net: gro: fix a potential crash in skb_gro_reset_offset
[deliverable/linux.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/if.h>
29 #include <linux/if_ether.h>
30 #include <linux/if_packet.h>
31 #include <linux/if_link.h>
32
33 #ifdef __KERNEL__
34 #include <linux/pm_qos.h>
35 #include <linux/timer.h>
36 #include <linux/bug.h>
37 #include <linux/delay.h>
38 #include <linux/atomic.h>
39 #include <asm/cache.h>
40 #include <asm/byteorder.h>
41
42 #include <linux/percpu.h>
43 #include <linux/rculist.h>
44 #include <linux/dmaengine.h>
45 #include <linux/workqueue.h>
46 #include <linux/dynamic_queue_limits.h>
47
48 #include <linux/ethtool.h>
49 #include <net/net_namespace.h>
50 #include <net/dsa.h>
51 #ifdef CONFIG_DCB
52 #include <net/dcbnl.h>
53 #endif
54 #include <net/netprio_cgroup.h>
55
56 #include <linux/netdev_features.h>
57 #include <linux/neighbour.h>
58
59 struct netpoll_info;
60 struct device;
61 struct phy_device;
62 /* 802.11 specific */
63 struct wireless_dev;
64 /* source back-compat hooks */
65 #define SET_ETHTOOL_OPS(netdev,ops) \
66 ( (netdev)->ethtool_ops = (ops) )
67
68 /* hardware address assignment types */
69 #define NET_ADDR_PERM 0 /* address is permanent (default) */
70 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
71 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
72
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
76
77 /*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
92 */
93
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
109
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
115 };
116 typedef enum netdev_tx netdev_tx_t;
117
118 /*
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121 */
122 static inline bool dev_xmit_complete(int rc)
123 {
124 /*
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 */
130 if (likely(rc < NET_XMIT_MASK))
131 return true;
132
133 return false;
134 }
135
136 #endif
137
138 #define MAX_ADDR_LEN 32 /* Largest hardware address length */
139
140 /* Initial net device group. All devices belong to group 0 by default. */
141 #define INIT_NETDEV_GROUP 0
142
143 #ifdef __KERNEL__
144 /*
145 * Compute the worst case header length according to the protocols
146 * used.
147 */
148
149 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
150 # if defined(CONFIG_MAC80211_MESH)
151 # define LL_MAX_HEADER 128
152 # else
153 # define LL_MAX_HEADER 96
154 # endif
155 #elif IS_ENABLED(CONFIG_TR)
156 # define LL_MAX_HEADER 48
157 #else
158 # define LL_MAX_HEADER 32
159 #endif
160
161 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
162 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
163 #define MAX_HEADER LL_MAX_HEADER
164 #else
165 #define MAX_HEADER (LL_MAX_HEADER + 48)
166 #endif
167
168 /*
169 * Old network device statistics. Fields are native words
170 * (unsigned long) so they can be read and written atomically.
171 */
172
173 struct net_device_stats {
174 unsigned long rx_packets;
175 unsigned long tx_packets;
176 unsigned long rx_bytes;
177 unsigned long tx_bytes;
178 unsigned long rx_errors;
179 unsigned long tx_errors;
180 unsigned long rx_dropped;
181 unsigned long tx_dropped;
182 unsigned long multicast;
183 unsigned long collisions;
184 unsigned long rx_length_errors;
185 unsigned long rx_over_errors;
186 unsigned long rx_crc_errors;
187 unsigned long rx_frame_errors;
188 unsigned long rx_fifo_errors;
189 unsigned long rx_missed_errors;
190 unsigned long tx_aborted_errors;
191 unsigned long tx_carrier_errors;
192 unsigned long tx_fifo_errors;
193 unsigned long tx_heartbeat_errors;
194 unsigned long tx_window_errors;
195 unsigned long rx_compressed;
196 unsigned long tx_compressed;
197 };
198
199 #endif /* __KERNEL__ */
200
201
202 /* Media selection options. */
203 enum {
204 IF_PORT_UNKNOWN = 0,
205 IF_PORT_10BASE2,
206 IF_PORT_10BASET,
207 IF_PORT_AUI,
208 IF_PORT_100BASET,
209 IF_PORT_100BASETX,
210 IF_PORT_100BASEFX
211 };
212
213 #ifdef __KERNEL__
214
215 #include <linux/cache.h>
216 #include <linux/skbuff.h>
217
218 #ifdef CONFIG_RPS
219 #include <linux/static_key.h>
220 extern struct static_key rps_needed;
221 #endif
222
223 struct neighbour;
224 struct neigh_parms;
225 struct sk_buff;
226
227 struct netdev_hw_addr {
228 struct list_head list;
229 unsigned char addr[MAX_ADDR_LEN];
230 unsigned char type;
231 #define NETDEV_HW_ADDR_T_LAN 1
232 #define NETDEV_HW_ADDR_T_SAN 2
233 #define NETDEV_HW_ADDR_T_SLAVE 3
234 #define NETDEV_HW_ADDR_T_UNICAST 4
235 #define NETDEV_HW_ADDR_T_MULTICAST 5
236 bool synced;
237 bool global_use;
238 int refcount;
239 struct rcu_head rcu_head;
240 };
241
242 struct netdev_hw_addr_list {
243 struct list_head list;
244 int count;
245 };
246
247 #define netdev_hw_addr_list_count(l) ((l)->count)
248 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
249 #define netdev_hw_addr_list_for_each(ha, l) \
250 list_for_each_entry(ha, &(l)->list, list)
251
252 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
253 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
254 #define netdev_for_each_uc_addr(ha, dev) \
255 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
256
257 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
258 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
259 #define netdev_for_each_mc_addr(ha, dev) \
260 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
261
262 struct hh_cache {
263 u16 hh_len;
264 u16 __pad;
265 seqlock_t hh_lock;
266
267 /* cached hardware header; allow for machine alignment needs. */
268 #define HH_DATA_MOD 16
269 #define HH_DATA_OFF(__len) \
270 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
271 #define HH_DATA_ALIGN(__len) \
272 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
273 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
274 };
275
276 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
277 * Alternative is:
278 * dev->hard_header_len ? (dev->hard_header_len +
279 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
280 *
281 * We could use other alignment values, but we must maintain the
282 * relationship HH alignment <= LL alignment.
283 */
284 #define LL_RESERVED_SPACE(dev) \
285 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
286 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
287 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
288
289 struct header_ops {
290 int (*create) (struct sk_buff *skb, struct net_device *dev,
291 unsigned short type, const void *daddr,
292 const void *saddr, unsigned int len);
293 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
294 int (*rebuild)(struct sk_buff *skb);
295 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
296 void (*cache_update)(struct hh_cache *hh,
297 const struct net_device *dev,
298 const unsigned char *haddr);
299 };
300
301 /* These flag bits are private to the generic network queueing
302 * layer, they may not be explicitly referenced by any other
303 * code.
304 */
305
306 enum netdev_state_t {
307 __LINK_STATE_START,
308 __LINK_STATE_PRESENT,
309 __LINK_STATE_NOCARRIER,
310 __LINK_STATE_LINKWATCH_PENDING,
311 __LINK_STATE_DORMANT,
312 };
313
314
315 /*
316 * This structure holds at boot time configured netdevice settings. They
317 * are then used in the device probing.
318 */
319 struct netdev_boot_setup {
320 char name[IFNAMSIZ];
321 struct ifmap map;
322 };
323 #define NETDEV_BOOT_SETUP_MAX 8
324
325 extern int __init netdev_boot_setup(char *str);
326
327 /*
328 * Structure for NAPI scheduling similar to tasklet but with weighting
329 */
330 struct napi_struct {
331 /* The poll_list must only be managed by the entity which
332 * changes the state of the NAPI_STATE_SCHED bit. This means
333 * whoever atomically sets that bit can add this napi_struct
334 * to the per-cpu poll_list, and whoever clears that bit
335 * can remove from the list right before clearing the bit.
336 */
337 struct list_head poll_list;
338
339 unsigned long state;
340 int weight;
341 unsigned int gro_count;
342 int (*poll)(struct napi_struct *, int);
343 #ifdef CONFIG_NETPOLL
344 spinlock_t poll_lock;
345 int poll_owner;
346 #endif
347 struct net_device *dev;
348 struct sk_buff *gro_list;
349 struct sk_buff *skb;
350 struct list_head dev_list;
351 };
352
353 enum {
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_DISABLE, /* Disable pending */
356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
357 };
358
359 enum gro_result {
360 GRO_MERGED,
361 GRO_MERGED_FREE,
362 GRO_HELD,
363 GRO_NORMAL,
364 GRO_DROP,
365 };
366 typedef enum gro_result gro_result_t;
367
368 /*
369 * enum rx_handler_result - Possible return values for rx_handlers.
370 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
371 * further.
372 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
373 * case skb->dev was changed by rx_handler.
374 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
375 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
376 *
377 * rx_handlers are functions called from inside __netif_receive_skb(), to do
378 * special processing of the skb, prior to delivery to protocol handlers.
379 *
380 * Currently, a net_device can only have a single rx_handler registered. Trying
381 * to register a second rx_handler will return -EBUSY.
382 *
383 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
384 * To unregister a rx_handler on a net_device, use
385 * netdev_rx_handler_unregister().
386 *
387 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
388 * do with the skb.
389 *
390 * If the rx_handler consumed to skb in some way, it should return
391 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
392 * the skb to be delivered in some other ways.
393 *
394 * If the rx_handler changed skb->dev, to divert the skb to another
395 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
396 * new device will be called if it exists.
397 *
398 * If the rx_handler consider the skb should be ignored, it should return
399 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
400 * are registred on exact device (ptype->dev == skb->dev).
401 *
402 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
403 * delivered, it should return RX_HANDLER_PASS.
404 *
405 * A device without a registered rx_handler will behave as if rx_handler
406 * returned RX_HANDLER_PASS.
407 */
408
409 enum rx_handler_result {
410 RX_HANDLER_CONSUMED,
411 RX_HANDLER_ANOTHER,
412 RX_HANDLER_EXACT,
413 RX_HANDLER_PASS,
414 };
415 typedef enum rx_handler_result rx_handler_result_t;
416 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
417
418 extern void __napi_schedule(struct napi_struct *n);
419
420 static inline bool napi_disable_pending(struct napi_struct *n)
421 {
422 return test_bit(NAPI_STATE_DISABLE, &n->state);
423 }
424
425 /**
426 * napi_schedule_prep - check if napi can be scheduled
427 * @n: napi context
428 *
429 * Test if NAPI routine is already running, and if not mark
430 * it as running. This is used as a condition variable
431 * insure only one NAPI poll instance runs. We also make
432 * sure there is no pending NAPI disable.
433 */
434 static inline bool napi_schedule_prep(struct napi_struct *n)
435 {
436 return !napi_disable_pending(n) &&
437 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
438 }
439
440 /**
441 * napi_schedule - schedule NAPI poll
442 * @n: napi context
443 *
444 * Schedule NAPI poll routine to be called if it is not already
445 * running.
446 */
447 static inline void napi_schedule(struct napi_struct *n)
448 {
449 if (napi_schedule_prep(n))
450 __napi_schedule(n);
451 }
452
453 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
454 static inline bool napi_reschedule(struct napi_struct *napi)
455 {
456 if (napi_schedule_prep(napi)) {
457 __napi_schedule(napi);
458 return true;
459 }
460 return false;
461 }
462
463 /**
464 * napi_complete - NAPI processing complete
465 * @n: napi context
466 *
467 * Mark NAPI processing as complete.
468 */
469 extern void __napi_complete(struct napi_struct *n);
470 extern void napi_complete(struct napi_struct *n);
471
472 /**
473 * napi_disable - prevent NAPI from scheduling
474 * @n: napi context
475 *
476 * Stop NAPI from being scheduled on this context.
477 * Waits till any outstanding processing completes.
478 */
479 static inline void napi_disable(struct napi_struct *n)
480 {
481 set_bit(NAPI_STATE_DISABLE, &n->state);
482 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
483 msleep(1);
484 clear_bit(NAPI_STATE_DISABLE, &n->state);
485 }
486
487 /**
488 * napi_enable - enable NAPI scheduling
489 * @n: napi context
490 *
491 * Resume NAPI from being scheduled on this context.
492 * Must be paired with napi_disable.
493 */
494 static inline void napi_enable(struct napi_struct *n)
495 {
496 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
497 smp_mb__before_clear_bit();
498 clear_bit(NAPI_STATE_SCHED, &n->state);
499 }
500
501 #ifdef CONFIG_SMP
502 /**
503 * napi_synchronize - wait until NAPI is not running
504 * @n: napi context
505 *
506 * Wait until NAPI is done being scheduled on this context.
507 * Waits till any outstanding processing completes but
508 * does not disable future activations.
509 */
510 static inline void napi_synchronize(const struct napi_struct *n)
511 {
512 while (test_bit(NAPI_STATE_SCHED, &n->state))
513 msleep(1);
514 }
515 #else
516 # define napi_synchronize(n) barrier()
517 #endif
518
519 enum netdev_queue_state_t {
520 __QUEUE_STATE_DRV_XOFF,
521 __QUEUE_STATE_STACK_XOFF,
522 __QUEUE_STATE_FROZEN,
523 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
524 (1 << __QUEUE_STATE_STACK_XOFF))
525 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
526 (1 << __QUEUE_STATE_FROZEN))
527 };
528 /*
529 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
530 * netif_tx_* functions below are used to manipulate this flag. The
531 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
532 * queue independently. The netif_xmit_*stopped functions below are called
533 * to check if the queue has been stopped by the driver or stack (either
534 * of the XOFF bits are set in the state). Drivers should not need to call
535 * netif_xmit*stopped functions, they should only be using netif_tx_*.
536 */
537
538 struct netdev_queue {
539 /*
540 * read mostly part
541 */
542 struct net_device *dev;
543 struct Qdisc *qdisc;
544 struct Qdisc *qdisc_sleeping;
545 #ifdef CONFIG_SYSFS
546 struct kobject kobj;
547 #endif
548 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
549 int numa_node;
550 #endif
551 /*
552 * write mostly part
553 */
554 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
555 int xmit_lock_owner;
556 /*
557 * please use this field instead of dev->trans_start
558 */
559 unsigned long trans_start;
560
561 /*
562 * Number of TX timeouts for this queue
563 * (/sys/class/net/DEV/Q/trans_timeout)
564 */
565 unsigned long trans_timeout;
566
567 unsigned long state;
568
569 #ifdef CONFIG_BQL
570 struct dql dql;
571 #endif
572 } ____cacheline_aligned_in_smp;
573
574 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
575 {
576 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
577 return q->numa_node;
578 #else
579 return NUMA_NO_NODE;
580 #endif
581 }
582
583 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
584 {
585 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
586 q->numa_node = node;
587 #endif
588 }
589
590 #ifdef CONFIG_RPS
591 /*
592 * This structure holds an RPS map which can be of variable length. The
593 * map is an array of CPUs.
594 */
595 struct rps_map {
596 unsigned int len;
597 struct rcu_head rcu;
598 u16 cpus[0];
599 };
600 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
601
602 /*
603 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
604 * tail pointer for that CPU's input queue at the time of last enqueue, and
605 * a hardware filter index.
606 */
607 struct rps_dev_flow {
608 u16 cpu;
609 u16 filter;
610 unsigned int last_qtail;
611 };
612 #define RPS_NO_FILTER 0xffff
613
614 /*
615 * The rps_dev_flow_table structure contains a table of flow mappings.
616 */
617 struct rps_dev_flow_table {
618 unsigned int mask;
619 struct rcu_head rcu;
620 struct work_struct free_work;
621 struct rps_dev_flow flows[0];
622 };
623 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
624 ((_num) * sizeof(struct rps_dev_flow)))
625
626 /*
627 * The rps_sock_flow_table contains mappings of flows to the last CPU
628 * on which they were processed by the application (set in recvmsg).
629 */
630 struct rps_sock_flow_table {
631 unsigned int mask;
632 u16 ents[0];
633 };
634 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
635 ((_num) * sizeof(u16)))
636
637 #define RPS_NO_CPU 0xffff
638
639 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
640 u32 hash)
641 {
642 if (table && hash) {
643 unsigned int cpu, index = hash & table->mask;
644
645 /* We only give a hint, preemption can change cpu under us */
646 cpu = raw_smp_processor_id();
647
648 if (table->ents[index] != cpu)
649 table->ents[index] = cpu;
650 }
651 }
652
653 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
654 u32 hash)
655 {
656 if (table && hash)
657 table->ents[hash & table->mask] = RPS_NO_CPU;
658 }
659
660 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
661
662 #ifdef CONFIG_RFS_ACCEL
663 extern bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
664 u32 flow_id, u16 filter_id);
665 #endif
666
667 /* This structure contains an instance of an RX queue. */
668 struct netdev_rx_queue {
669 struct rps_map __rcu *rps_map;
670 struct rps_dev_flow_table __rcu *rps_flow_table;
671 struct kobject kobj;
672 struct net_device *dev;
673 } ____cacheline_aligned_in_smp;
674 #endif /* CONFIG_RPS */
675
676 #ifdef CONFIG_XPS
677 /*
678 * This structure holds an XPS map which can be of variable length. The
679 * map is an array of queues.
680 */
681 struct xps_map {
682 unsigned int len;
683 unsigned int alloc_len;
684 struct rcu_head rcu;
685 u16 queues[0];
686 };
687 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
688 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
689 / sizeof(u16))
690
691 /*
692 * This structure holds all XPS maps for device. Maps are indexed by CPU.
693 */
694 struct xps_dev_maps {
695 struct rcu_head rcu;
696 struct xps_map __rcu *cpu_map[0];
697 };
698 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
699 (nr_cpu_ids * sizeof(struct xps_map *)))
700 #endif /* CONFIG_XPS */
701
702 #define TC_MAX_QUEUE 16
703 #define TC_BITMASK 15
704 /* HW offloaded queuing disciplines txq count and offset maps */
705 struct netdev_tc_txq {
706 u16 count;
707 u16 offset;
708 };
709
710 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
711 /*
712 * This structure is to hold information about the device
713 * configured to run FCoE protocol stack.
714 */
715 struct netdev_fcoe_hbainfo {
716 char manufacturer[64];
717 char serial_number[64];
718 char hardware_version[64];
719 char driver_version[64];
720 char optionrom_version[64];
721 char firmware_version[64];
722 char model[256];
723 char model_description[256];
724 };
725 #endif
726
727 /*
728 * This structure defines the management hooks for network devices.
729 * The following hooks can be defined; unless noted otherwise, they are
730 * optional and can be filled with a null pointer.
731 *
732 * int (*ndo_init)(struct net_device *dev);
733 * This function is called once when network device is registered.
734 * The network device can use this to any late stage initializaton
735 * or semantic validattion. It can fail with an error code which will
736 * be propogated back to register_netdev
737 *
738 * void (*ndo_uninit)(struct net_device *dev);
739 * This function is called when device is unregistered or when registration
740 * fails. It is not called if init fails.
741 *
742 * int (*ndo_open)(struct net_device *dev);
743 * This function is called when network device transistions to the up
744 * state.
745 *
746 * int (*ndo_stop)(struct net_device *dev);
747 * This function is called when network device transistions to the down
748 * state.
749 *
750 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
751 * struct net_device *dev);
752 * Called when a packet needs to be transmitted.
753 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
754 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
755 * Required can not be NULL.
756 *
757 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb);
758 * Called to decide which queue to when device supports multiple
759 * transmit queues.
760 *
761 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
762 * This function is called to allow device receiver to make
763 * changes to configuration when multicast or promiscious is enabled.
764 *
765 * void (*ndo_set_rx_mode)(struct net_device *dev);
766 * This function is called device changes address list filtering.
767 * If driver handles unicast address filtering, it should set
768 * IFF_UNICAST_FLT to its priv_flags.
769 *
770 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
771 * This function is called when the Media Access Control address
772 * needs to be changed. If this interface is not defined, the
773 * mac address can not be changed.
774 *
775 * int (*ndo_validate_addr)(struct net_device *dev);
776 * Test if Media Access Control address is valid for the device.
777 *
778 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
779 * Called when a user request an ioctl which can't be handled by
780 * the generic interface code. If not defined ioctl's return
781 * not supported error code.
782 *
783 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
784 * Used to set network devices bus interface parameters. This interface
785 * is retained for legacy reason, new devices should use the bus
786 * interface (PCI) for low level management.
787 *
788 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
789 * Called when a user wants to change the Maximum Transfer Unit
790 * of a device. If not defined, any request to change MTU will
791 * will return an error.
792 *
793 * void (*ndo_tx_timeout)(struct net_device *dev);
794 * Callback uses when the transmitter has not made any progress
795 * for dev->watchdog ticks.
796 *
797 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
798 * struct rtnl_link_stats64 *storage);
799 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
800 * Called when a user wants to get the network device usage
801 * statistics. Drivers must do one of the following:
802 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
803 * rtnl_link_stats64 structure passed by the caller.
804 * 2. Define @ndo_get_stats to update a net_device_stats structure
805 * (which should normally be dev->stats) and return a pointer to
806 * it. The structure may be changed asynchronously only if each
807 * field is written atomically.
808 * 3. Update dev->stats asynchronously and atomically, and define
809 * neither operation.
810 *
811 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, unsigned short vid);
812 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
813 * this function is called when a VLAN id is registered.
814 *
815 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
816 * If device support VLAN filtering (dev->features & NETIF_F_HW_VLAN_FILTER)
817 * this function is called when a VLAN id is unregistered.
818 *
819 * void (*ndo_poll_controller)(struct net_device *dev);
820 *
821 * SR-IOV management functions.
822 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
823 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
824 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
825 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
826 * int (*ndo_get_vf_config)(struct net_device *dev,
827 * int vf, struct ifla_vf_info *ivf);
828 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
829 * struct nlattr *port[]);
830 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
831 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
832 * Called to setup 'tc' number of traffic classes in the net device. This
833 * is always called from the stack with the rtnl lock held and netif tx
834 * queues stopped. This allows the netdevice to perform queue management
835 * safely.
836 *
837 * Fiber Channel over Ethernet (FCoE) offload functions.
838 * int (*ndo_fcoe_enable)(struct net_device *dev);
839 * Called when the FCoE protocol stack wants to start using LLD for FCoE
840 * so the underlying device can perform whatever needed configuration or
841 * initialization to support acceleration of FCoE traffic.
842 *
843 * int (*ndo_fcoe_disable)(struct net_device *dev);
844 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
845 * so the underlying device can perform whatever needed clean-ups to
846 * stop supporting acceleration of FCoE traffic.
847 *
848 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
849 * struct scatterlist *sgl, unsigned int sgc);
850 * Called when the FCoE Initiator wants to initialize an I/O that
851 * is a possible candidate for Direct Data Placement (DDP). The LLD can
852 * perform necessary setup and returns 1 to indicate the device is set up
853 * successfully to perform DDP on this I/O, otherwise this returns 0.
854 *
855 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
856 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
857 * indicated by the FC exchange id 'xid', so the underlying device can
858 * clean up and reuse resources for later DDP requests.
859 *
860 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
861 * struct scatterlist *sgl, unsigned int sgc);
862 * Called when the FCoE Target wants to initialize an I/O that
863 * is a possible candidate for Direct Data Placement (DDP). The LLD can
864 * perform necessary setup and returns 1 to indicate the device is set up
865 * successfully to perform DDP on this I/O, otherwise this returns 0.
866 *
867 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
868 * struct netdev_fcoe_hbainfo *hbainfo);
869 * Called when the FCoE Protocol stack wants information on the underlying
870 * device. This information is utilized by the FCoE protocol stack to
871 * register attributes with Fiber Channel management service as per the
872 * FC-GS Fabric Device Management Information(FDMI) specification.
873 *
874 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
875 * Called when the underlying device wants to override default World Wide
876 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
877 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
878 * protocol stack to use.
879 *
880 * RFS acceleration.
881 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
882 * u16 rxq_index, u32 flow_id);
883 * Set hardware filter for RFS. rxq_index is the target queue index;
884 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
885 * Return the filter ID on success, or a negative error code.
886 *
887 * Slave management functions (for bridge, bonding, etc). User should
888 * call netdev_set_master() to set dev->master properly.
889 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
890 * Called to make another netdev an underling.
891 *
892 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
893 * Called to release previously enslaved netdev.
894 *
895 * Feature/offload setting functions.
896 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
897 * netdev_features_t features);
898 * Adjusts the requested feature flags according to device-specific
899 * constraints, and returns the resulting flags. Must not modify
900 * the device state.
901 *
902 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
903 * Called to update device configuration to new features. Passed
904 * feature set might be less than what was returned by ndo_fix_features()).
905 * Must return >0 or -errno if it changed dev->features itself.
906 *
907 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
908 * struct net_device *dev,
909 * const unsigned char *addr, u16 flags)
910 * Adds an FDB entry to dev for addr.
911 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct net_device *dev,
912 * const unsigned char *addr)
913 * Deletes the FDB entry from dev coresponding to addr.
914 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
915 * struct net_device *dev, int idx)
916 * Used to add FDB entries to dump requests. Implementers should add
917 * entries to skb and update idx with the number of entries.
918 */
919 struct net_device_ops {
920 int (*ndo_init)(struct net_device *dev);
921 void (*ndo_uninit)(struct net_device *dev);
922 int (*ndo_open)(struct net_device *dev);
923 int (*ndo_stop)(struct net_device *dev);
924 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
925 struct net_device *dev);
926 u16 (*ndo_select_queue)(struct net_device *dev,
927 struct sk_buff *skb);
928 void (*ndo_change_rx_flags)(struct net_device *dev,
929 int flags);
930 void (*ndo_set_rx_mode)(struct net_device *dev);
931 int (*ndo_set_mac_address)(struct net_device *dev,
932 void *addr);
933 int (*ndo_validate_addr)(struct net_device *dev);
934 int (*ndo_do_ioctl)(struct net_device *dev,
935 struct ifreq *ifr, int cmd);
936 int (*ndo_set_config)(struct net_device *dev,
937 struct ifmap *map);
938 int (*ndo_change_mtu)(struct net_device *dev,
939 int new_mtu);
940 int (*ndo_neigh_setup)(struct net_device *dev,
941 struct neigh_parms *);
942 void (*ndo_tx_timeout) (struct net_device *dev);
943
944 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
945 struct rtnl_link_stats64 *storage);
946 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
947
948 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
949 unsigned short vid);
950 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
951 unsigned short vid);
952 #ifdef CONFIG_NET_POLL_CONTROLLER
953 void (*ndo_poll_controller)(struct net_device *dev);
954 int (*ndo_netpoll_setup)(struct net_device *dev,
955 struct netpoll_info *info,
956 gfp_t gfp);
957 void (*ndo_netpoll_cleanup)(struct net_device *dev);
958 #endif
959 int (*ndo_set_vf_mac)(struct net_device *dev,
960 int queue, u8 *mac);
961 int (*ndo_set_vf_vlan)(struct net_device *dev,
962 int queue, u16 vlan, u8 qos);
963 int (*ndo_set_vf_tx_rate)(struct net_device *dev,
964 int vf, int rate);
965 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
966 int vf, bool setting);
967 int (*ndo_get_vf_config)(struct net_device *dev,
968 int vf,
969 struct ifla_vf_info *ivf);
970 int (*ndo_set_vf_port)(struct net_device *dev,
971 int vf,
972 struct nlattr *port[]);
973 int (*ndo_get_vf_port)(struct net_device *dev,
974 int vf, struct sk_buff *skb);
975 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
976 #if IS_ENABLED(CONFIG_FCOE)
977 int (*ndo_fcoe_enable)(struct net_device *dev);
978 int (*ndo_fcoe_disable)(struct net_device *dev);
979 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
980 u16 xid,
981 struct scatterlist *sgl,
982 unsigned int sgc);
983 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
984 u16 xid);
985 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
986 u16 xid,
987 struct scatterlist *sgl,
988 unsigned int sgc);
989 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
990 struct netdev_fcoe_hbainfo *hbainfo);
991 #endif
992
993 #if IS_ENABLED(CONFIG_LIBFCOE)
994 #define NETDEV_FCOE_WWNN 0
995 #define NETDEV_FCOE_WWPN 1
996 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
997 u64 *wwn, int type);
998 #endif
999
1000 #ifdef CONFIG_RFS_ACCEL
1001 int (*ndo_rx_flow_steer)(struct net_device *dev,
1002 const struct sk_buff *skb,
1003 u16 rxq_index,
1004 u32 flow_id);
1005 #endif
1006 int (*ndo_add_slave)(struct net_device *dev,
1007 struct net_device *slave_dev);
1008 int (*ndo_del_slave)(struct net_device *dev,
1009 struct net_device *slave_dev);
1010 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1011 netdev_features_t features);
1012 int (*ndo_set_features)(struct net_device *dev,
1013 netdev_features_t features);
1014 int (*ndo_neigh_construct)(struct neighbour *n);
1015 void (*ndo_neigh_destroy)(struct neighbour *n);
1016
1017 int (*ndo_fdb_add)(struct ndmsg *ndm,
1018 struct nlattr *tb[],
1019 struct net_device *dev,
1020 const unsigned char *addr,
1021 u16 flags);
1022 int (*ndo_fdb_del)(struct ndmsg *ndm,
1023 struct net_device *dev,
1024 const unsigned char *addr);
1025 int (*ndo_fdb_dump)(struct sk_buff *skb,
1026 struct netlink_callback *cb,
1027 struct net_device *dev,
1028 int idx);
1029 };
1030
1031 /*
1032 * The DEVICE structure.
1033 * Actually, this whole structure is a big mistake. It mixes I/O
1034 * data with strictly "high-level" data, and it has to know about
1035 * almost every data structure used in the INET module.
1036 *
1037 * FIXME: cleanup struct net_device such that network protocol info
1038 * moves out.
1039 */
1040
1041 struct net_device {
1042
1043 /*
1044 * This is the first field of the "visible" part of this structure
1045 * (i.e. as seen by users in the "Space.c" file). It is the name
1046 * of the interface.
1047 */
1048 char name[IFNAMSIZ];
1049
1050 /* device name hash chain, please keep it close to name[] */
1051 struct hlist_node name_hlist;
1052
1053 /* snmp alias */
1054 char *ifalias;
1055
1056 /*
1057 * I/O specific fields
1058 * FIXME: Merge these and struct ifmap into one
1059 */
1060 unsigned long mem_end; /* shared mem end */
1061 unsigned long mem_start; /* shared mem start */
1062 unsigned long base_addr; /* device I/O address */
1063 unsigned int irq; /* device IRQ number */
1064
1065 /*
1066 * Some hardware also needs these fields, but they are not
1067 * part of the usual set specified in Space.c.
1068 */
1069
1070 unsigned long state;
1071
1072 struct list_head dev_list;
1073 struct list_head napi_list;
1074 struct list_head unreg_list;
1075
1076 /* currently active device features */
1077 netdev_features_t features;
1078 /* user-changeable features */
1079 netdev_features_t hw_features;
1080 /* user-requested features */
1081 netdev_features_t wanted_features;
1082 /* mask of features inheritable by VLAN devices */
1083 netdev_features_t vlan_features;
1084
1085 /* Interface index. Unique device identifier */
1086 int ifindex;
1087 int iflink;
1088
1089 struct net_device_stats stats;
1090 atomic_long_t rx_dropped; /* dropped packets by core network
1091 * Do not use this in drivers.
1092 */
1093
1094 #ifdef CONFIG_WIRELESS_EXT
1095 /* List of functions to handle Wireless Extensions (instead of ioctl).
1096 * See <net/iw_handler.h> for details. Jean II */
1097 const struct iw_handler_def * wireless_handlers;
1098 /* Instance data managed by the core of Wireless Extensions. */
1099 struct iw_public_data * wireless_data;
1100 #endif
1101 /* Management operations */
1102 const struct net_device_ops *netdev_ops;
1103 const struct ethtool_ops *ethtool_ops;
1104
1105 /* Hardware header description */
1106 const struct header_ops *header_ops;
1107
1108 unsigned int flags; /* interface flags (a la BSD) */
1109 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1110 * See if.h for definitions. */
1111 unsigned short gflags;
1112 unsigned short padded; /* How much padding added by alloc_netdev() */
1113
1114 unsigned char operstate; /* RFC2863 operstate */
1115 unsigned char link_mode; /* mapping policy to operstate */
1116
1117 unsigned char if_port; /* Selectable AUI, TP,..*/
1118 unsigned char dma; /* DMA channel */
1119
1120 unsigned int mtu; /* interface MTU value */
1121 unsigned short type; /* interface hardware type */
1122 unsigned short hard_header_len; /* hardware hdr length */
1123
1124 /* extra head- and tailroom the hardware may need, but not in all cases
1125 * can this be guaranteed, especially tailroom. Some cases also use
1126 * LL_MAX_HEADER instead to allocate the skb.
1127 */
1128 unsigned short needed_headroom;
1129 unsigned short needed_tailroom;
1130
1131 /* Interface address info. */
1132 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1133 unsigned char addr_assign_type; /* hw address assignment type */
1134 unsigned char addr_len; /* hardware address length */
1135 unsigned char neigh_priv_len;
1136 unsigned short dev_id; /* for shared network cards */
1137
1138 spinlock_t addr_list_lock;
1139 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1140 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1141 bool uc_promisc;
1142 unsigned int promiscuity;
1143 unsigned int allmulti;
1144
1145
1146 /* Protocol specific pointers */
1147
1148 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1149 struct vlan_info __rcu *vlan_info; /* VLAN info */
1150 #endif
1151 #if IS_ENABLED(CONFIG_NET_DSA)
1152 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1153 #endif
1154 void *atalk_ptr; /* AppleTalk link */
1155 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1156 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1157 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1158 void *ax25_ptr; /* AX.25 specific data */
1159 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1160 assign before registering */
1161
1162 /*
1163 * Cache lines mostly used on receive path (including eth_type_trans())
1164 */
1165 unsigned long last_rx; /* Time of last Rx
1166 * This should not be set in
1167 * drivers, unless really needed,
1168 * because network stack (bonding)
1169 * use it if/when necessary, to
1170 * avoid dirtying this cache line.
1171 */
1172
1173 struct net_device *master; /* Pointer to master device of a group,
1174 * which this device is member of.
1175 */
1176
1177 /* Interface address info used in eth_type_trans() */
1178 unsigned char *dev_addr; /* hw address, (before bcast
1179 because most packets are
1180 unicast) */
1181
1182 struct netdev_hw_addr_list dev_addrs; /* list of device
1183 hw addresses */
1184
1185 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1186
1187 #ifdef CONFIG_SYSFS
1188 struct kset *queues_kset;
1189 #endif
1190
1191 #ifdef CONFIG_RPS
1192 struct netdev_rx_queue *_rx;
1193
1194 /* Number of RX queues allocated at register_netdev() time */
1195 unsigned int num_rx_queues;
1196
1197 /* Number of RX queues currently active in device */
1198 unsigned int real_num_rx_queues;
1199
1200 #ifdef CONFIG_RFS_ACCEL
1201 /* CPU reverse-mapping for RX completion interrupts, indexed
1202 * by RX queue number. Assigned by driver. This must only be
1203 * set if the ndo_rx_flow_steer operation is defined. */
1204 struct cpu_rmap *rx_cpu_rmap;
1205 #endif
1206 #endif
1207
1208 rx_handler_func_t __rcu *rx_handler;
1209 void __rcu *rx_handler_data;
1210
1211 struct netdev_queue __rcu *ingress_queue;
1212
1213 /*
1214 * Cache lines mostly used on transmit path
1215 */
1216 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1217
1218 /* Number of TX queues allocated at alloc_netdev_mq() time */
1219 unsigned int num_tx_queues;
1220
1221 /* Number of TX queues currently active in device */
1222 unsigned int real_num_tx_queues;
1223
1224 /* root qdisc from userspace point of view */
1225 struct Qdisc *qdisc;
1226
1227 unsigned long tx_queue_len; /* Max frames per queue allowed */
1228 spinlock_t tx_global_lock;
1229
1230 #ifdef CONFIG_XPS
1231 struct xps_dev_maps __rcu *xps_maps;
1232 #endif
1233
1234 /* These may be needed for future network-power-down code. */
1235
1236 /*
1237 * trans_start here is expensive for high speed devices on SMP,
1238 * please use netdev_queue->trans_start instead.
1239 */
1240 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1241
1242 int watchdog_timeo; /* used by dev_watchdog() */
1243 struct timer_list watchdog_timer;
1244
1245 /* Number of references to this device */
1246 int __percpu *pcpu_refcnt;
1247
1248 /* delayed register/unregister */
1249 struct list_head todo_list;
1250 /* device index hash chain */
1251 struct hlist_node index_hlist;
1252
1253 struct list_head link_watch_list;
1254
1255 /* register/unregister state machine */
1256 enum { NETREG_UNINITIALIZED=0,
1257 NETREG_REGISTERED, /* completed register_netdevice */
1258 NETREG_UNREGISTERING, /* called unregister_netdevice */
1259 NETREG_UNREGISTERED, /* completed unregister todo */
1260 NETREG_RELEASED, /* called free_netdev */
1261 NETREG_DUMMY, /* dummy device for NAPI poll */
1262 } reg_state:8;
1263
1264 bool dismantle; /* device is going do be freed */
1265
1266 enum {
1267 RTNL_LINK_INITIALIZED,
1268 RTNL_LINK_INITIALIZING,
1269 } rtnl_link_state:16;
1270
1271 /* Called from unregister, can be used to call free_netdev */
1272 void (*destructor)(struct net_device *dev);
1273
1274 #ifdef CONFIG_NETPOLL
1275 struct netpoll_info *npinfo;
1276 #endif
1277
1278 #ifdef CONFIG_NET_NS
1279 /* Network namespace this network device is inside */
1280 struct net *nd_net;
1281 #endif
1282
1283 /* mid-layer private */
1284 union {
1285 void *ml_priv;
1286 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1287 struct pcpu_tstats __percpu *tstats; /* tunnel stats */
1288 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1289 };
1290 /* GARP */
1291 struct garp_port __rcu *garp_port;
1292
1293 /* class/net/name entry */
1294 struct device dev;
1295 /* space for optional device, statistics, and wireless sysfs groups */
1296 const struct attribute_group *sysfs_groups[4];
1297
1298 /* rtnetlink link ops */
1299 const struct rtnl_link_ops *rtnl_link_ops;
1300
1301 /* for setting kernel sock attribute on TCP connection setup */
1302 #define GSO_MAX_SIZE 65536
1303 unsigned int gso_max_size;
1304 #define GSO_MAX_SEGS 65535
1305 u16 gso_max_segs;
1306
1307 #ifdef CONFIG_DCB
1308 /* Data Center Bridging netlink ops */
1309 const struct dcbnl_rtnl_ops *dcbnl_ops;
1310 #endif
1311 u8 num_tc;
1312 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1313 u8 prio_tc_map[TC_BITMASK + 1];
1314
1315 #if IS_ENABLED(CONFIG_FCOE)
1316 /* max exchange id for FCoE LRO by ddp */
1317 unsigned int fcoe_ddp_xid;
1318 #endif
1319 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
1320 struct netprio_map __rcu *priomap;
1321 #endif
1322 /* phy device may attach itself for hardware timestamping */
1323 struct phy_device *phydev;
1324
1325 struct lock_class_key *qdisc_tx_busylock;
1326
1327 /* group the device belongs to */
1328 int group;
1329
1330 struct pm_qos_request pm_qos_req;
1331 };
1332 #define to_net_dev(d) container_of(d, struct net_device, dev)
1333
1334 #define NETDEV_ALIGN 32
1335
1336 static inline
1337 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1338 {
1339 return dev->prio_tc_map[prio & TC_BITMASK];
1340 }
1341
1342 static inline
1343 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1344 {
1345 if (tc >= dev->num_tc)
1346 return -EINVAL;
1347
1348 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1349 return 0;
1350 }
1351
1352 static inline
1353 void netdev_reset_tc(struct net_device *dev)
1354 {
1355 dev->num_tc = 0;
1356 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1357 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1358 }
1359
1360 static inline
1361 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1362 {
1363 if (tc >= dev->num_tc)
1364 return -EINVAL;
1365
1366 dev->tc_to_txq[tc].count = count;
1367 dev->tc_to_txq[tc].offset = offset;
1368 return 0;
1369 }
1370
1371 static inline
1372 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1373 {
1374 if (num_tc > TC_MAX_QUEUE)
1375 return -EINVAL;
1376
1377 dev->num_tc = num_tc;
1378 return 0;
1379 }
1380
1381 static inline
1382 int netdev_get_num_tc(struct net_device *dev)
1383 {
1384 return dev->num_tc;
1385 }
1386
1387 static inline
1388 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1389 unsigned int index)
1390 {
1391 return &dev->_tx[index];
1392 }
1393
1394 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1395 void (*f)(struct net_device *,
1396 struct netdev_queue *,
1397 void *),
1398 void *arg)
1399 {
1400 unsigned int i;
1401
1402 for (i = 0; i < dev->num_tx_queues; i++)
1403 f(dev, &dev->_tx[i], arg);
1404 }
1405
1406 extern struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1407 struct sk_buff *skb);
1408
1409 /*
1410 * Net namespace inlines
1411 */
1412 static inline
1413 struct net *dev_net(const struct net_device *dev)
1414 {
1415 return read_pnet(&dev->nd_net);
1416 }
1417
1418 static inline
1419 void dev_net_set(struct net_device *dev, struct net *net)
1420 {
1421 #ifdef CONFIG_NET_NS
1422 release_net(dev->nd_net);
1423 dev->nd_net = hold_net(net);
1424 #endif
1425 }
1426
1427 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1428 {
1429 #ifdef CONFIG_NET_DSA_TAG_DSA
1430 if (dev->dsa_ptr != NULL)
1431 return dsa_uses_dsa_tags(dev->dsa_ptr);
1432 #endif
1433
1434 return 0;
1435 }
1436
1437 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1438 {
1439 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1440 if (dev->dsa_ptr != NULL)
1441 return dsa_uses_trailer_tags(dev->dsa_ptr);
1442 #endif
1443
1444 return 0;
1445 }
1446
1447 /**
1448 * netdev_priv - access network device private data
1449 * @dev: network device
1450 *
1451 * Get network device private data
1452 */
1453 static inline void *netdev_priv(const struct net_device *dev)
1454 {
1455 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1456 }
1457
1458 /* Set the sysfs physical device reference for the network logical device
1459 * if set prior to registration will cause a symlink during initialization.
1460 */
1461 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1462
1463 /* Set the sysfs device type for the network logical device to allow
1464 * fin grained indentification of different network device types. For
1465 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1466 */
1467 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1468
1469 /**
1470 * netif_napi_add - initialize a napi context
1471 * @dev: network device
1472 * @napi: napi context
1473 * @poll: polling function
1474 * @weight: default weight
1475 *
1476 * netif_napi_add() must be used to initialize a napi context prior to calling
1477 * *any* of the other napi related functions.
1478 */
1479 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1480 int (*poll)(struct napi_struct *, int), int weight);
1481
1482 /**
1483 * netif_napi_del - remove a napi context
1484 * @napi: napi context
1485 *
1486 * netif_napi_del() removes a napi context from the network device napi list
1487 */
1488 void netif_napi_del(struct napi_struct *napi);
1489
1490 struct napi_gro_cb {
1491 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1492 void *frag0;
1493
1494 /* Length of frag0. */
1495 unsigned int frag0_len;
1496
1497 /* This indicates where we are processing relative to skb->data. */
1498 int data_offset;
1499
1500 /* This is non-zero if the packet may be of the same flow. */
1501 int same_flow;
1502
1503 /* This is non-zero if the packet cannot be merged with the new skb. */
1504 int flush;
1505
1506 /* Number of segments aggregated. */
1507 int count;
1508
1509 /* Free the skb? */
1510 int free;
1511 #define NAPI_GRO_FREE 1
1512 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1513 };
1514
1515 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1516
1517 struct packet_type {
1518 __be16 type; /* This is really htons(ether_type). */
1519 struct net_device *dev; /* NULL is wildcarded here */
1520 int (*func) (struct sk_buff *,
1521 struct net_device *,
1522 struct packet_type *,
1523 struct net_device *);
1524 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1525 netdev_features_t features);
1526 int (*gso_send_check)(struct sk_buff *skb);
1527 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1528 struct sk_buff *skb);
1529 int (*gro_complete)(struct sk_buff *skb);
1530 bool (*id_match)(struct packet_type *ptype,
1531 struct sock *sk);
1532 void *af_packet_priv;
1533 struct list_head list;
1534 };
1535
1536 #include <linux/notifier.h>
1537
1538 /* netdevice notifier chain. Please remember to update the rtnetlink
1539 * notification exclusion list in rtnetlink_event() when adding new
1540 * types.
1541 */
1542 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1543 #define NETDEV_DOWN 0x0002
1544 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1545 detected a hardware crash and restarted
1546 - we can use this eg to kick tcp sessions
1547 once done */
1548 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1549 #define NETDEV_REGISTER 0x0005
1550 #define NETDEV_UNREGISTER 0x0006
1551 #define NETDEV_CHANGEMTU 0x0007
1552 #define NETDEV_CHANGEADDR 0x0008
1553 #define NETDEV_GOING_DOWN 0x0009
1554 #define NETDEV_CHANGENAME 0x000A
1555 #define NETDEV_FEAT_CHANGE 0x000B
1556 #define NETDEV_BONDING_FAILOVER 0x000C
1557 #define NETDEV_PRE_UP 0x000D
1558 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1559 #define NETDEV_POST_TYPE_CHANGE 0x000F
1560 #define NETDEV_POST_INIT 0x0010
1561 #define NETDEV_UNREGISTER_FINAL 0x0011
1562 #define NETDEV_RELEASE 0x0012
1563 #define NETDEV_NOTIFY_PEERS 0x0013
1564 #define NETDEV_JOIN 0x0014
1565
1566 extern int register_netdevice_notifier(struct notifier_block *nb);
1567 extern int unregister_netdevice_notifier(struct notifier_block *nb);
1568 extern int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1569
1570
1571 extern rwlock_t dev_base_lock; /* Device list lock */
1572
1573
1574 #define for_each_netdev(net, d) \
1575 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1576 #define for_each_netdev_reverse(net, d) \
1577 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1578 #define for_each_netdev_rcu(net, d) \
1579 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1580 #define for_each_netdev_safe(net, d, n) \
1581 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1582 #define for_each_netdev_continue(net, d) \
1583 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1584 #define for_each_netdev_continue_rcu(net, d) \
1585 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1586 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1587
1588 static inline struct net_device *next_net_device(struct net_device *dev)
1589 {
1590 struct list_head *lh;
1591 struct net *net;
1592
1593 net = dev_net(dev);
1594 lh = dev->dev_list.next;
1595 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1596 }
1597
1598 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1599 {
1600 struct list_head *lh;
1601 struct net *net;
1602
1603 net = dev_net(dev);
1604 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1605 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1606 }
1607
1608 static inline struct net_device *first_net_device(struct net *net)
1609 {
1610 return list_empty(&net->dev_base_head) ? NULL :
1611 net_device_entry(net->dev_base_head.next);
1612 }
1613
1614 static inline struct net_device *first_net_device_rcu(struct net *net)
1615 {
1616 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1617
1618 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1619 }
1620
1621 extern int netdev_boot_setup_check(struct net_device *dev);
1622 extern unsigned long netdev_boot_base(const char *prefix, int unit);
1623 extern struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1624 const char *hwaddr);
1625 extern struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1626 extern struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1627 extern void dev_add_pack(struct packet_type *pt);
1628 extern void dev_remove_pack(struct packet_type *pt);
1629 extern void __dev_remove_pack(struct packet_type *pt);
1630
1631 extern struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1632 unsigned short mask);
1633 extern struct net_device *dev_get_by_name(struct net *net, const char *name);
1634 extern struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1635 extern struct net_device *__dev_get_by_name(struct net *net, const char *name);
1636 extern int dev_alloc_name(struct net_device *dev, const char *name);
1637 extern int dev_open(struct net_device *dev);
1638 extern int dev_close(struct net_device *dev);
1639 extern void dev_disable_lro(struct net_device *dev);
1640 extern int dev_loopback_xmit(struct sk_buff *newskb);
1641 extern int dev_queue_xmit(struct sk_buff *skb);
1642 extern int register_netdevice(struct net_device *dev);
1643 extern void unregister_netdevice_queue(struct net_device *dev,
1644 struct list_head *head);
1645 extern void unregister_netdevice_many(struct list_head *head);
1646 static inline void unregister_netdevice(struct net_device *dev)
1647 {
1648 unregister_netdevice_queue(dev, NULL);
1649 }
1650
1651 extern int netdev_refcnt_read(const struct net_device *dev);
1652 extern void free_netdev(struct net_device *dev);
1653 extern void synchronize_net(void);
1654 extern int init_dummy_netdev(struct net_device *dev);
1655 extern void netdev_resync_ops(struct net_device *dev);
1656
1657 extern struct net_device *dev_get_by_index(struct net *net, int ifindex);
1658 extern struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1659 extern struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1660 extern int dev_restart(struct net_device *dev);
1661 #ifdef CONFIG_NETPOLL_TRAP
1662 extern int netpoll_trap(void);
1663 #endif
1664 extern int skb_gro_receive(struct sk_buff **head,
1665 struct sk_buff *skb);
1666
1667 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1668 {
1669 return NAPI_GRO_CB(skb)->data_offset;
1670 }
1671
1672 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1673 {
1674 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1675 }
1676
1677 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
1678 {
1679 NAPI_GRO_CB(skb)->data_offset += len;
1680 }
1681
1682 static inline void *skb_gro_header_fast(struct sk_buff *skb,
1683 unsigned int offset)
1684 {
1685 return NAPI_GRO_CB(skb)->frag0 + offset;
1686 }
1687
1688 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
1689 {
1690 return NAPI_GRO_CB(skb)->frag0_len < hlen;
1691 }
1692
1693 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
1694 unsigned int offset)
1695 {
1696 if (!pskb_may_pull(skb, hlen))
1697 return NULL;
1698
1699 NAPI_GRO_CB(skb)->frag0 = NULL;
1700 NAPI_GRO_CB(skb)->frag0_len = 0;
1701 return skb->data + offset;
1702 }
1703
1704 static inline void *skb_gro_mac_header(struct sk_buff *skb)
1705 {
1706 return NAPI_GRO_CB(skb)->frag0 ?: skb_mac_header(skb);
1707 }
1708
1709 static inline void *skb_gro_network_header(struct sk_buff *skb)
1710 {
1711 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
1712 skb_network_offset(skb);
1713 }
1714
1715 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
1716 unsigned short type,
1717 const void *daddr, const void *saddr,
1718 unsigned int len)
1719 {
1720 if (!dev->header_ops || !dev->header_ops->create)
1721 return 0;
1722
1723 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
1724 }
1725
1726 static inline int dev_parse_header(const struct sk_buff *skb,
1727 unsigned char *haddr)
1728 {
1729 const struct net_device *dev = skb->dev;
1730
1731 if (!dev->header_ops || !dev->header_ops->parse)
1732 return 0;
1733 return dev->header_ops->parse(skb, haddr);
1734 }
1735
1736 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
1737 extern int register_gifconf(unsigned int family, gifconf_func_t * gifconf);
1738 static inline int unregister_gifconf(unsigned int family)
1739 {
1740 return register_gifconf(family, NULL);
1741 }
1742
1743 /*
1744 * Incoming packets are placed on per-cpu queues
1745 */
1746 struct softnet_data {
1747 struct Qdisc *output_queue;
1748 struct Qdisc **output_queue_tailp;
1749 struct list_head poll_list;
1750 struct sk_buff *completion_queue;
1751 struct sk_buff_head process_queue;
1752
1753 /* stats */
1754 unsigned int processed;
1755 unsigned int time_squeeze;
1756 unsigned int cpu_collision;
1757 unsigned int received_rps;
1758
1759 #ifdef CONFIG_RPS
1760 struct softnet_data *rps_ipi_list;
1761
1762 /* Elements below can be accessed between CPUs for RPS */
1763 struct call_single_data csd ____cacheline_aligned_in_smp;
1764 struct softnet_data *rps_ipi_next;
1765 unsigned int cpu;
1766 unsigned int input_queue_head;
1767 unsigned int input_queue_tail;
1768 #endif
1769 unsigned int dropped;
1770 struct sk_buff_head input_pkt_queue;
1771 struct napi_struct backlog;
1772 };
1773
1774 static inline void input_queue_head_incr(struct softnet_data *sd)
1775 {
1776 #ifdef CONFIG_RPS
1777 sd->input_queue_head++;
1778 #endif
1779 }
1780
1781 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
1782 unsigned int *qtail)
1783 {
1784 #ifdef CONFIG_RPS
1785 *qtail = ++sd->input_queue_tail;
1786 #endif
1787 }
1788
1789 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
1790
1791 extern void __netif_schedule(struct Qdisc *q);
1792
1793 static inline void netif_schedule_queue(struct netdev_queue *txq)
1794 {
1795 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
1796 __netif_schedule(txq->qdisc);
1797 }
1798
1799 static inline void netif_tx_schedule_all(struct net_device *dev)
1800 {
1801 unsigned int i;
1802
1803 for (i = 0; i < dev->num_tx_queues; i++)
1804 netif_schedule_queue(netdev_get_tx_queue(dev, i));
1805 }
1806
1807 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
1808 {
1809 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1810 }
1811
1812 /**
1813 * netif_start_queue - allow transmit
1814 * @dev: network device
1815 *
1816 * Allow upper layers to call the device hard_start_xmit routine.
1817 */
1818 static inline void netif_start_queue(struct net_device *dev)
1819 {
1820 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
1821 }
1822
1823 static inline void netif_tx_start_all_queues(struct net_device *dev)
1824 {
1825 unsigned int i;
1826
1827 for (i = 0; i < dev->num_tx_queues; i++) {
1828 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1829 netif_tx_start_queue(txq);
1830 }
1831 }
1832
1833 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
1834 {
1835 #ifdef CONFIG_NETPOLL_TRAP
1836 if (netpoll_trap()) {
1837 netif_tx_start_queue(dev_queue);
1838 return;
1839 }
1840 #endif
1841 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
1842 __netif_schedule(dev_queue->qdisc);
1843 }
1844
1845 /**
1846 * netif_wake_queue - restart transmit
1847 * @dev: network device
1848 *
1849 * Allow upper layers to call the device hard_start_xmit routine.
1850 * Used for flow control when transmit resources are available.
1851 */
1852 static inline void netif_wake_queue(struct net_device *dev)
1853 {
1854 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
1855 }
1856
1857 static inline void netif_tx_wake_all_queues(struct net_device *dev)
1858 {
1859 unsigned int i;
1860
1861 for (i = 0; i < dev->num_tx_queues; i++) {
1862 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1863 netif_tx_wake_queue(txq);
1864 }
1865 }
1866
1867 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
1868 {
1869 if (WARN_ON(!dev_queue)) {
1870 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
1871 return;
1872 }
1873 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1874 }
1875
1876 /**
1877 * netif_stop_queue - stop transmitted packets
1878 * @dev: network device
1879 *
1880 * Stop upper layers calling the device hard_start_xmit routine.
1881 * Used for flow control when transmit resources are unavailable.
1882 */
1883 static inline void netif_stop_queue(struct net_device *dev)
1884 {
1885 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
1886 }
1887
1888 static inline void netif_tx_stop_all_queues(struct net_device *dev)
1889 {
1890 unsigned int i;
1891
1892 for (i = 0; i < dev->num_tx_queues; i++) {
1893 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
1894 netif_tx_stop_queue(txq);
1895 }
1896 }
1897
1898 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
1899 {
1900 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
1901 }
1902
1903 /**
1904 * netif_queue_stopped - test if transmit queue is flowblocked
1905 * @dev: network device
1906 *
1907 * Test if transmit queue on device is currently unable to send.
1908 */
1909 static inline bool netif_queue_stopped(const struct net_device *dev)
1910 {
1911 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
1912 }
1913
1914 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
1915 {
1916 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
1917 }
1918
1919 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
1920 {
1921 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
1922 }
1923
1924 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
1925 unsigned int bytes)
1926 {
1927 #ifdef CONFIG_BQL
1928 dql_queued(&dev_queue->dql, bytes);
1929
1930 if (likely(dql_avail(&dev_queue->dql) >= 0))
1931 return;
1932
1933 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1934
1935 /*
1936 * The XOFF flag must be set before checking the dql_avail below,
1937 * because in netdev_tx_completed_queue we update the dql_completed
1938 * before checking the XOFF flag.
1939 */
1940 smp_mb();
1941
1942 /* check again in case another CPU has just made room avail */
1943 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
1944 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
1945 #endif
1946 }
1947
1948 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
1949 {
1950 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
1951 }
1952
1953 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
1954 unsigned int pkts, unsigned int bytes)
1955 {
1956 #ifdef CONFIG_BQL
1957 if (unlikely(!bytes))
1958 return;
1959
1960 dql_completed(&dev_queue->dql, bytes);
1961
1962 /*
1963 * Without the memory barrier there is a small possiblity that
1964 * netdev_tx_sent_queue will miss the update and cause the queue to
1965 * be stopped forever
1966 */
1967 smp_mb();
1968
1969 if (dql_avail(&dev_queue->dql) < 0)
1970 return;
1971
1972 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
1973 netif_schedule_queue(dev_queue);
1974 #endif
1975 }
1976
1977 static inline void netdev_completed_queue(struct net_device *dev,
1978 unsigned int pkts, unsigned int bytes)
1979 {
1980 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
1981 }
1982
1983 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
1984 {
1985 #ifdef CONFIG_BQL
1986 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
1987 dql_reset(&q->dql);
1988 #endif
1989 }
1990
1991 static inline void netdev_reset_queue(struct net_device *dev_queue)
1992 {
1993 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
1994 }
1995
1996 /**
1997 * netif_running - test if up
1998 * @dev: network device
1999 *
2000 * Test if the device has been brought up.
2001 */
2002 static inline bool netif_running(const struct net_device *dev)
2003 {
2004 return test_bit(__LINK_STATE_START, &dev->state);
2005 }
2006
2007 /*
2008 * Routines to manage the subqueues on a device. We only need start
2009 * stop, and a check if it's stopped. All other device management is
2010 * done at the overall netdevice level.
2011 * Also test the device if we're multiqueue.
2012 */
2013
2014 /**
2015 * netif_start_subqueue - allow sending packets on subqueue
2016 * @dev: network device
2017 * @queue_index: sub queue index
2018 *
2019 * Start individual transmit queue of a device with multiple transmit queues.
2020 */
2021 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2022 {
2023 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2024
2025 netif_tx_start_queue(txq);
2026 }
2027
2028 /**
2029 * netif_stop_subqueue - stop sending packets on subqueue
2030 * @dev: network device
2031 * @queue_index: sub queue index
2032 *
2033 * Stop individual transmit queue of a device with multiple transmit queues.
2034 */
2035 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2036 {
2037 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2038 #ifdef CONFIG_NETPOLL_TRAP
2039 if (netpoll_trap())
2040 return;
2041 #endif
2042 netif_tx_stop_queue(txq);
2043 }
2044
2045 /**
2046 * netif_subqueue_stopped - test status of subqueue
2047 * @dev: network device
2048 * @queue_index: sub queue index
2049 *
2050 * Check individual transmit queue of a device with multiple transmit queues.
2051 */
2052 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2053 u16 queue_index)
2054 {
2055 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2056
2057 return netif_tx_queue_stopped(txq);
2058 }
2059
2060 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2061 struct sk_buff *skb)
2062 {
2063 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2064 }
2065
2066 /**
2067 * netif_wake_subqueue - allow sending packets on subqueue
2068 * @dev: network device
2069 * @queue_index: sub queue index
2070 *
2071 * Resume individual transmit queue of a device with multiple transmit queues.
2072 */
2073 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2074 {
2075 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2076 #ifdef CONFIG_NETPOLL_TRAP
2077 if (netpoll_trap())
2078 return;
2079 #endif
2080 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2081 __netif_schedule(txq->qdisc);
2082 }
2083
2084 /*
2085 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2086 * as a distribution range limit for the returned value.
2087 */
2088 static inline u16 skb_tx_hash(const struct net_device *dev,
2089 const struct sk_buff *skb)
2090 {
2091 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2092 }
2093
2094 /**
2095 * netif_is_multiqueue - test if device has multiple transmit queues
2096 * @dev: network device
2097 *
2098 * Check if device has multiple transmit queues
2099 */
2100 static inline bool netif_is_multiqueue(const struct net_device *dev)
2101 {
2102 return dev->num_tx_queues > 1;
2103 }
2104
2105 extern int netif_set_real_num_tx_queues(struct net_device *dev,
2106 unsigned int txq);
2107
2108 #ifdef CONFIG_RPS
2109 extern int netif_set_real_num_rx_queues(struct net_device *dev,
2110 unsigned int rxq);
2111 #else
2112 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2113 unsigned int rxq)
2114 {
2115 return 0;
2116 }
2117 #endif
2118
2119 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2120 const struct net_device *from_dev)
2121 {
2122 int err;
2123
2124 err = netif_set_real_num_tx_queues(to_dev,
2125 from_dev->real_num_tx_queues);
2126 if (err)
2127 return err;
2128 #ifdef CONFIG_RPS
2129 return netif_set_real_num_rx_queues(to_dev,
2130 from_dev->real_num_rx_queues);
2131 #else
2132 return 0;
2133 #endif
2134 }
2135
2136 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2137 extern int netif_get_num_default_rss_queues(void);
2138
2139 /* Use this variant when it is known for sure that it
2140 * is executing from hardware interrupt context or with hardware interrupts
2141 * disabled.
2142 */
2143 extern void dev_kfree_skb_irq(struct sk_buff *skb);
2144
2145 /* Use this variant in places where it could be invoked
2146 * from either hardware interrupt or other context, with hardware interrupts
2147 * either disabled or enabled.
2148 */
2149 extern void dev_kfree_skb_any(struct sk_buff *skb);
2150
2151 extern int netif_rx(struct sk_buff *skb);
2152 extern int netif_rx_ni(struct sk_buff *skb);
2153 extern int netif_receive_skb(struct sk_buff *skb);
2154 extern gro_result_t dev_gro_receive(struct napi_struct *napi,
2155 struct sk_buff *skb);
2156 extern gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb);
2157 extern gro_result_t napi_gro_receive(struct napi_struct *napi,
2158 struct sk_buff *skb);
2159 extern void napi_gro_flush(struct napi_struct *napi);
2160 extern struct sk_buff * napi_get_frags(struct napi_struct *napi);
2161 extern gro_result_t napi_frags_finish(struct napi_struct *napi,
2162 struct sk_buff *skb,
2163 gro_result_t ret);
2164 extern gro_result_t napi_gro_frags(struct napi_struct *napi);
2165
2166 static inline void napi_free_frags(struct napi_struct *napi)
2167 {
2168 kfree_skb(napi->skb);
2169 napi->skb = NULL;
2170 }
2171
2172 extern int netdev_rx_handler_register(struct net_device *dev,
2173 rx_handler_func_t *rx_handler,
2174 void *rx_handler_data);
2175 extern void netdev_rx_handler_unregister(struct net_device *dev);
2176
2177 extern bool dev_valid_name(const char *name);
2178 extern int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2179 extern int dev_ethtool(struct net *net, struct ifreq *);
2180 extern unsigned int dev_get_flags(const struct net_device *);
2181 extern int __dev_change_flags(struct net_device *, unsigned int flags);
2182 extern int dev_change_flags(struct net_device *, unsigned int);
2183 extern void __dev_notify_flags(struct net_device *, unsigned int old_flags);
2184 extern int dev_change_name(struct net_device *, const char *);
2185 extern int dev_set_alias(struct net_device *, const char *, size_t);
2186 extern int dev_change_net_namespace(struct net_device *,
2187 struct net *, const char *);
2188 extern int dev_set_mtu(struct net_device *, int);
2189 extern void dev_set_group(struct net_device *, int);
2190 extern int dev_set_mac_address(struct net_device *,
2191 struct sockaddr *);
2192 extern int dev_hard_start_xmit(struct sk_buff *skb,
2193 struct net_device *dev,
2194 struct netdev_queue *txq);
2195 extern int dev_forward_skb(struct net_device *dev,
2196 struct sk_buff *skb);
2197
2198 extern int netdev_budget;
2199
2200 /* Called by rtnetlink.c:rtnl_unlock() */
2201 extern void netdev_run_todo(void);
2202
2203 /**
2204 * dev_put - release reference to device
2205 * @dev: network device
2206 *
2207 * Release reference to device to allow it to be freed.
2208 */
2209 static inline void dev_put(struct net_device *dev)
2210 {
2211 this_cpu_dec(*dev->pcpu_refcnt);
2212 }
2213
2214 /**
2215 * dev_hold - get reference to device
2216 * @dev: network device
2217 *
2218 * Hold reference to device to keep it from being freed.
2219 */
2220 static inline void dev_hold(struct net_device *dev)
2221 {
2222 this_cpu_inc(*dev->pcpu_refcnt);
2223 }
2224
2225 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2226 * and _off may be called from IRQ context, but it is caller
2227 * who is responsible for serialization of these calls.
2228 *
2229 * The name carrier is inappropriate, these functions should really be
2230 * called netif_lowerlayer_*() because they represent the state of any
2231 * kind of lower layer not just hardware media.
2232 */
2233
2234 extern void linkwatch_init_dev(struct net_device *dev);
2235 extern void linkwatch_fire_event(struct net_device *dev);
2236 extern void linkwatch_forget_dev(struct net_device *dev);
2237
2238 /**
2239 * netif_carrier_ok - test if carrier present
2240 * @dev: network device
2241 *
2242 * Check if carrier is present on device
2243 */
2244 static inline bool netif_carrier_ok(const struct net_device *dev)
2245 {
2246 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2247 }
2248
2249 extern unsigned long dev_trans_start(struct net_device *dev);
2250
2251 extern void __netdev_watchdog_up(struct net_device *dev);
2252
2253 extern void netif_carrier_on(struct net_device *dev);
2254
2255 extern void netif_carrier_off(struct net_device *dev);
2256
2257 /**
2258 * netif_dormant_on - mark device as dormant.
2259 * @dev: network device
2260 *
2261 * Mark device as dormant (as per RFC2863).
2262 *
2263 * The dormant state indicates that the relevant interface is not
2264 * actually in a condition to pass packets (i.e., it is not 'up') but is
2265 * in a "pending" state, waiting for some external event. For "on-
2266 * demand" interfaces, this new state identifies the situation where the
2267 * interface is waiting for events to place it in the up state.
2268 *
2269 */
2270 static inline void netif_dormant_on(struct net_device *dev)
2271 {
2272 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2273 linkwatch_fire_event(dev);
2274 }
2275
2276 /**
2277 * netif_dormant_off - set device as not dormant.
2278 * @dev: network device
2279 *
2280 * Device is not in dormant state.
2281 */
2282 static inline void netif_dormant_off(struct net_device *dev)
2283 {
2284 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2285 linkwatch_fire_event(dev);
2286 }
2287
2288 /**
2289 * netif_dormant - test if carrier present
2290 * @dev: network device
2291 *
2292 * Check if carrier is present on device
2293 */
2294 static inline bool netif_dormant(const struct net_device *dev)
2295 {
2296 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2297 }
2298
2299
2300 /**
2301 * netif_oper_up - test if device is operational
2302 * @dev: network device
2303 *
2304 * Check if carrier is operational
2305 */
2306 static inline bool netif_oper_up(const struct net_device *dev)
2307 {
2308 return (dev->operstate == IF_OPER_UP ||
2309 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2310 }
2311
2312 /**
2313 * netif_device_present - is device available or removed
2314 * @dev: network device
2315 *
2316 * Check if device has not been removed from system.
2317 */
2318 static inline bool netif_device_present(struct net_device *dev)
2319 {
2320 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2321 }
2322
2323 extern void netif_device_detach(struct net_device *dev);
2324
2325 extern void netif_device_attach(struct net_device *dev);
2326
2327 /*
2328 * Network interface message level settings
2329 */
2330
2331 enum {
2332 NETIF_MSG_DRV = 0x0001,
2333 NETIF_MSG_PROBE = 0x0002,
2334 NETIF_MSG_LINK = 0x0004,
2335 NETIF_MSG_TIMER = 0x0008,
2336 NETIF_MSG_IFDOWN = 0x0010,
2337 NETIF_MSG_IFUP = 0x0020,
2338 NETIF_MSG_RX_ERR = 0x0040,
2339 NETIF_MSG_TX_ERR = 0x0080,
2340 NETIF_MSG_TX_QUEUED = 0x0100,
2341 NETIF_MSG_INTR = 0x0200,
2342 NETIF_MSG_TX_DONE = 0x0400,
2343 NETIF_MSG_RX_STATUS = 0x0800,
2344 NETIF_MSG_PKTDATA = 0x1000,
2345 NETIF_MSG_HW = 0x2000,
2346 NETIF_MSG_WOL = 0x4000,
2347 };
2348
2349 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2350 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2351 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2352 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2353 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2354 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2355 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2356 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2357 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2358 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2359 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2360 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2361 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2362 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2363 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2364
2365 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2366 {
2367 /* use default */
2368 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2369 return default_msg_enable_bits;
2370 if (debug_value == 0) /* no output */
2371 return 0;
2372 /* set low N bits */
2373 return (1 << debug_value) - 1;
2374 }
2375
2376 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2377 {
2378 spin_lock(&txq->_xmit_lock);
2379 txq->xmit_lock_owner = cpu;
2380 }
2381
2382 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2383 {
2384 spin_lock_bh(&txq->_xmit_lock);
2385 txq->xmit_lock_owner = smp_processor_id();
2386 }
2387
2388 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2389 {
2390 bool ok = spin_trylock(&txq->_xmit_lock);
2391 if (likely(ok))
2392 txq->xmit_lock_owner = smp_processor_id();
2393 return ok;
2394 }
2395
2396 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2397 {
2398 txq->xmit_lock_owner = -1;
2399 spin_unlock(&txq->_xmit_lock);
2400 }
2401
2402 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2403 {
2404 txq->xmit_lock_owner = -1;
2405 spin_unlock_bh(&txq->_xmit_lock);
2406 }
2407
2408 static inline void txq_trans_update(struct netdev_queue *txq)
2409 {
2410 if (txq->xmit_lock_owner != -1)
2411 txq->trans_start = jiffies;
2412 }
2413
2414 /**
2415 * netif_tx_lock - grab network device transmit lock
2416 * @dev: network device
2417 *
2418 * Get network device transmit lock
2419 */
2420 static inline void netif_tx_lock(struct net_device *dev)
2421 {
2422 unsigned int i;
2423 int cpu;
2424
2425 spin_lock(&dev->tx_global_lock);
2426 cpu = smp_processor_id();
2427 for (i = 0; i < dev->num_tx_queues; i++) {
2428 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2429
2430 /* We are the only thread of execution doing a
2431 * freeze, but we have to grab the _xmit_lock in
2432 * order to synchronize with threads which are in
2433 * the ->hard_start_xmit() handler and already
2434 * checked the frozen bit.
2435 */
2436 __netif_tx_lock(txq, cpu);
2437 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2438 __netif_tx_unlock(txq);
2439 }
2440 }
2441
2442 static inline void netif_tx_lock_bh(struct net_device *dev)
2443 {
2444 local_bh_disable();
2445 netif_tx_lock(dev);
2446 }
2447
2448 static inline void netif_tx_unlock(struct net_device *dev)
2449 {
2450 unsigned int i;
2451
2452 for (i = 0; i < dev->num_tx_queues; i++) {
2453 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2454
2455 /* No need to grab the _xmit_lock here. If the
2456 * queue is not stopped for another reason, we
2457 * force a schedule.
2458 */
2459 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2460 netif_schedule_queue(txq);
2461 }
2462 spin_unlock(&dev->tx_global_lock);
2463 }
2464
2465 static inline void netif_tx_unlock_bh(struct net_device *dev)
2466 {
2467 netif_tx_unlock(dev);
2468 local_bh_enable();
2469 }
2470
2471 #define HARD_TX_LOCK(dev, txq, cpu) { \
2472 if ((dev->features & NETIF_F_LLTX) == 0) { \
2473 __netif_tx_lock(txq, cpu); \
2474 } \
2475 }
2476
2477 #define HARD_TX_UNLOCK(dev, txq) { \
2478 if ((dev->features & NETIF_F_LLTX) == 0) { \
2479 __netif_tx_unlock(txq); \
2480 } \
2481 }
2482
2483 static inline void netif_tx_disable(struct net_device *dev)
2484 {
2485 unsigned int i;
2486 int cpu;
2487
2488 local_bh_disable();
2489 cpu = smp_processor_id();
2490 for (i = 0; i < dev->num_tx_queues; i++) {
2491 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2492
2493 __netif_tx_lock(txq, cpu);
2494 netif_tx_stop_queue(txq);
2495 __netif_tx_unlock(txq);
2496 }
2497 local_bh_enable();
2498 }
2499
2500 static inline void netif_addr_lock(struct net_device *dev)
2501 {
2502 spin_lock(&dev->addr_list_lock);
2503 }
2504
2505 static inline void netif_addr_lock_nested(struct net_device *dev)
2506 {
2507 spin_lock_nested(&dev->addr_list_lock, SINGLE_DEPTH_NESTING);
2508 }
2509
2510 static inline void netif_addr_lock_bh(struct net_device *dev)
2511 {
2512 spin_lock_bh(&dev->addr_list_lock);
2513 }
2514
2515 static inline void netif_addr_unlock(struct net_device *dev)
2516 {
2517 spin_unlock(&dev->addr_list_lock);
2518 }
2519
2520 static inline void netif_addr_unlock_bh(struct net_device *dev)
2521 {
2522 spin_unlock_bh(&dev->addr_list_lock);
2523 }
2524
2525 /*
2526 * dev_addrs walker. Should be used only for read access. Call with
2527 * rcu_read_lock held.
2528 */
2529 #define for_each_dev_addr(dev, ha) \
2530 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2531
2532 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2533
2534 extern void ether_setup(struct net_device *dev);
2535
2536 /* Support for loadable net-drivers */
2537 extern struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2538 void (*setup)(struct net_device *),
2539 unsigned int txqs, unsigned int rxqs);
2540 #define alloc_netdev(sizeof_priv, name, setup) \
2541 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2542
2543 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2544 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2545
2546 extern int register_netdev(struct net_device *dev);
2547 extern void unregister_netdev(struct net_device *dev);
2548
2549 /* General hardware address lists handling functions */
2550 extern int __hw_addr_add_multiple(struct netdev_hw_addr_list *to_list,
2551 struct netdev_hw_addr_list *from_list,
2552 int addr_len, unsigned char addr_type);
2553 extern void __hw_addr_del_multiple(struct netdev_hw_addr_list *to_list,
2554 struct netdev_hw_addr_list *from_list,
2555 int addr_len, unsigned char addr_type);
2556 extern int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
2557 struct netdev_hw_addr_list *from_list,
2558 int addr_len);
2559 extern void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
2560 struct netdev_hw_addr_list *from_list,
2561 int addr_len);
2562 extern void __hw_addr_flush(struct netdev_hw_addr_list *list);
2563 extern void __hw_addr_init(struct netdev_hw_addr_list *list);
2564
2565 /* Functions used for device addresses handling */
2566 extern int dev_addr_add(struct net_device *dev, const unsigned char *addr,
2567 unsigned char addr_type);
2568 extern int dev_addr_del(struct net_device *dev, const unsigned char *addr,
2569 unsigned char addr_type);
2570 extern int dev_addr_add_multiple(struct net_device *to_dev,
2571 struct net_device *from_dev,
2572 unsigned char addr_type);
2573 extern int dev_addr_del_multiple(struct net_device *to_dev,
2574 struct net_device *from_dev,
2575 unsigned char addr_type);
2576 extern void dev_addr_flush(struct net_device *dev);
2577 extern int dev_addr_init(struct net_device *dev);
2578
2579 /* Functions used for unicast addresses handling */
2580 extern int dev_uc_add(struct net_device *dev, const unsigned char *addr);
2581 extern int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
2582 extern int dev_uc_del(struct net_device *dev, const unsigned char *addr);
2583 extern int dev_uc_sync(struct net_device *to, struct net_device *from);
2584 extern void dev_uc_unsync(struct net_device *to, struct net_device *from);
2585 extern void dev_uc_flush(struct net_device *dev);
2586 extern void dev_uc_init(struct net_device *dev);
2587
2588 /* Functions used for multicast addresses handling */
2589 extern int dev_mc_add(struct net_device *dev, const unsigned char *addr);
2590 extern int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
2591 extern int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
2592 extern int dev_mc_del(struct net_device *dev, const unsigned char *addr);
2593 extern int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
2594 extern int dev_mc_sync(struct net_device *to, struct net_device *from);
2595 extern void dev_mc_unsync(struct net_device *to, struct net_device *from);
2596 extern void dev_mc_flush(struct net_device *dev);
2597 extern void dev_mc_init(struct net_device *dev);
2598
2599 /* Functions used for secondary unicast and multicast support */
2600 extern void dev_set_rx_mode(struct net_device *dev);
2601 extern void __dev_set_rx_mode(struct net_device *dev);
2602 extern int dev_set_promiscuity(struct net_device *dev, int inc);
2603 extern int dev_set_allmulti(struct net_device *dev, int inc);
2604 extern void netdev_state_change(struct net_device *dev);
2605 extern void netdev_notify_peers(struct net_device *dev);
2606 extern void netdev_features_change(struct net_device *dev);
2607 /* Load a device via the kmod */
2608 extern void dev_load(struct net *net, const char *name);
2609 extern void dev_mcast_init(void);
2610 extern struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
2611 struct rtnl_link_stats64 *storage);
2612 extern void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
2613 const struct net_device_stats *netdev_stats);
2614
2615 extern int netdev_max_backlog;
2616 extern int netdev_tstamp_prequeue;
2617 extern int weight_p;
2618 extern int bpf_jit_enable;
2619 extern int netdev_set_master(struct net_device *dev, struct net_device *master);
2620 extern int netdev_set_bond_master(struct net_device *dev,
2621 struct net_device *master);
2622 extern int skb_checksum_help(struct sk_buff *skb);
2623 extern struct sk_buff *skb_gso_segment(struct sk_buff *skb,
2624 netdev_features_t features);
2625 #ifdef CONFIG_BUG
2626 extern void netdev_rx_csum_fault(struct net_device *dev);
2627 #else
2628 static inline void netdev_rx_csum_fault(struct net_device *dev)
2629 {
2630 }
2631 #endif
2632 /* rx skb timestamps */
2633 extern void net_enable_timestamp(void);
2634 extern void net_disable_timestamp(void);
2635
2636 #ifdef CONFIG_PROC_FS
2637 extern void *dev_seq_start(struct seq_file *seq, loff_t *pos);
2638 extern void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos);
2639 extern void dev_seq_stop(struct seq_file *seq, void *v);
2640 #endif
2641
2642 extern int netdev_class_create_file(struct class_attribute *class_attr);
2643 extern void netdev_class_remove_file(struct class_attribute *class_attr);
2644
2645 extern struct kobj_ns_type_operations net_ns_type_operations;
2646
2647 extern const char *netdev_drivername(const struct net_device *dev);
2648
2649 extern void linkwatch_run_queue(void);
2650
2651 static inline netdev_features_t netdev_get_wanted_features(
2652 struct net_device *dev)
2653 {
2654 return (dev->features & ~dev->hw_features) | dev->wanted_features;
2655 }
2656 netdev_features_t netdev_increment_features(netdev_features_t all,
2657 netdev_features_t one, netdev_features_t mask);
2658 int __netdev_update_features(struct net_device *dev);
2659 void netdev_update_features(struct net_device *dev);
2660 void netdev_change_features(struct net_device *dev);
2661
2662 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
2663 struct net_device *dev);
2664
2665 netdev_features_t netif_skb_features(struct sk_buff *skb);
2666
2667 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
2668 {
2669 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
2670
2671 /* check flags correspondence */
2672 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
2673 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
2674 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
2675 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
2676 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
2677 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
2678
2679 return (features & feature) == feature;
2680 }
2681
2682 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
2683 {
2684 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
2685 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
2686 }
2687
2688 static inline bool netif_needs_gso(struct sk_buff *skb,
2689 netdev_features_t features)
2690 {
2691 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
2692 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
2693 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
2694 }
2695
2696 static inline void netif_set_gso_max_size(struct net_device *dev,
2697 unsigned int size)
2698 {
2699 dev->gso_max_size = size;
2700 }
2701
2702 static inline bool netif_is_bond_slave(struct net_device *dev)
2703 {
2704 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
2705 }
2706
2707 static inline bool netif_supports_nofcs(struct net_device *dev)
2708 {
2709 return dev->priv_flags & IFF_SUPP_NOFCS;
2710 }
2711
2712 extern struct pernet_operations __net_initdata loopback_net_ops;
2713
2714 /* Logging, debugging and troubleshooting/diagnostic helpers. */
2715
2716 /* netdev_printk helpers, similar to dev_printk */
2717
2718 static inline const char *netdev_name(const struct net_device *dev)
2719 {
2720 if (dev->reg_state != NETREG_REGISTERED)
2721 return "(unregistered net_device)";
2722 return dev->name;
2723 }
2724
2725 extern __printf(3, 4)
2726 int netdev_printk(const char *level, const struct net_device *dev,
2727 const char *format, ...);
2728 extern __printf(2, 3)
2729 int netdev_emerg(const struct net_device *dev, const char *format, ...);
2730 extern __printf(2, 3)
2731 int netdev_alert(const struct net_device *dev, const char *format, ...);
2732 extern __printf(2, 3)
2733 int netdev_crit(const struct net_device *dev, const char *format, ...);
2734 extern __printf(2, 3)
2735 int netdev_err(const struct net_device *dev, const char *format, ...);
2736 extern __printf(2, 3)
2737 int netdev_warn(const struct net_device *dev, const char *format, ...);
2738 extern __printf(2, 3)
2739 int netdev_notice(const struct net_device *dev, const char *format, ...);
2740 extern __printf(2, 3)
2741 int netdev_info(const struct net_device *dev, const char *format, ...);
2742
2743 #define MODULE_ALIAS_NETDEV(device) \
2744 MODULE_ALIAS("netdev-" device)
2745
2746 #if defined(CONFIG_DYNAMIC_DEBUG)
2747 #define netdev_dbg(__dev, format, args...) \
2748 do { \
2749 dynamic_netdev_dbg(__dev, format, ##args); \
2750 } while (0)
2751 #elif defined(DEBUG)
2752 #define netdev_dbg(__dev, format, args...) \
2753 netdev_printk(KERN_DEBUG, __dev, format, ##args)
2754 #else
2755 #define netdev_dbg(__dev, format, args...) \
2756 ({ \
2757 if (0) \
2758 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
2759 0; \
2760 })
2761 #endif
2762
2763 #if defined(VERBOSE_DEBUG)
2764 #define netdev_vdbg netdev_dbg
2765 #else
2766
2767 #define netdev_vdbg(dev, format, args...) \
2768 ({ \
2769 if (0) \
2770 netdev_printk(KERN_DEBUG, dev, format, ##args); \
2771 0; \
2772 })
2773 #endif
2774
2775 /*
2776 * netdev_WARN() acts like dev_printk(), but with the key difference
2777 * of using a WARN/WARN_ON to get the message out, including the
2778 * file/line information and a backtrace.
2779 */
2780 #define netdev_WARN(dev, format, args...) \
2781 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args);
2782
2783 /* netif printk helpers, similar to netdev_printk */
2784
2785 #define netif_printk(priv, type, level, dev, fmt, args...) \
2786 do { \
2787 if (netif_msg_##type(priv)) \
2788 netdev_printk(level, (dev), fmt, ##args); \
2789 } while (0)
2790
2791 #define netif_level(level, priv, type, dev, fmt, args...) \
2792 do { \
2793 if (netif_msg_##type(priv)) \
2794 netdev_##level(dev, fmt, ##args); \
2795 } while (0)
2796
2797 #define netif_emerg(priv, type, dev, fmt, args...) \
2798 netif_level(emerg, priv, type, dev, fmt, ##args)
2799 #define netif_alert(priv, type, dev, fmt, args...) \
2800 netif_level(alert, priv, type, dev, fmt, ##args)
2801 #define netif_crit(priv, type, dev, fmt, args...) \
2802 netif_level(crit, priv, type, dev, fmt, ##args)
2803 #define netif_err(priv, type, dev, fmt, args...) \
2804 netif_level(err, priv, type, dev, fmt, ##args)
2805 #define netif_warn(priv, type, dev, fmt, args...) \
2806 netif_level(warn, priv, type, dev, fmt, ##args)
2807 #define netif_notice(priv, type, dev, fmt, args...) \
2808 netif_level(notice, priv, type, dev, fmt, ##args)
2809 #define netif_info(priv, type, dev, fmt, args...) \
2810 netif_level(info, priv, type, dev, fmt, ##args)
2811
2812 #if defined(CONFIG_DYNAMIC_DEBUG)
2813 #define netif_dbg(priv, type, netdev, format, args...) \
2814 do { \
2815 if (netif_msg_##type(priv)) \
2816 dynamic_netdev_dbg(netdev, format, ##args); \
2817 } while (0)
2818 #elif defined(DEBUG)
2819 #define netif_dbg(priv, type, dev, format, args...) \
2820 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
2821 #else
2822 #define netif_dbg(priv, type, dev, format, args...) \
2823 ({ \
2824 if (0) \
2825 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2826 0; \
2827 })
2828 #endif
2829
2830 #if defined(VERBOSE_DEBUG)
2831 #define netif_vdbg netif_dbg
2832 #else
2833 #define netif_vdbg(priv, type, dev, format, args...) \
2834 ({ \
2835 if (0) \
2836 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
2837 0; \
2838 })
2839 #endif
2840
2841 #endif /* __KERNEL__ */
2842
2843 #endif /* _LINUX_NETDEVICE_H */
This page took 0.09012 seconds and 6 git commands to generate.