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