1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
36
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
42
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
45 #include <net/dsa.h>
46 #ifdef CONFIG_DCB
47 #include <net/dcbnl.h>
48 #endif
49 #include <net/netprio_cgroup.h>
50
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 /* 802.11 specific */
59 struct wireless_dev;
60
61 void netdev_set_default_ethtool_ops(struct net_device *dev,
62 const struct ethtool_ops *ops);
63
64 /* Backlog congestion levels */
65 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
66 #define NET_RX_DROP 1 /* packet dropped */
67
68 /*
69 * Transmit return codes: transmit return codes originate from three different
70 * namespaces:
71 *
72 * - qdisc return codes
73 * - driver transmit return codes
74 * - errno values
75 *
76 * Drivers are allowed to return any one of those in their hard_start_xmit()
77 * function. Real network devices commonly used with qdiscs should only return
78 * the driver transmit return codes though - when qdiscs are used, the actual
79 * transmission happens asynchronously, so the value is not propagated to
80 * higher layers. Virtual network devices transmit synchronously, in this case
81 * the driver transmit return codes are consumed by dev_queue_xmit(), all
82 * others are propagated to higher layers.
83 */
84
85 /* qdisc ->enqueue() return codes. */
86 #define NET_XMIT_SUCCESS 0x00
87 #define NET_XMIT_DROP 0x01 /* skb dropped */
88 #define NET_XMIT_CN 0x02 /* congestion notification */
89 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
90 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
91
92 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
93 * indicates that the device will soon be dropping packets, or already drops
94 * some packets of the same priority; prompting us to send less aggressively. */
95 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
96 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
97
98 /* Driver transmit return codes */
99 #define NETDEV_TX_MASK 0xf0
100
101 enum netdev_tx {
102 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
103 NETDEV_TX_OK = 0x00, /* driver took care of packet */
104 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
105 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
106 };
107 typedef enum netdev_tx netdev_tx_t;
108
109 /*
110 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
111 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
112 */
dev_xmit_complete(int rc)113 static inline bool dev_xmit_complete(int rc)
114 {
115 /*
116 * Positive cases with an skb consumed by a driver:
117 * - successful transmission (rc == NETDEV_TX_OK)
118 * - error while transmitting (rc < 0)
119 * - error while queueing to a different device (rc & NET_XMIT_MASK)
120 */
121 if (likely(rc < NET_XMIT_MASK))
122 return true;
123
124 return false;
125 }
126
127 /*
128 * Compute the worst case header length according to the protocols
129 * used.
130 */
131
132 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
133 # if defined(CONFIG_MAC80211_MESH)
134 # define LL_MAX_HEADER 128
135 # else
136 # define LL_MAX_HEADER 96
137 # endif
138 #else
139 # define LL_MAX_HEADER 32
140 #endif
141
142 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
143 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
144 #define MAX_HEADER LL_MAX_HEADER
145 #else
146 #define MAX_HEADER (LL_MAX_HEADER + 48)
147 #endif
148
149 /*
150 * Old network device statistics. Fields are native words
151 * (unsigned long) so they can be read and written atomically.
152 */
153
154 struct net_device_stats {
155 unsigned long rx_packets;
156 unsigned long tx_packets;
157 unsigned long rx_bytes;
158 unsigned long tx_bytes;
159 unsigned long rx_errors;
160 unsigned long tx_errors;
161 unsigned long rx_dropped;
162 unsigned long tx_dropped;
163 unsigned long multicast;
164 unsigned long collisions;
165 unsigned long rx_length_errors;
166 unsigned long rx_over_errors;
167 unsigned long rx_crc_errors;
168 unsigned long rx_frame_errors;
169 unsigned long rx_fifo_errors;
170 unsigned long rx_missed_errors;
171 unsigned long tx_aborted_errors;
172 unsigned long tx_carrier_errors;
173 unsigned long tx_fifo_errors;
174 unsigned long tx_heartbeat_errors;
175 unsigned long tx_window_errors;
176 unsigned long rx_compressed;
177 unsigned long tx_compressed;
178 };
179
180
181 #include <linux/cache.h>
182 #include <linux/skbuff.h>
183
184 #ifdef CONFIG_RPS
185 #include <linux/static_key.h>
186 extern struct static_key rps_needed;
187 #endif
188
189 struct neighbour;
190 struct neigh_parms;
191 struct sk_buff;
192
193 struct netdev_hw_addr {
194 struct list_head list;
195 unsigned char addr[MAX_ADDR_LEN];
196 unsigned char type;
197 #define NETDEV_HW_ADDR_T_LAN 1
198 #define NETDEV_HW_ADDR_T_SAN 2
199 #define NETDEV_HW_ADDR_T_SLAVE 3
200 #define NETDEV_HW_ADDR_T_UNICAST 4
201 #define NETDEV_HW_ADDR_T_MULTICAST 5
202 bool global_use;
203 int sync_cnt;
204 int refcount;
205 int synced;
206 struct rcu_head rcu_head;
207 };
208
209 struct netdev_hw_addr_list {
210 struct list_head list;
211 int count;
212 };
213
214 #define netdev_hw_addr_list_count(l) ((l)->count)
215 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
216 #define netdev_hw_addr_list_for_each(ha, l) \
217 list_for_each_entry(ha, &(l)->list, list)
218
219 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
220 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
221 #define netdev_for_each_uc_addr(ha, dev) \
222 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
223
224 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
225 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
226 #define netdev_for_each_mc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
228
229 struct hh_cache {
230 u16 hh_len;
231 u16 __pad;
232 seqlock_t hh_lock;
233
234 /* cached hardware header; allow for machine alignment needs. */
235 #define HH_DATA_MOD 16
236 #define HH_DATA_OFF(__len) \
237 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
238 #define HH_DATA_ALIGN(__len) \
239 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
240 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
241 };
242
243 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
244 * Alternative is:
245 * dev->hard_header_len ? (dev->hard_header_len +
246 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
247 *
248 * We could use other alignment values, but we must maintain the
249 * relationship HH alignment <= LL alignment.
250 */
251 #define LL_RESERVED_SPACE(dev) \
252 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
253 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
254 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
255
256 struct header_ops {
257 int (*create) (struct sk_buff *skb, struct net_device *dev,
258 unsigned short type, const void *daddr,
259 const void *saddr, unsigned int len);
260 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
261 int (*rebuild)(struct sk_buff *skb);
262 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
263 void (*cache_update)(struct hh_cache *hh,
264 const struct net_device *dev,
265 const unsigned char *haddr);
266 };
267
268 /* These flag bits are private to the generic network queueing
269 * layer, they may not be explicitly referenced by any other
270 * code.
271 */
272
273 enum netdev_state_t {
274 __LINK_STATE_START,
275 __LINK_STATE_PRESENT,
276 __LINK_STATE_NOCARRIER,
277 __LINK_STATE_LINKWATCH_PENDING,
278 __LINK_STATE_DORMANT,
279 };
280
281
282 /*
283 * This structure holds at boot time configured netdevice settings. They
284 * are then used in the device probing.
285 */
286 struct netdev_boot_setup {
287 char name[IFNAMSIZ];
288 struct ifmap map;
289 };
290 #define NETDEV_BOOT_SETUP_MAX 8
291
292 int __init netdev_boot_setup(char *str);
293
294 /*
295 * Structure for NAPI scheduling similar to tasklet but with weighting
296 */
297 struct napi_struct {
298 /* The poll_list must only be managed by the entity which
299 * changes the state of the NAPI_STATE_SCHED bit. This means
300 * whoever atomically sets that bit can add this napi_struct
301 * to the per-cpu poll_list, and whoever clears that bit
302 * can remove from the list right before clearing the bit.
303 */
304 struct list_head poll_list;
305
306 unsigned long state;
307 int weight;
308 unsigned int gro_count;
309 int (*poll)(struct napi_struct *, int);
310 #ifdef CONFIG_NETPOLL
311 spinlock_t poll_lock;
312 int poll_owner;
313 #endif
314 struct net_device *dev;
315 struct sk_buff *gro_list;
316 struct sk_buff *skb;
317 struct list_head dev_list;
318 struct hlist_node napi_hash_node;
319 unsigned int napi_id;
320 };
321
322 enum {
323 NAPI_STATE_SCHED, /* Poll is scheduled */
324 NAPI_STATE_DISABLE, /* Disable pending */
325 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
326 NAPI_STATE_HASHED, /* In NAPI hash */
327 };
328
329 enum gro_result {
330 GRO_MERGED,
331 GRO_MERGED_FREE,
332 GRO_HELD,
333 GRO_NORMAL,
334 GRO_DROP,
335 };
336 typedef enum gro_result gro_result_t;
337
338 /*
339 * enum rx_handler_result - Possible return values for rx_handlers.
340 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
341 * further.
342 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
343 * case skb->dev was changed by rx_handler.
344 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
345 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
346 *
347 * rx_handlers are functions called from inside __netif_receive_skb(), to do
348 * special processing of the skb, prior to delivery to protocol handlers.
349 *
350 * Currently, a net_device can only have a single rx_handler registered. Trying
351 * to register a second rx_handler will return -EBUSY.
352 *
353 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
354 * To unregister a rx_handler on a net_device, use
355 * netdev_rx_handler_unregister().
356 *
357 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
358 * do with the skb.
359 *
360 * If the rx_handler consumed to skb in some way, it should return
361 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
362 * the skb to be delivered in some other ways.
363 *
364 * If the rx_handler changed skb->dev, to divert the skb to another
365 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
366 * new device will be called if it exists.
367 *
368 * If the rx_handler consider the skb should be ignored, it should return
369 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
370 * are registered on exact device (ptype->dev == skb->dev).
371 *
372 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
373 * delivered, it should return RX_HANDLER_PASS.
374 *
375 * A device without a registered rx_handler will behave as if rx_handler
376 * returned RX_HANDLER_PASS.
377 */
378
379 enum rx_handler_result {
380 RX_HANDLER_CONSUMED,
381 RX_HANDLER_ANOTHER,
382 RX_HANDLER_EXACT,
383 RX_HANDLER_PASS,
384 };
385 typedef enum rx_handler_result rx_handler_result_t;
386 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
387
388 void __napi_schedule(struct napi_struct *n);
389
napi_disable_pending(struct napi_struct * n)390 static inline bool napi_disable_pending(struct napi_struct *n)
391 {
392 return test_bit(NAPI_STATE_DISABLE, &n->state);
393 }
394
395 /**
396 * napi_schedule_prep - check if napi can be scheduled
397 * @n: napi context
398 *
399 * Test if NAPI routine is already running, and if not mark
400 * it as running. This is used as a condition variable
401 * insure only one NAPI poll instance runs. We also make
402 * sure there is no pending NAPI disable.
403 */
napi_schedule_prep(struct napi_struct * n)404 static inline bool napi_schedule_prep(struct napi_struct *n)
405 {
406 return !napi_disable_pending(n) &&
407 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
408 }
409
410 /**
411 * napi_schedule - schedule NAPI poll
412 * @n: napi context
413 *
414 * Schedule NAPI poll routine to be called if it is not already
415 * running.
416 */
napi_schedule(struct napi_struct * n)417 static inline void napi_schedule(struct napi_struct *n)
418 {
419 if (napi_schedule_prep(n))
420 __napi_schedule(n);
421 }
422
423 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)424 static inline bool napi_reschedule(struct napi_struct *napi)
425 {
426 if (napi_schedule_prep(napi)) {
427 __napi_schedule(napi);
428 return true;
429 }
430 return false;
431 }
432
433 /**
434 * napi_complete - NAPI processing complete
435 * @n: napi context
436 *
437 * Mark NAPI processing as complete.
438 */
439 void __napi_complete(struct napi_struct *n);
440 void napi_complete(struct napi_struct *n);
441
442 /**
443 * napi_by_id - lookup a NAPI by napi_id
444 * @napi_id: hashed napi_id
445 *
446 * lookup @napi_id in napi_hash table
447 * must be called under rcu_read_lock()
448 */
449 struct napi_struct *napi_by_id(unsigned int napi_id);
450
451 /**
452 * napi_hash_add - add a NAPI to global hashtable
453 * @napi: napi context
454 *
455 * generate a new napi_id and store a @napi under it in napi_hash
456 */
457 void napi_hash_add(struct napi_struct *napi);
458
459 /**
460 * napi_hash_del - remove a NAPI from global table
461 * @napi: napi context
462 *
463 * Warning: caller must observe rcu grace period
464 * before freeing memory containing @napi
465 */
466 void napi_hash_del(struct napi_struct *napi);
467
468 /**
469 * napi_disable - prevent NAPI from scheduling
470 * @n: napi context
471 *
472 * Stop NAPI from being scheduled on this context.
473 * Waits till any outstanding processing completes.
474 */
napi_disable(struct napi_struct * n)475 static inline void napi_disable(struct napi_struct *n)
476 {
477 might_sleep();
478 set_bit(NAPI_STATE_DISABLE, &n->state);
479 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
480 msleep(1);
481 clear_bit(NAPI_STATE_DISABLE, &n->state);
482 }
483
484 /**
485 * napi_enable - enable NAPI scheduling
486 * @n: napi context
487 *
488 * Resume NAPI from being scheduled on this context.
489 * Must be paired with napi_disable.
490 */
napi_enable(struct napi_struct * n)491 static inline void napi_enable(struct napi_struct *n)
492 {
493 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
494 smp_mb__before_atomic();
495 clear_bit(NAPI_STATE_SCHED, &n->state);
496 }
497
498 #ifdef CONFIG_SMP
499 /**
500 * napi_synchronize - wait until NAPI is not running
501 * @n: napi context
502 *
503 * Wait until NAPI is done being scheduled on this context.
504 * Waits till any outstanding processing completes but
505 * does not disable future activations.
506 */
napi_synchronize(const struct napi_struct * n)507 static inline void napi_synchronize(const struct napi_struct *n)
508 {
509 while (test_bit(NAPI_STATE_SCHED, &n->state))
510 msleep(1);
511 }
512 #else
513 # define napi_synchronize(n) barrier()
514 #endif
515
516 enum netdev_queue_state_t {
517 __QUEUE_STATE_DRV_XOFF,
518 __QUEUE_STATE_STACK_XOFF,
519 __QUEUE_STATE_FROZEN,
520 };
521
522 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
523 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
524 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
525
526 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
527 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
528 QUEUE_STATE_FROZEN)
529 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
530 QUEUE_STATE_FROZEN)
531
532 /*
533 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
534 * netif_tx_* functions below are used to manipulate this flag. The
535 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
536 * queue independently. The netif_xmit_*stopped functions below are called
537 * to check if the queue has been stopped by the driver or stack (either
538 * of the XOFF bits are set in the state). Drivers should not need to call
539 * netif_xmit*stopped functions, they should only be using netif_tx_*.
540 */
541
542 struct netdev_queue {
543 /*
544 * read mostly part
545 */
546 struct net_device *dev;
547 struct Qdisc __rcu *qdisc;
548 struct Qdisc *qdisc_sleeping;
549 #ifdef CONFIG_SYSFS
550 struct kobject kobj;
551 #endif
552 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
553 int numa_node;
554 #endif
555 /*
556 * write mostly part
557 */
558 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
559 int xmit_lock_owner;
560 /*
561 * please use this field instead of dev->trans_start
562 */
563 unsigned long trans_start;
564
565 /*
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
568 */
569 unsigned long trans_timeout;
570
571 unsigned long state;
572
573 #ifdef CONFIG_BQL
574 struct dql dql;
575 #endif
576 } ____cacheline_aligned_in_smp;
577
netdev_queue_numa_node_read(const struct netdev_queue * q)578 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
579 {
580 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
581 return q->numa_node;
582 #else
583 return NUMA_NO_NODE;
584 #endif
585 }
586
netdev_queue_numa_node_write(struct netdev_queue * q,int node)587 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
588 {
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 q->numa_node = node;
591 #endif
592 }
593
594 #ifdef CONFIG_RPS
595 /*
596 * This structure holds an RPS map which can be of variable length. The
597 * map is an array of CPUs.
598 */
599 struct rps_map {
600 unsigned int len;
601 struct rcu_head rcu;
602 u16 cpus[0];
603 };
604 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
605
606 /*
607 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
608 * tail pointer for that CPU's input queue at the time of last enqueue, and
609 * a hardware filter index.
610 */
611 struct rps_dev_flow {
612 u16 cpu;
613 u16 filter;
614 unsigned int last_qtail;
615 };
616 #define RPS_NO_FILTER 0xffff
617
618 /*
619 * The rps_dev_flow_table structure contains a table of flow mappings.
620 */
621 struct rps_dev_flow_table {
622 unsigned int mask;
623 struct rcu_head rcu;
624 struct rps_dev_flow flows[0];
625 };
626 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
627 ((_num) * sizeof(struct rps_dev_flow)))
628
629 /*
630 * The rps_sock_flow_table contains mappings of flows to the last CPU
631 * on which they were processed by the application (set in recvmsg).
632 */
633 struct rps_sock_flow_table {
634 unsigned int mask;
635 u16 ents[0];
636 };
637 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
638 ((_num) * sizeof(u16)))
639
640 #define RPS_NO_CPU 0xffff
641
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)642 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
643 u32 hash)
644 {
645 if (table && hash) {
646 unsigned int cpu, index = hash & table->mask;
647
648 /* We only give a hint, preemption can change cpu under us */
649 cpu = raw_smp_processor_id();
650
651 if (table->ents[index] != cpu)
652 table->ents[index] = cpu;
653 }
654 }
655
rps_reset_sock_flow(struct rps_sock_flow_table * table,u32 hash)656 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
657 u32 hash)
658 {
659 if (table && hash)
660 table->ents[hash & table->mask] = RPS_NO_CPU;
661 }
662
663 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
664
665 #ifdef CONFIG_RFS_ACCEL
666 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
667 u16 filter_id);
668 #endif
669 #endif /* CONFIG_RPS */
670
671 /* This structure contains an instance of an RX queue. */
672 struct netdev_rx_queue {
673 #ifdef CONFIG_RPS
674 struct rps_map __rcu *rps_map;
675 struct rps_dev_flow_table __rcu *rps_flow_table;
676 #endif
677 struct kobject kobj;
678 struct net_device *dev;
679 } ____cacheline_aligned_in_smp;
680
681 /*
682 * RX queue sysfs structures and functions.
683 */
684 struct rx_queue_attribute {
685 struct attribute attr;
686 ssize_t (*show)(struct netdev_rx_queue *queue,
687 struct rx_queue_attribute *attr, char *buf);
688 ssize_t (*store)(struct netdev_rx_queue *queue,
689 struct rx_queue_attribute *attr, const char *buf, size_t len);
690 };
691
692 #ifdef CONFIG_XPS
693 /*
694 * This structure holds an XPS map which can be of variable length. The
695 * map is an array of queues.
696 */
697 struct xps_map {
698 unsigned int len;
699 unsigned int alloc_len;
700 struct rcu_head rcu;
701 u16 queues[0];
702 };
703 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
704 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
705 / sizeof(u16))
706
707 /*
708 * This structure holds all XPS maps for device. Maps are indexed by CPU.
709 */
710 struct xps_dev_maps {
711 struct rcu_head rcu;
712 struct xps_map __rcu *cpu_map[0];
713 };
714 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
715 (nr_cpu_ids * sizeof(struct xps_map *)))
716 #endif /* CONFIG_XPS */
717
718 #define TC_MAX_QUEUE 16
719 #define TC_BITMASK 15
720 /* HW offloaded queuing disciplines txq count and offset maps */
721 struct netdev_tc_txq {
722 u16 count;
723 u16 offset;
724 };
725
726 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
727 /*
728 * This structure is to hold information about the device
729 * configured to run FCoE protocol stack.
730 */
731 struct netdev_fcoe_hbainfo {
732 char manufacturer[64];
733 char serial_number[64];
734 char hardware_version[64];
735 char driver_version[64];
736 char optionrom_version[64];
737 char firmware_version[64];
738 char model[256];
739 char model_description[256];
740 };
741 #endif
742
743 #define MAX_PHYS_PORT_ID_LEN 32
744
745 /* This structure holds a unique identifier to identify the
746 * physical port used by a netdevice.
747 */
748 struct netdev_phys_port_id {
749 unsigned char id[MAX_PHYS_PORT_ID_LEN];
750 unsigned char id_len;
751 };
752
753 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
754 struct sk_buff *skb);
755
756 /*
757 * This structure defines the management hooks for network devices.
758 * The following hooks can be defined; unless noted otherwise, they are
759 * optional and can be filled with a null pointer.
760 *
761 * int (*ndo_init)(struct net_device *dev);
762 * This function is called once when network device is registered.
763 * The network device can use this to any late stage initializaton
764 * or semantic validattion. It can fail with an error code which will
765 * be propogated back to register_netdev
766 *
767 * void (*ndo_uninit)(struct net_device *dev);
768 * This function is called when device is unregistered or when registration
769 * fails. It is not called if init fails.
770 *
771 * int (*ndo_open)(struct net_device *dev);
772 * This function is called when network device transistions to the up
773 * state.
774 *
775 * int (*ndo_stop)(struct net_device *dev);
776 * This function is called when network device transistions to the down
777 * state.
778 *
779 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
780 * struct net_device *dev);
781 * Called when a packet needs to be transmitted.
782 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
783 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
784 * Required can not be NULL.
785 *
786 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
787 * void *accel_priv, select_queue_fallback_t fallback);
788 * Called to decide which queue to when device supports multiple
789 * transmit queues.
790 *
791 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
792 * This function is called to allow device receiver to make
793 * changes to configuration when multicast or promiscious is enabled.
794 *
795 * void (*ndo_set_rx_mode)(struct net_device *dev);
796 * This function is called device changes address list filtering.
797 * If driver handles unicast address filtering, it should set
798 * IFF_UNICAST_FLT to its priv_flags.
799 *
800 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
801 * This function is called when the Media Access Control address
802 * needs to be changed. If this interface is not defined, the
803 * mac address can not be changed.
804 *
805 * int (*ndo_validate_addr)(struct net_device *dev);
806 * Test if Media Access Control address is valid for the device.
807 *
808 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
809 * Called when a user request an ioctl which can't be handled by
810 * the generic interface code. If not defined ioctl's return
811 * not supported error code.
812 *
813 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
814 * Used to set network devices bus interface parameters. This interface
815 * is retained for legacy reason, new devices should use the bus
816 * interface (PCI) for low level management.
817 *
818 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
819 * Called when a user wants to change the Maximum Transfer Unit
820 * of a device. If not defined, any request to change MTU will
821 * will return an error.
822 *
823 * void (*ndo_tx_timeout)(struct net_device *dev);
824 * Callback uses when the transmitter has not made any progress
825 * for dev->watchdog ticks.
826 *
827 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
828 * struct rtnl_link_stats64 *storage);
829 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
830 * Called when a user wants to get the network device usage
831 * statistics. Drivers must do one of the following:
832 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
833 * rtnl_link_stats64 structure passed by the caller.
834 * 2. Define @ndo_get_stats to update a net_device_stats structure
835 * (which should normally be dev->stats) and return a pointer to
836 * it. The structure may be changed asynchronously only if each
837 * field is written atomically.
838 * 3. Update dev->stats asynchronously and atomically, and define
839 * neither operation.
840 *
841 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
842 * If device support VLAN filtering this function is called when a
843 * VLAN id is registered.
844 *
845 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
846 * If device support VLAN filtering this function is called when a
847 * VLAN id is unregistered.
848 *
849 * void (*ndo_poll_controller)(struct net_device *dev);
850 *
851 * SR-IOV management functions.
852 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
853 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
854 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
855 * int max_tx_rate);
856 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
857 * int (*ndo_get_vf_config)(struct net_device *dev,
858 * int vf, struct ifla_vf_info *ivf);
859 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
860 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
861 * struct nlattr *port[]);
862 *
863 * Enable or disable the VF ability to query its RSS Redirection Table and
864 * Hash Key. This is needed since on some devices VF share this information
865 * with PF and querying it may adduce a theoretical security risk.
866 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
867 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
868 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
869 * Called to setup 'tc' number of traffic classes in the net device. This
870 * is always called from the stack with the rtnl lock held and netif tx
871 * queues stopped. This allows the netdevice to perform queue management
872 * safely.
873 *
874 * Fiber Channel over Ethernet (FCoE) offload functions.
875 * int (*ndo_fcoe_enable)(struct net_device *dev);
876 * Called when the FCoE protocol stack wants to start using LLD for FCoE
877 * so the underlying device can perform whatever needed configuration or
878 * initialization to support acceleration of FCoE traffic.
879 *
880 * int (*ndo_fcoe_disable)(struct net_device *dev);
881 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
882 * so the underlying device can perform whatever needed clean-ups to
883 * stop supporting acceleration of FCoE traffic.
884 *
885 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
886 * struct scatterlist *sgl, unsigned int sgc);
887 * Called when the FCoE Initiator wants to initialize an I/O that
888 * is a possible candidate for Direct Data Placement (DDP). The LLD can
889 * perform necessary setup and returns 1 to indicate the device is set up
890 * successfully to perform DDP on this I/O, otherwise this returns 0.
891 *
892 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
893 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
894 * indicated by the FC exchange id 'xid', so the underlying device can
895 * clean up and reuse resources for later DDP requests.
896 *
897 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
898 * struct scatterlist *sgl, unsigned int sgc);
899 * Called when the FCoE Target wants to initialize an I/O that
900 * is a possible candidate for Direct Data Placement (DDP). The LLD can
901 * perform necessary setup and returns 1 to indicate the device is set up
902 * successfully to perform DDP on this I/O, otherwise this returns 0.
903 *
904 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
905 * struct netdev_fcoe_hbainfo *hbainfo);
906 * Called when the FCoE Protocol stack wants information on the underlying
907 * device. This information is utilized by the FCoE protocol stack to
908 * register attributes with Fiber Channel management service as per the
909 * FC-GS Fabric Device Management Information(FDMI) specification.
910 *
911 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
912 * Called when the underlying device wants to override default World Wide
913 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
914 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
915 * protocol stack to use.
916 *
917 * RFS acceleration.
918 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
919 * u16 rxq_index, u32 flow_id);
920 * Set hardware filter for RFS. rxq_index is the target queue index;
921 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
922 * Return the filter ID on success, or a negative error code.
923 *
924 * Slave management functions (for bridge, bonding, etc).
925 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
926 * Called to make another netdev an underling.
927 *
928 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
929 * Called to release previously enslaved netdev.
930 *
931 * Feature/offload setting functions.
932 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
933 * netdev_features_t features);
934 * Adjusts the requested feature flags according to device-specific
935 * constraints, and returns the resulting flags. Must not modify
936 * the device state.
937 *
938 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
939 * Called to update device configuration to new features. Passed
940 * feature set might be less than what was returned by ndo_fix_features()).
941 * Must return >0 or -errno if it changed dev->features itself.
942 *
943 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
944 * struct net_device *dev,
945 * const unsigned char *addr, u16 flags)
946 * Adds an FDB entry to dev for addr.
947 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
948 * struct net_device *dev,
949 * const unsigned char *addr)
950 * Deletes the FDB entry from dev coresponding to addr.
951 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
952 * struct net_device *dev, struct net_device *filter_dev,
953 * int idx)
954 * Used to add FDB entries to dump requests. Implementers should add
955 * entries to skb and update idx with the number of entries.
956 *
957 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
958 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
959 * struct net_device *dev, u32 filter_mask)
960 *
961 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
962 * Called to change device carrier. Soft-devices (like dummy, team, etc)
963 * which do not represent real hardware may define this to allow their
964 * userspace components to manage their virtual carrier state. Devices
965 * that determine carrier state from physical hardware properties (eg
966 * network cables) or protocol-dependent mechanisms (eg
967 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
968 *
969 * int (*ndo_get_phys_port_id)(struct net_device *dev,
970 * struct netdev_phys_port_id *ppid);
971 * Called to get ID of physical port of this device. If driver does
972 * not implement this, it is assumed that the hw is not able to have
973 * multiple net devices on single physical port.
974 *
975 * void (*ndo_add_vxlan_port)(struct net_device *dev,
976 * sa_family_t sa_family, __be16 port);
977 * Called by vxlan to notiy a driver about the UDP port and socket
978 * address family that vxlan is listnening to. It is called only when
979 * a new port starts listening. The operation is protected by the
980 * vxlan_net->sock_lock.
981 *
982 * void (*ndo_del_vxlan_port)(struct net_device *dev,
983 * sa_family_t sa_family, __be16 port);
984 * Called by vxlan to notify the driver about a UDP port and socket
985 * address family that vxlan is not listening to anymore. The operation
986 * is protected by the vxlan_net->sock_lock.
987 *
988 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
989 * struct net_device *dev)
990 * Called by upper layer devices to accelerate switching or other
991 * station functionality into hardware. 'pdev is the lowerdev
992 * to use for the offload and 'dev' is the net device that will
993 * back the offload. Returns a pointer to the private structure
994 * the upper layer will maintain.
995 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
996 * Called by upper layer device to delete the station created
997 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
998 * the station and priv is the structure returned by the add
999 * operation.
1000 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1001 * struct net_device *dev,
1002 * void *priv);
1003 * Callback to use for xmit over the accelerated station. This
1004 * is used in place of ndo_start_xmit on accelerated net
1005 * devices.
1006 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1007 * struct net_device *dev
1008 * netdev_features_t features);
1009 * Called by core transmit path to determine if device is capable of
1010 * performing offload operations on a given packet. This is to give
1011 * the device an opportunity to implement any restrictions that cannot
1012 * be otherwise expressed by feature flags. The check is called with
1013 * the set of features that the stack has calculated and it returns
1014 * those the driver believes to be appropriate.
1015 */
1016 struct net_device_ops {
1017 int (*ndo_init)(struct net_device *dev);
1018 void (*ndo_uninit)(struct net_device *dev);
1019 int (*ndo_open)(struct net_device *dev);
1020 int (*ndo_stop)(struct net_device *dev);
1021 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1022 struct net_device *dev);
1023 u16 (*ndo_select_queue)(struct net_device *dev,
1024 struct sk_buff *skb,
1025 void *accel_priv,
1026 select_queue_fallback_t fallback);
1027 void (*ndo_change_rx_flags)(struct net_device *dev,
1028 int flags);
1029 void (*ndo_set_rx_mode)(struct net_device *dev);
1030 int (*ndo_set_mac_address)(struct net_device *dev,
1031 void *addr);
1032 int (*ndo_validate_addr)(struct net_device *dev);
1033 int (*ndo_do_ioctl)(struct net_device *dev,
1034 struct ifreq *ifr, int cmd);
1035 int (*ndo_set_config)(struct net_device *dev,
1036 struct ifmap *map);
1037 int (*ndo_change_mtu)(struct net_device *dev,
1038 int new_mtu);
1039 int (*ndo_neigh_setup)(struct net_device *dev,
1040 struct neigh_parms *);
1041 void (*ndo_tx_timeout) (struct net_device *dev);
1042
1043 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1044 struct rtnl_link_stats64 *storage);
1045 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1046
1047 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1048 __be16 proto, u16 vid);
1049 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1050 __be16 proto, u16 vid);
1051 #ifdef CONFIG_NET_POLL_CONTROLLER
1052 void (*ndo_poll_controller)(struct net_device *dev);
1053 int (*ndo_netpoll_setup)(struct net_device *dev,
1054 struct netpoll_info *info);
1055 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1056 #endif
1057 #ifdef CONFIG_NET_RX_BUSY_POLL
1058 int (*ndo_busy_poll)(struct napi_struct *dev);
1059 #endif
1060 int (*ndo_set_vf_mac)(struct net_device *dev,
1061 int queue, u8 *mac);
1062 int (*ndo_set_vf_vlan)(struct net_device *dev,
1063 int queue, u16 vlan, u8 qos);
1064 int (*ndo_set_vf_rate)(struct net_device *dev,
1065 int vf, int min_tx_rate,
1066 int max_tx_rate);
1067 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1068 int vf, bool setting);
1069 int (*ndo_get_vf_config)(struct net_device *dev,
1070 int vf,
1071 struct ifla_vf_info *ivf);
1072 int (*ndo_set_vf_link_state)(struct net_device *dev,
1073 int vf, int link_state);
1074 int (*ndo_set_vf_port)(struct net_device *dev,
1075 int vf,
1076 struct nlattr *port[]);
1077 int (*ndo_get_vf_port)(struct net_device *dev,
1078 int vf, struct sk_buff *skb);
1079 int (*ndo_set_vf_rss_query_en)(
1080 struct net_device *dev,
1081 int vf, bool setting);
1082 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1083 #if IS_ENABLED(CONFIG_FCOE)
1084 int (*ndo_fcoe_enable)(struct net_device *dev);
1085 int (*ndo_fcoe_disable)(struct net_device *dev);
1086 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1087 u16 xid,
1088 struct scatterlist *sgl,
1089 unsigned int sgc);
1090 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1091 u16 xid);
1092 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1093 u16 xid,
1094 struct scatterlist *sgl,
1095 unsigned int sgc);
1096 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1097 struct netdev_fcoe_hbainfo *hbainfo);
1098 #endif
1099
1100 #if IS_ENABLED(CONFIG_LIBFCOE)
1101 #define NETDEV_FCOE_WWNN 0
1102 #define NETDEV_FCOE_WWPN 1
1103 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1104 u64 *wwn, int type);
1105 #endif
1106
1107 #ifdef CONFIG_RFS_ACCEL
1108 int (*ndo_rx_flow_steer)(struct net_device *dev,
1109 const struct sk_buff *skb,
1110 u16 rxq_index,
1111 u32 flow_id);
1112 #endif
1113 int (*ndo_add_slave)(struct net_device *dev,
1114 struct net_device *slave_dev);
1115 int (*ndo_del_slave)(struct net_device *dev,
1116 struct net_device *slave_dev);
1117 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1118 netdev_features_t features);
1119 int (*ndo_set_features)(struct net_device *dev,
1120 netdev_features_t features);
1121 int (*ndo_neigh_construct)(struct neighbour *n);
1122 void (*ndo_neigh_destroy)(struct neighbour *n);
1123
1124 int (*ndo_fdb_add)(struct ndmsg *ndm,
1125 struct nlattr *tb[],
1126 struct net_device *dev,
1127 const unsigned char *addr,
1128 u16 flags);
1129 int (*ndo_fdb_del)(struct ndmsg *ndm,
1130 struct nlattr *tb[],
1131 struct net_device *dev,
1132 const unsigned char *addr);
1133 int (*ndo_fdb_dump)(struct sk_buff *skb,
1134 struct netlink_callback *cb,
1135 struct net_device *dev,
1136 struct net_device *filter_dev,
1137 int idx);
1138
1139 int (*ndo_bridge_setlink)(struct net_device *dev,
1140 struct nlmsghdr *nlh);
1141 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1142 u32 pid, u32 seq,
1143 struct net_device *dev,
1144 u32 filter_mask);
1145 int (*ndo_bridge_dellink)(struct net_device *dev,
1146 struct nlmsghdr *nlh);
1147 int (*ndo_change_carrier)(struct net_device *dev,
1148 bool new_carrier);
1149 int (*ndo_get_phys_port_id)(struct net_device *dev,
1150 struct netdev_phys_port_id *ppid);
1151 void (*ndo_add_vxlan_port)(struct net_device *dev,
1152 sa_family_t sa_family,
1153 __be16 port);
1154 void (*ndo_del_vxlan_port)(struct net_device *dev,
1155 sa_family_t sa_family,
1156 __be16 port);
1157
1158 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1159 struct net_device *dev);
1160 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1161 void *priv);
1162
1163 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1164 struct net_device *dev,
1165 void *priv);
1166 int (*ndo_get_lock_subclass)(struct net_device *dev);
1167 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1168 struct net_device *dev,
1169 netdev_features_t features);
1170 };
1171
1172 /**
1173 * enum net_device_priv_flags - &struct net_device priv_flags
1174 *
1175 * These are the &struct net_device, they are only set internally
1176 * by drivers and used in the kernel. These flags are invisible to
1177 * userspace, this means that the order of these flags can change
1178 * during any kernel release.
1179 *
1180 * You should have a pretty good reason to be extending these flags.
1181 *
1182 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1183 * @IFF_EBRIDGE: Ethernet bridging device
1184 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1185 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1186 * @IFF_MASTER_ALB: bonding master, balance-alb
1187 * @IFF_BONDING: bonding master or slave
1188 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1189 * @IFF_ISATAP: ISATAP interface (RFC4214)
1190 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1191 * @IFF_WAN_HDLC: WAN HDLC device
1192 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1193 * release skb->dst
1194 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1195 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1196 * @IFF_MACVLAN_PORT: device used as macvlan port
1197 * @IFF_BRIDGE_PORT: device used as bridge port
1198 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1199 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1200 * @IFF_UNICAST_FLT: Supports unicast filtering
1201 * @IFF_TEAM_PORT: device used as team port
1202 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1203 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1204 * change when it's running
1205 * @IFF_MACVLAN: Macvlan device
1206 */
1207 enum netdev_priv_flags {
1208 IFF_802_1Q_VLAN = 1<<0,
1209 IFF_EBRIDGE = 1<<1,
1210 IFF_SLAVE_INACTIVE = 1<<2,
1211 IFF_MASTER_8023AD = 1<<3,
1212 IFF_MASTER_ALB = 1<<4,
1213 IFF_BONDING = 1<<5,
1214 IFF_SLAVE_NEEDARP = 1<<6,
1215 IFF_ISATAP = 1<<7,
1216 IFF_MASTER_ARPMON = 1<<8,
1217 IFF_WAN_HDLC = 1<<9,
1218 IFF_XMIT_DST_RELEASE = 1<<10,
1219 IFF_DONT_BRIDGE = 1<<11,
1220 IFF_DISABLE_NETPOLL = 1<<12,
1221 IFF_MACVLAN_PORT = 1<<13,
1222 IFF_BRIDGE_PORT = 1<<14,
1223 IFF_OVS_DATAPATH = 1<<15,
1224 IFF_TX_SKB_SHARING = 1<<16,
1225 IFF_UNICAST_FLT = 1<<17,
1226 IFF_TEAM_PORT = 1<<18,
1227 IFF_SUPP_NOFCS = 1<<19,
1228 IFF_LIVE_ADDR_CHANGE = 1<<20,
1229 IFF_MACVLAN = 1<<21,
1230 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1231 };
1232
1233 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1234 #define IFF_EBRIDGE IFF_EBRIDGE
1235 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1236 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1237 #define IFF_MASTER_ALB IFF_MASTER_ALB
1238 #define IFF_BONDING IFF_BONDING
1239 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1240 #define IFF_ISATAP IFF_ISATAP
1241 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1242 #define IFF_WAN_HDLC IFF_WAN_HDLC
1243 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1244 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1245 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1246 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1247 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1248 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1249 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1250 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1251 #define IFF_TEAM_PORT IFF_TEAM_PORT
1252 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1253 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1254 #define IFF_MACVLAN IFF_MACVLAN
1255 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1256
1257 /**
1258 * struct net_device - The DEVICE structure.
1259 * Actually, this whole structure is a big mistake. It mixes I/O
1260 * data with strictly "high-level" data, and it has to know about
1261 * almost every data structure used in the INET module.
1262 *
1263 * @name: This is the first field of the "visible" part of this structure
1264 * (i.e. as seen by users in the "Space.c" file). It is the name
1265 * of the interface.
1266 *
1267 * @name_hlist: Device name hash chain, please keep it close to name[]
1268 * @ifalias: SNMP alias
1269 * @mem_end: Shared memory end
1270 * @mem_start: Shared memory start
1271 * @base_addr: Device I/O address
1272 * @irq: Device IRQ number
1273 *
1274 * @state: Generic network queuing layer state, see netdev_state_t
1275 * @dev_list: The global list of network devices
1276 * @napi_list: List entry, that is used for polling napi devices
1277 * @unreg_list: List entry, that is used, when we are unregistering the
1278 * device, see the function unregister_netdev
1279 * @close_list: List entry, that is used, when we are closing the device
1280 *
1281 * @adj_list: Directly linked devices, like slaves for bonding
1282 * @all_adj_list: All linked devices, *including* neighbours
1283 * @features: Currently active device features
1284 * @hw_features: User-changeable features
1285 *
1286 * @wanted_features: User-requested features
1287 * @vlan_features: Mask of features inheritable by VLAN devices
1288 *
1289 * @hw_enc_features: Mask of features inherited by encapsulating devices
1290 * This field indicates what encapsulation
1291 * offloads the hardware is capable of doing,
1292 * and drivers will need to set them appropriately.
1293 *
1294 * @mpls_features: Mask of features inheritable by MPLS
1295 *
1296 * @ifindex: interface index
1297 * @iflink: unique device identifier
1298 *
1299 * @stats: Statistics struct, which was left as a legacy, use
1300 * rtnl_link_stats64 instead
1301 *
1302 * @rx_dropped: Dropped packets by core network,
1303 * do not use this in drivers
1304 * @tx_dropped: Dropped packets by core network,
1305 * do not use this in drivers
1306 *
1307 * @carrier_changes: Stats to monitor carrier on<->off transitions
1308 *
1309 * @wireless_handlers: List of functions to handle Wireless Extensions,
1310 * instead of ioctl,
1311 * see <net/iw_handler.h> for details.
1312 * @wireless_data: Instance data managed by the core of wireless extensions
1313 *
1314 * @netdev_ops: Includes several pointers to callbacks,
1315 * if one wants to override the ndo_*() functions
1316 * @ethtool_ops: Management operations
1317 * @fwd_ops: Management operations
1318 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1319 * of Layer 2 headers.
1320 *
1321 * @flags: Interface flags (a la BSD)
1322 * @priv_flags: Like 'flags' but invisible to userspace,
1323 * see if.h for the definitions
1324 * @gflags: Global flags ( kept as legacy )
1325 * @padded: How much padding added by alloc_netdev()
1326 * @operstate: RFC2863 operstate
1327 * @link_mode: Mapping policy to operstate
1328 * @if_port: Selectable AUI, TP, ...
1329 * @dma: DMA channel
1330 * @mtu: Interface MTU value
1331 * @type: Interface hardware type
1332 * @hard_header_len: Hardware header length
1333 *
1334 * @needed_headroom: Extra headroom the hardware may need, but not in all
1335 * cases can this be guaranteed
1336 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1337 * cases can this be guaranteed. Some cases also use
1338 * LL_MAX_HEADER instead to allocate the skb
1339 *
1340 * interface address info:
1341 *
1342 * @perm_addr: Permanent hw address
1343 * @addr_assign_type: Hw address assignment type
1344 * @addr_len: Hardware address length
1345 * @neigh_priv_len; Used in neigh_alloc(),
1346 * initialized only in atm/clip.c
1347 * @dev_id: Used to differentiate devices that share
1348 * the same link layer address
1349 * @dev_port: Used to differentiate devices that share
1350 * the same function
1351 * @addr_list_lock: XXX: need comments on this one
1352 * @uc: unicast mac addresses
1353 * @mc: multicast mac addresses
1354 * @dev_addrs: list of device hw addresses
1355 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1356 * @uc_promisc: Counter, that indicates, that promiscuous mode
1357 * has been enabled due to the need to listen to
1358 * additional unicast addresses in a device that
1359 * does not implement ndo_set_rx_mode()
1360 * @promiscuity: Number of times, the NIC is told to work in
1361 * Promiscuous mode, if it becomes 0 the NIC will
1362 * exit from working in Promiscuous mode
1363 * @allmulti: Counter, enables or disables allmulticast mode
1364 *
1365 * @vlan_info: VLAN info
1366 * @dsa_ptr: dsa specific data
1367 * @tipc_ptr: TIPC specific data
1368 * @atalk_ptr: AppleTalk link
1369 * @ip_ptr: IPv4 specific data
1370 * @dn_ptr: DECnet specific data
1371 * @ip6_ptr: IPv6 specific data
1372 * @ax25_ptr: AX.25 specific data
1373 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1374 *
1375 * @last_rx: Time of last Rx
1376 * @dev_addr: Hw address (before bcast,
1377 * because most packets are unicast)
1378 *
1379 * @_rx: Array of RX queues
1380 * @num_rx_queues: Number of RX queues
1381 * allocated at register_netdev() time
1382 * @real_num_rx_queues: Number of RX queues currently active in device
1383 *
1384 * @rx_handler: handler for received packets
1385 * @rx_handler_data: XXX: need comments on this one
1386 * @ingress_queue: XXX: need comments on this one
1387 * @broadcast: hw bcast address
1388 *
1389 * @_tx: Array of TX queues
1390 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1391 * @real_num_tx_queues: Number of TX queues currently active in device
1392 * @qdisc: Root qdisc from userspace point of view
1393 * @tx_queue_len: Max frames per queue allowed
1394 * @tx_global_lock: XXX: need comments on this one
1395 *
1396 * @xps_maps: XXX: need comments on this one
1397 *
1398 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1399 * indexed by RX queue number. Assigned by driver.
1400 * This must only be set if the ndo_rx_flow_steer
1401 * operation is defined
1402 *
1403 * @trans_start: Time (in jiffies) of last Tx
1404 * @watchdog_timeo: Represents the timeout that is used by
1405 * the watchdog ( see dev_watchdog() )
1406 * @watchdog_timer: List of timers
1407 *
1408 * @pcpu_refcnt: Number of references to this device
1409 * @todo_list: Delayed register/unregister
1410 * @index_hlist: Device index hash chain
1411 * @link_watch_list: XXX: need comments on this one
1412 *
1413 * @reg_state: Register/unregister state machine
1414 * @dismantle: Device is going to be freed
1415 * @rtnl_link_state: This enum represents the phases of creating
1416 * a new link
1417 *
1418 * @destructor: Called from unregister,
1419 * can be used to call free_netdev
1420 * @npinfo: XXX: need comments on this one
1421 * @nd_net: Network namespace this network device is inside
1422 *
1423 * @ml_priv: Mid-layer private
1424 * @lstats: Loopback statistics
1425 * @tstats: Tunnel statistics
1426 * @dstats: Dummy statistics
1427 * @vstats: Virtual ethernet statistics
1428 *
1429 * @garp_port: GARP
1430 * @mrp_port: MRP
1431 *
1432 * @dev: Class/net/name entry
1433 * @sysfs_groups: Space for optional device, statistics and wireless
1434 * sysfs groups
1435 *
1436 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1437 * @rtnl_link_ops: Rtnl_link_ops
1438 *
1439 * @gso_max_size: Maximum size of generic segmentation offload
1440 * @gso_max_segs: Maximum number of segments that can be passed to the
1441 * NIC for GSO
1442 * @gso_min_segs: Minimum number of segments that can be passed to the
1443 * NIC for GSO
1444 *
1445 * @dcbnl_ops: Data Center Bridging netlink ops
1446 * @num_tc: Number of traffic classes in the net device
1447 * @tc_to_txq: XXX: need comments on this one
1448 * @prio_tc_map XXX: need comments on this one
1449 *
1450 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1451 *
1452 * @priomap: XXX: need comments on this one
1453 * @phydev: Physical device may attach itself
1454 * for hardware timestamping
1455 *
1456 * @qdisc_tx_busylock: XXX: need comments on this one
1457 *
1458 * @group: The group, that the device belongs to
1459 * @pm_qos_req: Power Management QoS object
1460 *
1461 * FIXME: cleanup struct net_device such that network protocol info
1462 * moves out.
1463 */
1464
1465 struct net_device {
1466 char name[IFNAMSIZ];
1467 struct hlist_node name_hlist;
1468 char *ifalias;
1469 /*
1470 * I/O specific fields
1471 * FIXME: Merge these and struct ifmap into one
1472 */
1473 unsigned long mem_end;
1474 unsigned long mem_start;
1475 unsigned long base_addr;
1476 int irq;
1477
1478 /*
1479 * Some hardware also needs these fields (state,dev_list,
1480 * napi_list,unreg_list,close_list) but they are not
1481 * part of the usual set specified in Space.c.
1482 */
1483
1484 unsigned long state;
1485
1486 struct list_head dev_list;
1487 struct list_head napi_list;
1488 struct list_head unreg_list;
1489 struct list_head close_list;
1490
1491 struct {
1492 struct list_head upper;
1493 struct list_head lower;
1494 } adj_list;
1495
1496 struct {
1497 struct list_head upper;
1498 struct list_head lower;
1499 } all_adj_list;
1500
1501 netdev_features_t features;
1502 netdev_features_t hw_features;
1503 netdev_features_t wanted_features;
1504 netdev_features_t vlan_features;
1505 netdev_features_t hw_enc_features;
1506 netdev_features_t mpls_features;
1507
1508 int ifindex;
1509 int iflink;
1510
1511 struct net_device_stats stats;
1512
1513 atomic_long_t rx_dropped;
1514 atomic_long_t tx_dropped;
1515
1516 atomic_t carrier_changes;
1517
1518 #ifdef CONFIG_WIRELESS_EXT
1519 const struct iw_handler_def * wireless_handlers;
1520 struct iw_public_data * wireless_data;
1521 #endif
1522 const struct net_device_ops *netdev_ops;
1523 const struct ethtool_ops *ethtool_ops;
1524 const struct forwarding_accel_ops *fwd_ops;
1525
1526 const struct header_ops *header_ops;
1527
1528 unsigned int flags;
1529 unsigned int priv_flags;
1530
1531 unsigned short gflags;
1532 unsigned short padded;
1533
1534 unsigned char operstate;
1535 unsigned char link_mode;
1536
1537 unsigned char if_port;
1538 unsigned char dma;
1539
1540 unsigned int mtu;
1541 unsigned short type;
1542 unsigned short hard_header_len;
1543
1544 unsigned short needed_headroom;
1545 unsigned short needed_tailroom;
1546
1547 /* Interface address info. */
1548 unsigned char perm_addr[MAX_ADDR_LEN];
1549 unsigned char addr_assign_type;
1550 unsigned char addr_len;
1551 unsigned short neigh_priv_len;
1552 unsigned short dev_id;
1553 unsigned short dev_port;
1554 spinlock_t addr_list_lock;
1555 struct netdev_hw_addr_list uc;
1556 struct netdev_hw_addr_list mc;
1557 struct netdev_hw_addr_list dev_addrs;
1558
1559 #ifdef CONFIG_SYSFS
1560 struct kset *queues_kset;
1561 #endif
1562
1563 unsigned char name_assign_type;
1564
1565 bool uc_promisc;
1566 unsigned int promiscuity;
1567 unsigned int allmulti;
1568
1569
1570 /* Protocol specific pointers */
1571
1572 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1573 struct vlan_info __rcu *vlan_info;
1574 #endif
1575 #if IS_ENABLED(CONFIG_NET_DSA)
1576 struct dsa_switch_tree *dsa_ptr;
1577 #endif
1578 #if IS_ENABLED(CONFIG_TIPC)
1579 struct tipc_bearer __rcu *tipc_ptr;
1580 #endif
1581 void *atalk_ptr;
1582 struct in_device __rcu *ip_ptr;
1583 struct dn_dev __rcu *dn_ptr;
1584 struct inet6_dev __rcu *ip6_ptr;
1585 void *ax25_ptr;
1586 struct wireless_dev *ieee80211_ptr;
1587
1588 /*
1589 * Cache lines mostly used on receive path (including eth_type_trans())
1590 */
1591 unsigned long last_rx;
1592
1593 /* Interface address info used in eth_type_trans() */
1594 unsigned char *dev_addr;
1595
1596
1597 #ifdef CONFIG_SYSFS
1598 struct netdev_rx_queue *_rx;
1599
1600 unsigned int num_rx_queues;
1601 unsigned int real_num_rx_queues;
1602
1603 #endif
1604
1605 rx_handler_func_t __rcu *rx_handler;
1606 void __rcu *rx_handler_data;
1607
1608 struct netdev_queue __rcu *ingress_queue;
1609 unsigned char broadcast[MAX_ADDR_LEN];
1610
1611
1612 /*
1613 * Cache lines mostly used on transmit path
1614 */
1615 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1616 unsigned int num_tx_queues;
1617 unsigned int real_num_tx_queues;
1618 struct Qdisc *qdisc;
1619 unsigned long tx_queue_len;
1620 spinlock_t tx_global_lock;
1621
1622 #ifdef CONFIG_XPS
1623 struct xps_dev_maps __rcu *xps_maps;
1624 #endif
1625 #ifdef CONFIG_RFS_ACCEL
1626 struct cpu_rmap *rx_cpu_rmap;
1627 #endif
1628
1629 /* These may be needed for future network-power-down code. */
1630
1631 /*
1632 * trans_start here is expensive for high speed devices on SMP,
1633 * please use netdev_queue->trans_start instead.
1634 */
1635 unsigned long trans_start;
1636
1637 int watchdog_timeo;
1638 struct timer_list watchdog_timer;
1639
1640 int __percpu *pcpu_refcnt;
1641 struct list_head todo_list;
1642
1643 struct hlist_node index_hlist;
1644 struct list_head link_watch_list;
1645
1646 enum { NETREG_UNINITIALIZED=0,
1647 NETREG_REGISTERED, /* completed register_netdevice */
1648 NETREG_UNREGISTERING, /* called unregister_netdevice */
1649 NETREG_UNREGISTERED, /* completed unregister todo */
1650 NETREG_RELEASED, /* called free_netdev */
1651 NETREG_DUMMY, /* dummy device for NAPI poll */
1652 } reg_state:8;
1653
1654 bool dismantle;
1655
1656 enum {
1657 RTNL_LINK_INITIALIZED,
1658 RTNL_LINK_INITIALIZING,
1659 } rtnl_link_state:16;
1660
1661 void (*destructor)(struct net_device *dev);
1662
1663 #ifdef CONFIG_NETPOLL
1664 struct netpoll_info __rcu *npinfo;
1665 #endif
1666
1667 #ifdef CONFIG_NET_NS
1668 struct net *nd_net;
1669 #endif
1670
1671 /* mid-layer private */
1672 union {
1673 void *ml_priv;
1674 struct pcpu_lstats __percpu *lstats;
1675 struct pcpu_sw_netstats __percpu *tstats;
1676 struct pcpu_dstats __percpu *dstats;
1677 struct pcpu_vstats __percpu *vstats;
1678 };
1679
1680 struct garp_port __rcu *garp_port;
1681 struct mrp_port __rcu *mrp_port;
1682
1683 struct device dev;
1684 const struct attribute_group *sysfs_groups[4];
1685 const struct attribute_group *sysfs_rx_queue_group;
1686
1687 const struct rtnl_link_ops *rtnl_link_ops;
1688
1689 /* for setting kernel sock attribute on TCP connection setup */
1690 #define GSO_MAX_SIZE 65536
1691 unsigned int gso_max_size;
1692 #define GSO_MAX_SEGS 65535
1693 u16 gso_max_segs;
1694 u16 gso_min_segs;
1695 #ifdef CONFIG_DCB
1696 const struct dcbnl_rtnl_ops *dcbnl_ops;
1697 #endif
1698 u8 num_tc;
1699 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1700 u8 prio_tc_map[TC_BITMASK + 1];
1701
1702 #if IS_ENABLED(CONFIG_FCOE)
1703 unsigned int fcoe_ddp_xid;
1704 #endif
1705 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1706 struct netprio_map __rcu *priomap;
1707 #endif
1708 struct phy_device *phydev;
1709 struct lock_class_key *qdisc_tx_busylock;
1710 int group;
1711 struct pm_qos_request pm_qos_req;
1712 };
1713 #define to_net_dev(d) container_of(d, struct net_device, dev)
1714
1715 #define NETDEV_ALIGN 32
1716
1717 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)1718 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1719 {
1720 return dev->prio_tc_map[prio & TC_BITMASK];
1721 }
1722
1723 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)1724 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1725 {
1726 if (tc >= dev->num_tc)
1727 return -EINVAL;
1728
1729 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1730 return 0;
1731 }
1732
1733 static inline
netdev_reset_tc(struct net_device * dev)1734 void netdev_reset_tc(struct net_device *dev)
1735 {
1736 dev->num_tc = 0;
1737 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1738 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1739 }
1740
1741 static inline
netdev_set_tc_queue(struct net_device * dev,u8 tc,u16 count,u16 offset)1742 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1743 {
1744 if (tc >= dev->num_tc)
1745 return -EINVAL;
1746
1747 dev->tc_to_txq[tc].count = count;
1748 dev->tc_to_txq[tc].offset = offset;
1749 return 0;
1750 }
1751
1752 static inline
netdev_set_num_tc(struct net_device * dev,u8 num_tc)1753 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1754 {
1755 if (num_tc > TC_MAX_QUEUE)
1756 return -EINVAL;
1757
1758 dev->num_tc = num_tc;
1759 return 0;
1760 }
1761
1762 static inline
netdev_get_num_tc(struct net_device * dev)1763 int netdev_get_num_tc(struct net_device *dev)
1764 {
1765 return dev->num_tc;
1766 }
1767
1768 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)1769 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1770 unsigned int index)
1771 {
1772 return &dev->_tx[index];
1773 }
1774
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)1775 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1776 const struct sk_buff *skb)
1777 {
1778 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1779 }
1780
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)1781 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1782 void (*f)(struct net_device *,
1783 struct netdev_queue *,
1784 void *),
1785 void *arg)
1786 {
1787 unsigned int i;
1788
1789 for (i = 0; i < dev->num_tx_queues; i++)
1790 f(dev, &dev->_tx[i], arg);
1791 }
1792
1793 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1794 struct sk_buff *skb,
1795 void *accel_priv);
1796
1797 /*
1798 * Net namespace inlines
1799 */
1800 static inline
dev_net(const struct net_device * dev)1801 struct net *dev_net(const struct net_device *dev)
1802 {
1803 return read_pnet(&dev->nd_net);
1804 }
1805
1806 static inline
dev_net_set(struct net_device * dev,struct net * net)1807 void dev_net_set(struct net_device *dev, struct net *net)
1808 {
1809 #ifdef CONFIG_NET_NS
1810 release_net(dev->nd_net);
1811 dev->nd_net = hold_net(net);
1812 #endif
1813 }
1814
netdev_uses_dsa(struct net_device * dev)1815 static inline bool netdev_uses_dsa(struct net_device *dev)
1816 {
1817 #if IS_ENABLED(CONFIG_NET_DSA)
1818 if (dev->dsa_ptr != NULL)
1819 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1820 #endif
1821 return false;
1822 }
1823
1824 /**
1825 * netdev_priv - access network device private data
1826 * @dev: network device
1827 *
1828 * Get network device private data
1829 */
netdev_priv(const struct net_device * dev)1830 static inline void *netdev_priv(const struct net_device *dev)
1831 {
1832 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1833 }
1834
1835 /* Set the sysfs physical device reference for the network logical device
1836 * if set prior to registration will cause a symlink during initialization.
1837 */
1838 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1839
1840 /* Set the sysfs device type for the network logical device to allow
1841 * fine-grained identification of different network device types. For
1842 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1843 */
1844 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1845
1846 /* Default NAPI poll() weight
1847 * Device drivers are strongly advised to not use bigger value
1848 */
1849 #define NAPI_POLL_WEIGHT 64
1850
1851 /**
1852 * netif_napi_add - initialize a napi context
1853 * @dev: network device
1854 * @napi: napi context
1855 * @poll: polling function
1856 * @weight: default weight
1857 *
1858 * netif_napi_add() must be used to initialize a napi context prior to calling
1859 * *any* of the other napi related functions.
1860 */
1861 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1862 int (*poll)(struct napi_struct *, int), int weight);
1863
1864 /**
1865 * netif_napi_del - remove a napi context
1866 * @napi: napi context
1867 *
1868 * netif_napi_del() removes a napi context from the network device napi list
1869 */
1870 void netif_napi_del(struct napi_struct *napi);
1871
1872 struct napi_gro_cb {
1873 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1874 void *frag0;
1875
1876 /* Length of frag0. */
1877 unsigned int frag0_len;
1878
1879 /* This indicates where we are processing relative to skb->data. */
1880 int data_offset;
1881
1882 /* This is non-zero if the packet cannot be merged with the new skb. */
1883 u16 flush;
1884
1885 /* Save the IP ID here and check when we get to the transport layer */
1886 u16 flush_id;
1887
1888 /* Number of segments aggregated. */
1889 u16 count;
1890
1891 /* jiffies when first packet was created/queued */
1892 unsigned long age;
1893
1894 /* Used in ipv6_gro_receive() and foo-over-udp */
1895 u16 proto;
1896
1897 /* This is non-zero if the packet may be of the same flow. */
1898 u8 same_flow:1;
1899
1900 /* Used in tunnel GRO receive */
1901 u8 encap_mark:1;
1902
1903 /* GRO checksum is valid */
1904 u8 csum_valid:1;
1905
1906 /* Number of checksums via CHECKSUM_UNNECESSARY */
1907 u8 csum_cnt:3;
1908
1909 /* Free the skb? */
1910 u8 free:2;
1911 #define NAPI_GRO_FREE 1
1912 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1913
1914 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1915 u8 is_ipv6:1;
1916
1917 /* 7 bit hole */
1918
1919 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1920 __wsum csum;
1921
1922 /* used in skb_gro_receive() slow path */
1923 struct sk_buff *last;
1924 };
1925
1926 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1927
1928 struct packet_type {
1929 __be16 type; /* This is really htons(ether_type). */
1930 struct net_device *dev; /* NULL is wildcarded here */
1931 int (*func) (struct sk_buff *,
1932 struct net_device *,
1933 struct packet_type *,
1934 struct net_device *);
1935 bool (*id_match)(struct packet_type *ptype,
1936 struct sock *sk);
1937 void *af_packet_priv;
1938 struct list_head list;
1939 };
1940
1941 struct offload_callbacks {
1942 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1943 netdev_features_t features);
1944 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1945 struct sk_buff *skb);
1946 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1947 };
1948
1949 struct packet_offload {
1950 __be16 type; /* This is really htons(ether_type). */
1951 struct offload_callbacks callbacks;
1952 struct list_head list;
1953 };
1954
1955 struct udp_offload {
1956 __be16 port;
1957 u8 ipproto;
1958 struct offload_callbacks callbacks;
1959 };
1960
1961 /* often modified stats are per cpu, other are shared (netdev->stats) */
1962 struct pcpu_sw_netstats {
1963 u64 rx_packets;
1964 u64 rx_bytes;
1965 u64 tx_packets;
1966 u64 tx_bytes;
1967 struct u64_stats_sync syncp;
1968 };
1969
1970 #define netdev_alloc_pcpu_stats(type) \
1971 ({ \
1972 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1973 if (pcpu_stats) { \
1974 int i; \
1975 for_each_possible_cpu(i) { \
1976 typeof(type) *stat; \
1977 stat = per_cpu_ptr(pcpu_stats, i); \
1978 u64_stats_init(&stat->syncp); \
1979 } \
1980 } \
1981 pcpu_stats; \
1982 })
1983
1984 #include <linux/notifier.h>
1985
1986 /* netdevice notifier chain. Please remember to update the rtnetlink
1987 * notification exclusion list in rtnetlink_event() when adding new
1988 * types.
1989 */
1990 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1991 #define NETDEV_DOWN 0x0002
1992 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1993 detected a hardware crash and restarted
1994 - we can use this eg to kick tcp sessions
1995 once done */
1996 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1997 #define NETDEV_REGISTER 0x0005
1998 #define NETDEV_UNREGISTER 0x0006
1999 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2000 #define NETDEV_CHANGEADDR 0x0008
2001 #define NETDEV_GOING_DOWN 0x0009
2002 #define NETDEV_CHANGENAME 0x000A
2003 #define NETDEV_FEAT_CHANGE 0x000B
2004 #define NETDEV_BONDING_FAILOVER 0x000C
2005 #define NETDEV_PRE_UP 0x000D
2006 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2007 #define NETDEV_POST_TYPE_CHANGE 0x000F
2008 #define NETDEV_POST_INIT 0x0010
2009 #define NETDEV_UNREGISTER_FINAL 0x0011
2010 #define NETDEV_RELEASE 0x0012
2011 #define NETDEV_NOTIFY_PEERS 0x0013
2012 #define NETDEV_JOIN 0x0014
2013 #define NETDEV_CHANGEUPPER 0x0015
2014 #define NETDEV_RESEND_IGMP 0x0016
2015 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2016 #define NETDEV_CHANGEINFODATA 0x0018
2017
2018 int register_netdevice_notifier(struct notifier_block *nb);
2019 int unregister_netdevice_notifier(struct notifier_block *nb);
2020
2021 struct netdev_notifier_info {
2022 struct net_device *dev;
2023 };
2024
2025 struct netdev_notifier_change_info {
2026 struct netdev_notifier_info info; /* must be first */
2027 unsigned int flags_changed;
2028 };
2029
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2030 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2031 struct net_device *dev)
2032 {
2033 info->dev = dev;
2034 }
2035
2036 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2037 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2038 {
2039 return info->dev;
2040 }
2041
2042 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2043
2044
2045 extern rwlock_t dev_base_lock; /* Device list lock */
2046
2047 #define for_each_netdev(net, d) \
2048 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2049 #define for_each_netdev_reverse(net, d) \
2050 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2051 #define for_each_netdev_rcu(net, d) \
2052 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2053 #define for_each_netdev_safe(net, d, n) \
2054 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2055 #define for_each_netdev_continue(net, d) \
2056 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2057 #define for_each_netdev_continue_rcu(net, d) \
2058 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2059 #define for_each_netdev_in_bond_rcu(bond, slave) \
2060 for_each_netdev_rcu(&init_net, slave) \
2061 if (netdev_master_upper_dev_get_rcu(slave) == bond)
2062 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2063
next_net_device(struct net_device * dev)2064 static inline struct net_device *next_net_device(struct net_device *dev)
2065 {
2066 struct list_head *lh;
2067 struct net *net;
2068
2069 net = dev_net(dev);
2070 lh = dev->dev_list.next;
2071 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2072 }
2073
next_net_device_rcu(struct net_device * dev)2074 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2075 {
2076 struct list_head *lh;
2077 struct net *net;
2078
2079 net = dev_net(dev);
2080 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2081 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2082 }
2083
first_net_device(struct net * net)2084 static inline struct net_device *first_net_device(struct net *net)
2085 {
2086 return list_empty(&net->dev_base_head) ? NULL :
2087 net_device_entry(net->dev_base_head.next);
2088 }
2089
first_net_device_rcu(struct net * net)2090 static inline struct net_device *first_net_device_rcu(struct net *net)
2091 {
2092 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2093
2094 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2095 }
2096
2097 int netdev_boot_setup_check(struct net_device *dev);
2098 unsigned long netdev_boot_base(const char *prefix, int unit);
2099 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2100 const char *hwaddr);
2101 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2102 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2103 void dev_add_pack(struct packet_type *pt);
2104 void dev_remove_pack(struct packet_type *pt);
2105 void __dev_remove_pack(struct packet_type *pt);
2106 void dev_add_offload(struct packet_offload *po);
2107 void dev_remove_offload(struct packet_offload *po);
2108
2109 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2110 unsigned short mask);
2111 struct net_device *dev_get_by_name(struct net *net, const char *name);
2112 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2113 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2114 int dev_alloc_name(struct net_device *dev, const char *name);
2115 int dev_open(struct net_device *dev);
2116 int dev_close(struct net_device *dev);
2117 void dev_disable_lro(struct net_device *dev);
2118 int dev_loopback_xmit(struct sk_buff *newskb);
2119 int dev_queue_xmit(struct sk_buff *skb);
2120 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2121 int register_netdevice(struct net_device *dev);
2122 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2123 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2124 static inline void unregister_netdevice(struct net_device *dev)
2125 {
2126 unregister_netdevice_queue(dev, NULL);
2127 }
2128
2129 int netdev_refcnt_read(const struct net_device *dev);
2130 void free_netdev(struct net_device *dev);
2131 void netdev_freemem(struct net_device *dev);
2132 void synchronize_net(void);
2133 int init_dummy_netdev(struct net_device *dev);
2134
2135 DECLARE_PER_CPU(int, xmit_recursion);
dev_recursion_level(void)2136 static inline int dev_recursion_level(void)
2137 {
2138 return this_cpu_read(xmit_recursion);
2139 }
2140
2141 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2142 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2143 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2144 int netdev_get_name(struct net *net, char *name, int ifindex);
2145 int dev_restart(struct net_device *dev);
2146 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2147
skb_gro_offset(const struct sk_buff * skb)2148 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2149 {
2150 return NAPI_GRO_CB(skb)->data_offset;
2151 }
2152
skb_gro_len(const struct sk_buff * skb)2153 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2154 {
2155 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2156 }
2157
skb_gro_pull(struct sk_buff * skb,unsigned int len)2158 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2159 {
2160 NAPI_GRO_CB(skb)->data_offset += len;
2161 }
2162
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2163 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2164 unsigned int offset)
2165 {
2166 return NAPI_GRO_CB(skb)->frag0 + offset;
2167 }
2168
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2169 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2170 {
2171 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2172 }
2173
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2174 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2175 unsigned int offset)
2176 {
2177 if (!pskb_may_pull(skb, hlen))
2178 return NULL;
2179
2180 NAPI_GRO_CB(skb)->frag0 = NULL;
2181 NAPI_GRO_CB(skb)->frag0_len = 0;
2182 return skb->data + offset;
2183 }
2184
skb_gro_network_header(struct sk_buff * skb)2185 static inline void *skb_gro_network_header(struct sk_buff *skb)
2186 {
2187 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2188 skb_network_offset(skb);
2189 }
2190
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2191 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2192 const void *start, unsigned int len)
2193 {
2194 if (NAPI_GRO_CB(skb)->csum_valid)
2195 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2196 csum_partial(start, len, 0));
2197 }
2198
2199 /* GRO checksum functions. These are logical equivalents of the normal
2200 * checksum functions (in skbuff.h) except that they operate on the GRO
2201 * offsets and fields in sk_buff.
2202 */
2203
2204 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2205
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2206 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2207 bool zero_okay,
2208 __sum16 check)
2209 {
2210 return (skb->ip_summed != CHECKSUM_PARTIAL &&
2211 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2212 (!zero_okay || check));
2213 }
2214
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2215 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2216 __wsum psum)
2217 {
2218 if (NAPI_GRO_CB(skb)->csum_valid &&
2219 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2220 return 0;
2221
2222 NAPI_GRO_CB(skb)->csum = psum;
2223
2224 return __skb_gro_checksum_complete(skb);
2225 }
2226
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)2227 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2228 {
2229 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2230 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2231 NAPI_GRO_CB(skb)->csum_cnt--;
2232 } else {
2233 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2234 * verified a new top level checksum or an encapsulated one
2235 * during GRO. This saves work if we fallback to normal path.
2236 */
2237 __skb_incr_checksum_unnecessary(skb);
2238 }
2239 }
2240
2241 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2242 compute_pseudo) \
2243 ({ \
2244 __sum16 __ret = 0; \
2245 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2246 __ret = __skb_gro_checksum_validate_complete(skb, \
2247 compute_pseudo(skb, proto)); \
2248 if (__ret) \
2249 __skb_mark_checksum_bad(skb); \
2250 else \
2251 skb_gro_incr_csum_unnecessary(skb); \
2252 __ret; \
2253 })
2254
2255 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2256 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2257
2258 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2259 compute_pseudo) \
2260 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2261
2262 #define skb_gro_checksum_simple_validate(skb) \
2263 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2264
__skb_gro_checksum_convert_check(struct sk_buff * skb)2265 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2266 {
2267 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2268 !NAPI_GRO_CB(skb)->csum_valid);
2269 }
2270
__skb_gro_checksum_convert(struct sk_buff * skb,__sum16 check,__wsum pseudo)2271 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2272 __sum16 check, __wsum pseudo)
2273 {
2274 NAPI_GRO_CB(skb)->csum = ~pseudo;
2275 NAPI_GRO_CB(skb)->csum_valid = 1;
2276 }
2277
2278 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2279 do { \
2280 if (__skb_gro_checksum_convert_check(skb)) \
2281 __skb_gro_checksum_convert(skb, check, \
2282 compute_pseudo(skb, proto)); \
2283 } while (0)
2284
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)2285 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2286 unsigned short type,
2287 const void *daddr, const void *saddr,
2288 unsigned int len)
2289 {
2290 if (!dev->header_ops || !dev->header_ops->create)
2291 return 0;
2292
2293 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2294 }
2295
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)2296 static inline int dev_parse_header(const struct sk_buff *skb,
2297 unsigned char *haddr)
2298 {
2299 const struct net_device *dev = skb->dev;
2300
2301 if (!dev->header_ops || !dev->header_ops->parse)
2302 return 0;
2303 return dev->header_ops->parse(skb, haddr);
2304 }
2305
dev_rebuild_header(struct sk_buff * skb)2306 static inline int dev_rebuild_header(struct sk_buff *skb)
2307 {
2308 const struct net_device *dev = skb->dev;
2309
2310 if (!dev->header_ops || !dev->header_ops->rebuild)
2311 return 0;
2312 return dev->header_ops->rebuild(skb);
2313 }
2314
2315 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2316 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
unregister_gifconf(unsigned int family)2317 static inline int unregister_gifconf(unsigned int family)
2318 {
2319 return register_gifconf(family, NULL);
2320 }
2321
2322 #ifdef CONFIG_NET_FLOW_LIMIT
2323 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2324 struct sd_flow_limit {
2325 u64 count;
2326 unsigned int num_buckets;
2327 unsigned int history_head;
2328 u16 history[FLOW_LIMIT_HISTORY];
2329 u8 buckets[];
2330 };
2331
2332 extern int netdev_flow_limit_table_len;
2333 #endif /* CONFIG_NET_FLOW_LIMIT */
2334
2335 /*
2336 * Incoming packets are placed on per-cpu queues
2337 */
2338 struct softnet_data {
2339 struct Qdisc *output_queue;
2340 struct Qdisc **output_queue_tailp;
2341 struct list_head poll_list;
2342 struct sk_buff *completion_queue;
2343 struct sk_buff_head process_queue;
2344
2345 /* stats */
2346 unsigned int processed;
2347 unsigned int time_squeeze;
2348 unsigned int cpu_collision;
2349 unsigned int received_rps;
2350
2351 #ifdef CONFIG_RPS
2352 struct softnet_data *rps_ipi_list;
2353
2354 /* Elements below can be accessed between CPUs for RPS */
2355 struct call_single_data csd ____cacheline_aligned_in_smp;
2356 struct softnet_data *rps_ipi_next;
2357 unsigned int cpu;
2358 unsigned int input_queue_head;
2359 unsigned int input_queue_tail;
2360 #endif
2361 unsigned int dropped;
2362 struct sk_buff_head input_pkt_queue;
2363 struct napi_struct backlog;
2364
2365 #ifdef CONFIG_NET_FLOW_LIMIT
2366 struct sd_flow_limit __rcu *flow_limit;
2367 #endif
2368 };
2369
input_queue_head_incr(struct softnet_data * sd)2370 static inline void input_queue_head_incr(struct softnet_data *sd)
2371 {
2372 #ifdef CONFIG_RPS
2373 sd->input_queue_head++;
2374 #endif
2375 }
2376
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)2377 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2378 unsigned int *qtail)
2379 {
2380 #ifdef CONFIG_RPS
2381 *qtail = ++sd->input_queue_tail;
2382 #endif
2383 }
2384
2385 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2386
2387 void __netif_schedule(struct Qdisc *q);
2388 void netif_schedule_queue(struct netdev_queue *txq);
2389
netif_tx_schedule_all(struct net_device * dev)2390 static inline void netif_tx_schedule_all(struct net_device *dev)
2391 {
2392 unsigned int i;
2393
2394 for (i = 0; i < dev->num_tx_queues; i++)
2395 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2396 }
2397
netif_tx_start_queue(struct netdev_queue * dev_queue)2398 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2399 {
2400 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2401 }
2402
2403 /**
2404 * netif_start_queue - allow transmit
2405 * @dev: network device
2406 *
2407 * Allow upper layers to call the device hard_start_xmit routine.
2408 */
netif_start_queue(struct net_device * dev)2409 static inline void netif_start_queue(struct net_device *dev)
2410 {
2411 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2412 }
2413
netif_tx_start_all_queues(struct net_device * dev)2414 static inline void netif_tx_start_all_queues(struct net_device *dev)
2415 {
2416 unsigned int i;
2417
2418 for (i = 0; i < dev->num_tx_queues; i++) {
2419 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2420 netif_tx_start_queue(txq);
2421 }
2422 }
2423
2424 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2425
2426 /**
2427 * netif_wake_queue - restart transmit
2428 * @dev: network device
2429 *
2430 * Allow upper layers to call the device hard_start_xmit routine.
2431 * Used for flow control when transmit resources are available.
2432 */
netif_wake_queue(struct net_device * dev)2433 static inline void netif_wake_queue(struct net_device *dev)
2434 {
2435 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2436 }
2437
netif_tx_wake_all_queues(struct net_device * dev)2438 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2439 {
2440 unsigned int i;
2441
2442 for (i = 0; i < dev->num_tx_queues; i++) {
2443 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2444 netif_tx_wake_queue(txq);
2445 }
2446 }
2447
netif_tx_stop_queue(struct netdev_queue * dev_queue)2448 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2449 {
2450 if (WARN_ON(!dev_queue)) {
2451 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2452 return;
2453 }
2454 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2455 }
2456
2457 /**
2458 * netif_stop_queue - stop transmitted packets
2459 * @dev: network device
2460 *
2461 * Stop upper layers calling the device hard_start_xmit routine.
2462 * Used for flow control when transmit resources are unavailable.
2463 */
netif_stop_queue(struct net_device * dev)2464 static inline void netif_stop_queue(struct net_device *dev)
2465 {
2466 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2467 }
2468
netif_tx_stop_all_queues(struct net_device * dev)2469 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2470 {
2471 unsigned int i;
2472
2473 for (i = 0; i < dev->num_tx_queues; i++) {
2474 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2475 netif_tx_stop_queue(txq);
2476 }
2477 }
2478
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)2479 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2480 {
2481 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2482 }
2483
2484 /**
2485 * netif_queue_stopped - test if transmit queue is flowblocked
2486 * @dev: network device
2487 *
2488 * Test if transmit queue on device is currently unable to send.
2489 */
netif_queue_stopped(const struct net_device * dev)2490 static inline bool netif_queue_stopped(const struct net_device *dev)
2491 {
2492 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2493 }
2494
netif_xmit_stopped(const struct netdev_queue * dev_queue)2495 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2496 {
2497 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2498 }
2499
2500 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)2501 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2502 {
2503 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2504 }
2505
2506 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)2507 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2508 {
2509 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2510 }
2511
2512 /**
2513 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2514 * @dev_queue: pointer to transmit queue
2515 *
2516 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2517 * to give appropriate hint to the cpu.
2518 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)2519 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2520 {
2521 #ifdef CONFIG_BQL
2522 prefetchw(&dev_queue->dql.num_queued);
2523 #endif
2524 }
2525
2526 /**
2527 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2528 * @dev_queue: pointer to transmit queue
2529 *
2530 * BQL enabled drivers might use this helper in their TX completion path,
2531 * to give appropriate hint to the cpu.
2532 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)2533 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2534 {
2535 #ifdef CONFIG_BQL
2536 prefetchw(&dev_queue->dql.limit);
2537 #endif
2538 }
2539
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)2540 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2541 unsigned int bytes)
2542 {
2543 #ifdef CONFIG_BQL
2544 dql_queued(&dev_queue->dql, bytes);
2545
2546 if (likely(dql_avail(&dev_queue->dql) >= 0))
2547 return;
2548
2549 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2550
2551 /*
2552 * The XOFF flag must be set before checking the dql_avail below,
2553 * because in netdev_tx_completed_queue we update the dql_completed
2554 * before checking the XOFF flag.
2555 */
2556 smp_mb();
2557
2558 /* check again in case another CPU has just made room avail */
2559 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2560 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2561 #endif
2562 }
2563
2564 /**
2565 * netdev_sent_queue - report the number of bytes queued to hardware
2566 * @dev: network device
2567 * @bytes: number of bytes queued to the hardware device queue
2568 *
2569 * Report the number of bytes queued for sending/completion to the network
2570 * device hardware queue. @bytes should be a good approximation and should
2571 * exactly match netdev_completed_queue() @bytes
2572 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)2573 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2574 {
2575 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2576 }
2577
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)2578 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2579 unsigned int pkts, unsigned int bytes)
2580 {
2581 #ifdef CONFIG_BQL
2582 if (unlikely(!bytes))
2583 return;
2584
2585 dql_completed(&dev_queue->dql, bytes);
2586
2587 /*
2588 * Without the memory barrier there is a small possiblity that
2589 * netdev_tx_sent_queue will miss the update and cause the queue to
2590 * be stopped forever
2591 */
2592 smp_mb();
2593
2594 if (dql_avail(&dev_queue->dql) < 0)
2595 return;
2596
2597 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2598 netif_schedule_queue(dev_queue);
2599 #endif
2600 }
2601
2602 /**
2603 * netdev_completed_queue - report bytes and packets completed by device
2604 * @dev: network device
2605 * @pkts: actual number of packets sent over the medium
2606 * @bytes: actual number of bytes sent over the medium
2607 *
2608 * Report the number of bytes and packets transmitted by the network device
2609 * hardware queue over the physical medium, @bytes must exactly match the
2610 * @bytes amount passed to netdev_sent_queue()
2611 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)2612 static inline void netdev_completed_queue(struct net_device *dev,
2613 unsigned int pkts, unsigned int bytes)
2614 {
2615 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2616 }
2617
netdev_tx_reset_queue(struct netdev_queue * q)2618 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2619 {
2620 #ifdef CONFIG_BQL
2621 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2622 dql_reset(&q->dql);
2623 #endif
2624 }
2625
2626 /**
2627 * netdev_reset_queue - reset the packets and bytes count of a network device
2628 * @dev_queue: network device
2629 *
2630 * Reset the bytes and packet count of a network device and clear the
2631 * software flow control OFF bit for this network device
2632 */
netdev_reset_queue(struct net_device * dev_queue)2633 static inline void netdev_reset_queue(struct net_device *dev_queue)
2634 {
2635 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2636 }
2637
2638 /**
2639 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2640 * @dev: network device
2641 * @queue_index: given tx queue index
2642 *
2643 * Returns 0 if given tx queue index >= number of device tx queues,
2644 * otherwise returns the originally passed tx queue index.
2645 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)2646 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2647 {
2648 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2649 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2650 dev->name, queue_index,
2651 dev->real_num_tx_queues);
2652 return 0;
2653 }
2654
2655 return queue_index;
2656 }
2657
2658 /**
2659 * netif_running - test if up
2660 * @dev: network device
2661 *
2662 * Test if the device has been brought up.
2663 */
netif_running(const struct net_device * dev)2664 static inline bool netif_running(const struct net_device *dev)
2665 {
2666 return test_bit(__LINK_STATE_START, &dev->state);
2667 }
2668
2669 /*
2670 * Routines to manage the subqueues on a device. We only need start
2671 * stop, and a check if it's stopped. All other device management is
2672 * done at the overall netdevice level.
2673 * Also test the device if we're multiqueue.
2674 */
2675
2676 /**
2677 * netif_start_subqueue - allow sending packets on subqueue
2678 * @dev: network device
2679 * @queue_index: sub queue index
2680 *
2681 * Start individual transmit queue of a device with multiple transmit queues.
2682 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)2683 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2684 {
2685 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2686
2687 netif_tx_start_queue(txq);
2688 }
2689
2690 /**
2691 * netif_stop_subqueue - stop sending packets on subqueue
2692 * @dev: network device
2693 * @queue_index: sub queue index
2694 *
2695 * Stop individual transmit queue of a device with multiple transmit queues.
2696 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)2697 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2698 {
2699 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2700 netif_tx_stop_queue(txq);
2701 }
2702
2703 /**
2704 * netif_subqueue_stopped - test status of subqueue
2705 * @dev: network device
2706 * @queue_index: sub queue index
2707 *
2708 * Check individual transmit queue of a device with multiple transmit queues.
2709 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)2710 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2711 u16 queue_index)
2712 {
2713 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2714
2715 return netif_tx_queue_stopped(txq);
2716 }
2717
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)2718 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2719 struct sk_buff *skb)
2720 {
2721 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2722 }
2723
2724 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2725
2726 #ifdef CONFIG_XPS
2727 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2728 u16 index);
2729 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)2730 static inline int netif_set_xps_queue(struct net_device *dev,
2731 const struct cpumask *mask,
2732 u16 index)
2733 {
2734 return 0;
2735 }
2736 #endif
2737
2738 /*
2739 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2740 * as a distribution range limit for the returned value.
2741 */
skb_tx_hash(const struct net_device * dev,struct sk_buff * skb)2742 static inline u16 skb_tx_hash(const struct net_device *dev,
2743 struct sk_buff *skb)
2744 {
2745 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2746 }
2747
2748 /**
2749 * netif_is_multiqueue - test if device has multiple transmit queues
2750 * @dev: network device
2751 *
2752 * Check if device has multiple transmit queues
2753 */
netif_is_multiqueue(const struct net_device * dev)2754 static inline bool netif_is_multiqueue(const struct net_device *dev)
2755 {
2756 return dev->num_tx_queues > 1;
2757 }
2758
2759 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2760
2761 #ifdef CONFIG_SYSFS
2762 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2763 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxq)2764 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2765 unsigned int rxq)
2766 {
2767 return 0;
2768 }
2769 #endif
2770
netif_copy_real_num_queues(struct net_device * to_dev,const struct net_device * from_dev)2771 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2772 const struct net_device *from_dev)
2773 {
2774 int err;
2775
2776 err = netif_set_real_num_tx_queues(to_dev,
2777 from_dev->real_num_tx_queues);
2778 if (err)
2779 return err;
2780 #ifdef CONFIG_SYSFS
2781 return netif_set_real_num_rx_queues(to_dev,
2782 from_dev->real_num_rx_queues);
2783 #else
2784 return 0;
2785 #endif
2786 }
2787
2788 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)2789 static inline unsigned int get_netdev_rx_queue_index(
2790 struct netdev_rx_queue *queue)
2791 {
2792 struct net_device *dev = queue->dev;
2793 int index = queue - dev->_rx;
2794
2795 BUG_ON(index >= dev->num_rx_queues);
2796 return index;
2797 }
2798 #endif
2799
2800 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2801 int netif_get_num_default_rss_queues(void);
2802
2803 enum skb_free_reason {
2804 SKB_REASON_CONSUMED,
2805 SKB_REASON_DROPPED,
2806 };
2807
2808 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2809 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2810
2811 /*
2812 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2813 * interrupt context or with hardware interrupts being disabled.
2814 * (in_irq() || irqs_disabled())
2815 *
2816 * We provide four helpers that can be used in following contexts :
2817 *
2818 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2819 * replacing kfree_skb(skb)
2820 *
2821 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2822 * Typically used in place of consume_skb(skb) in TX completion path
2823 *
2824 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2825 * replacing kfree_skb(skb)
2826 *
2827 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2828 * and consumed a packet. Used in place of consume_skb(skb)
2829 */
dev_kfree_skb_irq(struct sk_buff * skb)2830 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2831 {
2832 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2833 }
2834
dev_consume_skb_irq(struct sk_buff * skb)2835 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2836 {
2837 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2838 }
2839
dev_kfree_skb_any(struct sk_buff * skb)2840 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2841 {
2842 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2843 }
2844
dev_consume_skb_any(struct sk_buff * skb)2845 static inline void dev_consume_skb_any(struct sk_buff *skb)
2846 {
2847 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2848 }
2849
2850 int netif_rx(struct sk_buff *skb);
2851 int netif_rx_ni(struct sk_buff *skb);
2852 int netif_receive_skb(struct sk_buff *skb);
2853 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2854 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2855 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2856 gro_result_t napi_gro_frags(struct napi_struct *napi);
2857 struct packet_offload *gro_find_receive_by_type(__be16 type);
2858 struct packet_offload *gro_find_complete_by_type(__be16 type);
2859
napi_free_frags(struct napi_struct * napi)2860 static inline void napi_free_frags(struct napi_struct *napi)
2861 {
2862 kfree_skb(napi->skb);
2863 napi->skb = NULL;
2864 }
2865
2866 int netdev_rx_handler_register(struct net_device *dev,
2867 rx_handler_func_t *rx_handler,
2868 void *rx_handler_data);
2869 void netdev_rx_handler_unregister(struct net_device *dev);
2870
2871 bool dev_valid_name(const char *name);
2872 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2873 int dev_ethtool(struct net *net, struct ifreq *);
2874 unsigned int dev_get_flags(const struct net_device *);
2875 int __dev_change_flags(struct net_device *, unsigned int flags);
2876 int dev_change_flags(struct net_device *, unsigned int);
2877 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2878 unsigned int gchanges);
2879 int dev_change_name(struct net_device *, const char *);
2880 int dev_set_alias(struct net_device *, const char *, size_t);
2881 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2882 int dev_set_mtu(struct net_device *, int);
2883 void dev_set_group(struct net_device *, int);
2884 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2885 int dev_change_carrier(struct net_device *, bool new_carrier);
2886 int dev_get_phys_port_id(struct net_device *dev,
2887 struct netdev_phys_port_id *ppid);
2888 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2889 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2890 struct netdev_queue *txq, int *ret);
2891 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2892 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2893 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2894
2895 extern int netdev_budget;
2896
2897 /* Called by rtnetlink.c:rtnl_unlock() */
2898 void netdev_run_todo(void);
2899
2900 /**
2901 * dev_put - release reference to device
2902 * @dev: network device
2903 *
2904 * Release reference to device to allow it to be freed.
2905 */
dev_put(struct net_device * dev)2906 static inline void dev_put(struct net_device *dev)
2907 {
2908 this_cpu_dec(*dev->pcpu_refcnt);
2909 }
2910
2911 /**
2912 * dev_hold - get reference to device
2913 * @dev: network device
2914 *
2915 * Hold reference to device to keep it from being freed.
2916 */
dev_hold(struct net_device * dev)2917 static inline void dev_hold(struct net_device *dev)
2918 {
2919 this_cpu_inc(*dev->pcpu_refcnt);
2920 }
2921
2922 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2923 * and _off may be called from IRQ context, but it is caller
2924 * who is responsible for serialization of these calls.
2925 *
2926 * The name carrier is inappropriate, these functions should really be
2927 * called netif_lowerlayer_*() because they represent the state of any
2928 * kind of lower layer not just hardware media.
2929 */
2930
2931 void linkwatch_init_dev(struct net_device *dev);
2932 void linkwatch_fire_event(struct net_device *dev);
2933 void linkwatch_forget_dev(struct net_device *dev);
2934
2935 /**
2936 * netif_carrier_ok - test if carrier present
2937 * @dev: network device
2938 *
2939 * Check if carrier is present on device
2940 */
netif_carrier_ok(const struct net_device * dev)2941 static inline bool netif_carrier_ok(const struct net_device *dev)
2942 {
2943 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2944 }
2945
2946 unsigned long dev_trans_start(struct net_device *dev);
2947
2948 void __netdev_watchdog_up(struct net_device *dev);
2949
2950 void netif_carrier_on(struct net_device *dev);
2951
2952 void netif_carrier_off(struct net_device *dev);
2953
2954 /**
2955 * netif_dormant_on - mark device as dormant.
2956 * @dev: network device
2957 *
2958 * Mark device as dormant (as per RFC2863).
2959 *
2960 * The dormant state indicates that the relevant interface is not
2961 * actually in a condition to pass packets (i.e., it is not 'up') but is
2962 * in a "pending" state, waiting for some external event. For "on-
2963 * demand" interfaces, this new state identifies the situation where the
2964 * interface is waiting for events to place it in the up state.
2965 *
2966 */
netif_dormant_on(struct net_device * dev)2967 static inline void netif_dormant_on(struct net_device *dev)
2968 {
2969 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2970 linkwatch_fire_event(dev);
2971 }
2972
2973 /**
2974 * netif_dormant_off - set device as not dormant.
2975 * @dev: network device
2976 *
2977 * Device is not in dormant state.
2978 */
netif_dormant_off(struct net_device * dev)2979 static inline void netif_dormant_off(struct net_device *dev)
2980 {
2981 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2982 linkwatch_fire_event(dev);
2983 }
2984
2985 /**
2986 * netif_dormant - test if carrier present
2987 * @dev: network device
2988 *
2989 * Check if carrier is present on device
2990 */
netif_dormant(const struct net_device * dev)2991 static inline bool netif_dormant(const struct net_device *dev)
2992 {
2993 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2994 }
2995
2996
2997 /**
2998 * netif_oper_up - test if device is operational
2999 * @dev: network device
3000 *
3001 * Check if carrier is operational
3002 */
netif_oper_up(const struct net_device * dev)3003 static inline bool netif_oper_up(const struct net_device *dev)
3004 {
3005 return (dev->operstate == IF_OPER_UP ||
3006 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3007 }
3008
3009 /**
3010 * netif_device_present - is device available or removed
3011 * @dev: network device
3012 *
3013 * Check if device has not been removed from system.
3014 */
netif_device_present(struct net_device * dev)3015 static inline bool netif_device_present(struct net_device *dev)
3016 {
3017 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3018 }
3019
3020 void netif_device_detach(struct net_device *dev);
3021
3022 void netif_device_attach(struct net_device *dev);
3023
3024 /*
3025 * Network interface message level settings
3026 */
3027
3028 enum {
3029 NETIF_MSG_DRV = 0x0001,
3030 NETIF_MSG_PROBE = 0x0002,
3031 NETIF_MSG_LINK = 0x0004,
3032 NETIF_MSG_TIMER = 0x0008,
3033 NETIF_MSG_IFDOWN = 0x0010,
3034 NETIF_MSG_IFUP = 0x0020,
3035 NETIF_MSG_RX_ERR = 0x0040,
3036 NETIF_MSG_TX_ERR = 0x0080,
3037 NETIF_MSG_TX_QUEUED = 0x0100,
3038 NETIF_MSG_INTR = 0x0200,
3039 NETIF_MSG_TX_DONE = 0x0400,
3040 NETIF_MSG_RX_STATUS = 0x0800,
3041 NETIF_MSG_PKTDATA = 0x1000,
3042 NETIF_MSG_HW = 0x2000,
3043 NETIF_MSG_WOL = 0x4000,
3044 };
3045
3046 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3047 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3048 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3049 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3050 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3051 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3052 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3053 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3054 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3055 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3056 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3057 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3058 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3059 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3060 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3061
netif_msg_init(int debug_value,int default_msg_enable_bits)3062 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3063 {
3064 /* use default */
3065 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3066 return default_msg_enable_bits;
3067 if (debug_value == 0) /* no output */
3068 return 0;
3069 /* set low N bits */
3070 return (1 << debug_value) - 1;
3071 }
3072
__netif_tx_lock(struct netdev_queue * txq,int cpu)3073 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3074 {
3075 spin_lock(&txq->_xmit_lock);
3076 txq->xmit_lock_owner = cpu;
3077 }
3078
__netif_tx_lock_bh(struct netdev_queue * txq)3079 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3080 {
3081 spin_lock_bh(&txq->_xmit_lock);
3082 txq->xmit_lock_owner = smp_processor_id();
3083 }
3084
__netif_tx_trylock(struct netdev_queue * txq)3085 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3086 {
3087 bool ok = spin_trylock(&txq->_xmit_lock);
3088 if (likely(ok))
3089 txq->xmit_lock_owner = smp_processor_id();
3090 return ok;
3091 }
3092
__netif_tx_unlock(struct netdev_queue * txq)3093 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3094 {
3095 txq->xmit_lock_owner = -1;
3096 spin_unlock(&txq->_xmit_lock);
3097 }
3098
__netif_tx_unlock_bh(struct netdev_queue * txq)3099 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3100 {
3101 txq->xmit_lock_owner = -1;
3102 spin_unlock_bh(&txq->_xmit_lock);
3103 }
3104
txq_trans_update(struct netdev_queue * txq)3105 static inline void txq_trans_update(struct netdev_queue *txq)
3106 {
3107 if (txq->xmit_lock_owner != -1)
3108 txq->trans_start = jiffies;
3109 }
3110
3111 /**
3112 * netif_tx_lock - grab network device transmit lock
3113 * @dev: network device
3114 *
3115 * Get network device transmit lock
3116 */
netif_tx_lock(struct net_device * dev)3117 static inline void netif_tx_lock(struct net_device *dev)
3118 {
3119 unsigned int i;
3120 int cpu;
3121
3122 spin_lock(&dev->tx_global_lock);
3123 cpu = smp_processor_id();
3124 for (i = 0; i < dev->num_tx_queues; i++) {
3125 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3126
3127 /* We are the only thread of execution doing a
3128 * freeze, but we have to grab the _xmit_lock in
3129 * order to synchronize with threads which are in
3130 * the ->hard_start_xmit() handler and already
3131 * checked the frozen bit.
3132 */
3133 __netif_tx_lock(txq, cpu);
3134 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3135 __netif_tx_unlock(txq);
3136 }
3137 }
3138
netif_tx_lock_bh(struct net_device * dev)3139 static inline void netif_tx_lock_bh(struct net_device *dev)
3140 {
3141 local_bh_disable();
3142 netif_tx_lock(dev);
3143 }
3144
netif_tx_unlock(struct net_device * dev)3145 static inline void netif_tx_unlock(struct net_device *dev)
3146 {
3147 unsigned int i;
3148
3149 for (i = 0; i < dev->num_tx_queues; i++) {
3150 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3151
3152 /* No need to grab the _xmit_lock here. If the
3153 * queue is not stopped for another reason, we
3154 * force a schedule.
3155 */
3156 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3157 netif_schedule_queue(txq);
3158 }
3159 spin_unlock(&dev->tx_global_lock);
3160 }
3161
netif_tx_unlock_bh(struct net_device * dev)3162 static inline void netif_tx_unlock_bh(struct net_device *dev)
3163 {
3164 netif_tx_unlock(dev);
3165 local_bh_enable();
3166 }
3167
3168 #define HARD_TX_LOCK(dev, txq, cpu) { \
3169 if ((dev->features & NETIF_F_LLTX) == 0) { \
3170 __netif_tx_lock(txq, cpu); \
3171 } \
3172 }
3173
3174 #define HARD_TX_TRYLOCK(dev, txq) \
3175 (((dev->features & NETIF_F_LLTX) == 0) ? \
3176 __netif_tx_trylock(txq) : \
3177 true )
3178
3179 #define HARD_TX_UNLOCK(dev, txq) { \
3180 if ((dev->features & NETIF_F_LLTX) == 0) { \
3181 __netif_tx_unlock(txq); \
3182 } \
3183 }
3184
netif_tx_disable(struct net_device * dev)3185 static inline void netif_tx_disable(struct net_device *dev)
3186 {
3187 unsigned int i;
3188 int cpu;
3189
3190 local_bh_disable();
3191 cpu = smp_processor_id();
3192 for (i = 0; i < dev->num_tx_queues; i++) {
3193 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3194
3195 __netif_tx_lock(txq, cpu);
3196 netif_tx_stop_queue(txq);
3197 __netif_tx_unlock(txq);
3198 }
3199 local_bh_enable();
3200 }
3201
netif_addr_lock(struct net_device * dev)3202 static inline void netif_addr_lock(struct net_device *dev)
3203 {
3204 spin_lock(&dev->addr_list_lock);
3205 }
3206
netif_addr_lock_nested(struct net_device * dev)3207 static inline void netif_addr_lock_nested(struct net_device *dev)
3208 {
3209 int subclass = SINGLE_DEPTH_NESTING;
3210
3211 if (dev->netdev_ops->ndo_get_lock_subclass)
3212 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3213
3214 spin_lock_nested(&dev->addr_list_lock, subclass);
3215 }
3216
netif_addr_lock_bh(struct net_device * dev)3217 static inline void netif_addr_lock_bh(struct net_device *dev)
3218 {
3219 spin_lock_bh(&dev->addr_list_lock);
3220 }
3221
netif_addr_unlock(struct net_device * dev)3222 static inline void netif_addr_unlock(struct net_device *dev)
3223 {
3224 spin_unlock(&dev->addr_list_lock);
3225 }
3226
netif_addr_unlock_bh(struct net_device * dev)3227 static inline void netif_addr_unlock_bh(struct net_device *dev)
3228 {
3229 spin_unlock_bh(&dev->addr_list_lock);
3230 }
3231
3232 /*
3233 * dev_addrs walker. Should be used only for read access. Call with
3234 * rcu_read_lock held.
3235 */
3236 #define for_each_dev_addr(dev, ha) \
3237 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3238
3239 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3240
3241 void ether_setup(struct net_device *dev);
3242
3243 /* Support for loadable net-drivers */
3244 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3245 unsigned char name_assign_type,
3246 void (*setup)(struct net_device *),
3247 unsigned int txqs, unsigned int rxqs);
3248 int dev_get_valid_name(struct net *net, struct net_device *dev,
3249 const char *name);
3250
3251 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3252 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3253
3254 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3255 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3256 count)
3257
3258 int register_netdev(struct net_device *dev);
3259 void unregister_netdev(struct net_device *dev);
3260
3261 /* General hardware address lists handling functions */
3262 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3263 struct netdev_hw_addr_list *from_list, int addr_len);
3264 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3265 struct netdev_hw_addr_list *from_list, int addr_len);
3266 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3267 struct net_device *dev,
3268 int (*sync)(struct net_device *, const unsigned char *),
3269 int (*unsync)(struct net_device *,
3270 const unsigned char *));
3271 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3272 struct net_device *dev,
3273 int (*unsync)(struct net_device *,
3274 const unsigned char *));
3275 void __hw_addr_init(struct netdev_hw_addr_list *list);
3276
3277 /* Functions used for device addresses handling */
3278 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3279 unsigned char addr_type);
3280 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3281 unsigned char addr_type);
3282 void dev_addr_flush(struct net_device *dev);
3283 int dev_addr_init(struct net_device *dev);
3284
3285 /* Functions used for unicast addresses handling */
3286 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3287 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3288 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3289 int dev_uc_sync(struct net_device *to, struct net_device *from);
3290 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3291 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3292 void dev_uc_flush(struct net_device *dev);
3293 void dev_uc_init(struct net_device *dev);
3294
3295 /**
3296 * __dev_uc_sync - Synchonize device's unicast list
3297 * @dev: device to sync
3298 * @sync: function to call if address should be added
3299 * @unsync: function to call if address should be removed
3300 *
3301 * Add newly added addresses to the interface, and release
3302 * addresses that have been deleted.
3303 **/
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))3304 static inline int __dev_uc_sync(struct net_device *dev,
3305 int (*sync)(struct net_device *,
3306 const unsigned char *),
3307 int (*unsync)(struct net_device *,
3308 const unsigned char *))
3309 {
3310 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3311 }
3312
3313 /**
3314 * __dev_uc_unsync - Remove synchronized addresses from device
3315 * @dev: device to sync
3316 * @unsync: function to call if address should be removed
3317 *
3318 * Remove all addresses that were added to the device by dev_uc_sync().
3319 **/
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))3320 static inline void __dev_uc_unsync(struct net_device *dev,
3321 int (*unsync)(struct net_device *,
3322 const unsigned char *))
3323 {
3324 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3325 }
3326
3327 /* Functions used for multicast addresses handling */
3328 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3329 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3330 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3331 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3332 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3333 int dev_mc_sync(struct net_device *to, struct net_device *from);
3334 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3335 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3336 void dev_mc_flush(struct net_device *dev);
3337 void dev_mc_init(struct net_device *dev);
3338
3339 /**
3340 * __dev_mc_sync - Synchonize device's multicast list
3341 * @dev: device to sync
3342 * @sync: function to call if address should be added
3343 * @unsync: function to call if address should be removed
3344 *
3345 * Add newly added addresses to the interface, and release
3346 * addresses that have been deleted.
3347 **/
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))3348 static inline int __dev_mc_sync(struct net_device *dev,
3349 int (*sync)(struct net_device *,
3350 const unsigned char *),
3351 int (*unsync)(struct net_device *,
3352 const unsigned char *))
3353 {
3354 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3355 }
3356
3357 /**
3358 * __dev_mc_unsync - Remove synchronized addresses from device
3359 * @dev: device to sync
3360 * @unsync: function to call if address should be removed
3361 *
3362 * Remove all addresses that were added to the device by dev_mc_sync().
3363 **/
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))3364 static inline void __dev_mc_unsync(struct net_device *dev,
3365 int (*unsync)(struct net_device *,
3366 const unsigned char *))
3367 {
3368 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3369 }
3370
3371 /* Functions used for secondary unicast and multicast support */
3372 void dev_set_rx_mode(struct net_device *dev);
3373 void __dev_set_rx_mode(struct net_device *dev);
3374 int dev_set_promiscuity(struct net_device *dev, int inc);
3375 int dev_set_allmulti(struct net_device *dev, int inc);
3376 void netdev_state_change(struct net_device *dev);
3377 void netdev_notify_peers(struct net_device *dev);
3378 void netdev_features_change(struct net_device *dev);
3379 /* Load a device via the kmod */
3380 void dev_load(struct net *net, const char *name);
3381 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3382 struct rtnl_link_stats64 *storage);
3383 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3384 const struct net_device_stats *netdev_stats);
3385
3386 extern int netdev_max_backlog;
3387 extern int netdev_tstamp_prequeue;
3388 extern int weight_p;
3389 extern int bpf_jit_enable;
3390
3391 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3392 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3393 struct list_head **iter);
3394 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3395 struct list_head **iter);
3396
3397 /* iterate through upper list, must be called under RCU read lock */
3398 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3399 for (iter = &(dev)->adj_list.upper, \
3400 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3401 updev; \
3402 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3403
3404 /* iterate through upper list, must be called under RCU read lock */
3405 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3406 for (iter = &(dev)->all_adj_list.upper, \
3407 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3408 updev; \
3409 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3410
3411 void *netdev_lower_get_next_private(struct net_device *dev,
3412 struct list_head **iter);
3413 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3414 struct list_head **iter);
3415
3416 #define netdev_for_each_lower_private(dev, priv, iter) \
3417 for (iter = (dev)->adj_list.lower.next, \
3418 priv = netdev_lower_get_next_private(dev, &(iter)); \
3419 priv; \
3420 priv = netdev_lower_get_next_private(dev, &(iter)))
3421
3422 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3423 for (iter = &(dev)->adj_list.lower, \
3424 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3425 priv; \
3426 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3427
3428 void *netdev_lower_get_next(struct net_device *dev,
3429 struct list_head **iter);
3430 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3431 for (iter = &(dev)->adj_list.lower, \
3432 ldev = netdev_lower_get_next(dev, &(iter)); \
3433 ldev; \
3434 ldev = netdev_lower_get_next(dev, &(iter)))
3435
3436 void *netdev_adjacent_get_private(struct list_head *adj_list);
3437 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3438 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3439 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3440 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3441 int netdev_master_upper_dev_link(struct net_device *dev,
3442 struct net_device *upper_dev);
3443 int netdev_master_upper_dev_link_private(struct net_device *dev,
3444 struct net_device *upper_dev,
3445 void *private);
3446 void netdev_upper_dev_unlink(struct net_device *dev,
3447 struct net_device *upper_dev);
3448 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3449 void *netdev_lower_dev_get_private(struct net_device *dev,
3450 struct net_device *lower_dev);
3451 int dev_get_nest_level(struct net_device *dev,
3452 bool (*type_check)(struct net_device *dev));
3453 int skb_checksum_help(struct sk_buff *skb);
3454 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3455 netdev_features_t features, bool tx_path);
3456 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3457 netdev_features_t features);
3458
3459 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)3460 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3461 {
3462 return __skb_gso_segment(skb, features, true);
3463 }
3464 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3465
can_checksum_protocol(netdev_features_t features,__be16 protocol)3466 static inline bool can_checksum_protocol(netdev_features_t features,
3467 __be16 protocol)
3468 {
3469 return ((features & NETIF_F_GEN_CSUM) ||
3470 ((features & NETIF_F_V4_CSUM) &&
3471 protocol == htons(ETH_P_IP)) ||
3472 ((features & NETIF_F_V6_CSUM) &&
3473 protocol == htons(ETH_P_IPV6)) ||
3474 ((features & NETIF_F_FCOE_CRC) &&
3475 protocol == htons(ETH_P_FCOE)));
3476 }
3477
3478 #ifdef CONFIG_BUG
3479 void netdev_rx_csum_fault(struct net_device *dev);
3480 #else
netdev_rx_csum_fault(struct net_device * dev)3481 static inline void netdev_rx_csum_fault(struct net_device *dev)
3482 {
3483 }
3484 #endif
3485 /* rx skb timestamps */
3486 void net_enable_timestamp(void);
3487 void net_disable_timestamp(void);
3488
3489 #ifdef CONFIG_PROC_FS
3490 int __init dev_proc_init(void);
3491 #else
3492 #define dev_proc_init() 0
3493 #endif
3494
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)3495 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3496 struct sk_buff *skb, struct net_device *dev,
3497 bool more)
3498 {
3499 skb->xmit_more = more ? 1 : 0;
3500 return ops->ndo_start_xmit(skb, dev);
3501 }
3502
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)3503 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3504 struct netdev_queue *txq, bool more)
3505 {
3506 const struct net_device_ops *ops = dev->netdev_ops;
3507 int rc;
3508
3509 rc = __netdev_start_xmit(ops, skb, dev, more);
3510 if (rc == NETDEV_TX_OK)
3511 txq_trans_update(txq);
3512
3513 return rc;
3514 }
3515
3516 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3517 const void *ns);
3518 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3519 const void *ns);
3520
netdev_class_create_file(struct class_attribute * class_attr)3521 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3522 {
3523 return netdev_class_create_file_ns(class_attr, NULL);
3524 }
3525
netdev_class_remove_file(struct class_attribute * class_attr)3526 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3527 {
3528 netdev_class_remove_file_ns(class_attr, NULL);
3529 }
3530
3531 extern struct kobj_ns_type_operations net_ns_type_operations;
3532
3533 const char *netdev_drivername(const struct net_device *dev);
3534
3535 void linkwatch_run_queue(void);
3536
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)3537 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3538 netdev_features_t f2)
3539 {
3540 if (f1 & NETIF_F_GEN_CSUM)
3541 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3542 if (f2 & NETIF_F_GEN_CSUM)
3543 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3544 f1 &= f2;
3545 if (f1 & NETIF_F_GEN_CSUM)
3546 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3547
3548 return f1;
3549 }
3550
netdev_get_wanted_features(struct net_device * dev)3551 static inline netdev_features_t netdev_get_wanted_features(
3552 struct net_device *dev)
3553 {
3554 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3555 }
3556 netdev_features_t netdev_increment_features(netdev_features_t all,
3557 netdev_features_t one, netdev_features_t mask);
3558
3559 /* Allow TSO being used on stacked device :
3560 * Performing the GSO segmentation before last device
3561 * is a performance improvement.
3562 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)3563 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3564 netdev_features_t mask)
3565 {
3566 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3567 }
3568
3569 int __netdev_update_features(struct net_device *dev);
3570 void netdev_update_features(struct net_device *dev);
3571 void netdev_change_features(struct net_device *dev);
3572
3573 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3574 struct net_device *dev);
3575
3576 netdev_features_t netif_skb_features(struct sk_buff *skb);
3577
net_gso_ok(netdev_features_t features,int gso_type)3578 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3579 {
3580 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3581
3582 /* check flags correspondence */
3583 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3584 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3585 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3586 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3587 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3588 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3589 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3590 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3591 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3592 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3593 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3594 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3595 BUILD_BUG_ON(SKB_GSO_MPLS != (NETIF_F_GSO_MPLS >> NETIF_F_GSO_SHIFT));
3596
3597 return (features & feature) == feature;
3598 }
3599
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)3600 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3601 {
3602 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3603 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3604 }
3605
netif_needs_gso(struct net_device * dev,struct sk_buff * skb,netdev_features_t features)3606 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3607 netdev_features_t features)
3608 {
3609 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3610 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3611 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3612 }
3613
netif_set_gso_max_size(struct net_device * dev,unsigned int size)3614 static inline void netif_set_gso_max_size(struct net_device *dev,
3615 unsigned int size)
3616 {
3617 dev->gso_max_size = size;
3618 }
3619
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)3620 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3621 int pulled_hlen, u16 mac_offset,
3622 int mac_len)
3623 {
3624 skb->protocol = protocol;
3625 skb->encapsulation = 1;
3626 skb_push(skb, pulled_hlen);
3627 skb_reset_transport_header(skb);
3628 skb->mac_header = mac_offset;
3629 skb->network_header = skb->mac_header + mac_len;
3630 skb->mac_len = mac_len;
3631 }
3632
netif_is_macvlan(struct net_device * dev)3633 static inline bool netif_is_macvlan(struct net_device *dev)
3634 {
3635 return dev->priv_flags & IFF_MACVLAN;
3636 }
3637
netif_is_bond_master(struct net_device * dev)3638 static inline bool netif_is_bond_master(struct net_device *dev)
3639 {
3640 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3641 }
3642
netif_is_bond_slave(struct net_device * dev)3643 static inline bool netif_is_bond_slave(struct net_device *dev)
3644 {
3645 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3646 }
3647
netif_supports_nofcs(struct net_device * dev)3648 static inline bool netif_supports_nofcs(struct net_device *dev)
3649 {
3650 return dev->priv_flags & IFF_SUPP_NOFCS;
3651 }
3652
3653 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)3654 static inline void netif_keep_dst(struct net_device *dev)
3655 {
3656 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3657 }
3658
3659 extern struct pernet_operations __net_initdata loopback_net_ops;
3660
3661 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3662
3663 /* netdev_printk helpers, similar to dev_printk */
3664
netdev_name(const struct net_device * dev)3665 static inline const char *netdev_name(const struct net_device *dev)
3666 {
3667 if (!dev->name[0] || strchr(dev->name, '%'))
3668 return "(unnamed net_device)";
3669 return dev->name;
3670 }
3671
netdev_reg_state(const struct net_device * dev)3672 static inline const char *netdev_reg_state(const struct net_device *dev)
3673 {
3674 switch (dev->reg_state) {
3675 case NETREG_UNINITIALIZED: return " (uninitialized)";
3676 case NETREG_REGISTERED: return "";
3677 case NETREG_UNREGISTERING: return " (unregistering)";
3678 case NETREG_UNREGISTERED: return " (unregistered)";
3679 case NETREG_RELEASED: return " (released)";
3680 case NETREG_DUMMY: return " (dummy)";
3681 }
3682
3683 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3684 return " (unknown)";
3685 }
3686
3687 __printf(3, 4)
3688 void netdev_printk(const char *level, const struct net_device *dev,
3689 const char *format, ...);
3690 __printf(2, 3)
3691 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3692 __printf(2, 3)
3693 void netdev_alert(const struct net_device *dev, const char *format, ...);
3694 __printf(2, 3)
3695 void netdev_crit(const struct net_device *dev, const char *format, ...);
3696 __printf(2, 3)
3697 void netdev_err(const struct net_device *dev, const char *format, ...);
3698 __printf(2, 3)
3699 void netdev_warn(const struct net_device *dev, const char *format, ...);
3700 __printf(2, 3)
3701 void netdev_notice(const struct net_device *dev, const char *format, ...);
3702 __printf(2, 3)
3703 void netdev_info(const struct net_device *dev, const char *format, ...);
3704
3705 #define MODULE_ALIAS_NETDEV(device) \
3706 MODULE_ALIAS("netdev-" device)
3707
3708 #if defined(CONFIG_DYNAMIC_DEBUG)
3709 #define netdev_dbg(__dev, format, args...) \
3710 do { \
3711 dynamic_netdev_dbg(__dev, format, ##args); \
3712 } while (0)
3713 #elif defined(DEBUG)
3714 #define netdev_dbg(__dev, format, args...) \
3715 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3716 #else
3717 #define netdev_dbg(__dev, format, args...) \
3718 ({ \
3719 if (0) \
3720 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3721 })
3722 #endif
3723
3724 #if defined(VERBOSE_DEBUG)
3725 #define netdev_vdbg netdev_dbg
3726 #else
3727
3728 #define netdev_vdbg(dev, format, args...) \
3729 ({ \
3730 if (0) \
3731 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3732 0; \
3733 })
3734 #endif
3735
3736 /*
3737 * netdev_WARN() acts like dev_printk(), but with the key difference
3738 * of using a WARN/WARN_ON to get the message out, including the
3739 * file/line information and a backtrace.
3740 */
3741 #define netdev_WARN(dev, format, args...) \
3742 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3743 netdev_reg_state(dev), ##args)
3744
3745 /* netif printk helpers, similar to netdev_printk */
3746
3747 #define netif_printk(priv, type, level, dev, fmt, args...) \
3748 do { \
3749 if (netif_msg_##type(priv)) \
3750 netdev_printk(level, (dev), fmt, ##args); \
3751 } while (0)
3752
3753 #define netif_level(level, priv, type, dev, fmt, args...) \
3754 do { \
3755 if (netif_msg_##type(priv)) \
3756 netdev_##level(dev, fmt, ##args); \
3757 } while (0)
3758
3759 #define netif_emerg(priv, type, dev, fmt, args...) \
3760 netif_level(emerg, priv, type, dev, fmt, ##args)
3761 #define netif_alert(priv, type, dev, fmt, args...) \
3762 netif_level(alert, priv, type, dev, fmt, ##args)
3763 #define netif_crit(priv, type, dev, fmt, args...) \
3764 netif_level(crit, priv, type, dev, fmt, ##args)
3765 #define netif_err(priv, type, dev, fmt, args...) \
3766 netif_level(err, priv, type, dev, fmt, ##args)
3767 #define netif_warn(priv, type, dev, fmt, args...) \
3768 netif_level(warn, priv, type, dev, fmt, ##args)
3769 #define netif_notice(priv, type, dev, fmt, args...) \
3770 netif_level(notice, priv, type, dev, fmt, ##args)
3771 #define netif_info(priv, type, dev, fmt, args...) \
3772 netif_level(info, priv, type, dev, fmt, ##args)
3773
3774 #if defined(CONFIG_DYNAMIC_DEBUG)
3775 #define netif_dbg(priv, type, netdev, format, args...) \
3776 do { \
3777 if (netif_msg_##type(priv)) \
3778 dynamic_netdev_dbg(netdev, format, ##args); \
3779 } while (0)
3780 #elif defined(DEBUG)
3781 #define netif_dbg(priv, type, dev, format, args...) \
3782 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3783 #else
3784 #define netif_dbg(priv, type, dev, format, args...) \
3785 ({ \
3786 if (0) \
3787 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3788 0; \
3789 })
3790 #endif
3791
3792 #if defined(VERBOSE_DEBUG)
3793 #define netif_vdbg netif_dbg
3794 #else
3795 #define netif_vdbg(priv, type, dev, format, args...) \
3796 ({ \
3797 if (0) \
3798 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3799 0; \
3800 })
3801 #endif
3802
3803 /*
3804 * The list of packet types we will receive (as opposed to discard)
3805 * and the routines to invoke.
3806 *
3807 * Why 16. Because with 16 the only overlap we get on a hash of the
3808 * low nibble of the protocol value is RARP/SNAP/X.25.
3809 *
3810 * NOTE: That is no longer true with the addition of VLAN tags. Not
3811 * sure which should go first, but I bet it won't make much
3812 * difference if we are running VLANs. The good news is that
3813 * this protocol won't be in the list unless compiled in, so
3814 * the average user (w/out VLANs) will not be adversely affected.
3815 * --BLG
3816 *
3817 * 0800 IP
3818 * 8100 802.1Q VLAN
3819 * 0001 802.3
3820 * 0002 AX.25
3821 * 0004 802.2
3822 * 8035 RARP
3823 * 0005 SNAP
3824 * 0805 X.25
3825 * 0806 ARP
3826 * 8137 IPX
3827 * 0009 Localtalk
3828 * 86DD IPv6
3829 */
3830 #define PTYPE_HASH_SIZE (16)
3831 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3832
3833 #endif /* _LINUX_NETDEVICE_H */
3834