]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - include/linux/netdevice.h
bpf: move tmp variable into ax register in interpreter
[thirdparty/kernel/stable.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
40
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
43 #ifdef CONFIG_DCB
44 #include <net/dcbnl.h>
45 #endif
46 #include <net/netprio_cgroup.h>
47
48 #include <linux/netdev_features.h>
49 #include <linux/neighbour.h>
50 #include <uapi/linux/netdevice.h>
51 #include <uapi/linux/if_bonding.h>
52 #include <uapi/linux/pkt_cls.h>
53 #include <linux/hashtable.h>
54
55 struct netpoll_info;
56 struct device;
57 struct phy_device;
58 struct dsa_switch_tree;
59
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
68 struct xdp_buff;
69
70 void netdev_set_default_ethtool_ops(struct net_device *dev,
71 const struct ethtool_ops *ops);
72
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
76
77 /*
78 * Transmit return codes: transmit return codes originate from three different
79 * namespaces:
80 *
81 * - qdisc return codes
82 * - driver transmit return codes
83 * - errno values
84 *
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
92 */
93
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
108
109 enum netdev_tx {
110 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
111 NETDEV_TX_OK = 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 };
114 typedef enum netdev_tx netdev_tx_t;
115
116 /*
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 */
120 static inline bool dev_xmit_complete(int rc)
121 {
122 /*
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 */
128 if (likely(rc < NET_XMIT_MASK))
129 return true;
130
131 return false;
132 }
133
134 /*
135 * Compute the worst-case header length according to the protocols
136 * used.
137 */
138
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
144 # else
145 # define LL_MAX_HEADER 96
146 # endif
147 #else
148 # define LL_MAX_HEADER 32
149 #endif
150
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
154 #else
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 #endif
157
158 /*
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
161 */
162
163 struct net_device_stats {
164 unsigned long rx_packets;
165 unsigned long tx_packets;
166 unsigned long rx_bytes;
167 unsigned long tx_bytes;
168 unsigned long rx_errors;
169 unsigned long tx_errors;
170 unsigned long rx_dropped;
171 unsigned long tx_dropped;
172 unsigned long multicast;
173 unsigned long collisions;
174 unsigned long rx_length_errors;
175 unsigned long rx_over_errors;
176 unsigned long rx_crc_errors;
177 unsigned long rx_frame_errors;
178 unsigned long rx_fifo_errors;
179 unsigned long rx_missed_errors;
180 unsigned long tx_aborted_errors;
181 unsigned long tx_carrier_errors;
182 unsigned long tx_fifo_errors;
183 unsigned long tx_heartbeat_errors;
184 unsigned long tx_window_errors;
185 unsigned long rx_compressed;
186 unsigned long tx_compressed;
187 };
188
189
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
192
193 #ifdef CONFIG_RPS
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed;
196 extern struct static_key rfs_needed;
197 #endif
198
199 struct neighbour;
200 struct neigh_parms;
201 struct sk_buff;
202
203 struct netdev_hw_addr {
204 struct list_head list;
205 unsigned char addr[MAX_ADDR_LEN];
206 unsigned char type;
207 #define NETDEV_HW_ADDR_T_LAN 1
208 #define NETDEV_HW_ADDR_T_SAN 2
209 #define NETDEV_HW_ADDR_T_SLAVE 3
210 #define NETDEV_HW_ADDR_T_UNICAST 4
211 #define NETDEV_HW_ADDR_T_MULTICAST 5
212 bool global_use;
213 int sync_cnt;
214 int refcount;
215 int synced;
216 struct rcu_head rcu_head;
217 };
218
219 struct netdev_hw_addr_list {
220 struct list_head list;
221 int count;
222 };
223
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
228
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
238
239 struct hh_cache {
240 unsigned int hh_len;
241 seqlock_t hh_lock;
242
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * Alternative is:
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 *
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
259 */
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264
265 struct header_ops {
266 int (*create) (struct sk_buff *skb, struct net_device *dev,
267 unsigned short type, const void *daddr,
268 const void *saddr, unsigned int len);
269 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 void (*cache_update)(struct hh_cache *hh,
272 const struct net_device *dev,
273 const unsigned char *haddr);
274 bool (*validate)(const char *ll_header, unsigned int len);
275 };
276
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
279 * code.
280 */
281
282 enum netdev_state_t {
283 __LINK_STATE_START,
284 __LINK_STATE_PRESENT,
285 __LINK_STATE_NOCARRIER,
286 __LINK_STATE_LINKWATCH_PENDING,
287 __LINK_STATE_DORMANT,
288 };
289
290
291 /*
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
294 */
295 struct netdev_boot_setup {
296 char name[IFNAMSIZ];
297 struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300
301 int __init netdev_boot_setup(char *str);
302
303 /*
304 * Structure for NAPI scheduling similar to tasklet but with weighting
305 */
306 struct napi_struct {
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
312 */
313 struct list_head poll_list;
314
315 unsigned long state;
316 int weight;
317 unsigned int gro_count;
318 int (*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 int poll_owner;
321 #endif
322 struct net_device *dev;
323 struct sk_buff *gro_list;
324 struct sk_buff *skb;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
329 };
330
331 enum {
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_MISSED, /* reschedule a napi */
334 NAPI_STATE_DISABLE, /* Disable pending */
335 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
336 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
338 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
339 };
340
341 enum {
342 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
343 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
344 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
345 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
346 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
347 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
348 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
349 };
350
351 enum gro_result {
352 GRO_MERGED,
353 GRO_MERGED_FREE,
354 GRO_HELD,
355 GRO_NORMAL,
356 GRO_DROP,
357 GRO_CONSUMED,
358 };
359 typedef enum gro_result gro_result_t;
360
361 /*
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
364 * further.
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
369 *
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
372 *
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
375 *
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
379 *
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
381 * do with the skb.
382 *
383 * If the rx_handler consumed the skb in some way, it should return
384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385 * the skb to be delivered in some other way.
386 *
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
390 *
391 * If the rx_handler decides the skb should be ignored, it should return
392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393 * are registered on exact device (ptype->dev == skb->dev).
394 *
395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396 * delivered, it should return RX_HANDLER_PASS.
397 *
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
400 */
401
402 enum rx_handler_result {
403 RX_HANDLER_CONSUMED,
404 RX_HANDLER_ANOTHER,
405 RX_HANDLER_EXACT,
406 RX_HANDLER_PASS,
407 };
408 typedef enum rx_handler_result rx_handler_result_t;
409 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
410
411 void __napi_schedule(struct napi_struct *n);
412 void __napi_schedule_irqoff(struct napi_struct *n);
413
414 static inline bool napi_disable_pending(struct napi_struct *n)
415 {
416 return test_bit(NAPI_STATE_DISABLE, &n->state);
417 }
418
419 bool napi_schedule_prep(struct napi_struct *n);
420
421 /**
422 * napi_schedule - schedule NAPI poll
423 * @n: NAPI context
424 *
425 * Schedule NAPI poll routine to be called if it is not already
426 * running.
427 */
428 static inline void napi_schedule(struct napi_struct *n)
429 {
430 if (napi_schedule_prep(n))
431 __napi_schedule(n);
432 }
433
434 /**
435 * napi_schedule_irqoff - schedule NAPI poll
436 * @n: NAPI context
437 *
438 * Variant of napi_schedule(), assuming hard irqs are masked.
439 */
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
441 {
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
444 }
445
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct *napi)
448 {
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
451 return true;
452 }
453 return false;
454 }
455
456 bool napi_complete_done(struct napi_struct *n, int work_done);
457 /**
458 * napi_complete - NAPI processing complete
459 * @n: NAPI context
460 *
461 * Mark NAPI processing as complete.
462 * Consider using napi_complete_done() instead.
463 * Return false if device should avoid rearming interrupts.
464 */
465 static inline bool napi_complete(struct napi_struct *n)
466 {
467 return napi_complete_done(n, 0);
468 }
469
470 /**
471 * napi_hash_del - remove a NAPI from global table
472 * @napi: NAPI context
473 *
474 * Warning: caller must observe RCU grace period
475 * before freeing memory containing @napi, if
476 * this function returns true.
477 * Note: core networking stack automatically calls it
478 * from netif_napi_del().
479 * Drivers might want to call this helper to combine all
480 * the needed RCU grace periods into a single one.
481 */
482 bool napi_hash_del(struct napi_struct *napi);
483
484 /**
485 * napi_disable - prevent NAPI from scheduling
486 * @n: NAPI context
487 *
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
490 */
491 void napi_disable(struct napi_struct *n);
492
493 /**
494 * napi_enable - enable NAPI scheduling
495 * @n: NAPI context
496 *
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
499 */
500 static inline void napi_enable(struct napi_struct *n)
501 {
502 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
503 smp_mb__before_atomic();
504 clear_bit(NAPI_STATE_SCHED, &n->state);
505 clear_bit(NAPI_STATE_NPSVC, &n->state);
506 }
507
508 /**
509 * napi_synchronize - wait until NAPI is not running
510 * @n: NAPI context
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
516 static inline void napi_synchronize(const struct napi_struct *n)
517 {
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
523 }
524
525 enum netdev_queue_state_t {
526 __QUEUE_STATE_DRV_XOFF,
527 __QUEUE_STATE_STACK_XOFF,
528 __QUEUE_STATE_FROZEN,
529 };
530
531 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
534
535 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
537 QUEUE_STATE_FROZEN)
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
539 QUEUE_STATE_FROZEN)
540
541 /*
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
549 */
550
551 struct netdev_queue {
552 /*
553 * read-mostly part
554 */
555 struct net_device *dev;
556 struct Qdisc __rcu *qdisc;
557 struct Qdisc *qdisc_sleeping;
558 #ifdef CONFIG_SYSFS
559 struct kobject kobj;
560 #endif
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
562 int numa_node;
563 #endif
564 unsigned long tx_maxrate;
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 * write-mostly part
572 */
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
574 int xmit_lock_owner;
575 /*
576 * Time (in jiffies) of last Tx
577 */
578 unsigned long trans_start;
579
580 unsigned long state;
581
582 #ifdef CONFIG_BQL
583 struct dql dql;
584 #endif
585 } ____cacheline_aligned_in_smp;
586
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
588 {
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
590 return q->numa_node;
591 #else
592 return NUMA_NO_NODE;
593 #endif
594 }
595
596 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
597 {
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 q->numa_node = node;
600 #endif
601 }
602
603 #ifdef CONFIG_RPS
604 /*
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
607 */
608 struct rps_map {
609 unsigned int len;
610 struct rcu_head rcu;
611 u16 cpus[0];
612 };
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
614
615 /*
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
619 */
620 struct rps_dev_flow {
621 u16 cpu;
622 u16 filter;
623 unsigned int last_qtail;
624 };
625 #define RPS_NO_FILTER 0xffff
626
627 /*
628 * The rps_dev_flow_table structure contains a table of flow mappings.
629 */
630 struct rps_dev_flow_table {
631 unsigned int mask;
632 struct rcu_head rcu;
633 struct rps_dev_flow flows[0];
634 };
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636 ((_num) * sizeof(struct rps_dev_flow)))
637
638 /*
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
643 * rps_cpu_mask is used to partition the space, depending on number of
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646 * meaning we use 32-6=26 bits for the hash.
647 */
648 struct rps_sock_flow_table {
649 u32 mask;
650
651 u32 ents[0] ____cacheline_aligned_in_smp;
652 };
653 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
654
655 #define RPS_NO_CPU 0xffff
656
657 extern u32 rps_cpu_mask;
658 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
659
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
661 u32 hash)
662 {
663 if (table && hash) {
664 unsigned int index = hash & table->mask;
665 u32 val = hash & ~rps_cpu_mask;
666
667 /* We only give a hint, preemption can change CPU under us */
668 val |= raw_smp_processor_id();
669
670 if (table->ents[index] != val)
671 table->ents[index] = val;
672 }
673 }
674
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
677 u16 filter_id);
678 #endif
679 #endif /* CONFIG_RPS */
680
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue {
683 #ifdef CONFIG_RPS
684 struct rps_map __rcu *rps_map;
685 struct rps_dev_flow_table __rcu *rps_flow_table;
686 #endif
687 struct kobject kobj;
688 struct net_device *dev;
689 } ____cacheline_aligned_in_smp;
690
691 /*
692 * RX queue sysfs structures and functions.
693 */
694 struct rx_queue_attribute {
695 struct attribute attr;
696 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
697 ssize_t (*store)(struct netdev_rx_queue *queue,
698 const char *buf, size_t len);
699 };
700
701 #ifdef CONFIG_XPS
702 /*
703 * This structure holds an XPS map which can be of variable length. The
704 * map is an array of queues.
705 */
706 struct xps_map {
707 unsigned int len;
708 unsigned int alloc_len;
709 struct rcu_head rcu;
710 u16 queues[0];
711 };
712 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
713 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714 - sizeof(struct xps_map)) / sizeof(u16))
715
716 /*
717 * This structure holds all XPS maps for device. Maps are indexed by CPU.
718 */
719 struct xps_dev_maps {
720 struct rcu_head rcu;
721 struct xps_map __rcu *cpu_map[0];
722 };
723 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
724 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
725 #endif /* CONFIG_XPS */
726
727 #define TC_MAX_QUEUE 16
728 #define TC_BITMASK 15
729 /* HW offloaded queuing disciplines txq count and offset maps */
730 struct netdev_tc_txq {
731 u16 count;
732 u16 offset;
733 };
734
735 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
736 /*
737 * This structure is to hold information about the device
738 * configured to run FCoE protocol stack.
739 */
740 struct netdev_fcoe_hbainfo {
741 char manufacturer[64];
742 char serial_number[64];
743 char hardware_version[64];
744 char driver_version[64];
745 char optionrom_version[64];
746 char firmware_version[64];
747 char model[256];
748 char model_description[256];
749 };
750 #endif
751
752 #define MAX_PHYS_ITEM_ID_LEN 32
753
754 /* This structure holds a unique identifier to identify some
755 * physical item (port for example) used by a netdevice.
756 */
757 struct netdev_phys_item_id {
758 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
759 unsigned char id_len;
760 };
761
762 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
763 struct netdev_phys_item_id *b)
764 {
765 return a->id_len == b->id_len &&
766 memcmp(a->id, b->id, a->id_len) == 0;
767 }
768
769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
772 enum tc_setup_type {
773 TC_SETUP_MQPRIO,
774 TC_SETUP_CLSU32,
775 TC_SETUP_CLSFLOWER,
776 TC_SETUP_CLSMATCHALL,
777 TC_SETUP_CLSBPF,
778 };
779
780 /* These structures hold the attributes of xdp state that are being passed
781 * to the netdevice through the xdp op.
782 */
783 enum xdp_netdev_command {
784 /* Set or clear a bpf program used in the earliest stages of packet
785 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
786 * is responsible for calling bpf_prog_put on any old progs that are
787 * stored. In case of error, the callee need not release the new prog
788 * reference, but on success it takes ownership and must bpf_prog_put
789 * when it is no longer used.
790 */
791 XDP_SETUP_PROG,
792 XDP_SETUP_PROG_HW,
793 /* Check if a bpf program is set on the device. The callee should
794 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
795 * is equivalent to XDP_ATTACHED_DRV.
796 */
797 XDP_QUERY_PROG,
798 };
799
800 struct netlink_ext_ack;
801
802 struct netdev_xdp {
803 enum xdp_netdev_command command;
804 union {
805 /* XDP_SETUP_PROG */
806 struct {
807 u32 flags;
808 struct bpf_prog *prog;
809 struct netlink_ext_ack *extack;
810 };
811 /* XDP_QUERY_PROG */
812 struct {
813 u8 prog_attached;
814 u32 prog_id;
815 };
816 };
817 };
818
819 #ifdef CONFIG_XFRM_OFFLOAD
820 struct xfrmdev_ops {
821 int (*xdo_dev_state_add) (struct xfrm_state *x);
822 void (*xdo_dev_state_delete) (struct xfrm_state *x);
823 void (*xdo_dev_state_free) (struct xfrm_state *x);
824 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
825 struct xfrm_state *x);
826 };
827 #endif
828
829 /*
830 * This structure defines the management hooks for network devices.
831 * The following hooks can be defined; unless noted otherwise, they are
832 * optional and can be filled with a null pointer.
833 *
834 * int (*ndo_init)(struct net_device *dev);
835 * This function is called once when a network device is registered.
836 * The network device can use this for any late stage initialization
837 * or semantic validation. It can fail with an error code which will
838 * be propagated back to register_netdev.
839 *
840 * void (*ndo_uninit)(struct net_device *dev);
841 * This function is called when device is unregistered or when registration
842 * fails. It is not called if init fails.
843 *
844 * int (*ndo_open)(struct net_device *dev);
845 * This function is called when a network device transitions to the up
846 * state.
847 *
848 * int (*ndo_stop)(struct net_device *dev);
849 * This function is called when a network device transitions to the down
850 * state.
851 *
852 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
853 * struct net_device *dev);
854 * Called when a packet needs to be transmitted.
855 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
856 * the queue before that can happen; it's for obsolete devices and weird
857 * corner cases, but the stack really does a non-trivial amount
858 * of useless work if you return NETDEV_TX_BUSY.
859 * Required; cannot be NULL.
860 *
861 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
862 * struct net_device *dev
863 * netdev_features_t features);
864 * Called by core transmit path to determine if device is capable of
865 * performing offload operations on a given packet. This is to give
866 * the device an opportunity to implement any restrictions that cannot
867 * be otherwise expressed by feature flags. The check is called with
868 * the set of features that the stack has calculated and it returns
869 * those the driver believes to be appropriate.
870 *
871 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
872 * void *accel_priv, select_queue_fallback_t fallback);
873 * Called to decide which queue to use when device supports multiple
874 * transmit queues.
875 *
876 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
877 * This function is called to allow device receiver to make
878 * changes to configuration when multicast or promiscuous is enabled.
879 *
880 * void (*ndo_set_rx_mode)(struct net_device *dev);
881 * This function is called device changes address list filtering.
882 * If driver handles unicast address filtering, it should set
883 * IFF_UNICAST_FLT in its priv_flags.
884 *
885 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
886 * This function is called when the Media Access Control address
887 * needs to be changed. If this interface is not defined, the
888 * MAC address can not be changed.
889 *
890 * int (*ndo_validate_addr)(struct net_device *dev);
891 * Test if Media Access Control address is valid for the device.
892 *
893 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
894 * Called when a user requests an ioctl which can't be handled by
895 * the generic interface code. If not defined ioctls return
896 * not supported error code.
897 *
898 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
899 * Used to set network devices bus interface parameters. This interface
900 * is retained for legacy reasons; new devices should use the bus
901 * interface (PCI) for low level management.
902 *
903 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
904 * Called when a user wants to change the Maximum Transfer Unit
905 * of a device.
906 *
907 * void (*ndo_tx_timeout)(struct net_device *dev);
908 * Callback used when the transmitter has not made any progress
909 * for dev->watchdog ticks.
910 *
911 * void (*ndo_get_stats64)(struct net_device *dev,
912 * struct rtnl_link_stats64 *storage);
913 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
914 * Called when a user wants to get the network device usage
915 * statistics. Drivers must do one of the following:
916 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
917 * rtnl_link_stats64 structure passed by the caller.
918 * 2. Define @ndo_get_stats to update a net_device_stats structure
919 * (which should normally be dev->stats) and return a pointer to
920 * it. The structure may be changed asynchronously only if each
921 * field is written atomically.
922 * 3. Update dev->stats asynchronously and atomically, and define
923 * neither operation.
924 *
925 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
926 * Return true if this device supports offload stats of this attr_id.
927 *
928 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
929 * void *attr_data)
930 * Get statistics for offload operations by attr_id. Write it into the
931 * attr_data pointer.
932 *
933 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
934 * If device supports VLAN filtering this function is called when a
935 * VLAN id is registered.
936 *
937 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
938 * If device supports VLAN filtering this function is called when a
939 * VLAN id is unregistered.
940 *
941 * void (*ndo_poll_controller)(struct net_device *dev);
942 *
943 * SR-IOV management functions.
944 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
945 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
946 * u8 qos, __be16 proto);
947 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
948 * int max_tx_rate);
949 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
950 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
951 * int (*ndo_get_vf_config)(struct net_device *dev,
952 * int vf, struct ifla_vf_info *ivf);
953 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
954 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
955 * struct nlattr *port[]);
956 *
957 * Enable or disable the VF ability to query its RSS Redirection Table and
958 * Hash Key. This is needed since on some devices VF share this information
959 * with PF and querying it may introduce a theoretical security risk.
960 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
961 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
962 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
963 * void *type_data);
964 * Called to setup any 'tc' scheduler, classifier or action on @dev.
965 * This is always called from the stack with the rtnl lock held and netif
966 * tx queues stopped. This allows the netdevice to perform queue
967 * management safely.
968 *
969 * Fiber Channel over Ethernet (FCoE) offload functions.
970 * int (*ndo_fcoe_enable)(struct net_device *dev);
971 * Called when the FCoE protocol stack wants to start using LLD for FCoE
972 * so the underlying device can perform whatever needed configuration or
973 * initialization to support acceleration of FCoE traffic.
974 *
975 * int (*ndo_fcoe_disable)(struct net_device *dev);
976 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
977 * so the underlying device can perform whatever needed clean-ups to
978 * stop supporting acceleration of FCoE traffic.
979 *
980 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
981 * struct scatterlist *sgl, unsigned int sgc);
982 * Called when the FCoE Initiator wants to initialize an I/O that
983 * is a possible candidate for Direct Data Placement (DDP). The LLD can
984 * perform necessary setup and returns 1 to indicate the device is set up
985 * successfully to perform DDP on this I/O, otherwise this returns 0.
986 *
987 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
988 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
989 * indicated by the FC exchange id 'xid', so the underlying device can
990 * clean up and reuse resources for later DDP requests.
991 *
992 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
993 * struct scatterlist *sgl, unsigned int sgc);
994 * Called when the FCoE Target wants to initialize an I/O that
995 * is a possible candidate for Direct Data Placement (DDP). The LLD can
996 * perform necessary setup and returns 1 to indicate the device is set up
997 * successfully to perform DDP on this I/O, otherwise this returns 0.
998 *
999 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1000 * struct netdev_fcoe_hbainfo *hbainfo);
1001 * Called when the FCoE Protocol stack wants information on the underlying
1002 * device. This information is utilized by the FCoE protocol stack to
1003 * register attributes with Fiber Channel management service as per the
1004 * FC-GS Fabric Device Management Information(FDMI) specification.
1005 *
1006 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1007 * Called when the underlying device wants to override default World Wide
1008 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1009 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1010 * protocol stack to use.
1011 *
1012 * RFS acceleration.
1013 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1014 * u16 rxq_index, u32 flow_id);
1015 * Set hardware filter for RFS. rxq_index is the target queue index;
1016 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1017 * Return the filter ID on success, or a negative error code.
1018 *
1019 * Slave management functions (for bridge, bonding, etc).
1020 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1021 * Called to make another netdev an underling.
1022 *
1023 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1024 * Called to release previously enslaved netdev.
1025 *
1026 * Feature/offload setting functions.
1027 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1028 * netdev_features_t features);
1029 * Adjusts the requested feature flags according to device-specific
1030 * constraints, and returns the resulting flags. Must not modify
1031 * the device state.
1032 *
1033 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1034 * Called to update device configuration to new features. Passed
1035 * feature set might be less than what was returned by ndo_fix_features()).
1036 * Must return >0 or -errno if it changed dev->features itself.
1037 *
1038 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1039 * struct net_device *dev,
1040 * const unsigned char *addr, u16 vid, u16 flags)
1041 * Adds an FDB entry to dev for addr.
1042 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1043 * struct net_device *dev,
1044 * const unsigned char *addr, u16 vid)
1045 * Deletes the FDB entry from dev coresponding to addr.
1046 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1047 * struct net_device *dev, struct net_device *filter_dev,
1048 * int *idx)
1049 * Used to add FDB entries to dump requests. Implementers should add
1050 * entries to skb and update idx with the number of entries.
1051 *
1052 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1053 * u16 flags)
1054 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1055 * struct net_device *dev, u32 filter_mask,
1056 * int nlflags)
1057 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1058 * u16 flags);
1059 *
1060 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1061 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1062 * which do not represent real hardware may define this to allow their
1063 * userspace components to manage their virtual carrier state. Devices
1064 * that determine carrier state from physical hardware properties (eg
1065 * network cables) or protocol-dependent mechanisms (eg
1066 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1067 *
1068 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1069 * struct netdev_phys_item_id *ppid);
1070 * Called to get ID of physical port of this device. If driver does
1071 * not implement this, it is assumed that the hw is not able to have
1072 * multiple net devices on single physical port.
1073 *
1074 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1075 * struct udp_tunnel_info *ti);
1076 * Called by UDP tunnel to notify a driver about the UDP port and socket
1077 * address family that a UDP tunnel is listnening to. It is called only
1078 * when a new port starts listening. The operation is protected by the
1079 * RTNL.
1080 *
1081 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1082 * struct udp_tunnel_info *ti);
1083 * Called by UDP tunnel to notify the driver about a UDP port and socket
1084 * address family that the UDP tunnel is not listening to anymore. The
1085 * operation is protected by the RTNL.
1086 *
1087 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1088 * struct net_device *dev)
1089 * Called by upper layer devices to accelerate switching or other
1090 * station functionality into hardware. 'pdev is the lowerdev
1091 * to use for the offload and 'dev' is the net device that will
1092 * back the offload. Returns a pointer to the private structure
1093 * the upper layer will maintain.
1094 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1095 * Called by upper layer device to delete the station created
1096 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1097 * the station and priv is the structure returned by the add
1098 * operation.
1099 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1100 * int queue_index, u32 maxrate);
1101 * Called when a user wants to set a max-rate limitation of specific
1102 * TX queue.
1103 * int (*ndo_get_iflink)(const struct net_device *dev);
1104 * Called to get the iflink value of this device.
1105 * void (*ndo_change_proto_down)(struct net_device *dev,
1106 * bool proto_down);
1107 * This function is used to pass protocol port error state information
1108 * to the switch driver. The switch driver can react to the proto_down
1109 * by doing a phys down on the associated switch port.
1110 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1111 * This function is used to get egress tunnel information for given skb.
1112 * This is useful for retrieving outer tunnel header parameters while
1113 * sampling packet.
1114 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1115 * This function is used to specify the headroom that the skb must
1116 * consider when allocation skb during packet reception. Setting
1117 * appropriate rx headroom value allows avoiding skb head copy on
1118 * forward. Setting a negative value resets the rx headroom to the
1119 * default value.
1120 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1121 * This function is used to set or query state related to XDP on the
1122 * netdevice. See definition of enum xdp_netdev_command for details.
1123 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1124 * This function is used to submit a XDP packet for transmit on a
1125 * netdevice.
1126 * void (*ndo_xdp_flush)(struct net_device *dev);
1127 * This function is used to inform the driver to flush a particular
1128 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1129 */
1130 struct net_device_ops {
1131 int (*ndo_init)(struct net_device *dev);
1132 void (*ndo_uninit)(struct net_device *dev);
1133 int (*ndo_open)(struct net_device *dev);
1134 int (*ndo_stop)(struct net_device *dev);
1135 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1136 struct net_device *dev);
1137 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1138 struct net_device *dev,
1139 netdev_features_t features);
1140 u16 (*ndo_select_queue)(struct net_device *dev,
1141 struct sk_buff *skb,
1142 void *accel_priv,
1143 select_queue_fallback_t fallback);
1144 void (*ndo_change_rx_flags)(struct net_device *dev,
1145 int flags);
1146 void (*ndo_set_rx_mode)(struct net_device *dev);
1147 int (*ndo_set_mac_address)(struct net_device *dev,
1148 void *addr);
1149 int (*ndo_validate_addr)(struct net_device *dev);
1150 int (*ndo_do_ioctl)(struct net_device *dev,
1151 struct ifreq *ifr, int cmd);
1152 int (*ndo_set_config)(struct net_device *dev,
1153 struct ifmap *map);
1154 int (*ndo_change_mtu)(struct net_device *dev,
1155 int new_mtu);
1156 int (*ndo_neigh_setup)(struct net_device *dev,
1157 struct neigh_parms *);
1158 void (*ndo_tx_timeout) (struct net_device *dev);
1159
1160 void (*ndo_get_stats64)(struct net_device *dev,
1161 struct rtnl_link_stats64 *storage);
1162 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1163 int (*ndo_get_offload_stats)(int attr_id,
1164 const struct net_device *dev,
1165 void *attr_data);
1166 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1167
1168 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1169 __be16 proto, u16 vid);
1170 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1171 __be16 proto, u16 vid);
1172 #ifdef CONFIG_NET_POLL_CONTROLLER
1173 void (*ndo_poll_controller)(struct net_device *dev);
1174 int (*ndo_netpoll_setup)(struct net_device *dev,
1175 struct netpoll_info *info);
1176 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1177 #endif
1178 int (*ndo_set_vf_mac)(struct net_device *dev,
1179 int queue, u8 *mac);
1180 int (*ndo_set_vf_vlan)(struct net_device *dev,
1181 int queue, u16 vlan,
1182 u8 qos, __be16 proto);
1183 int (*ndo_set_vf_rate)(struct net_device *dev,
1184 int vf, int min_tx_rate,
1185 int max_tx_rate);
1186 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1187 int vf, bool setting);
1188 int (*ndo_set_vf_trust)(struct net_device *dev,
1189 int vf, bool setting);
1190 int (*ndo_get_vf_config)(struct net_device *dev,
1191 int vf,
1192 struct ifla_vf_info *ivf);
1193 int (*ndo_set_vf_link_state)(struct net_device *dev,
1194 int vf, int link_state);
1195 int (*ndo_get_vf_stats)(struct net_device *dev,
1196 int vf,
1197 struct ifla_vf_stats
1198 *vf_stats);
1199 int (*ndo_set_vf_port)(struct net_device *dev,
1200 int vf,
1201 struct nlattr *port[]);
1202 int (*ndo_get_vf_port)(struct net_device *dev,
1203 int vf, struct sk_buff *skb);
1204 int (*ndo_set_vf_guid)(struct net_device *dev,
1205 int vf, u64 guid,
1206 int guid_type);
1207 int (*ndo_set_vf_rss_query_en)(
1208 struct net_device *dev,
1209 int vf, bool setting);
1210 int (*ndo_setup_tc)(struct net_device *dev,
1211 enum tc_setup_type type,
1212 void *type_data);
1213 #if IS_ENABLED(CONFIG_FCOE)
1214 int (*ndo_fcoe_enable)(struct net_device *dev);
1215 int (*ndo_fcoe_disable)(struct net_device *dev);
1216 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1217 u16 xid,
1218 struct scatterlist *sgl,
1219 unsigned int sgc);
1220 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1221 u16 xid);
1222 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1223 u16 xid,
1224 struct scatterlist *sgl,
1225 unsigned int sgc);
1226 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1227 struct netdev_fcoe_hbainfo *hbainfo);
1228 #endif
1229
1230 #if IS_ENABLED(CONFIG_LIBFCOE)
1231 #define NETDEV_FCOE_WWNN 0
1232 #define NETDEV_FCOE_WWPN 1
1233 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1234 u64 *wwn, int type);
1235 #endif
1236
1237 #ifdef CONFIG_RFS_ACCEL
1238 int (*ndo_rx_flow_steer)(struct net_device *dev,
1239 const struct sk_buff *skb,
1240 u16 rxq_index,
1241 u32 flow_id);
1242 #endif
1243 int (*ndo_add_slave)(struct net_device *dev,
1244 struct net_device *slave_dev);
1245 int (*ndo_del_slave)(struct net_device *dev,
1246 struct net_device *slave_dev);
1247 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1248 netdev_features_t features);
1249 int (*ndo_set_features)(struct net_device *dev,
1250 netdev_features_t features);
1251 int (*ndo_neigh_construct)(struct net_device *dev,
1252 struct neighbour *n);
1253 void (*ndo_neigh_destroy)(struct net_device *dev,
1254 struct neighbour *n);
1255
1256 int (*ndo_fdb_add)(struct ndmsg *ndm,
1257 struct nlattr *tb[],
1258 struct net_device *dev,
1259 const unsigned char *addr,
1260 u16 vid,
1261 u16 flags);
1262 int (*ndo_fdb_del)(struct ndmsg *ndm,
1263 struct nlattr *tb[],
1264 struct net_device *dev,
1265 const unsigned char *addr,
1266 u16 vid);
1267 int (*ndo_fdb_dump)(struct sk_buff *skb,
1268 struct netlink_callback *cb,
1269 struct net_device *dev,
1270 struct net_device *filter_dev,
1271 int *idx);
1272
1273 int (*ndo_bridge_setlink)(struct net_device *dev,
1274 struct nlmsghdr *nlh,
1275 u16 flags);
1276 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1277 u32 pid, u32 seq,
1278 struct net_device *dev,
1279 u32 filter_mask,
1280 int nlflags);
1281 int (*ndo_bridge_dellink)(struct net_device *dev,
1282 struct nlmsghdr *nlh,
1283 u16 flags);
1284 int (*ndo_change_carrier)(struct net_device *dev,
1285 bool new_carrier);
1286 int (*ndo_get_phys_port_id)(struct net_device *dev,
1287 struct netdev_phys_item_id *ppid);
1288 int (*ndo_get_phys_port_name)(struct net_device *dev,
1289 char *name, size_t len);
1290 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1291 struct udp_tunnel_info *ti);
1292 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1293 struct udp_tunnel_info *ti);
1294 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1295 struct net_device *dev);
1296 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1297 void *priv);
1298
1299 int (*ndo_get_lock_subclass)(struct net_device *dev);
1300 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1301 int queue_index,
1302 u32 maxrate);
1303 int (*ndo_get_iflink)(const struct net_device *dev);
1304 int (*ndo_change_proto_down)(struct net_device *dev,
1305 bool proto_down);
1306 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1307 struct sk_buff *skb);
1308 void (*ndo_set_rx_headroom)(struct net_device *dev,
1309 int needed_headroom);
1310 int (*ndo_xdp)(struct net_device *dev,
1311 struct netdev_xdp *xdp);
1312 int (*ndo_xdp_xmit)(struct net_device *dev,
1313 struct xdp_buff *xdp);
1314 void (*ndo_xdp_flush)(struct net_device *dev);
1315 };
1316
1317 /**
1318 * enum net_device_priv_flags - &struct net_device priv_flags
1319 *
1320 * These are the &struct net_device, they are only set internally
1321 * by drivers and used in the kernel. These flags are invisible to
1322 * userspace; this means that the order of these flags can change
1323 * during any kernel release.
1324 *
1325 * You should have a pretty good reason to be extending these flags.
1326 *
1327 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1328 * @IFF_EBRIDGE: Ethernet bridging device
1329 * @IFF_BONDING: bonding master or slave
1330 * @IFF_ISATAP: ISATAP interface (RFC4214)
1331 * @IFF_WAN_HDLC: WAN HDLC device
1332 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1333 * release skb->dst
1334 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1335 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1336 * @IFF_MACVLAN_PORT: device used as macvlan port
1337 * @IFF_BRIDGE_PORT: device used as bridge port
1338 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1339 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1340 * @IFF_UNICAST_FLT: Supports unicast filtering
1341 * @IFF_TEAM_PORT: device used as team port
1342 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1343 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1344 * change when it's running
1345 * @IFF_MACVLAN: Macvlan device
1346 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1347 * underlying stacked devices
1348 * @IFF_IPVLAN_MASTER: IPvlan master device
1349 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1350 * @IFF_L3MDEV_MASTER: device is an L3 master device
1351 * @IFF_NO_QUEUE: device can run without qdisc attached
1352 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1353 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1354 * @IFF_TEAM: device is a team device
1355 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1356 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1357 * entity (i.e. the master device for bridged veth)
1358 * @IFF_MACSEC: device is a MACsec device
1359 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1360 */
1361 enum netdev_priv_flags {
1362 IFF_802_1Q_VLAN = 1<<0,
1363 IFF_EBRIDGE = 1<<1,
1364 IFF_BONDING = 1<<2,
1365 IFF_ISATAP = 1<<3,
1366 IFF_WAN_HDLC = 1<<4,
1367 IFF_XMIT_DST_RELEASE = 1<<5,
1368 IFF_DONT_BRIDGE = 1<<6,
1369 IFF_DISABLE_NETPOLL = 1<<7,
1370 IFF_MACVLAN_PORT = 1<<8,
1371 IFF_BRIDGE_PORT = 1<<9,
1372 IFF_OVS_DATAPATH = 1<<10,
1373 IFF_TX_SKB_SHARING = 1<<11,
1374 IFF_UNICAST_FLT = 1<<12,
1375 IFF_TEAM_PORT = 1<<13,
1376 IFF_SUPP_NOFCS = 1<<14,
1377 IFF_LIVE_ADDR_CHANGE = 1<<15,
1378 IFF_MACVLAN = 1<<16,
1379 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1380 IFF_IPVLAN_MASTER = 1<<18,
1381 IFF_IPVLAN_SLAVE = 1<<19,
1382 IFF_L3MDEV_MASTER = 1<<20,
1383 IFF_NO_QUEUE = 1<<21,
1384 IFF_OPENVSWITCH = 1<<22,
1385 IFF_L3MDEV_SLAVE = 1<<23,
1386 IFF_TEAM = 1<<24,
1387 IFF_RXFH_CONFIGURED = 1<<25,
1388 IFF_PHONY_HEADROOM = 1<<26,
1389 IFF_MACSEC = 1<<27,
1390 IFF_L3MDEV_RX_HANDLER = 1<<28,
1391 };
1392
1393 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1394 #define IFF_EBRIDGE IFF_EBRIDGE
1395 #define IFF_BONDING IFF_BONDING
1396 #define IFF_ISATAP IFF_ISATAP
1397 #define IFF_WAN_HDLC IFF_WAN_HDLC
1398 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1399 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1400 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1401 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1402 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1403 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1404 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1405 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1406 #define IFF_TEAM_PORT IFF_TEAM_PORT
1407 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1408 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1409 #define IFF_MACVLAN IFF_MACVLAN
1410 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1411 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1412 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1413 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1414 #define IFF_NO_QUEUE IFF_NO_QUEUE
1415 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1416 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1417 #define IFF_TEAM IFF_TEAM
1418 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1419 #define IFF_MACSEC IFF_MACSEC
1420 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1421
1422 /**
1423 * struct net_device - The DEVICE structure.
1424 *
1425 * Actually, this whole structure is a big mistake. It mixes I/O
1426 * data with strictly "high-level" data, and it has to know about
1427 * almost every data structure used in the INET module.
1428 *
1429 * @name: This is the first field of the "visible" part of this structure
1430 * (i.e. as seen by users in the "Space.c" file). It is the name
1431 * of the interface.
1432 *
1433 * @name_hlist: Device name hash chain, please keep it close to name[]
1434 * @ifalias: SNMP alias
1435 * @mem_end: Shared memory end
1436 * @mem_start: Shared memory start
1437 * @base_addr: Device I/O address
1438 * @irq: Device IRQ number
1439 *
1440 * @carrier_changes: Stats to monitor carrier on<->off transitions
1441 *
1442 * @state: Generic network queuing layer state, see netdev_state_t
1443 * @dev_list: The global list of network devices
1444 * @napi_list: List entry used for polling NAPI devices
1445 * @unreg_list: List entry when we are unregistering the
1446 * device; see the function unregister_netdev
1447 * @close_list: List entry used when we are closing the device
1448 * @ptype_all: Device-specific packet handlers for all protocols
1449 * @ptype_specific: Device-specific, protocol-specific packet handlers
1450 *
1451 * @adj_list: Directly linked devices, like slaves for bonding
1452 * @features: Currently active device features
1453 * @hw_features: User-changeable features
1454 *
1455 * @wanted_features: User-requested features
1456 * @vlan_features: Mask of features inheritable by VLAN devices
1457 *
1458 * @hw_enc_features: Mask of features inherited by encapsulating devices
1459 * This field indicates what encapsulation
1460 * offloads the hardware is capable of doing,
1461 * and drivers will need to set them appropriately.
1462 *
1463 * @mpls_features: Mask of features inheritable by MPLS
1464 *
1465 * @ifindex: interface index
1466 * @group: The group the device belongs to
1467 *
1468 * @stats: Statistics struct, which was left as a legacy, use
1469 * rtnl_link_stats64 instead
1470 *
1471 * @rx_dropped: Dropped packets by core network,
1472 * do not use this in drivers
1473 * @tx_dropped: Dropped packets by core network,
1474 * do not use this in drivers
1475 * @rx_nohandler: nohandler dropped packets by core network on
1476 * inactive devices, do not use this in drivers
1477 *
1478 * @wireless_handlers: List of functions to handle Wireless Extensions,
1479 * instead of ioctl,
1480 * see <net/iw_handler.h> for details.
1481 * @wireless_data: Instance data managed by the core of wireless extensions
1482 *
1483 * @netdev_ops: Includes several pointers to callbacks,
1484 * if one wants to override the ndo_*() functions
1485 * @ethtool_ops: Management operations
1486 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1487 * discovery handling. Necessary for e.g. 6LoWPAN.
1488 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1489 * of Layer 2 headers.
1490 *
1491 * @flags: Interface flags (a la BSD)
1492 * @priv_flags: Like 'flags' but invisible to userspace,
1493 * see if.h for the definitions
1494 * @gflags: Global flags ( kept as legacy )
1495 * @padded: How much padding added by alloc_netdev()
1496 * @operstate: RFC2863 operstate
1497 * @link_mode: Mapping policy to operstate
1498 * @if_port: Selectable AUI, TP, ...
1499 * @dma: DMA channel
1500 * @mtu: Interface MTU value
1501 * @min_mtu: Interface Minimum MTU value
1502 * @max_mtu: Interface Maximum MTU value
1503 * @type: Interface hardware type
1504 * @hard_header_len: Maximum hardware header length.
1505 * @min_header_len: Minimum hardware header length
1506 *
1507 * @needed_headroom: Extra headroom the hardware may need, but not in all
1508 * cases can this be guaranteed
1509 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1510 * cases can this be guaranteed. Some cases also use
1511 * LL_MAX_HEADER instead to allocate the skb
1512 *
1513 * interface address info:
1514 *
1515 * @perm_addr: Permanent hw address
1516 * @addr_assign_type: Hw address assignment type
1517 * @addr_len: Hardware address length
1518 * @neigh_priv_len: Used in neigh_alloc()
1519 * @dev_id: Used to differentiate devices that share
1520 * the same link layer address
1521 * @dev_port: Used to differentiate devices that share
1522 * the same function
1523 * @addr_list_lock: XXX: need comments on this one
1524 * @uc_promisc: Counter that indicates promiscuous mode
1525 * has been enabled due to the need to listen to
1526 * additional unicast addresses in a device that
1527 * does not implement ndo_set_rx_mode()
1528 * @uc: unicast mac addresses
1529 * @mc: multicast mac addresses
1530 * @dev_addrs: list of device hw addresses
1531 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1532 * @promiscuity: Number of times the NIC is told to work in
1533 * promiscuous mode; if it becomes 0 the NIC will
1534 * exit promiscuous mode
1535 * @allmulti: Counter, enables or disables allmulticast mode
1536 *
1537 * @vlan_info: VLAN info
1538 * @dsa_ptr: dsa specific data
1539 * @tipc_ptr: TIPC specific data
1540 * @atalk_ptr: AppleTalk link
1541 * @ip_ptr: IPv4 specific data
1542 * @dn_ptr: DECnet specific data
1543 * @ip6_ptr: IPv6 specific data
1544 * @ax25_ptr: AX.25 specific data
1545 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1546 *
1547 * @dev_addr: Hw address (before bcast,
1548 * because most packets are unicast)
1549 *
1550 * @_rx: Array of RX queues
1551 * @num_rx_queues: Number of RX queues
1552 * allocated at register_netdev() time
1553 * @real_num_rx_queues: Number of RX queues currently active in device
1554 *
1555 * @rx_handler: handler for received packets
1556 * @rx_handler_data: XXX: need comments on this one
1557 * @ingress_queue: XXX: need comments on this one
1558 * @broadcast: hw bcast address
1559 *
1560 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1561 * indexed by RX queue number. Assigned by driver.
1562 * This must only be set if the ndo_rx_flow_steer
1563 * operation is defined
1564 * @index_hlist: Device index hash chain
1565 *
1566 * @_tx: Array of TX queues
1567 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1568 * @real_num_tx_queues: Number of TX queues currently active in device
1569 * @qdisc: Root qdisc from userspace point of view
1570 * @tx_queue_len: Max frames per queue allowed
1571 * @tx_global_lock: XXX: need comments on this one
1572 *
1573 * @xps_maps: XXX: need comments on this one
1574 *
1575 * @watchdog_timeo: Represents the timeout that is used by
1576 * the watchdog (see dev_watchdog())
1577 * @watchdog_timer: List of timers
1578 *
1579 * @pcpu_refcnt: Number of references to this device
1580 * @todo_list: Delayed register/unregister
1581 * @link_watch_list: XXX: need comments on this one
1582 *
1583 * @reg_state: Register/unregister state machine
1584 * @dismantle: Device is going to be freed
1585 * @rtnl_link_state: This enum represents the phases of creating
1586 * a new link
1587 *
1588 * @needs_free_netdev: Should unregister perform free_netdev?
1589 * @priv_destructor: Called from unregister
1590 * @npinfo: XXX: need comments on this one
1591 * @nd_net: Network namespace this network device is inside
1592 *
1593 * @ml_priv: Mid-layer private
1594 * @lstats: Loopback statistics
1595 * @tstats: Tunnel statistics
1596 * @dstats: Dummy statistics
1597 * @vstats: Virtual ethernet statistics
1598 *
1599 * @garp_port: GARP
1600 * @mrp_port: MRP
1601 *
1602 * @dev: Class/net/name entry
1603 * @sysfs_groups: Space for optional device, statistics and wireless
1604 * sysfs groups
1605 *
1606 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1607 * @rtnl_link_ops: Rtnl_link_ops
1608 *
1609 * @gso_max_size: Maximum size of generic segmentation offload
1610 * @gso_max_segs: Maximum number of segments that can be passed to the
1611 * NIC for GSO
1612 *
1613 * @dcbnl_ops: Data Center Bridging netlink ops
1614 * @num_tc: Number of traffic classes in the net device
1615 * @tc_to_txq: XXX: need comments on this one
1616 * @prio_tc_map: XXX: need comments on this one
1617 *
1618 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1619 *
1620 * @priomap: XXX: need comments on this one
1621 * @phydev: Physical device may attach itself
1622 * for hardware timestamping
1623 *
1624 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1625 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1626 *
1627 * @proto_down: protocol port state information can be sent to the
1628 * switch driver and used to set the phys state of the
1629 * switch port.
1630 *
1631 * FIXME: cleanup struct net_device such that network protocol info
1632 * moves out.
1633 */
1634
1635 struct net_device {
1636 char name[IFNAMSIZ];
1637 struct hlist_node name_hlist;
1638 char *ifalias;
1639 /*
1640 * I/O specific fields
1641 * FIXME: Merge these and struct ifmap into one
1642 */
1643 unsigned long mem_end;
1644 unsigned long mem_start;
1645 unsigned long base_addr;
1646 int irq;
1647
1648 atomic_t carrier_changes;
1649
1650 /*
1651 * Some hardware also needs these fields (state,dev_list,
1652 * napi_list,unreg_list,close_list) but they are not
1653 * part of the usual set specified in Space.c.
1654 */
1655
1656 unsigned long state;
1657
1658 struct list_head dev_list;
1659 struct list_head napi_list;
1660 struct list_head unreg_list;
1661 struct list_head close_list;
1662 struct list_head ptype_all;
1663 struct list_head ptype_specific;
1664
1665 struct {
1666 struct list_head upper;
1667 struct list_head lower;
1668 } adj_list;
1669
1670 netdev_features_t features;
1671 netdev_features_t hw_features;
1672 netdev_features_t wanted_features;
1673 netdev_features_t vlan_features;
1674 netdev_features_t hw_enc_features;
1675 netdev_features_t mpls_features;
1676 netdev_features_t gso_partial_features;
1677
1678 int ifindex;
1679 int group;
1680
1681 struct net_device_stats stats;
1682
1683 atomic_long_t rx_dropped;
1684 atomic_long_t tx_dropped;
1685 atomic_long_t rx_nohandler;
1686
1687 #ifdef CONFIG_WIRELESS_EXT
1688 const struct iw_handler_def *wireless_handlers;
1689 struct iw_public_data *wireless_data;
1690 #endif
1691 const struct net_device_ops *netdev_ops;
1692 const struct ethtool_ops *ethtool_ops;
1693 #ifdef CONFIG_NET_SWITCHDEV
1694 const struct switchdev_ops *switchdev_ops;
1695 #endif
1696 #ifdef CONFIG_NET_L3_MASTER_DEV
1697 const struct l3mdev_ops *l3mdev_ops;
1698 #endif
1699 #if IS_ENABLED(CONFIG_IPV6)
1700 const struct ndisc_ops *ndisc_ops;
1701 #endif
1702
1703 #ifdef CONFIG_XFRM
1704 const struct xfrmdev_ops *xfrmdev_ops;
1705 #endif
1706
1707 const struct header_ops *header_ops;
1708
1709 unsigned int flags;
1710 unsigned int priv_flags;
1711
1712 unsigned short gflags;
1713 unsigned short padded;
1714
1715 unsigned char operstate;
1716 unsigned char link_mode;
1717
1718 unsigned char if_port;
1719 unsigned char dma;
1720
1721 unsigned int mtu;
1722 unsigned int min_mtu;
1723 unsigned int max_mtu;
1724 unsigned short type;
1725 unsigned short hard_header_len;
1726 unsigned char min_header_len;
1727
1728 unsigned short needed_headroom;
1729 unsigned short needed_tailroom;
1730
1731 /* Interface address info. */
1732 unsigned char perm_addr[MAX_ADDR_LEN];
1733 unsigned char addr_assign_type;
1734 unsigned char addr_len;
1735 unsigned short neigh_priv_len;
1736 unsigned short dev_id;
1737 unsigned short dev_port;
1738 spinlock_t addr_list_lock;
1739 unsigned char name_assign_type;
1740 bool uc_promisc;
1741 struct netdev_hw_addr_list uc;
1742 struct netdev_hw_addr_list mc;
1743 struct netdev_hw_addr_list dev_addrs;
1744
1745 #ifdef CONFIG_SYSFS
1746 struct kset *queues_kset;
1747 #endif
1748 unsigned int promiscuity;
1749 unsigned int allmulti;
1750
1751
1752 /* Protocol-specific pointers */
1753
1754 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1755 struct vlan_info __rcu *vlan_info;
1756 #endif
1757 #if IS_ENABLED(CONFIG_NET_DSA)
1758 struct dsa_switch_tree *dsa_ptr;
1759 #endif
1760 #if IS_ENABLED(CONFIG_TIPC)
1761 struct tipc_bearer __rcu *tipc_ptr;
1762 #endif
1763 void *atalk_ptr;
1764 struct in_device __rcu *ip_ptr;
1765 struct dn_dev __rcu *dn_ptr;
1766 struct inet6_dev __rcu *ip6_ptr;
1767 void *ax25_ptr;
1768 struct wireless_dev *ieee80211_ptr;
1769 struct wpan_dev *ieee802154_ptr;
1770 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1771 struct mpls_dev __rcu *mpls_ptr;
1772 #endif
1773
1774 /*
1775 * Cache lines mostly used on receive path (including eth_type_trans())
1776 */
1777 /* Interface address info used in eth_type_trans() */
1778 unsigned char *dev_addr;
1779
1780 #ifdef CONFIG_SYSFS
1781 struct netdev_rx_queue *_rx;
1782
1783 unsigned int num_rx_queues;
1784 unsigned int real_num_rx_queues;
1785 #endif
1786
1787 struct bpf_prog __rcu *xdp_prog;
1788 unsigned long gro_flush_timeout;
1789 rx_handler_func_t __rcu *rx_handler;
1790 void __rcu *rx_handler_data;
1791
1792 #ifdef CONFIG_NET_CLS_ACT
1793 struct tcf_proto __rcu *ingress_cl_list;
1794 #endif
1795 struct netdev_queue __rcu *ingress_queue;
1796 #ifdef CONFIG_NETFILTER_INGRESS
1797 struct nf_hook_entries __rcu *nf_hooks_ingress;
1798 #endif
1799
1800 unsigned char broadcast[MAX_ADDR_LEN];
1801 #ifdef CONFIG_RFS_ACCEL
1802 struct cpu_rmap *rx_cpu_rmap;
1803 #endif
1804 struct hlist_node index_hlist;
1805
1806 /*
1807 * Cache lines mostly used on transmit path
1808 */
1809 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1810 unsigned int num_tx_queues;
1811 unsigned int real_num_tx_queues;
1812 struct Qdisc *qdisc;
1813 #ifdef CONFIG_NET_SCHED
1814 DECLARE_HASHTABLE (qdisc_hash, 4);
1815 #endif
1816 unsigned int tx_queue_len;
1817 spinlock_t tx_global_lock;
1818 int watchdog_timeo;
1819
1820 #ifdef CONFIG_XPS
1821 struct xps_dev_maps __rcu *xps_maps;
1822 #endif
1823 #ifdef CONFIG_NET_CLS_ACT
1824 struct tcf_proto __rcu *egress_cl_list;
1825 #endif
1826
1827 /* These may be needed for future network-power-down code. */
1828 struct timer_list watchdog_timer;
1829
1830 int __percpu *pcpu_refcnt;
1831 struct list_head todo_list;
1832
1833 struct list_head link_watch_list;
1834
1835 enum { NETREG_UNINITIALIZED=0,
1836 NETREG_REGISTERED, /* completed register_netdevice */
1837 NETREG_UNREGISTERING, /* called unregister_netdevice */
1838 NETREG_UNREGISTERED, /* completed unregister todo */
1839 NETREG_RELEASED, /* called free_netdev */
1840 NETREG_DUMMY, /* dummy device for NAPI poll */
1841 } reg_state:8;
1842
1843 bool dismantle;
1844
1845 enum {
1846 RTNL_LINK_INITIALIZED,
1847 RTNL_LINK_INITIALIZING,
1848 } rtnl_link_state:16;
1849
1850 bool needs_free_netdev;
1851 void (*priv_destructor)(struct net_device *dev);
1852
1853 #ifdef CONFIG_NETPOLL
1854 struct netpoll_info __rcu *npinfo;
1855 #endif
1856
1857 possible_net_t nd_net;
1858
1859 /* mid-layer private */
1860 union {
1861 void *ml_priv;
1862 struct pcpu_lstats __percpu *lstats;
1863 struct pcpu_sw_netstats __percpu *tstats;
1864 struct pcpu_dstats __percpu *dstats;
1865 struct pcpu_vstats __percpu *vstats;
1866 };
1867
1868 #if IS_ENABLED(CONFIG_GARP)
1869 struct garp_port __rcu *garp_port;
1870 #endif
1871 #if IS_ENABLED(CONFIG_MRP)
1872 struct mrp_port __rcu *mrp_port;
1873 #endif
1874
1875 struct device dev;
1876 const struct attribute_group *sysfs_groups[4];
1877 const struct attribute_group *sysfs_rx_queue_group;
1878
1879 const struct rtnl_link_ops *rtnl_link_ops;
1880
1881 /* for setting kernel sock attribute on TCP connection setup */
1882 #define GSO_MAX_SIZE 65536
1883 unsigned int gso_max_size;
1884 #define GSO_MAX_SEGS 65535
1885 u16 gso_max_segs;
1886
1887 #ifdef CONFIG_DCB
1888 const struct dcbnl_rtnl_ops *dcbnl_ops;
1889 #endif
1890 u8 num_tc;
1891 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1892 u8 prio_tc_map[TC_BITMASK + 1];
1893
1894 #if IS_ENABLED(CONFIG_FCOE)
1895 unsigned int fcoe_ddp_xid;
1896 #endif
1897 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1898 struct netprio_map __rcu *priomap;
1899 #endif
1900 struct phy_device *phydev;
1901 struct lock_class_key *qdisc_tx_busylock;
1902 struct lock_class_key *qdisc_running_key;
1903 bool proto_down;
1904 };
1905 #define to_net_dev(d) container_of(d, struct net_device, dev)
1906
1907 static inline bool netif_elide_gro(const struct net_device *dev)
1908 {
1909 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1910 return true;
1911 return false;
1912 }
1913
1914 #define NETDEV_ALIGN 32
1915
1916 static inline
1917 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1918 {
1919 return dev->prio_tc_map[prio & TC_BITMASK];
1920 }
1921
1922 static inline
1923 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1924 {
1925 if (tc >= dev->num_tc)
1926 return -EINVAL;
1927
1928 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1929 return 0;
1930 }
1931
1932 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1933 void netdev_reset_tc(struct net_device *dev);
1934 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1935 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1936
1937 static inline
1938 int netdev_get_num_tc(struct net_device *dev)
1939 {
1940 return dev->num_tc;
1941 }
1942
1943 static inline
1944 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1945 unsigned int index)
1946 {
1947 return &dev->_tx[index];
1948 }
1949
1950 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1951 const struct sk_buff *skb)
1952 {
1953 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1954 }
1955
1956 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1957 void (*f)(struct net_device *,
1958 struct netdev_queue *,
1959 void *),
1960 void *arg)
1961 {
1962 unsigned int i;
1963
1964 for (i = 0; i < dev->num_tx_queues; i++)
1965 f(dev, &dev->_tx[i], arg);
1966 }
1967
1968 #define netdev_lockdep_set_classes(dev) \
1969 { \
1970 static struct lock_class_key qdisc_tx_busylock_key; \
1971 static struct lock_class_key qdisc_running_key; \
1972 static struct lock_class_key qdisc_xmit_lock_key; \
1973 static struct lock_class_key dev_addr_list_lock_key; \
1974 unsigned int i; \
1975 \
1976 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1977 (dev)->qdisc_running_key = &qdisc_running_key; \
1978 lockdep_set_class(&(dev)->addr_list_lock, \
1979 &dev_addr_list_lock_key); \
1980 for (i = 0; i < (dev)->num_tx_queues; i++) \
1981 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1982 &qdisc_xmit_lock_key); \
1983 }
1984
1985 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1986 struct sk_buff *skb,
1987 void *accel_priv);
1988
1989 /* returns the headroom that the master device needs to take in account
1990 * when forwarding to this dev
1991 */
1992 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1993 {
1994 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1995 }
1996
1997 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1998 {
1999 if (dev->netdev_ops->ndo_set_rx_headroom)
2000 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2001 }
2002
2003 /* set the device rx headroom to the dev's default */
2004 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2005 {
2006 netdev_set_rx_headroom(dev, -1);
2007 }
2008
2009 /*
2010 * Net namespace inlines
2011 */
2012 static inline
2013 struct net *dev_net(const struct net_device *dev)
2014 {
2015 return read_pnet(&dev->nd_net);
2016 }
2017
2018 static inline
2019 void dev_net_set(struct net_device *dev, struct net *net)
2020 {
2021 write_pnet(&dev->nd_net, net);
2022 }
2023
2024 /**
2025 * netdev_priv - access network device private data
2026 * @dev: network device
2027 *
2028 * Get network device private data
2029 */
2030 static inline void *netdev_priv(const struct net_device *dev)
2031 {
2032 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2033 }
2034
2035 /* Set the sysfs physical device reference for the network logical device
2036 * if set prior to registration will cause a symlink during initialization.
2037 */
2038 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2039
2040 /* Set the sysfs device type for the network logical device to allow
2041 * fine-grained identification of different network device types. For
2042 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2043 */
2044 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2045
2046 /* Default NAPI poll() weight
2047 * Device drivers are strongly advised to not use bigger value
2048 */
2049 #define NAPI_POLL_WEIGHT 64
2050
2051 /**
2052 * netif_napi_add - initialize a NAPI context
2053 * @dev: network device
2054 * @napi: NAPI context
2055 * @poll: polling function
2056 * @weight: default weight
2057 *
2058 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2059 * *any* of the other NAPI-related functions.
2060 */
2061 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2062 int (*poll)(struct napi_struct *, int), int weight);
2063
2064 /**
2065 * netif_tx_napi_add - initialize a NAPI context
2066 * @dev: network device
2067 * @napi: NAPI context
2068 * @poll: polling function
2069 * @weight: default weight
2070 *
2071 * This variant of netif_napi_add() should be used from drivers using NAPI
2072 * to exclusively poll a TX queue.
2073 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2074 */
2075 static inline void netif_tx_napi_add(struct net_device *dev,
2076 struct napi_struct *napi,
2077 int (*poll)(struct napi_struct *, int),
2078 int weight)
2079 {
2080 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2081 netif_napi_add(dev, napi, poll, weight);
2082 }
2083
2084 /**
2085 * netif_napi_del - remove a NAPI context
2086 * @napi: NAPI context
2087 *
2088 * netif_napi_del() removes a NAPI context from the network device NAPI list
2089 */
2090 void netif_napi_del(struct napi_struct *napi);
2091
2092 struct napi_gro_cb {
2093 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2094 void *frag0;
2095
2096 /* Length of frag0. */
2097 unsigned int frag0_len;
2098
2099 /* This indicates where we are processing relative to skb->data. */
2100 int data_offset;
2101
2102 /* This is non-zero if the packet cannot be merged with the new skb. */
2103 u16 flush;
2104
2105 /* Save the IP ID here and check when we get to the transport layer */
2106 u16 flush_id;
2107
2108 /* Number of segments aggregated. */
2109 u16 count;
2110
2111 /* Start offset for remote checksum offload */
2112 u16 gro_remcsum_start;
2113
2114 /* jiffies when first packet was created/queued */
2115 unsigned long age;
2116
2117 /* Used in ipv6_gro_receive() and foo-over-udp */
2118 u16 proto;
2119
2120 /* This is non-zero if the packet may be of the same flow. */
2121 u8 same_flow:1;
2122
2123 /* Used in tunnel GRO receive */
2124 u8 encap_mark:1;
2125
2126 /* GRO checksum is valid */
2127 u8 csum_valid:1;
2128
2129 /* Number of checksums via CHECKSUM_UNNECESSARY */
2130 u8 csum_cnt:3;
2131
2132 /* Free the skb? */
2133 u8 free:2;
2134 #define NAPI_GRO_FREE 1
2135 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2136
2137 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2138 u8 is_ipv6:1;
2139
2140 /* Used in GRE, set in fou/gue_gro_receive */
2141 u8 is_fou:1;
2142
2143 /* Used to determine if flush_id can be ignored */
2144 u8 is_atomic:1;
2145
2146 /* Number of gro_receive callbacks this packet already went through */
2147 u8 recursion_counter:4;
2148
2149 /* 1 bit hole */
2150
2151 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2152 __wsum csum;
2153
2154 /* used in skb_gro_receive() slow path */
2155 struct sk_buff *last;
2156 };
2157
2158 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2159
2160 #define GRO_RECURSION_LIMIT 15
2161 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2162 {
2163 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2164 }
2165
2166 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2167 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2168 struct sk_buff **head,
2169 struct sk_buff *skb)
2170 {
2171 if (unlikely(gro_recursion_inc_test(skb))) {
2172 NAPI_GRO_CB(skb)->flush |= 1;
2173 return NULL;
2174 }
2175
2176 return cb(head, skb);
2177 }
2178
2179 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2180 struct sk_buff *);
2181 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2182 struct sock *sk,
2183 struct sk_buff **head,
2184 struct sk_buff *skb)
2185 {
2186 if (unlikely(gro_recursion_inc_test(skb))) {
2187 NAPI_GRO_CB(skb)->flush |= 1;
2188 return NULL;
2189 }
2190
2191 return cb(sk, head, skb);
2192 }
2193
2194 struct packet_type {
2195 __be16 type; /* This is really htons(ether_type). */
2196 struct net_device *dev; /* NULL is wildcarded here */
2197 int (*func) (struct sk_buff *,
2198 struct net_device *,
2199 struct packet_type *,
2200 struct net_device *);
2201 bool (*id_match)(struct packet_type *ptype,
2202 struct sock *sk);
2203 void *af_packet_priv;
2204 struct list_head list;
2205 };
2206
2207 struct offload_callbacks {
2208 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2209 netdev_features_t features);
2210 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2211 struct sk_buff *skb);
2212 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2213 };
2214
2215 struct packet_offload {
2216 __be16 type; /* This is really htons(ether_type). */
2217 u16 priority;
2218 struct offload_callbacks callbacks;
2219 struct list_head list;
2220 };
2221
2222 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2223 struct pcpu_sw_netstats {
2224 u64 rx_packets;
2225 u64 rx_bytes;
2226 u64 tx_packets;
2227 u64 tx_bytes;
2228 struct u64_stats_sync syncp;
2229 };
2230
2231 #define __netdev_alloc_pcpu_stats(type, gfp) \
2232 ({ \
2233 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2234 if (pcpu_stats) { \
2235 int __cpu; \
2236 for_each_possible_cpu(__cpu) { \
2237 typeof(type) *stat; \
2238 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2239 u64_stats_init(&stat->syncp); \
2240 } \
2241 } \
2242 pcpu_stats; \
2243 })
2244
2245 #define netdev_alloc_pcpu_stats(type) \
2246 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2247
2248 enum netdev_lag_tx_type {
2249 NETDEV_LAG_TX_TYPE_UNKNOWN,
2250 NETDEV_LAG_TX_TYPE_RANDOM,
2251 NETDEV_LAG_TX_TYPE_BROADCAST,
2252 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2253 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2254 NETDEV_LAG_TX_TYPE_HASH,
2255 };
2256
2257 struct netdev_lag_upper_info {
2258 enum netdev_lag_tx_type tx_type;
2259 };
2260
2261 struct netdev_lag_lower_state_info {
2262 u8 link_up : 1,
2263 tx_enabled : 1;
2264 };
2265
2266 #include <linux/notifier.h>
2267
2268 /* netdevice notifier chain. Please remember to update the rtnetlink
2269 * notification exclusion list in rtnetlink_event() when adding new
2270 * types.
2271 */
2272 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2273 #define NETDEV_DOWN 0x0002
2274 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2275 detected a hardware crash and restarted
2276 - we can use this eg to kick tcp sessions
2277 once done */
2278 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2279 #define NETDEV_REGISTER 0x0005
2280 #define NETDEV_UNREGISTER 0x0006
2281 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2282 #define NETDEV_CHANGEADDR 0x0008
2283 #define NETDEV_GOING_DOWN 0x0009
2284 #define NETDEV_CHANGENAME 0x000A
2285 #define NETDEV_FEAT_CHANGE 0x000B
2286 #define NETDEV_BONDING_FAILOVER 0x000C
2287 #define NETDEV_PRE_UP 0x000D
2288 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2289 #define NETDEV_POST_TYPE_CHANGE 0x000F
2290 #define NETDEV_POST_INIT 0x0010
2291 #define NETDEV_UNREGISTER_FINAL 0x0011
2292 #define NETDEV_RELEASE 0x0012
2293 #define NETDEV_NOTIFY_PEERS 0x0013
2294 #define NETDEV_JOIN 0x0014
2295 #define NETDEV_CHANGEUPPER 0x0015
2296 #define NETDEV_RESEND_IGMP 0x0016
2297 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2298 #define NETDEV_CHANGEINFODATA 0x0018
2299 #define NETDEV_BONDING_INFO 0x0019
2300 #define NETDEV_PRECHANGEUPPER 0x001A
2301 #define NETDEV_CHANGELOWERSTATE 0x001B
2302 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2303 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2304 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2305
2306 int register_netdevice_notifier(struct notifier_block *nb);
2307 int unregister_netdevice_notifier(struct notifier_block *nb);
2308
2309 struct netdev_notifier_info {
2310 struct net_device *dev;
2311 };
2312
2313 struct netdev_notifier_info_ext {
2314 struct netdev_notifier_info info; /* must be first */
2315 union {
2316 u32 mtu;
2317 } ext;
2318 };
2319
2320 struct netdev_notifier_change_info {
2321 struct netdev_notifier_info info; /* must be first */
2322 unsigned int flags_changed;
2323 };
2324
2325 struct netdev_notifier_changeupper_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 struct net_device *upper_dev; /* new upper dev */
2328 bool master; /* is upper dev master */
2329 bool linking; /* is the notification for link or unlink */
2330 void *upper_info; /* upper dev info */
2331 };
2332
2333 struct netdev_notifier_changelowerstate_info {
2334 struct netdev_notifier_info info; /* must be first */
2335 void *lower_state_info; /* is lower dev state */
2336 };
2337
2338 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2339 struct net_device *dev)
2340 {
2341 info->dev = dev;
2342 }
2343
2344 static inline struct net_device *
2345 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2346 {
2347 return info->dev;
2348 }
2349
2350 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2351
2352
2353 extern rwlock_t dev_base_lock; /* Device list lock */
2354
2355 #define for_each_netdev(net, d) \
2356 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2357 #define for_each_netdev_reverse(net, d) \
2358 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2359 #define for_each_netdev_rcu(net, d) \
2360 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2361 #define for_each_netdev_safe(net, d, n) \
2362 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2363 #define for_each_netdev_continue(net, d) \
2364 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2365 #define for_each_netdev_continue_rcu(net, d) \
2366 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2367 #define for_each_netdev_in_bond_rcu(bond, slave) \
2368 for_each_netdev_rcu(&init_net, slave) \
2369 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2370 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2371
2372 static inline struct net_device *next_net_device(struct net_device *dev)
2373 {
2374 struct list_head *lh;
2375 struct net *net;
2376
2377 net = dev_net(dev);
2378 lh = dev->dev_list.next;
2379 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2380 }
2381
2382 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2383 {
2384 struct list_head *lh;
2385 struct net *net;
2386
2387 net = dev_net(dev);
2388 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2389 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2390 }
2391
2392 static inline struct net_device *first_net_device(struct net *net)
2393 {
2394 return list_empty(&net->dev_base_head) ? NULL :
2395 net_device_entry(net->dev_base_head.next);
2396 }
2397
2398 static inline struct net_device *first_net_device_rcu(struct net *net)
2399 {
2400 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2401
2402 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2403 }
2404
2405 int netdev_boot_setup_check(struct net_device *dev);
2406 unsigned long netdev_boot_base(const char *prefix, int unit);
2407 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2408 const char *hwaddr);
2409 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2410 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2411 void dev_add_pack(struct packet_type *pt);
2412 void dev_remove_pack(struct packet_type *pt);
2413 void __dev_remove_pack(struct packet_type *pt);
2414 void dev_add_offload(struct packet_offload *po);
2415 void dev_remove_offload(struct packet_offload *po);
2416
2417 int dev_get_iflink(const struct net_device *dev);
2418 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2419 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2420 unsigned short mask);
2421 struct net_device *dev_get_by_name(struct net *net, const char *name);
2422 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2423 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2424 int dev_alloc_name(struct net_device *dev, const char *name);
2425 int dev_open(struct net_device *dev);
2426 void dev_close(struct net_device *dev);
2427 void dev_close_many(struct list_head *head, bool unlink);
2428 void dev_disable_lro(struct net_device *dev);
2429 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2430 int dev_queue_xmit(struct sk_buff *skb);
2431 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2432 int register_netdevice(struct net_device *dev);
2433 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2434 void unregister_netdevice_many(struct list_head *head);
2435 static inline void unregister_netdevice(struct net_device *dev)
2436 {
2437 unregister_netdevice_queue(dev, NULL);
2438 }
2439
2440 int netdev_refcnt_read(const struct net_device *dev);
2441 void free_netdev(struct net_device *dev);
2442 void netdev_freemem(struct net_device *dev);
2443 void synchronize_net(void);
2444 int init_dummy_netdev(struct net_device *dev);
2445
2446 DECLARE_PER_CPU(int, xmit_recursion);
2447 #define XMIT_RECURSION_LIMIT 10
2448
2449 static inline int dev_recursion_level(void)
2450 {
2451 return this_cpu_read(xmit_recursion);
2452 }
2453
2454 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2455 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2456 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2457 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2458 int netdev_get_name(struct net *net, char *name, int ifindex);
2459 int dev_restart(struct net_device *dev);
2460 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2461
2462 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2463 {
2464 return NAPI_GRO_CB(skb)->data_offset;
2465 }
2466
2467 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2468 {
2469 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2470 }
2471
2472 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2473 {
2474 NAPI_GRO_CB(skb)->data_offset += len;
2475 }
2476
2477 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2478 unsigned int offset)
2479 {
2480 return NAPI_GRO_CB(skb)->frag0 + offset;
2481 }
2482
2483 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2484 {
2485 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2486 }
2487
2488 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2489 {
2490 NAPI_GRO_CB(skb)->frag0 = NULL;
2491 NAPI_GRO_CB(skb)->frag0_len = 0;
2492 }
2493
2494 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2495 unsigned int offset)
2496 {
2497 if (!pskb_may_pull(skb, hlen))
2498 return NULL;
2499
2500 skb_gro_frag0_invalidate(skb);
2501 return skb->data + offset;
2502 }
2503
2504 static inline void *skb_gro_network_header(struct sk_buff *skb)
2505 {
2506 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2507 skb_network_offset(skb);
2508 }
2509
2510 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2511 const void *start, unsigned int len)
2512 {
2513 if (NAPI_GRO_CB(skb)->csum_valid)
2514 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2515 csum_partial(start, len, 0));
2516 }
2517
2518 /* GRO checksum functions. These are logical equivalents of the normal
2519 * checksum functions (in skbuff.h) except that they operate on the GRO
2520 * offsets and fields in sk_buff.
2521 */
2522
2523 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2524
2525 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2526 {
2527 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2528 }
2529
2530 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2531 bool zero_okay,
2532 __sum16 check)
2533 {
2534 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2535 skb_checksum_start_offset(skb) <
2536 skb_gro_offset(skb)) &&
2537 !skb_at_gro_remcsum_start(skb) &&
2538 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2539 (!zero_okay || check));
2540 }
2541
2542 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2543 __wsum psum)
2544 {
2545 if (NAPI_GRO_CB(skb)->csum_valid &&
2546 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2547 return 0;
2548
2549 NAPI_GRO_CB(skb)->csum = psum;
2550
2551 return __skb_gro_checksum_complete(skb);
2552 }
2553
2554 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2555 {
2556 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2557 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2558 NAPI_GRO_CB(skb)->csum_cnt--;
2559 } else {
2560 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2561 * verified a new top level checksum or an encapsulated one
2562 * during GRO. This saves work if we fallback to normal path.
2563 */
2564 __skb_incr_checksum_unnecessary(skb);
2565 }
2566 }
2567
2568 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2569 compute_pseudo) \
2570 ({ \
2571 __sum16 __ret = 0; \
2572 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2573 __ret = __skb_gro_checksum_validate_complete(skb, \
2574 compute_pseudo(skb, proto)); \
2575 if (!__ret) \
2576 skb_gro_incr_csum_unnecessary(skb); \
2577 __ret; \
2578 })
2579
2580 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2581 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2582
2583 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2584 compute_pseudo) \
2585 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2586
2587 #define skb_gro_checksum_simple_validate(skb) \
2588 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2589
2590 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2591 {
2592 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2593 !NAPI_GRO_CB(skb)->csum_valid);
2594 }
2595
2596 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2597 __sum16 check, __wsum pseudo)
2598 {
2599 NAPI_GRO_CB(skb)->csum = ~pseudo;
2600 NAPI_GRO_CB(skb)->csum_valid = 1;
2601 }
2602
2603 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2604 do { \
2605 if (__skb_gro_checksum_convert_check(skb)) \
2606 __skb_gro_checksum_convert(skb, check, \
2607 compute_pseudo(skb, proto)); \
2608 } while (0)
2609
2610 struct gro_remcsum {
2611 int offset;
2612 __wsum delta;
2613 };
2614
2615 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2616 {
2617 grc->offset = 0;
2618 grc->delta = 0;
2619 }
2620
2621 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2622 unsigned int off, size_t hdrlen,
2623 int start, int offset,
2624 struct gro_remcsum *grc,
2625 bool nopartial)
2626 {
2627 __wsum delta;
2628 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2629
2630 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2631
2632 if (!nopartial) {
2633 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2634 return ptr;
2635 }
2636
2637 ptr = skb_gro_header_fast(skb, off);
2638 if (skb_gro_header_hard(skb, off + plen)) {
2639 ptr = skb_gro_header_slow(skb, off + plen, off);
2640 if (!ptr)
2641 return NULL;
2642 }
2643
2644 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2645 start, offset);
2646
2647 /* Adjust skb->csum since we changed the packet */
2648 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2649
2650 grc->offset = off + hdrlen + offset;
2651 grc->delta = delta;
2652
2653 return ptr;
2654 }
2655
2656 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2657 struct gro_remcsum *grc)
2658 {
2659 void *ptr;
2660 size_t plen = grc->offset + sizeof(u16);
2661
2662 if (!grc->delta)
2663 return;
2664
2665 ptr = skb_gro_header_fast(skb, grc->offset);
2666 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2667 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2668 if (!ptr)
2669 return;
2670 }
2671
2672 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2673 }
2674
2675 #ifdef CONFIG_XFRM_OFFLOAD
2676 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2677 {
2678 if (PTR_ERR(pp) != -EINPROGRESS)
2679 NAPI_GRO_CB(skb)->flush |= flush;
2680 }
2681 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2682 struct sk_buff **pp,
2683 int flush,
2684 struct gro_remcsum *grc)
2685 {
2686 if (PTR_ERR(pp) != -EINPROGRESS) {
2687 NAPI_GRO_CB(skb)->flush |= flush;
2688 skb_gro_remcsum_cleanup(skb, grc);
2689 skb->remcsum_offload = 0;
2690 }
2691 }
2692 #else
2693 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2694 {
2695 NAPI_GRO_CB(skb)->flush |= flush;
2696 }
2697 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2698 struct sk_buff **pp,
2699 int flush,
2700 struct gro_remcsum *grc)
2701 {
2702 NAPI_GRO_CB(skb)->flush |= flush;
2703 skb_gro_remcsum_cleanup(skb, grc);
2704 skb->remcsum_offload = 0;
2705 }
2706 #endif
2707
2708 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2709 unsigned short type,
2710 const void *daddr, const void *saddr,
2711 unsigned int len)
2712 {
2713 if (!dev->header_ops || !dev->header_ops->create)
2714 return 0;
2715
2716 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2717 }
2718
2719 static inline int dev_parse_header(const struct sk_buff *skb,
2720 unsigned char *haddr)
2721 {
2722 const struct net_device *dev = skb->dev;
2723
2724 if (!dev->header_ops || !dev->header_ops->parse)
2725 return 0;
2726 return dev->header_ops->parse(skb, haddr);
2727 }
2728
2729 /* ll_header must have at least hard_header_len allocated */
2730 static inline bool dev_validate_header(const struct net_device *dev,
2731 char *ll_header, int len)
2732 {
2733 if (likely(len >= dev->hard_header_len))
2734 return true;
2735 if (len < dev->min_header_len)
2736 return false;
2737
2738 if (capable(CAP_SYS_RAWIO)) {
2739 memset(ll_header + len, 0, dev->hard_header_len - len);
2740 return true;
2741 }
2742
2743 if (dev->header_ops && dev->header_ops->validate)
2744 return dev->header_ops->validate(ll_header, len);
2745
2746 return false;
2747 }
2748
2749 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2750 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2751 static inline int unregister_gifconf(unsigned int family)
2752 {
2753 return register_gifconf(family, NULL);
2754 }
2755
2756 #ifdef CONFIG_NET_FLOW_LIMIT
2757 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2758 struct sd_flow_limit {
2759 u64 count;
2760 unsigned int num_buckets;
2761 unsigned int history_head;
2762 u16 history[FLOW_LIMIT_HISTORY];
2763 u8 buckets[];
2764 };
2765
2766 extern int netdev_flow_limit_table_len;
2767 #endif /* CONFIG_NET_FLOW_LIMIT */
2768
2769 /*
2770 * Incoming packets are placed on per-CPU queues
2771 */
2772 struct softnet_data {
2773 struct list_head poll_list;
2774 struct sk_buff_head process_queue;
2775
2776 /* stats */
2777 unsigned int processed;
2778 unsigned int time_squeeze;
2779 unsigned int received_rps;
2780 #ifdef CONFIG_RPS
2781 struct softnet_data *rps_ipi_list;
2782 #endif
2783 #ifdef CONFIG_NET_FLOW_LIMIT
2784 struct sd_flow_limit __rcu *flow_limit;
2785 #endif
2786 struct Qdisc *output_queue;
2787 struct Qdisc **output_queue_tailp;
2788 struct sk_buff *completion_queue;
2789
2790 #ifdef CONFIG_RPS
2791 /* input_queue_head should be written by cpu owning this struct,
2792 * and only read by other cpus. Worth using a cache line.
2793 */
2794 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2795
2796 /* Elements below can be accessed between CPUs for RPS/RFS */
2797 call_single_data_t csd ____cacheline_aligned_in_smp;
2798 struct softnet_data *rps_ipi_next;
2799 unsigned int cpu;
2800 unsigned int input_queue_tail;
2801 #endif
2802 unsigned int dropped;
2803 struct sk_buff_head input_pkt_queue;
2804 struct napi_struct backlog;
2805
2806 };
2807
2808 static inline void input_queue_head_incr(struct softnet_data *sd)
2809 {
2810 #ifdef CONFIG_RPS
2811 sd->input_queue_head++;
2812 #endif
2813 }
2814
2815 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2816 unsigned int *qtail)
2817 {
2818 #ifdef CONFIG_RPS
2819 *qtail = ++sd->input_queue_tail;
2820 #endif
2821 }
2822
2823 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2824
2825 void __netif_schedule(struct Qdisc *q);
2826 void netif_schedule_queue(struct netdev_queue *txq);
2827
2828 static inline void netif_tx_schedule_all(struct net_device *dev)
2829 {
2830 unsigned int i;
2831
2832 for (i = 0; i < dev->num_tx_queues; i++)
2833 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2834 }
2835
2836 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2837 {
2838 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2839 }
2840
2841 /**
2842 * netif_start_queue - allow transmit
2843 * @dev: network device
2844 *
2845 * Allow upper layers to call the device hard_start_xmit routine.
2846 */
2847 static inline void netif_start_queue(struct net_device *dev)
2848 {
2849 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2850 }
2851
2852 static inline void netif_tx_start_all_queues(struct net_device *dev)
2853 {
2854 unsigned int i;
2855
2856 for (i = 0; i < dev->num_tx_queues; i++) {
2857 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2858 netif_tx_start_queue(txq);
2859 }
2860 }
2861
2862 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2863
2864 /**
2865 * netif_wake_queue - restart transmit
2866 * @dev: network device
2867 *
2868 * Allow upper layers to call the device hard_start_xmit routine.
2869 * Used for flow control when transmit resources are available.
2870 */
2871 static inline void netif_wake_queue(struct net_device *dev)
2872 {
2873 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2874 }
2875
2876 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2877 {
2878 unsigned int i;
2879
2880 for (i = 0; i < dev->num_tx_queues; i++) {
2881 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2882 netif_tx_wake_queue(txq);
2883 }
2884 }
2885
2886 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2887 {
2888 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2889 }
2890
2891 /**
2892 * netif_stop_queue - stop transmitted packets
2893 * @dev: network device
2894 *
2895 * Stop upper layers calling the device hard_start_xmit routine.
2896 * Used for flow control when transmit resources are unavailable.
2897 */
2898 static inline void netif_stop_queue(struct net_device *dev)
2899 {
2900 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2901 }
2902
2903 void netif_tx_stop_all_queues(struct net_device *dev);
2904
2905 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2906 {
2907 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2908 }
2909
2910 /**
2911 * netif_queue_stopped - test if transmit queue is flowblocked
2912 * @dev: network device
2913 *
2914 * Test if transmit queue on device is currently unable to send.
2915 */
2916 static inline bool netif_queue_stopped(const struct net_device *dev)
2917 {
2918 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2919 }
2920
2921 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2922 {
2923 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2924 }
2925
2926 static inline bool
2927 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2928 {
2929 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2930 }
2931
2932 static inline bool
2933 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2934 {
2935 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2936 }
2937
2938 /**
2939 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2940 * @dev_queue: pointer to transmit queue
2941 *
2942 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2943 * to give appropriate hint to the CPU.
2944 */
2945 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2946 {
2947 #ifdef CONFIG_BQL
2948 prefetchw(&dev_queue->dql.num_queued);
2949 #endif
2950 }
2951
2952 /**
2953 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2954 * @dev_queue: pointer to transmit queue
2955 *
2956 * BQL enabled drivers might use this helper in their TX completion path,
2957 * to give appropriate hint to the CPU.
2958 */
2959 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2960 {
2961 #ifdef CONFIG_BQL
2962 prefetchw(&dev_queue->dql.limit);
2963 #endif
2964 }
2965
2966 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2967 unsigned int bytes)
2968 {
2969 #ifdef CONFIG_BQL
2970 dql_queued(&dev_queue->dql, bytes);
2971
2972 if (likely(dql_avail(&dev_queue->dql) >= 0))
2973 return;
2974
2975 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2976
2977 /*
2978 * The XOFF flag must be set before checking the dql_avail below,
2979 * because in netdev_tx_completed_queue we update the dql_completed
2980 * before checking the XOFF flag.
2981 */
2982 smp_mb();
2983
2984 /* check again in case another CPU has just made room avail */
2985 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2986 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2987 #endif
2988 }
2989
2990 /**
2991 * netdev_sent_queue - report the number of bytes queued to hardware
2992 * @dev: network device
2993 * @bytes: number of bytes queued to the hardware device queue
2994 *
2995 * Report the number of bytes queued for sending/completion to the network
2996 * device hardware queue. @bytes should be a good approximation and should
2997 * exactly match netdev_completed_queue() @bytes
2998 */
2999 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3000 {
3001 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3002 }
3003
3004 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3005 unsigned int pkts, unsigned int bytes)
3006 {
3007 #ifdef CONFIG_BQL
3008 if (unlikely(!bytes))
3009 return;
3010
3011 dql_completed(&dev_queue->dql, bytes);
3012
3013 /*
3014 * Without the memory barrier there is a small possiblity that
3015 * netdev_tx_sent_queue will miss the update and cause the queue to
3016 * be stopped forever
3017 */
3018 smp_mb();
3019
3020 if (dql_avail(&dev_queue->dql) < 0)
3021 return;
3022
3023 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3024 netif_schedule_queue(dev_queue);
3025 #endif
3026 }
3027
3028 /**
3029 * netdev_completed_queue - report bytes and packets completed by device
3030 * @dev: network device
3031 * @pkts: actual number of packets sent over the medium
3032 * @bytes: actual number of bytes sent over the medium
3033 *
3034 * Report the number of bytes and packets transmitted by the network device
3035 * hardware queue over the physical medium, @bytes must exactly match the
3036 * @bytes amount passed to netdev_sent_queue()
3037 */
3038 static inline void netdev_completed_queue(struct net_device *dev,
3039 unsigned int pkts, unsigned int bytes)
3040 {
3041 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3042 }
3043
3044 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3045 {
3046 #ifdef CONFIG_BQL
3047 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3048 dql_reset(&q->dql);
3049 #endif
3050 }
3051
3052 /**
3053 * netdev_reset_queue - reset the packets and bytes count of a network device
3054 * @dev_queue: network device
3055 *
3056 * Reset the bytes and packet count of a network device and clear the
3057 * software flow control OFF bit for this network device
3058 */
3059 static inline void netdev_reset_queue(struct net_device *dev_queue)
3060 {
3061 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3062 }
3063
3064 /**
3065 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3066 * @dev: network device
3067 * @queue_index: given tx queue index
3068 *
3069 * Returns 0 if given tx queue index >= number of device tx queues,
3070 * otherwise returns the originally passed tx queue index.
3071 */
3072 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3073 {
3074 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3075 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3076 dev->name, queue_index,
3077 dev->real_num_tx_queues);
3078 return 0;
3079 }
3080
3081 return queue_index;
3082 }
3083
3084 /**
3085 * netif_running - test if up
3086 * @dev: network device
3087 *
3088 * Test if the device has been brought up.
3089 */
3090 static inline bool netif_running(const struct net_device *dev)
3091 {
3092 return test_bit(__LINK_STATE_START, &dev->state);
3093 }
3094
3095 /*
3096 * Routines to manage the subqueues on a device. We only need start,
3097 * stop, and a check if it's stopped. All other device management is
3098 * done at the overall netdevice level.
3099 * Also test the device if we're multiqueue.
3100 */
3101
3102 /**
3103 * netif_start_subqueue - allow sending packets on subqueue
3104 * @dev: network device
3105 * @queue_index: sub queue index
3106 *
3107 * Start individual transmit queue of a device with multiple transmit queues.
3108 */
3109 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3110 {
3111 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3112
3113 netif_tx_start_queue(txq);
3114 }
3115
3116 /**
3117 * netif_stop_subqueue - stop sending packets on subqueue
3118 * @dev: network device
3119 * @queue_index: sub queue index
3120 *
3121 * Stop individual transmit queue of a device with multiple transmit queues.
3122 */
3123 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3124 {
3125 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3126 netif_tx_stop_queue(txq);
3127 }
3128
3129 /**
3130 * netif_subqueue_stopped - test status of subqueue
3131 * @dev: network device
3132 * @queue_index: sub queue index
3133 *
3134 * Check individual transmit queue of a device with multiple transmit queues.
3135 */
3136 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3137 u16 queue_index)
3138 {
3139 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3140
3141 return netif_tx_queue_stopped(txq);
3142 }
3143
3144 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3145 struct sk_buff *skb)
3146 {
3147 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3148 }
3149
3150 /**
3151 * netif_wake_subqueue - allow sending packets on subqueue
3152 * @dev: network device
3153 * @queue_index: sub queue index
3154 *
3155 * Resume individual transmit queue of a device with multiple transmit queues.
3156 */
3157 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3158 {
3159 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3160
3161 netif_tx_wake_queue(txq);
3162 }
3163
3164 #ifdef CONFIG_XPS
3165 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3166 u16 index);
3167 #else
3168 static inline int netif_set_xps_queue(struct net_device *dev,
3169 const struct cpumask *mask,
3170 u16 index)
3171 {
3172 return 0;
3173 }
3174 #endif
3175
3176 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3177 unsigned int num_tx_queues);
3178
3179 /*
3180 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3181 * as a distribution range limit for the returned value.
3182 */
3183 static inline u16 skb_tx_hash(const struct net_device *dev,
3184 struct sk_buff *skb)
3185 {
3186 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3187 }
3188
3189 /**
3190 * netif_is_multiqueue - test if device has multiple transmit queues
3191 * @dev: network device
3192 *
3193 * Check if device has multiple transmit queues
3194 */
3195 static inline bool netif_is_multiqueue(const struct net_device *dev)
3196 {
3197 return dev->num_tx_queues > 1;
3198 }
3199
3200 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3201
3202 #ifdef CONFIG_SYSFS
3203 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3204 #else
3205 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3206 unsigned int rxq)
3207 {
3208 return 0;
3209 }
3210 #endif
3211
3212 #ifdef CONFIG_SYSFS
3213 static inline unsigned int get_netdev_rx_queue_index(
3214 struct netdev_rx_queue *queue)
3215 {
3216 struct net_device *dev = queue->dev;
3217 int index = queue - dev->_rx;
3218
3219 BUG_ON(index >= dev->num_rx_queues);
3220 return index;
3221 }
3222 #endif
3223
3224 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3225 int netif_get_num_default_rss_queues(void);
3226
3227 enum skb_free_reason {
3228 SKB_REASON_CONSUMED,
3229 SKB_REASON_DROPPED,
3230 };
3231
3232 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3233 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3234
3235 /*
3236 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3237 * interrupt context or with hardware interrupts being disabled.
3238 * (in_irq() || irqs_disabled())
3239 *
3240 * We provide four helpers that can be used in following contexts :
3241 *
3242 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3243 * replacing kfree_skb(skb)
3244 *
3245 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3246 * Typically used in place of consume_skb(skb) in TX completion path
3247 *
3248 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3249 * replacing kfree_skb(skb)
3250 *
3251 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3252 * and consumed a packet. Used in place of consume_skb(skb)
3253 */
3254 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3255 {
3256 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3257 }
3258
3259 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3260 {
3261 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3262 }
3263
3264 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3265 {
3266 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3267 }
3268
3269 static inline void dev_consume_skb_any(struct sk_buff *skb)
3270 {
3271 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3272 }
3273
3274 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3275 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3276 int netif_rx(struct sk_buff *skb);
3277 int netif_rx_ni(struct sk_buff *skb);
3278 int netif_receive_skb(struct sk_buff *skb);
3279 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3280 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3281 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3282 gro_result_t napi_gro_frags(struct napi_struct *napi);
3283 struct packet_offload *gro_find_receive_by_type(__be16 type);
3284 struct packet_offload *gro_find_complete_by_type(__be16 type);
3285
3286 static inline void napi_free_frags(struct napi_struct *napi)
3287 {
3288 kfree_skb(napi->skb);
3289 napi->skb = NULL;
3290 }
3291
3292 bool netdev_is_rx_handler_busy(struct net_device *dev);
3293 int netdev_rx_handler_register(struct net_device *dev,
3294 rx_handler_func_t *rx_handler,
3295 void *rx_handler_data);
3296 void netdev_rx_handler_unregister(struct net_device *dev);
3297
3298 bool dev_valid_name(const char *name);
3299 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3300 int dev_ethtool(struct net *net, struct ifreq *);
3301 unsigned int dev_get_flags(const struct net_device *);
3302 int __dev_change_flags(struct net_device *, unsigned int flags);
3303 int dev_change_flags(struct net_device *, unsigned int);
3304 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3305 unsigned int gchanges);
3306 int dev_change_name(struct net_device *, const char *);
3307 int dev_set_alias(struct net_device *, const char *, size_t);
3308 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3309 int __dev_set_mtu(struct net_device *, int);
3310 int dev_set_mtu(struct net_device *, int);
3311 void dev_set_group(struct net_device *, int);
3312 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3313 int dev_change_carrier(struct net_device *, bool new_carrier);
3314 int dev_get_phys_port_id(struct net_device *dev,
3315 struct netdev_phys_item_id *ppid);
3316 int dev_get_phys_port_name(struct net_device *dev,
3317 char *name, size_t len);
3318 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3319 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3320 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3321 struct netdev_queue *txq, int *ret);
3322
3323 typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3324 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3325 int fd, u32 flags);
3326 u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
3327
3328 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3329 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3330 bool is_skb_forwardable(const struct net_device *dev,
3331 const struct sk_buff *skb);
3332
3333 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3334 struct sk_buff *skb)
3335 {
3336 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3337 unlikely(!is_skb_forwardable(dev, skb))) {
3338 atomic_long_inc(&dev->rx_dropped);
3339 kfree_skb(skb);
3340 return NET_RX_DROP;
3341 }
3342
3343 skb_scrub_packet(skb, true);
3344 skb->priority = 0;
3345 return 0;
3346 }
3347
3348 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3349
3350 extern int netdev_budget;
3351 extern unsigned int netdev_budget_usecs;
3352
3353 /* Called by rtnetlink.c:rtnl_unlock() */
3354 void netdev_run_todo(void);
3355
3356 /**
3357 * dev_put - release reference to device
3358 * @dev: network device
3359 *
3360 * Release reference to device to allow it to be freed.
3361 */
3362 static inline void dev_put(struct net_device *dev)
3363 {
3364 this_cpu_dec(*dev->pcpu_refcnt);
3365 }
3366
3367 /**
3368 * dev_hold - get reference to device
3369 * @dev: network device
3370 *
3371 * Hold reference to device to keep it from being freed.
3372 */
3373 static inline void dev_hold(struct net_device *dev)
3374 {
3375 this_cpu_inc(*dev->pcpu_refcnt);
3376 }
3377
3378 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3379 * and _off may be called from IRQ context, but it is caller
3380 * who is responsible for serialization of these calls.
3381 *
3382 * The name carrier is inappropriate, these functions should really be
3383 * called netif_lowerlayer_*() because they represent the state of any
3384 * kind of lower layer not just hardware media.
3385 */
3386
3387 void linkwatch_init_dev(struct net_device *dev);
3388 void linkwatch_fire_event(struct net_device *dev);
3389 void linkwatch_forget_dev(struct net_device *dev);
3390
3391 /**
3392 * netif_carrier_ok - test if carrier present
3393 * @dev: network device
3394 *
3395 * Check if carrier is present on device
3396 */
3397 static inline bool netif_carrier_ok(const struct net_device *dev)
3398 {
3399 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3400 }
3401
3402 unsigned long dev_trans_start(struct net_device *dev);
3403
3404 void __netdev_watchdog_up(struct net_device *dev);
3405
3406 void netif_carrier_on(struct net_device *dev);
3407
3408 void netif_carrier_off(struct net_device *dev);
3409
3410 /**
3411 * netif_dormant_on - mark device as dormant.
3412 * @dev: network device
3413 *
3414 * Mark device as dormant (as per RFC2863).
3415 *
3416 * The dormant state indicates that the relevant interface is not
3417 * actually in a condition to pass packets (i.e., it is not 'up') but is
3418 * in a "pending" state, waiting for some external event. For "on-
3419 * demand" interfaces, this new state identifies the situation where the
3420 * interface is waiting for events to place it in the up state.
3421 */
3422 static inline void netif_dormant_on(struct net_device *dev)
3423 {
3424 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3425 linkwatch_fire_event(dev);
3426 }
3427
3428 /**
3429 * netif_dormant_off - set device as not dormant.
3430 * @dev: network device
3431 *
3432 * Device is not in dormant state.
3433 */
3434 static inline void netif_dormant_off(struct net_device *dev)
3435 {
3436 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3437 linkwatch_fire_event(dev);
3438 }
3439
3440 /**
3441 * netif_dormant - test if device is dormant
3442 * @dev: network device
3443 *
3444 * Check if device is dormant.
3445 */
3446 static inline bool netif_dormant(const struct net_device *dev)
3447 {
3448 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3449 }
3450
3451
3452 /**
3453 * netif_oper_up - test if device is operational
3454 * @dev: network device
3455 *
3456 * Check if carrier is operational
3457 */
3458 static inline bool netif_oper_up(const struct net_device *dev)
3459 {
3460 return (dev->operstate == IF_OPER_UP ||
3461 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3462 }
3463
3464 /**
3465 * netif_device_present - is device available or removed
3466 * @dev: network device
3467 *
3468 * Check if device has not been removed from system.
3469 */
3470 static inline bool netif_device_present(struct net_device *dev)
3471 {
3472 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3473 }
3474
3475 void netif_device_detach(struct net_device *dev);
3476
3477 void netif_device_attach(struct net_device *dev);
3478
3479 /*
3480 * Network interface message level settings
3481 */
3482
3483 enum {
3484 NETIF_MSG_DRV = 0x0001,
3485 NETIF_MSG_PROBE = 0x0002,
3486 NETIF_MSG_LINK = 0x0004,
3487 NETIF_MSG_TIMER = 0x0008,
3488 NETIF_MSG_IFDOWN = 0x0010,
3489 NETIF_MSG_IFUP = 0x0020,
3490 NETIF_MSG_RX_ERR = 0x0040,
3491 NETIF_MSG_TX_ERR = 0x0080,
3492 NETIF_MSG_TX_QUEUED = 0x0100,
3493 NETIF_MSG_INTR = 0x0200,
3494 NETIF_MSG_TX_DONE = 0x0400,
3495 NETIF_MSG_RX_STATUS = 0x0800,
3496 NETIF_MSG_PKTDATA = 0x1000,
3497 NETIF_MSG_HW = 0x2000,
3498 NETIF_MSG_WOL = 0x4000,
3499 };
3500
3501 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3502 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3503 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3504 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3505 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3506 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3507 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3508 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3509 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3510 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3511 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3512 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3513 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3514 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3515 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3516
3517 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3518 {
3519 /* use default */
3520 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3521 return default_msg_enable_bits;
3522 if (debug_value == 0) /* no output */
3523 return 0;
3524 /* set low N bits */
3525 return (1 << debug_value) - 1;
3526 }
3527
3528 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3529 {
3530 spin_lock(&txq->_xmit_lock);
3531 txq->xmit_lock_owner = cpu;
3532 }
3533
3534 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3535 {
3536 __acquire(&txq->_xmit_lock);
3537 return true;
3538 }
3539
3540 static inline void __netif_tx_release(struct netdev_queue *txq)
3541 {
3542 __release(&txq->_xmit_lock);
3543 }
3544
3545 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3546 {
3547 spin_lock_bh(&txq->_xmit_lock);
3548 txq->xmit_lock_owner = smp_processor_id();
3549 }
3550
3551 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3552 {
3553 bool ok = spin_trylock(&txq->_xmit_lock);
3554 if (likely(ok))
3555 txq->xmit_lock_owner = smp_processor_id();
3556 return ok;
3557 }
3558
3559 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3560 {
3561 txq->xmit_lock_owner = -1;
3562 spin_unlock(&txq->_xmit_lock);
3563 }
3564
3565 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3566 {
3567 txq->xmit_lock_owner = -1;
3568 spin_unlock_bh(&txq->_xmit_lock);
3569 }
3570
3571 static inline void txq_trans_update(struct netdev_queue *txq)
3572 {
3573 if (txq->xmit_lock_owner != -1)
3574 txq->trans_start = jiffies;
3575 }
3576
3577 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3578 static inline void netif_trans_update(struct net_device *dev)
3579 {
3580 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3581
3582 if (txq->trans_start != jiffies)
3583 txq->trans_start = jiffies;
3584 }
3585
3586 /**
3587 * netif_tx_lock - grab network device transmit lock
3588 * @dev: network device
3589 *
3590 * Get network device transmit lock
3591 */
3592 static inline void netif_tx_lock(struct net_device *dev)
3593 {
3594 unsigned int i;
3595 int cpu;
3596
3597 spin_lock(&dev->tx_global_lock);
3598 cpu = smp_processor_id();
3599 for (i = 0; i < dev->num_tx_queues; i++) {
3600 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3601
3602 /* We are the only thread of execution doing a
3603 * freeze, but we have to grab the _xmit_lock in
3604 * order to synchronize with threads which are in
3605 * the ->hard_start_xmit() handler and already
3606 * checked the frozen bit.
3607 */
3608 __netif_tx_lock(txq, cpu);
3609 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3610 __netif_tx_unlock(txq);
3611 }
3612 }
3613
3614 static inline void netif_tx_lock_bh(struct net_device *dev)
3615 {
3616 local_bh_disable();
3617 netif_tx_lock(dev);
3618 }
3619
3620 static inline void netif_tx_unlock(struct net_device *dev)
3621 {
3622 unsigned int i;
3623
3624 for (i = 0; i < dev->num_tx_queues; i++) {
3625 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3626
3627 /* No need to grab the _xmit_lock here. If the
3628 * queue is not stopped for another reason, we
3629 * force a schedule.
3630 */
3631 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3632 netif_schedule_queue(txq);
3633 }
3634 spin_unlock(&dev->tx_global_lock);
3635 }
3636
3637 static inline void netif_tx_unlock_bh(struct net_device *dev)
3638 {
3639 netif_tx_unlock(dev);
3640 local_bh_enable();
3641 }
3642
3643 #define HARD_TX_LOCK(dev, txq, cpu) { \
3644 if ((dev->features & NETIF_F_LLTX) == 0) { \
3645 __netif_tx_lock(txq, cpu); \
3646 } else { \
3647 __netif_tx_acquire(txq); \
3648 } \
3649 }
3650
3651 #define HARD_TX_TRYLOCK(dev, txq) \
3652 (((dev->features & NETIF_F_LLTX) == 0) ? \
3653 __netif_tx_trylock(txq) : \
3654 __netif_tx_acquire(txq))
3655
3656 #define HARD_TX_UNLOCK(dev, txq) { \
3657 if ((dev->features & NETIF_F_LLTX) == 0) { \
3658 __netif_tx_unlock(txq); \
3659 } else { \
3660 __netif_tx_release(txq); \
3661 } \
3662 }
3663
3664 static inline void netif_tx_disable(struct net_device *dev)
3665 {
3666 unsigned int i;
3667 int cpu;
3668
3669 local_bh_disable();
3670 cpu = smp_processor_id();
3671 for (i = 0; i < dev->num_tx_queues; i++) {
3672 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3673
3674 __netif_tx_lock(txq, cpu);
3675 netif_tx_stop_queue(txq);
3676 __netif_tx_unlock(txq);
3677 }
3678 local_bh_enable();
3679 }
3680
3681 static inline void netif_addr_lock(struct net_device *dev)
3682 {
3683 spin_lock(&dev->addr_list_lock);
3684 }
3685
3686 static inline void netif_addr_lock_nested(struct net_device *dev)
3687 {
3688 int subclass = SINGLE_DEPTH_NESTING;
3689
3690 if (dev->netdev_ops->ndo_get_lock_subclass)
3691 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3692
3693 spin_lock_nested(&dev->addr_list_lock, subclass);
3694 }
3695
3696 static inline void netif_addr_lock_bh(struct net_device *dev)
3697 {
3698 spin_lock_bh(&dev->addr_list_lock);
3699 }
3700
3701 static inline void netif_addr_unlock(struct net_device *dev)
3702 {
3703 spin_unlock(&dev->addr_list_lock);
3704 }
3705
3706 static inline void netif_addr_unlock_bh(struct net_device *dev)
3707 {
3708 spin_unlock_bh(&dev->addr_list_lock);
3709 }
3710
3711 /*
3712 * dev_addrs walker. Should be used only for read access. Call with
3713 * rcu_read_lock held.
3714 */
3715 #define for_each_dev_addr(dev, ha) \
3716 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3717
3718 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3719
3720 void ether_setup(struct net_device *dev);
3721
3722 /* Support for loadable net-drivers */
3723 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3724 unsigned char name_assign_type,
3725 void (*setup)(struct net_device *),
3726 unsigned int txqs, unsigned int rxqs);
3727 int dev_get_valid_name(struct net *net, struct net_device *dev,
3728 const char *name);
3729
3730 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3731 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3732
3733 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3734 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3735 count)
3736
3737 int register_netdev(struct net_device *dev);
3738 void unregister_netdev(struct net_device *dev);
3739
3740 /* General hardware address lists handling functions */
3741 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3742 struct netdev_hw_addr_list *from_list, int addr_len);
3743 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3744 struct netdev_hw_addr_list *from_list, int addr_len);
3745 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3746 struct net_device *dev,
3747 int (*sync)(struct net_device *, const unsigned char *),
3748 int (*unsync)(struct net_device *,
3749 const unsigned char *));
3750 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3751 struct net_device *dev,
3752 int (*unsync)(struct net_device *,
3753 const unsigned char *));
3754 void __hw_addr_init(struct netdev_hw_addr_list *list);
3755
3756 /* Functions used for device addresses handling */
3757 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3758 unsigned char addr_type);
3759 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3760 unsigned char addr_type);
3761 void dev_addr_flush(struct net_device *dev);
3762 int dev_addr_init(struct net_device *dev);
3763
3764 /* Functions used for unicast addresses handling */
3765 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3766 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3767 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3768 int dev_uc_sync(struct net_device *to, struct net_device *from);
3769 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3770 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3771 void dev_uc_flush(struct net_device *dev);
3772 void dev_uc_init(struct net_device *dev);
3773
3774 /**
3775 * __dev_uc_sync - Synchonize device's unicast list
3776 * @dev: device to sync
3777 * @sync: function to call if address should be added
3778 * @unsync: function to call if address should be removed
3779 *
3780 * Add newly added addresses to the interface, and release
3781 * addresses that have been deleted.
3782 */
3783 static inline int __dev_uc_sync(struct net_device *dev,
3784 int (*sync)(struct net_device *,
3785 const unsigned char *),
3786 int (*unsync)(struct net_device *,
3787 const unsigned char *))
3788 {
3789 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3790 }
3791
3792 /**
3793 * __dev_uc_unsync - Remove synchronized addresses from device
3794 * @dev: device to sync
3795 * @unsync: function to call if address should be removed
3796 *
3797 * Remove all addresses that were added to the device by dev_uc_sync().
3798 */
3799 static inline void __dev_uc_unsync(struct net_device *dev,
3800 int (*unsync)(struct net_device *,
3801 const unsigned char *))
3802 {
3803 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3804 }
3805
3806 /* Functions used for multicast addresses handling */
3807 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3808 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3809 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3810 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3811 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3812 int dev_mc_sync(struct net_device *to, struct net_device *from);
3813 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3814 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3815 void dev_mc_flush(struct net_device *dev);
3816 void dev_mc_init(struct net_device *dev);
3817
3818 /**
3819 * __dev_mc_sync - Synchonize device's multicast list
3820 * @dev: device to sync
3821 * @sync: function to call if address should be added
3822 * @unsync: function to call if address should be removed
3823 *
3824 * Add newly added addresses to the interface, and release
3825 * addresses that have been deleted.
3826 */
3827 static inline int __dev_mc_sync(struct net_device *dev,
3828 int (*sync)(struct net_device *,
3829 const unsigned char *),
3830 int (*unsync)(struct net_device *,
3831 const unsigned char *))
3832 {
3833 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3834 }
3835
3836 /**
3837 * __dev_mc_unsync - Remove synchronized addresses from device
3838 * @dev: device to sync
3839 * @unsync: function to call if address should be removed
3840 *
3841 * Remove all addresses that were added to the device by dev_mc_sync().
3842 */
3843 static inline void __dev_mc_unsync(struct net_device *dev,
3844 int (*unsync)(struct net_device *,
3845 const unsigned char *))
3846 {
3847 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3848 }
3849
3850 /* Functions used for secondary unicast and multicast support */
3851 void dev_set_rx_mode(struct net_device *dev);
3852 void __dev_set_rx_mode(struct net_device *dev);
3853 int dev_set_promiscuity(struct net_device *dev, int inc);
3854 int dev_set_allmulti(struct net_device *dev, int inc);
3855 void netdev_state_change(struct net_device *dev);
3856 void netdev_notify_peers(struct net_device *dev);
3857 void netdev_features_change(struct net_device *dev);
3858 /* Load a device via the kmod */
3859 void dev_load(struct net *net, const char *name);
3860 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3861 struct rtnl_link_stats64 *storage);
3862 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3863 const struct net_device_stats *netdev_stats);
3864
3865 extern int netdev_max_backlog;
3866 extern int netdev_tstamp_prequeue;
3867 extern int weight_p;
3868 extern int dev_weight_rx_bias;
3869 extern int dev_weight_tx_bias;
3870 extern int dev_rx_weight;
3871 extern int dev_tx_weight;
3872
3873 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3874 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3875 struct list_head **iter);
3876 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3877 struct list_head **iter);
3878
3879 /* iterate through upper list, must be called under RCU read lock */
3880 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3881 for (iter = &(dev)->adj_list.upper, \
3882 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3883 updev; \
3884 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3885
3886 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3887 int (*fn)(struct net_device *upper_dev,
3888 void *data),
3889 void *data);
3890
3891 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3892 struct net_device *upper_dev);
3893
3894 bool netdev_has_any_upper_dev(struct net_device *dev);
3895
3896 void *netdev_lower_get_next_private(struct net_device *dev,
3897 struct list_head **iter);
3898 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3899 struct list_head **iter);
3900
3901 #define netdev_for_each_lower_private(dev, priv, iter) \
3902 for (iter = (dev)->adj_list.lower.next, \
3903 priv = netdev_lower_get_next_private(dev, &(iter)); \
3904 priv; \
3905 priv = netdev_lower_get_next_private(dev, &(iter)))
3906
3907 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3908 for (iter = &(dev)->adj_list.lower, \
3909 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3910 priv; \
3911 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3912
3913 void *netdev_lower_get_next(struct net_device *dev,
3914 struct list_head **iter);
3915
3916 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3917 for (iter = (dev)->adj_list.lower.next, \
3918 ldev = netdev_lower_get_next(dev, &(iter)); \
3919 ldev; \
3920 ldev = netdev_lower_get_next(dev, &(iter)))
3921
3922 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3923 struct list_head **iter);
3924 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3925 struct list_head **iter);
3926
3927 int netdev_walk_all_lower_dev(struct net_device *dev,
3928 int (*fn)(struct net_device *lower_dev,
3929 void *data),
3930 void *data);
3931 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3932 int (*fn)(struct net_device *lower_dev,
3933 void *data),
3934 void *data);
3935
3936 void *netdev_adjacent_get_private(struct list_head *adj_list);
3937 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3938 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3939 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3940 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3941 int netdev_master_upper_dev_link(struct net_device *dev,
3942 struct net_device *upper_dev,
3943 void *upper_priv, void *upper_info);
3944 void netdev_upper_dev_unlink(struct net_device *dev,
3945 struct net_device *upper_dev);
3946 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3947 void *netdev_lower_dev_get_private(struct net_device *dev,
3948 struct net_device *lower_dev);
3949 void netdev_lower_state_changed(struct net_device *lower_dev,
3950 void *lower_state_info);
3951
3952 /* RSS keys are 40 or 52 bytes long */
3953 #define NETDEV_RSS_KEY_LEN 52
3954 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3955 void netdev_rss_key_fill(void *buffer, size_t len);
3956
3957 int dev_get_nest_level(struct net_device *dev);
3958 int skb_checksum_help(struct sk_buff *skb);
3959 int skb_crc32c_csum_help(struct sk_buff *skb);
3960 int skb_csum_hwoffload_help(struct sk_buff *skb,
3961 const netdev_features_t features);
3962
3963 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3964 netdev_features_t features, bool tx_path);
3965 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3966 netdev_features_t features);
3967
3968 struct netdev_bonding_info {
3969 ifslave slave;
3970 ifbond master;
3971 };
3972
3973 struct netdev_notifier_bonding_info {
3974 struct netdev_notifier_info info; /* must be first */
3975 struct netdev_bonding_info bonding_info;
3976 };
3977
3978 void netdev_bonding_info_change(struct net_device *dev,
3979 struct netdev_bonding_info *bonding_info);
3980
3981 static inline
3982 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3983 {
3984 return __skb_gso_segment(skb, features, true);
3985 }
3986 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3987
3988 static inline bool can_checksum_protocol(netdev_features_t features,
3989 __be16 protocol)
3990 {
3991 if (protocol == htons(ETH_P_FCOE))
3992 return !!(features & NETIF_F_FCOE_CRC);
3993
3994 /* Assume this is an IP checksum (not SCTP CRC) */
3995
3996 if (features & NETIF_F_HW_CSUM) {
3997 /* Can checksum everything */
3998 return true;
3999 }
4000
4001 switch (protocol) {
4002 case htons(ETH_P_IP):
4003 return !!(features & NETIF_F_IP_CSUM);
4004 case htons(ETH_P_IPV6):
4005 return !!(features & NETIF_F_IPV6_CSUM);
4006 default:
4007 return false;
4008 }
4009 }
4010
4011 #ifdef CONFIG_BUG
4012 void netdev_rx_csum_fault(struct net_device *dev);
4013 #else
4014 static inline void netdev_rx_csum_fault(struct net_device *dev)
4015 {
4016 }
4017 #endif
4018 /* rx skb timestamps */
4019 void net_enable_timestamp(void);
4020 void net_disable_timestamp(void);
4021
4022 #ifdef CONFIG_PROC_FS
4023 int __init dev_proc_init(void);
4024 #else
4025 #define dev_proc_init() 0
4026 #endif
4027
4028 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4029 struct sk_buff *skb, struct net_device *dev,
4030 bool more)
4031 {
4032 skb->xmit_more = more ? 1 : 0;
4033 return ops->ndo_start_xmit(skb, dev);
4034 }
4035
4036 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4037 struct netdev_queue *txq, bool more)
4038 {
4039 const struct net_device_ops *ops = dev->netdev_ops;
4040 int rc;
4041
4042 rc = __netdev_start_xmit(ops, skb, dev, more);
4043 if (rc == NETDEV_TX_OK)
4044 txq_trans_update(txq);
4045
4046 return rc;
4047 }
4048
4049 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4050 const void *ns);
4051 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4052 const void *ns);
4053
4054 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4055 {
4056 return netdev_class_create_file_ns(class_attr, NULL);
4057 }
4058
4059 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4060 {
4061 netdev_class_remove_file_ns(class_attr, NULL);
4062 }
4063
4064 extern const struct kobj_ns_type_operations net_ns_type_operations;
4065
4066 const char *netdev_drivername(const struct net_device *dev);
4067
4068 void linkwatch_run_queue(void);
4069
4070 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4071 netdev_features_t f2)
4072 {
4073 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4074 if (f1 & NETIF_F_HW_CSUM)
4075 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4076 else
4077 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4078 }
4079
4080 return f1 & f2;
4081 }
4082
4083 static inline netdev_features_t netdev_get_wanted_features(
4084 struct net_device *dev)
4085 {
4086 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4087 }
4088 netdev_features_t netdev_increment_features(netdev_features_t all,
4089 netdev_features_t one, netdev_features_t mask);
4090
4091 /* Allow TSO being used on stacked device :
4092 * Performing the GSO segmentation before last device
4093 * is a performance improvement.
4094 */
4095 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4096 netdev_features_t mask)
4097 {
4098 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4099 }
4100
4101 int __netdev_update_features(struct net_device *dev);
4102 void netdev_update_features(struct net_device *dev);
4103 void netdev_change_features(struct net_device *dev);
4104
4105 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4106 struct net_device *dev);
4107
4108 netdev_features_t passthru_features_check(struct sk_buff *skb,
4109 struct net_device *dev,
4110 netdev_features_t features);
4111 netdev_features_t netif_skb_features(struct sk_buff *skb);
4112
4113 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4114 {
4115 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4116
4117 /* check flags correspondence */
4118 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4119 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4120 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4121 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4122 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4123 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4124 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4125 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4126 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4127 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4128 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4129 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4130 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4131 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4132 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4133 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4134 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4135
4136 return (features & feature) == feature;
4137 }
4138
4139 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4140 {
4141 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4142 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4143 }
4144
4145 static inline bool netif_needs_gso(struct sk_buff *skb,
4146 netdev_features_t features)
4147 {
4148 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4149 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4150 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4151 }
4152
4153 static inline void netif_set_gso_max_size(struct net_device *dev,
4154 unsigned int size)
4155 {
4156 dev->gso_max_size = size;
4157 }
4158
4159 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4160 int pulled_hlen, u16 mac_offset,
4161 int mac_len)
4162 {
4163 skb->protocol = protocol;
4164 skb->encapsulation = 1;
4165 skb_push(skb, pulled_hlen);
4166 skb_reset_transport_header(skb);
4167 skb->mac_header = mac_offset;
4168 skb->network_header = skb->mac_header + mac_len;
4169 skb->mac_len = mac_len;
4170 }
4171
4172 static inline bool netif_is_macsec(const struct net_device *dev)
4173 {
4174 return dev->priv_flags & IFF_MACSEC;
4175 }
4176
4177 static inline bool netif_is_macvlan(const struct net_device *dev)
4178 {
4179 return dev->priv_flags & IFF_MACVLAN;
4180 }
4181
4182 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4183 {
4184 return dev->priv_flags & IFF_MACVLAN_PORT;
4185 }
4186
4187 static inline bool netif_is_ipvlan(const struct net_device *dev)
4188 {
4189 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4190 }
4191
4192 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4193 {
4194 return dev->priv_flags & IFF_IPVLAN_MASTER;
4195 }
4196
4197 static inline bool netif_is_bond_master(const struct net_device *dev)
4198 {
4199 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4200 }
4201
4202 static inline bool netif_is_bond_slave(const struct net_device *dev)
4203 {
4204 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4205 }
4206
4207 static inline bool netif_supports_nofcs(struct net_device *dev)
4208 {
4209 return dev->priv_flags & IFF_SUPP_NOFCS;
4210 }
4211
4212 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4213 {
4214 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4215 }
4216
4217 static inline bool netif_is_l3_master(const struct net_device *dev)
4218 {
4219 return dev->priv_flags & IFF_L3MDEV_MASTER;
4220 }
4221
4222 static inline bool netif_is_l3_slave(const struct net_device *dev)
4223 {
4224 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4225 }
4226
4227 static inline bool netif_is_bridge_master(const struct net_device *dev)
4228 {
4229 return dev->priv_flags & IFF_EBRIDGE;
4230 }
4231
4232 static inline bool netif_is_bridge_port(const struct net_device *dev)
4233 {
4234 return dev->priv_flags & IFF_BRIDGE_PORT;
4235 }
4236
4237 static inline bool netif_is_ovs_master(const struct net_device *dev)
4238 {
4239 return dev->priv_flags & IFF_OPENVSWITCH;
4240 }
4241
4242 static inline bool netif_is_ovs_port(const struct net_device *dev)
4243 {
4244 return dev->priv_flags & IFF_OVS_DATAPATH;
4245 }
4246
4247 static inline bool netif_is_team_master(const struct net_device *dev)
4248 {
4249 return dev->priv_flags & IFF_TEAM;
4250 }
4251
4252 static inline bool netif_is_team_port(const struct net_device *dev)
4253 {
4254 return dev->priv_flags & IFF_TEAM_PORT;
4255 }
4256
4257 static inline bool netif_is_lag_master(const struct net_device *dev)
4258 {
4259 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4260 }
4261
4262 static inline bool netif_is_lag_port(const struct net_device *dev)
4263 {
4264 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4265 }
4266
4267 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4268 {
4269 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4270 }
4271
4272 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4273 static inline void netif_keep_dst(struct net_device *dev)
4274 {
4275 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4276 }
4277
4278 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4279 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4280 {
4281 /* TODO: reserve and use an additional IFF bit, if we get more users */
4282 return dev->priv_flags & IFF_MACSEC;
4283 }
4284
4285 extern struct pernet_operations __net_initdata loopback_net_ops;
4286
4287 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4288
4289 /* netdev_printk helpers, similar to dev_printk */
4290
4291 static inline const char *netdev_name(const struct net_device *dev)
4292 {
4293 if (!dev->name[0] || strchr(dev->name, '%'))
4294 return "(unnamed net_device)";
4295 return dev->name;
4296 }
4297
4298 static inline bool netdev_unregistering(const struct net_device *dev)
4299 {
4300 return dev->reg_state == NETREG_UNREGISTERING;
4301 }
4302
4303 static inline const char *netdev_reg_state(const struct net_device *dev)
4304 {
4305 switch (dev->reg_state) {
4306 case NETREG_UNINITIALIZED: return " (uninitialized)";
4307 case NETREG_REGISTERED: return "";
4308 case NETREG_UNREGISTERING: return " (unregistering)";
4309 case NETREG_UNREGISTERED: return " (unregistered)";
4310 case NETREG_RELEASED: return " (released)";
4311 case NETREG_DUMMY: return " (dummy)";
4312 }
4313
4314 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4315 return " (unknown)";
4316 }
4317
4318 __printf(3, 4)
4319 void netdev_printk(const char *level, const struct net_device *dev,
4320 const char *format, ...);
4321 __printf(2, 3)
4322 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4323 __printf(2, 3)
4324 void netdev_alert(const struct net_device *dev, const char *format, ...);
4325 __printf(2, 3)
4326 void netdev_crit(const struct net_device *dev, const char *format, ...);
4327 __printf(2, 3)
4328 void netdev_err(const struct net_device *dev, const char *format, ...);
4329 __printf(2, 3)
4330 void netdev_warn(const struct net_device *dev, const char *format, ...);
4331 __printf(2, 3)
4332 void netdev_notice(const struct net_device *dev, const char *format, ...);
4333 __printf(2, 3)
4334 void netdev_info(const struct net_device *dev, const char *format, ...);
4335
4336 #define MODULE_ALIAS_NETDEV(device) \
4337 MODULE_ALIAS("netdev-" device)
4338
4339 #if defined(CONFIG_DYNAMIC_DEBUG)
4340 #define netdev_dbg(__dev, format, args...) \
4341 do { \
4342 dynamic_netdev_dbg(__dev, format, ##args); \
4343 } while (0)
4344 #elif defined(DEBUG)
4345 #define netdev_dbg(__dev, format, args...) \
4346 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4347 #else
4348 #define netdev_dbg(__dev, format, args...) \
4349 ({ \
4350 if (0) \
4351 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4352 })
4353 #endif
4354
4355 #if defined(VERBOSE_DEBUG)
4356 #define netdev_vdbg netdev_dbg
4357 #else
4358
4359 #define netdev_vdbg(dev, format, args...) \
4360 ({ \
4361 if (0) \
4362 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4363 0; \
4364 })
4365 #endif
4366
4367 /*
4368 * netdev_WARN() acts like dev_printk(), but with the key difference
4369 * of using a WARN/WARN_ON to get the message out, including the
4370 * file/line information and a backtrace.
4371 */
4372 #define netdev_WARN(dev, format, args...) \
4373 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4374 netdev_reg_state(dev), ##args)
4375
4376 /* netif printk helpers, similar to netdev_printk */
4377
4378 #define netif_printk(priv, type, level, dev, fmt, args...) \
4379 do { \
4380 if (netif_msg_##type(priv)) \
4381 netdev_printk(level, (dev), fmt, ##args); \
4382 } while (0)
4383
4384 #define netif_level(level, priv, type, dev, fmt, args...) \
4385 do { \
4386 if (netif_msg_##type(priv)) \
4387 netdev_##level(dev, fmt, ##args); \
4388 } while (0)
4389
4390 #define netif_emerg(priv, type, dev, fmt, args...) \
4391 netif_level(emerg, priv, type, dev, fmt, ##args)
4392 #define netif_alert(priv, type, dev, fmt, args...) \
4393 netif_level(alert, priv, type, dev, fmt, ##args)
4394 #define netif_crit(priv, type, dev, fmt, args...) \
4395 netif_level(crit, priv, type, dev, fmt, ##args)
4396 #define netif_err(priv, type, dev, fmt, args...) \
4397 netif_level(err, priv, type, dev, fmt, ##args)
4398 #define netif_warn(priv, type, dev, fmt, args...) \
4399 netif_level(warn, priv, type, dev, fmt, ##args)
4400 #define netif_notice(priv, type, dev, fmt, args...) \
4401 netif_level(notice, priv, type, dev, fmt, ##args)
4402 #define netif_info(priv, type, dev, fmt, args...) \
4403 netif_level(info, priv, type, dev, fmt, ##args)
4404
4405 #if defined(CONFIG_DYNAMIC_DEBUG)
4406 #define netif_dbg(priv, type, netdev, format, args...) \
4407 do { \
4408 if (netif_msg_##type(priv)) \
4409 dynamic_netdev_dbg(netdev, format, ##args); \
4410 } while (0)
4411 #elif defined(DEBUG)
4412 #define netif_dbg(priv, type, dev, format, args...) \
4413 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4414 #else
4415 #define netif_dbg(priv, type, dev, format, args...) \
4416 ({ \
4417 if (0) \
4418 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4419 0; \
4420 })
4421 #endif
4422
4423 /* if @cond then downgrade to debug, else print at @level */
4424 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4425 do { \
4426 if (cond) \
4427 netif_dbg(priv, type, netdev, fmt, ##args); \
4428 else \
4429 netif_ ## level(priv, type, netdev, fmt, ##args); \
4430 } while (0)
4431
4432 #if defined(VERBOSE_DEBUG)
4433 #define netif_vdbg netif_dbg
4434 #else
4435 #define netif_vdbg(priv, type, dev, format, args...) \
4436 ({ \
4437 if (0) \
4438 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4439 0; \
4440 })
4441 #endif
4442
4443 /*
4444 * The list of packet types we will receive (as opposed to discard)
4445 * and the routines to invoke.
4446 *
4447 * Why 16. Because with 16 the only overlap we get on a hash of the
4448 * low nibble of the protocol value is RARP/SNAP/X.25.
4449 *
4450 * NOTE: That is no longer true with the addition of VLAN tags. Not
4451 * sure which should go first, but I bet it won't make much
4452 * difference if we are running VLANs. The good news is that
4453 * this protocol won't be in the list unless compiled in, so
4454 * the average user (w/out VLANs) will not be adversely affected.
4455 * --BLG
4456 *
4457 * 0800 IP
4458 * 8100 802.1Q VLAN
4459 * 0001 802.3
4460 * 0002 AX.25
4461 * 0004 802.2
4462 * 8035 RARP
4463 * 0005 SNAP
4464 * 0805 X.25
4465 * 0806 ARP
4466 * 8137 IPX
4467 * 0009 Localtalk
4468 * 86DD IPv6
4469 */
4470 #define PTYPE_HASH_SIZE (16)
4471 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4472
4473 #endif /* _LINUX_NETDEVICE_H */