 * otherwise returns the originally passed tx queue index.
2642 */
2643 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2644 {
2645 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2646 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2647 dev->name, queue_index,
2648 dev->real_num_tx_queues);
2649 return 0;
2650 }
2651
2652 return queue_index;
2653 }
2654
2655 /**
2656 * netif_running - test if up
2657 * @dev: network device
2658 *
2659 * Test if the device has been brought up.
2660 */
2661 static inline bool netif_running(const struct net_device *dev)
2662 {
2663 return test_bit(__LINK_STATE_START, &dev->state);
2664 }
2665
2666 /*
2667 * Routines to manage the subqueues on a device. We only need start
2668 * stop, and a check if it's stopped. All other device management is
2669 * done at the overall netdevice level.
2670 * Also test the device if we're multiqueue.
2671 */
2672
2673 /**
2674 * netif_start_subqueue - allow sending packets on subqueue
2675 * @dev: network device
2676 * @queue_index: sub queue index
2677 *
2678 * Start individual transmit queue of a device with multiple transmit queues.
2679 */
2680 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2681 {
2682 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2683
2684 netif_tx_start_queue(txq);
2685 }
2686
2687 /**
2688 * netif_stop_subqueue - stop sending packets on subqueue
2689 * @dev: network device
2690 * @queue_index: sub queue index
2691 *
2692 * Stop individual transmit queue of a device with multiple transmit queues.
2693 */
2694 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2695 {
2696 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2697 netif_tx_stop_queue(txq);
2698 }
2699
2700 /**
2701 * netif_subqueue_stopped - test status of subqueue
2702 * @dev: network device
2703 * @queue_index: sub queue index
2704 *
2705 * Check individual transmit queue of a device with multiple transmit queues.
2706 */
2707 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2708 u16 queue_index)
2709 {
2710 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2711
2712 return netif_tx_queue_stopped(txq);
2713 }
2714
2715 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2716 struct sk_buff *skb)
2717 {
2718 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2719 }
2720
2721 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2722
2723 #ifdef CONFIG_XPS
2724 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2725 u16 index);
2726 #else
2727 static inline int netif_set_xps_queue(struct net_device *dev,
2728 const struct cpumask *mask,
2729 u16 index)
2730 {
2731 return 0;
2732 }
2733 #endif
2734
2735 /*
2736 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2737 * as a distribution range limit for the returned value.
2738 */
2739 static inline u16 skb_tx_hash(const struct net_device *dev,
2740 struct sk_buff *skb)
2741 {
2742 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2743 }
2744
2745 /**
2746 * netif_is_multiqueue - test if device has multiple transmit queues
2747 * @dev: network device
2748 *
2749 * Check if device has multiple transmit queues
2750 */
2751 static inline bool netif_is_multiqueue(const struct net_device *dev)
2752 {
2753 return dev->num_tx_queues > 1;
2754 }
2755
2756 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2757
2758 #ifdef CONFIG_SYSFS
2759 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2760 #else
2761 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2762 unsigned int rxq)
2763 {
2764 return 0;
2765 }
2766 #endif
2767
2768 #ifdef CONFIG_SYSFS
2769 static inline unsigned int get_netdev_rx_queue_index(
2770 struct netdev_rx_queue *queue)
2771 {
2772 struct net_device *dev = queue->dev;
2773 int index = queue - dev->_rx;
2774
2775 BUG_ON(index >= dev->num_rx_queues);
2776 return index;
2777 }
2778 #endif
2779
2780 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2781 int netif_get_num_default_rss_queues(void);
2782
2783 enum skb_free_reason {
2784 SKB_REASON_CONSUMED,
2785 SKB_REASON_DROPPED,
2786 };
2787
2788 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2789 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2790
2791 /*
2792 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2793 * interrupt context or with hardware interrupts being disabled.
2794 * (in_irq() || irqs_disabled())
2795 *
2796 * We provide four helpers that can be used in following contexts :
2797 *
2798 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2799 * replacing kfree_skb(skb)
2800 *
2801 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2802 * Typically used in place of consume_skb(skb) in TX completion path
2803 *
2804 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2805 * replacing kfree_skb(skb)
2806 *
2807 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2808 * and consumed a packet. Used in place of consume_skb(skb)
2809 */
2810 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2811 {
2812 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2813 }
2814
2815 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2816 {
2817 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2818 }
2819
2820 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2821 {
2822 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2823 }
2824
2825 static inline void dev_consume_skb_any(struct sk_buff *skb)
2826 {
2827 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2828 }
2829
2830 int netif_rx(struct sk_buff *skb);
2831 int netif_rx_ni(struct sk_buff *skb);
2832 int netif_receive_skb(struct sk_buff *skb);
2833 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2834 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2835 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2836 gro_result_t napi_gro_frags(struct napi_struct *napi);
2837 struct packet_offload *gro_find_receive_by_type(__be16 type);
2838 struct packet_offload *gro_find_complete_by_type(__be16 type);
2839
2840 static inline void napi_free_frags(struct napi_struct *napi)
2841 {
2842 kfree_skb(napi->skb);
2843 napi->skb = NULL;
2844 }
2845
2846 int netdev_rx_handler_register(struct net_device *dev,
2847 rx_handler_func_t *rx_handler,
2848 void *rx_handler_data);
2849 void netdev_rx_handler_unregister(struct net_device *dev);
2850
2851 bool dev_valid_name(const char *name);
2852 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2853 int dev_ethtool(struct net *net, struct ifreq *);
2854 unsigned int dev_get_flags(const struct net_device *);
2855 int __dev_change_flags(struct net_device *, unsigned int flags);
2856 int dev_change_flags(struct net_device *, unsigned int);
2857 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2858 unsigned int gchanges);
2859 int dev_change_name(struct net_device *, const char *);
2860 int dev_set_alias(struct net_device *, const char *, size_t);
2861 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2862 int dev_set_mtu(struct net_device *, int);
2863 void dev_set_group(struct net_device *, int);
2864 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2865 int dev_change_carrier(struct net_device *, bool new_carrier);
2866 int dev_get_phys_port_id(struct net_device *dev,
2867 struct netdev_phys_port_id *ppid);
2868 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2869 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2870 struct netdev_queue *txq, int *ret);
2871 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2872 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2873 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2874
2875 extern int netdev_budget;
2876
2877 /* Called by rtnetlink.c:rtnl_unlock() */
2878 void netdev_run_todo(void);
2879
2880 /**
2881 * dev_put - release reference to device
2882 * @dev: network device
2883 *
2884 * Release reference to device to allow it to be freed.
2885 */
2886 static inline void dev_put(struct net_device *dev)
2887 {
2888 this_cpu_dec(*dev->pcpu_refcnt);
2889 }
2890
2891 /**
2892 * dev_hold - get reference to device
2893 * @dev: network device
2894 *
2895 * Hold reference to device to keep it from being freed.
2896 */
2897 static inline void dev_hold(struct net_device *dev)
2898 {
2899 this_cpu_inc(*dev->pcpu_refcnt);
2900 }
2901
2902 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2903 * and _off may be called from IRQ context, but it is caller
2904 * who is responsible for serialization of these calls.
2905 *
2906 * The name carrier is inappropriate, these functions should really be
2907 * called netif_lowerlayer_*() because they represent the state of any
2908 * kind of lower layer not just hardware media.
2909 */
2910
2911 void linkwatch_init_dev(struct net_device *dev);
2912 void linkwatch_fire_event(struct net_device *dev);
2913 void linkwatch_forget_dev(struct net_device *dev);
2914
2915 /**
2916 * netif_carrier_ok - test if carrier present
2917 * @dev: network device
2918 *
2919 * Check if carrier is present on device
2920 */
2921 static inline bool netif_carrier_ok(const struct net_device *dev)
2922 {
2923 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2924 }
2925
2926 unsigned long dev_trans_start(struct net_device *dev);
2927
2928 void __netdev_watchdog_up(struct net_device *dev);
2929
2930 void netif_carrier_on(struct net_device *dev);
2931
2932 void netif_carrier_off(struct net_device *dev);
2933
2934 /**
2935 * netif_dormant_on - mark device as dormant.
2936 * @dev: network device
2937 *
2938 * Mark device as dormant (as per RFC2863).
2939 *
2940 * The dormant state indicates that the relevant interface is not
2941 * actually in a condition to pass packets (i.e., it is not 'up') but is
2942 * in a "pending" state, waiting for some external event. For "on-
2943 * demand" interfaces, this new state identifies the situation where the
2944 * interface is waiting for events to place it in the up state.
2945 *
2946 */
2947 static inline void netif_dormant_on(struct net_device *dev)
2948 {
2949 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2950 linkwatch_fire_event(dev);
2951 }
2952
2953 /**
2954 * netif_dormant_off - set device as not dormant.
2955 * @dev: network device
2956 *
2957 * Device is not in dormant state.
2958 */
2959 static inline void netif_dormant_off(struct net_device *dev)
2960 {
2961 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2962 linkwatch_fire_event(dev);
2963 }
2964
2965 /**
2966 * netif_dormant - test if carrier present
2967 * @dev: network device
2968 *
2969 * Check if carrier is present on device
2970 */
2971 static inline bool netif_dormant(const struct net_device *dev)
2972 {
2973 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2974 }
2975
2976
2977 /**
2978 * netif_oper_up - test if device is operational
2979 * @dev: network device
2980 *
2981 * Check if carrier is operational
2982 */
2983 static inline bool netif_oper_up(const struct net_device *dev)
2984 {
2985 return (dev->operstate == IF_OPER_UP ||
2986 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2987 }
2988
2989 /**
2990 * netif_device_present - is device available or removed
2991 * @dev: network device
2992 *
2993 * Check if device has not been removed from system.
2994 */
2995 static inline bool netif_device_present(struct net_device *dev)
2996 {
2997 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2998 }
2999
3000 void netif_device_detach(struct net_device *dev);
3001
3002 void netif_device_attach(struct net_device *dev);
3003
3004 /*
3005 * Network interface message level settings
3006 */
3007
3008 enum {
3009 NETIF_MSG_DRV = 0x0001,
3010 NETIF_MSG_PROBE = 0x0002,
3011 NETIF_MSG_LINK = 0x0004,
3012 NETIF_MSG_TIMER = 0x0008,
3013 NETIF_MSG_IFDOWN = 0x0010,
3014 NETIF_MSG_IFUP = 0x0020,
3015 NETIF_MSG_RX_ERR = 0x0040,
3016 NETIF_MSG_TX_ERR = 0x0080,
3017 NETIF_MSG_TX_QUEUED = 0x0100,
3018 NETIF_MSG_INTR = 0x0200,
3019 NETIF_MSG_TX_DONE = 0x0400,
3020 NETIF_MSG_RX_STATUS = 0x0800,
3021 NETIF_MSG_PKTDATA = 0x1000,
3022 NETIF_MSG_HW = 0x2000,
3023 NETIF_MSG_WOL = 0x4000,
3024 };
3025
3026 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3027 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3028 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3029 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3030 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3031 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3032 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3033 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3034 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3035 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3036 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3037 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3038 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3039 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3040 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3041
3042 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3043 {
3044 /* use default */
3045 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3046 return default_msg_enable_bits;
3047 if (debug_value == 0) /* no output */
3048 return 0;
3049 /* set low N bits */
3050 return (1 << debug_value) - 1;
3051 }
3052
3053 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3054 {
3055 spin_lock(&txq->_xmit_lock);
3056 txq->xmit_lock_owner = cpu;
3057 }
3058
3059 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3060 {
3061 spin_lock_bh(&txq->_xmit_lock);
3062 txq->xmit_lock_owner = smp_processor_id();
3063 }
3064
3065 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3066 {
3067 bool ok = spin_trylock(&txq->_xmit_lock);
3068 if (likely(ok))
3069 txq->xmit_lock_owner = smp_processor_id();
3070 return ok;
3071 }
3072
3073 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3074 {
3075 txq->xmit_lock_owner = -1;
3076 spin_unlock(&txq->_xmit_lock);
3077 }
3078
3079 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3080 {
3081 txq->xmit_lock_owner = -1;
3082 spin_unlock_bh(&txq->_xmit_lock);
3083 }
3084
3085 static inline void txq_trans_update(struct netdev_queue *txq)
3086 {
3087 if (txq->xmit_lock_owner != -1)
3088 txq->trans_start = jiffies;
3089 }
3090
3091 /**
3092 * netif_tx_lock - grab network device transmit lock
3093 * @dev: network device
3094 *
3095 * Get network device transmit lock
3096 */
3097 static inline void netif_tx_lock(struct net_device *dev)
3098 {
3099 unsigned int i;
3100 int cpu;
3101
3102 spin_lock(&dev->tx_global_lock);
3103 cpu = smp_processor_id();
3104 for (i = 0; i < dev->num_tx_queues; i++) {
3105 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3106
3107 /* We are the only thread of execution doing a
3108 * freeze, but we have to grab the _xmit_lock in
3109 * order to synchronize with threads which are in
3110 * the ->hard_start_xmit() handler and already
3111 * checked the frozen bit.
3112 */
3113 __netif_tx_lock(txq, cpu);
3114 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3115 __netif_tx_unlock(txq);
3116 }
3117 }
3118
3119 static inline void netif_tx_lock_bh(struct net_device *dev)
3120 {
3121 local_bh_disable();
3122 netif_tx_lock(dev);
3123 }
3124
3125 static inline void netif_tx_unlock(struct net_device *dev)
3126 {
3127 unsigned int i;
3128
3129 for (i = 0; i < dev->num_tx_queues; i++) {
3130 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3131
3132 /* No need to grab the _xmit_lock here. If the
3133 * queue is not stopped for another reason, we
3134 * force a schedule.
3135 */
3136 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3137 netif_schedule_queue(txq);
3138 }
3139 spin_unlock(&dev->tx_global_lock);
3140 }
3141
3142 static inline void netif_tx_unlock_bh(struct net_device *dev)
3143 {
3144 netif_tx_unlock(dev);
3145 local_bh_enable();
3146 }
3147
3148 #define HARD_TX_LOCK(dev, txq, cpu) { \
3149 if ((dev->features & NETIF_F_LLTX) == 0) { \
3150 __netif_tx_lock(txq, cpu); \
3151 } \
3152 }
3153
3154 #define HARD_TX_TRYLOCK(dev, txq) \
3155 (((dev->features & NETIF_F_LLTX) == 0) ? \
3156 __netif_tx_trylock(txq) : \
3157 true )
3158
3159 #define HARD_TX_UNLOCK(dev, txq) { \
3160 if ((dev->features & NETIF_F_LLTX) == 0) { \
3161 __netif_tx_unlock(txq); \
3162 } \
3163 }
3164
3165 static inline void netif_tx_disable(struct net_device *dev)
3166 {
3167 unsigned int i;
3168 int cpu;
3169
3170 local_bh_disable();
3171 cpu = smp_processor_id();
3172 for (i = 0; i < dev->num_tx_queues; i++) {
3173 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3174
3175 __netif_tx_lock(txq, cpu);
3176 netif_tx_stop_queue(txq);
3177 __netif_tx_unlock(txq);
3178 }
3179 local_bh_enable();
3180 }
3181
3182 static inline void netif_addr_lock(struct net_device *dev)
3183 {
3184 spin_lock(&dev->addr_list_lock);
3185 }
3186
3187 static inline void netif_addr_lock_nested(struct net_device *dev)
3188 {
3189 int subclass = SINGLE_DEPTH_NESTING;
3190
3191 if (dev->netdev_ops->ndo_get_lock_subclass)
3192 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3193
3194 spin_lock_nested(&dev->addr_list_lock, subclass);
3195 }
3196
3197 static inline void netif_addr_lock_bh(struct net_device *dev)
3198 {
3199 spin_lock_bh(&dev->addr_list_lock);
3200 }
3201
3202 static inline void netif_addr_unlock(struct net_device *dev)
3203 {
3204 spin_unlock(&dev->addr_list_lock);
3205 }
3206
3207 static inline void netif_addr_unlock_bh(struct net_device *dev)
3208 {
3209 spin_unlock_bh(&dev->addr_list_lock);
3210 }
3211
3212 /*
3213 * dev_addrs walker. Should be used only for read access. Call with
3214 * rcu_read_lock held.
3215 */
3216 #define for_each_dev_addr(dev, ha) \
3217 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3218
3219 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3220
3221 void ether_setup(struct net_device *dev);
3222
3223 /* Support for loadable net-drivers */
3224 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3225 unsigned char name_assign_type,
3226 void (*setup)(struct net_device *),
3227 unsigned int txqs, unsigned int rxqs);
3228 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3229 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3230
3231 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3232 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3233 count)
3234
3235 int register_netdev(struct net_device *dev);
3236 void unregister_netdev(struct net_device *dev);
3237
3238 /* General hardware address lists handling functions */
3239 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3240 struct netdev_hw_addr_list *from_list, int addr_len);
3241 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3242 struct netdev_hw_addr_list *from_list, int addr_len);
3243 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3244 struct net_device *dev,
3245 int (*sync)(struct net_device *, const unsigned char *),
3246 int (*unsync)(struct net_device *,
3247 const unsigned char *));
3248 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3249 struct net_device *dev,
3250 int (*unsync)(struct net_device *,
3251 const unsigned char *));
3252 void __hw_addr_init(struct netdev_hw_addr_list *list);
3253
3254 /* Functions used for device addresses handling */
3255 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3256 unsigned char addr_type);
3257 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3258 unsigned char addr_type);
3259 void dev_addr_flush(struct net_device *dev);
3260 int dev_addr_init(struct net_device *dev);
3261
3262 /* Functions used for unicast addresses handling */
3263 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3264 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3265 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3266 int dev_uc_sync(struct net_device *to, struct net_device *from);
3267 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3268 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3269 void dev_uc_flush(struct net_device *dev);
3270 void dev_uc_init(struct net_device *dev);
3271
3272 /**
3273 * __dev_uc_sync - Synchonize device's unicast list
3274 * @dev: device to sync
3275 * @sync: function to call if address should be added
3276 * @unsync: function to call if address should be removed
3277 *
3278 * Add newly added addresses to the interface, and release
3279 * addresses that have been deleted.
3280 **/
3281 static inline int __dev_uc_sync(struct net_device *dev,
3282 int (*sync)(struct net_device *,
3283 const unsigned char *),
3284 int (*unsync)(struct net_device *,
3285 const unsigned char *))
3286 {
3287 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3288 }
3289
3290 /**
3291 * __dev_uc_unsync - Remove synchronized addresses from device
3292 * @dev: device to sync
3293 * @unsync: function to call if address should be removed
3294 *
3295 * Remove all addresses that were added to the device by dev_uc_sync().
3296 **/
3297 static inline void __dev_uc_unsync(struct net_device *dev,
3298 int (*unsync)(struct net_device *,
3299 const unsigned char *))
3300 {
3301 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3302 }
3303
3304 /* Functions used for multicast addresses handling */
3305 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3306 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3307 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3308 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3309 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3310 int dev_mc_sync(struct net_device *to, struct net_device *from);
3311 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3312 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3313 void dev_mc_flush(struct net_device *dev);
3314 void dev_mc_init(struct net_device *dev);
3315
3316 /**
3317 * __dev_mc_sync - Synchonize device's multicast list
3318 * @dev: device to sync
3319 * @sync: function to call if address should be added
3320 * @unsync: function to call if address should be removed
3321 *
3322 * Add newly added addresses to the interface, and release
3323 * addresses that have been deleted.
3324 **/
3325 static inline int __dev_mc_sync(struct net_device *dev,
3326 int (*sync)(struct net_device *,
3327 const unsigned char *),
3328 int (*unsync)(struct net_device *,
3329 const unsigned char *))
3330 {
3331 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3332 }
3333
3334 /**
3335 * __dev_mc_unsync - Remove synchronized addresses from device
3336 * @dev: device to sync
3337 * @unsync: function to call if address should be removed
3338 *
3339 * Remove all addresses that were added to the device by dev_mc_sync().
3340 **/
3341 static inline void __dev_mc_unsync(struct net_device *dev,
3342 int (*unsync)(struct net_device *,
3343 const unsigned char *))
3344 {
3345 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3346 }
3347
3348 /* Functions used for secondary unicast and multicast support */
3349 void dev_set_rx_mode(struct net_device *dev);
3350 void __dev_set_rx_mode(struct net_device *dev);
3351 int dev_set_promiscuity(struct net_device *dev, int inc);
3352 int dev_set_allmulti(struct net_device *dev, int inc);
3353 void netdev_state_change(struct net_device *dev);
3354 void netdev_notify_peers(struct net_device *dev);
3355 void netdev_features_change(struct net_device *dev);
3356 /* Load a device via the kmod */
3357 void dev_load(struct net *net, const char *name);
3358 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3359 struct rtnl_link_stats64 *storage);
3360 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3361 const struct net_device_stats *netdev_stats);
3362
3363 extern int netdev_max_backlog;
3364 extern int netdev_tstamp_prequeue;
3365 extern int weight_p;
3366 extern int bpf_jit_enable;
3367
3368 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3369 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3370 struct list_head **iter);
3371 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3372 struct list_head **iter);
3373
3374 /* iterate through upper list, must be called under RCU read lock */
3375 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3376 for (iter = &(dev)->adj_list.upper, \
3377 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3378 updev; \
3379 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3380
3381 /* iterate through upper list, must be called under RCU read lock */
3382 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3383 for (iter = &(dev)->all_adj_list.upper, \
3384 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3385 updev; \
3386 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3387
3388 void *netdev_lower_get_next_private(struct net_device *dev,
3389 struct list_head **iter);
3390 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3391 struct list_head **iter);
3392
3393 #define netdev_for_each_lower_private(dev, priv, iter) \
3394 for (iter = (dev)->adj_list.lower.next, \
3395 priv = netdev_lower_get_next_private(dev, &(iter)); \
3396 priv; \
3397 priv = netdev_lower_get_next_private(dev, &(iter)))
3398
3399 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3400 for (iter = &(dev)->adj_list.lower, \
3401 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3402 priv; \
3403 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3404
3405 void *netdev_lower_get_next(struct net_device *dev,
3406 struct list_head **iter);
3407 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3408 for (iter = &(dev)->adj_list.lower, \
3409 ldev = netdev_lower_get_next(dev, &(iter)); \
3410 ldev; \
3411 ldev = netdev_lower_get_next(dev, &(iter)))
3412
3413 void *netdev_adjacent_get_private(struct list_head *adj_list);
3414 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3415 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3416 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3417 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3418 int netdev_master_upper_dev_link(struct net_device *dev,
3419 struct net_device *upper_dev);
3420 int netdev_master_upper_dev_link_private(struct net_device *dev,
3421 struct net_device *upper_dev,
3422 void *private);
3423 void netdev_upper_dev_unlink(struct net_device *dev,
3424 struct net_device *upper_dev);
3425 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3426 void *netdev_lower_dev_get_private(struct net_device *dev,
3427 struct net_device *lower_dev);
3428
3429 /* RSS keys are 40 or 52 bytes long */
3430 #define NETDEV_RSS_KEY_LEN 52
3431 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3432 void netdev_rss_key_fill(void *buffer, size_t len);
3433
3434 int dev_get_nest_level(struct net_device *dev,
3435 bool (*type_check)(struct net_device *dev));
3436 int skb_checksum_help(struct sk_buff *skb);
3437 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3438 netdev_features_t features, bool tx_path);
3439 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3440 netdev_features_t features);
3441
3442 static inline
3443 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3444 {
3445 return __skb_gso_segment(skb, features, true);
3446 }
3447 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3448
3449 static inline bool can_checksum_protocol(netdev_features_t features,
3450 __be16 protocol)
3451 {
3452 return ((features & NETIF_F_GEN_CSUM) ||
3453 ((features & NETIF_F_V4_CSUM) &&
3454 protocol == htons(ETH_P_IP)) ||
3455 ((features & NETIF_F_V6_CSUM) &&
3456 protocol == htons(ETH_P_IPV6)) ||
3457 ((features & NETIF_F_FCOE_CRC) &&
3458 protocol == htons(ETH_P_FCOE)));
3459 }
3460
3461 #ifdef CONFIG_BUG
3462 void netdev_rx_csum_fault(struct net_device *dev);
3463 #else
3464 static inline void netdev_rx_csum_fault(struct net_device *dev)
3465 {
3466 }
3467 #endif
3468 /* rx skb timestamps */
3469 void net_enable_timestamp(void);
3470 void net_disable_timestamp(void);
3471
3472 #ifdef CONFIG_PROC_FS
3473 int __init dev_proc_init(void);
3474 #else
3475 #define dev_proc_init() 0
3476 #endif
3477
3478 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3479 struct sk_buff *skb, struct net_device *dev,
3480 bool more)
3481 {
3482 skb->xmit_more = more ? 1 : 0;
3483 return ops->ndo_start_xmit(skb, dev);
3484 }
3485
3486 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3487 struct netdev_queue *txq, bool more)
3488 {
3489 const struct net_device_ops *ops = dev->netdev_ops;
3490 int rc;
3491
3492 rc = __netdev_start_xmit(ops, skb, dev, more);
3493 if (rc == NETDEV_TX_OK)
3494 txq_trans_update(txq);
3495
3496 return rc;
3497 }
3498
3499 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3500 const void *ns);
3501 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3502 const void *ns);
3503
3504 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3505 {
3506 return netdev_class_create_file_ns(class_attr, NULL);
3507 }
3508
3509 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3510 {
3511 netdev_class_remove_file_ns(class_attr, NULL);
3512 }
3513
3514 extern struct kobj_ns_type_operations net_ns_type_operations;
3515
3516 const char *netdev_drivername(const struct net_device *dev);
3517
3518 void linkwatch_run_queue(void);
3519
3520 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3521 netdev_features_t f2)
3522 {
3523 if (f1 & NETIF_F_GEN_CSUM)
3524 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3525 if (f2 & NETIF_F_GEN_CSUM)
3526 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3527 f1 &= f2;
3528 if (f1 & NETIF_F_GEN_CSUM)
3529 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3530
3531 return f1;
3532 }
3533
3534 static inline netdev_features_t netdev_get_wanted_features(
3535 struct net_device *dev)
3536 {
3537 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3538 }
3539 netdev_features_t netdev_increment_features(netdev_features_t all,
3540 netdev_features_t one, netdev_features_t mask);
3541
3542 /* Allow TSO being used on stacked device :
3543 * Performing the GSO segmentation before last device
3544 * is a performance improvement.
3545 */
3546 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3547 netdev_features_t mask)
3548 {
3549 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3550 }
3551
3552 int __netdev_update_features(struct net_device *dev);
3553 void netdev_update_features(struct net_device *dev);
3554 void netdev_change_features(struct net_device *dev);
3555
3556 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3557 struct net_device *dev);
3558
3559 netdev_features_t netif_skb_features(struct sk_buff *skb);
3560
3561 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3562 {
3563 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3564
3565 /* check flags correspondence */
3566 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3567 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3568 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3569 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3570 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3571 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3572 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3573 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3574 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3575 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3576 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3577 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3578 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3579
3580 return (features & feature) == feature;
3581 }
3582
3583 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3584 {
3585 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3586 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3587 }
3588
3589 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3590 netdev_features_t features)
3591 {
3592 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3593 (dev->netdev_ops->ndo_gso_check &&
3594 !dev->netdev_ops->ndo_gso_check(skb, dev)) ||
3595 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3596 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3597 }
3598
3599 static inline void netif_set_gso_max_size(struct net_device *dev,
3600 unsigned int size)
3601 {
3602 dev->gso_max_size = size;
3603 }
3604
3605 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3606 int pulled_hlen, u16 mac_offset,
3607 int mac_len)
3608 {
3609 skb->protocol = protocol;
3610 skb->encapsulation = 1;
3611 skb_push(skb, pulled_hlen);
3612 skb_reset_transport_header(skb);
3613 skb->mac_header = mac_offset;
3614 skb->network_header = skb->mac_header + mac_len;
3615 skb->mac_len = mac_len;
3616 }
3617
3618 static inline bool netif_is_macvlan(struct net_device *dev)
3619 {
3620 return dev->priv_flags & IFF_MACVLAN;
3621 }
3622
3623 static inline bool netif_is_bond_master(struct net_device *dev)
3624 {
3625 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3626 }
3627
3628 static inline bool netif_is_bond_slave(struct net_device *dev)
3629 {
3630 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3631 }
3632
3633 static inline bool netif_supports_nofcs(struct net_device *dev)
3634 {
3635 return dev->priv_flags & IFF_SUPP_NOFCS;
3636 }
3637
3638 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3639 static inline void netif_keep_dst(struct net_device *dev)
3640 {
3641 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3642 }
3643
3644 extern struct pernet_operations __net_initdata loopback_net_ops;
3645
3646 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3647
3648 /* netdev_printk helpers, similar to dev_printk */
3649
3650 static inline const char *netdev_name(const struct net_device *dev)
3651 {
3652 if (!dev->name[0] || strchr(dev->name, '%'))
3653 return "(unnamed net_device)";
3654 return dev->name;
3655 }
3656
3657 static inline const char *netdev_reg_state(const struct net_device *dev)
3658 {
3659 switch (dev->reg_state) {
3660 case NETREG_UNINITIALIZED: return " (uninitialized)";
3661 case NETREG_REGISTERED: return "";
3662 case NETREG_UNREGISTERING: return " (unregistering)";
3663 case NETREG_UNREGISTERED: return " (unregistered)";
3664 case NETREG_RELEASED: return " (released)";
3665 case NETREG_DUMMY: return " (dummy)";
3666 }
3667
3668 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3669 return " (unknown)";
3670 }
3671
3672 __printf(3, 4)
3673 void netdev_printk(const char *level, const struct net_device *dev,
3674 const char *format, ...);
3675 __printf(2, 3)
3676 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3677 __printf(2, 3)
3678 void netdev_alert(const struct net_device *dev, const char *format, ...);
3679 __printf(2, 3)
3680 void netdev_crit(const struct net_device *dev, const char *format, ...);
3681 __printf(2, 3)
3682 void netdev_err(const struct net_device *dev, const char *format, ...);
3683 __printf(2, 3)
3684 void netdev_warn(const struct net_device *dev, const char *format, ...);
3685 __printf(2, 3)
3686 void netdev_notice(const struct net_device *dev, const char *format, ...);
3687 __printf(2, 3)
3688 void netdev_info(const struct net_device *dev, const char *format, ...);
3689
3690 #define MODULE_ALIAS_NETDEV(device) \
3691 MODULE_ALIAS("netdev-" device)
3692
3693 #if defined(CONFIG_DYNAMIC_DEBUG)
3694 #define netdev_dbg(__dev, format, args...) \
3695 do { \
3696 dynamic_netdev_dbg(__dev, format, ##args); \
3697 } while (0)
3698 #elif defined(DEBUG)
3699 #define netdev_dbg(__dev, format, args...) \
3700 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3701 #else
3702 #define netdev_dbg(__dev, format, args...) \
3703 ({ \
3704 if (0) \
3705 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3706 })
3707 #endif
3708
3709 #if defined(VERBOSE_DEBUG)
3710 #define netdev_vdbg netdev_dbg
3711 #else
3712
3713 #define netdev_vdbg(dev, format, args...) \
3714 ({ \
3715 if (0) \
3716 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3717 0; \
3718 })
3719 #endif
3720
3721 /*
3722 * netdev_WARN() acts like dev_printk(), but with the key difference
3723 * of using a WARN/WARN_ON to get the message out, including the
3724 * file/line information and a backtrace.
3725 */
3726 #define netdev_WARN(dev, format, args...) \
3727 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3728 netdev_reg_state(dev), ##args)
3729
3730 /* netif printk helpers, similar to netdev_printk */
3731
3732 #define netif_printk(priv, type, level, dev, fmt, args...) \
3733 do { \
3734 if (netif_msg_##type(priv)) \
3735 netdev_printk(level, (dev), fmt, ##args); \
3736 } while (0)
3737
3738 #define netif_level(level, priv, type, dev, fmt, args...) \
3739 do { \
3740 if (netif_msg_##type(priv)) \
3741 netdev_##level(dev, fmt, ##args); \
3742 } while (0)
3743
3744 #define netif_emerg(priv, type, dev, fmt, args...) \
3745 netif_level(emerg, priv, type, dev, fmt, ##args)
3746 #define netif_alert(priv, type, dev, fmt, args...) \
3747 netif_level(alert, priv, type, dev, fmt, ##args)
3748 #define netif_crit(priv, type, dev, fmt, args...) \
3749 netif_level(crit, priv, type, dev, fmt, ##args)
3750 #define netif_err(priv, type, dev, fmt, args...) \
3751 netif_level(err, priv, type, dev, fmt, ##args)
3752 #define netif_warn(priv, type, dev, fmt, args...) \
3753 netif_level(warn, priv, type, dev, fmt, ##args)
3754 #define netif_notice(priv, type, dev, fmt, args...) \
3755 netif_level(notice, priv, type, dev, fmt, ##args)
3756 #define netif_info(priv, type, dev, fmt, args...) \
3757 netif_level(info, priv, type, dev, fmt, ##args)
3758
3759 #if defined(CONFIG_DYNAMIC_DEBUG)
3760 #define netif_dbg(priv, type, netdev, format, args...) \
3761 do { \
3762 if (netif_msg_##type(priv)) \
3763 dynamic_netdev_dbg(netdev, format, ##args); \
3764 } while (0)
3765 #elif defined(DEBUG)
3766 #define netif_dbg(priv, type, dev, format, args...) \
3767 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3768 #else
3769 #define netif_dbg(priv, type, dev, format, args...) \
3770 ({ \
3771 if (0) \
3772 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3773 0; \
3774 })
3775 #endif
3776
3777 #if defined(VERBOSE_DEBUG)
3778 #define netif_vdbg netif_dbg
3779 #else
3780 #define netif_vdbg(priv, type, dev, format, args...) \
3781 ({ \
3782 if (0) \
3783 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3784 0; \
3785 })
3786 #endif
3787
3788 /*
3789 * The list of packet types we will receive (as opposed to discard)
3790 * and the routines to invoke.
3791 *
3792 * Why 16. Because with 16 the only overlap we get on a hash of the
3793 * low nibble of the protocol value is RARP/SNAP/X.25.
3794 *
3795 * NOTE: That is no longer true with the addition of VLAN tags. Not
3796 * sure which should go first, but I bet it won't make much
3797 * difference if we are running VLANs. The good news is that
3798 * this protocol won't be in the list unless compiled in, so
3799 * the average user (w/out VLANs) will not be adversely affected.
3800 * --BLG
3801 *
3802 * 0800 IP
3803 * 8100 802.1Q VLAN
3804 * 0001 802.3
3805 * 0002 AX.25
3806 * 0004 802.2
3807 * 8035 RARP
3808 * 0005 SNAP
3809 * 0805 X.25
3810 * 0806 ARP
3811 * 8137 IPX
3812 * 0009 Localtalk
3813 * 86DD IPv6
3814 */
3815 #define PTYPE_HASH_SIZE (16)
3816 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3817
3818 #endif /* _LINUX_NETDEVICE_H */