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