]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/netdevice.h
ipv4: lock mtu in fnhe when received PMTU < net.ipv4.route.min_pmtu
[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_QDISC_MQPRIO,
774 TC_SETUP_CLSU32,
775 TC_SETUP_CLSFLOWER,
776 TC_SETUP_CLSMATCHALL,
777 TC_SETUP_CLSBPF,
778 TC_SETUP_BLOCK,
779 TC_SETUP_QDISC_CBS,
780 TC_SETUP_QDISC_RED,
781 };
782
783 /* These structures hold the attributes of bpf state that are being passed
784 * to the netdevice through the bpf op.
785 */
786 enum bpf_netdev_command {
787 /* Set or clear a bpf program used in the earliest stages of packet
788 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
789 * is responsible for calling bpf_prog_put on any old progs that are
790 * stored. In case of error, the callee need not release the new prog
791 * reference, but on success it takes ownership and must bpf_prog_put
792 * when it is no longer used.
793 */
794 XDP_SETUP_PROG,
795 XDP_SETUP_PROG_HW,
796 /* Check if a bpf program is set on the device. The callee should
797 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
798 * is equivalent to XDP_ATTACHED_DRV.
799 */
800 XDP_QUERY_PROG,
801 /* BPF program for offload callbacks, invoked at program load time. */
802 BPF_OFFLOAD_VERIFIER_PREP,
803 BPF_OFFLOAD_TRANSLATE,
804 BPF_OFFLOAD_DESTROY,
805 };
806
807 struct bpf_ext_analyzer_ops;
808 struct netlink_ext_ack;
809
810 struct netdev_bpf {
811 enum bpf_netdev_command command;
812 union {
813 /* XDP_SETUP_PROG */
814 struct {
815 u32 flags;
816 struct bpf_prog *prog;
817 struct netlink_ext_ack *extack;
818 };
819 /* XDP_QUERY_PROG */
820 struct {
821 u8 prog_attached;
822 u32 prog_id;
823 };
824 /* BPF_OFFLOAD_VERIFIER_PREP */
825 struct {
826 struct bpf_prog *prog;
827 const struct bpf_ext_analyzer_ops *ops; /* callee set */
828 } verifier;
829 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
830 struct {
831 struct bpf_prog *prog;
832 } offload;
833 };
834 };
835
836 #ifdef CONFIG_XFRM_OFFLOAD
837 struct xfrmdev_ops {
838 int (*xdo_dev_state_add) (struct xfrm_state *x);
839 void (*xdo_dev_state_delete) (struct xfrm_state *x);
840 void (*xdo_dev_state_free) (struct xfrm_state *x);
841 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
842 struct xfrm_state *x);
843 };
844 #endif
845
846 struct dev_ifalias {
847 struct rcu_head rcuhead;
848 char ifalias[];
849 };
850
851 /*
852 * This structure defines the management hooks for network devices.
853 * The following hooks can be defined; unless noted otherwise, they are
854 * optional and can be filled with a null pointer.
855 *
856 * int (*ndo_init)(struct net_device *dev);
857 * This function is called once when a network device is registered.
858 * The network device can use this for any late stage initialization
859 * or semantic validation. It can fail with an error code which will
860 * be propagated back to register_netdev.
861 *
862 * void (*ndo_uninit)(struct net_device *dev);
863 * This function is called when device is unregistered or when registration
864 * fails. It is not called if init fails.
865 *
866 * int (*ndo_open)(struct net_device *dev);
867 * This function is called when a network device transitions to the up
868 * state.
869 *
870 * int (*ndo_stop)(struct net_device *dev);
871 * This function is called when a network device transitions to the down
872 * state.
873 *
874 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
875 * struct net_device *dev);
876 * Called when a packet needs to be transmitted.
877 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
878 * the queue before that can happen; it's for obsolete devices and weird
879 * corner cases, but the stack really does a non-trivial amount
880 * of useless work if you return NETDEV_TX_BUSY.
881 * Required; cannot be NULL.
882 *
883 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
884 * struct net_device *dev
885 * netdev_features_t features);
886 * Called by core transmit path to determine if device is capable of
887 * performing offload operations on a given packet. This is to give
888 * the device an opportunity to implement any restrictions that cannot
889 * be otherwise expressed by feature flags. The check is called with
890 * the set of features that the stack has calculated and it returns
891 * those the driver believes to be appropriate.
892 *
893 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
894 * void *accel_priv, select_queue_fallback_t fallback);
895 * Called to decide which queue to use when device supports multiple
896 * transmit queues.
897 *
898 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
899 * This function is called to allow device receiver to make
900 * changes to configuration when multicast or promiscuous is enabled.
901 *
902 * void (*ndo_set_rx_mode)(struct net_device *dev);
903 * This function is called device changes address list filtering.
904 * If driver handles unicast address filtering, it should set
905 * IFF_UNICAST_FLT in its priv_flags.
906 *
907 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
908 * This function is called when the Media Access Control address
909 * needs to be changed. If this interface is not defined, the
910 * MAC address can not be changed.
911 *
912 * int (*ndo_validate_addr)(struct net_device *dev);
913 * Test if Media Access Control address is valid for the device.
914 *
915 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
916 * Called when a user requests an ioctl which can't be handled by
917 * the generic interface code. If not defined ioctls return
918 * not supported error code.
919 *
920 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
921 * Used to set network devices bus interface parameters. This interface
922 * is retained for legacy reasons; new devices should use the bus
923 * interface (PCI) for low level management.
924 *
925 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
926 * Called when a user wants to change the Maximum Transfer Unit
927 * of a device.
928 *
929 * void (*ndo_tx_timeout)(struct net_device *dev);
930 * Callback used when the transmitter has not made any progress
931 * for dev->watchdog ticks.
932 *
933 * void (*ndo_get_stats64)(struct net_device *dev,
934 * struct rtnl_link_stats64 *storage);
935 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
936 * Called when a user wants to get the network device usage
937 * statistics. Drivers must do one of the following:
938 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
939 * rtnl_link_stats64 structure passed by the caller.
940 * 2. Define @ndo_get_stats to update a net_device_stats structure
941 * (which should normally be dev->stats) and return a pointer to
942 * it. The structure may be changed asynchronously only if each
943 * field is written atomically.
944 * 3. Update dev->stats asynchronously and atomically, and define
945 * neither operation.
946 *
947 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
948 * Return true if this device supports offload stats of this attr_id.
949 *
950 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
951 * void *attr_data)
952 * Get statistics for offload operations by attr_id. Write it into the
953 * attr_data pointer.
954 *
955 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
956 * If device supports VLAN filtering this function is called when a
957 * VLAN id is registered.
958 *
959 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
960 * If device supports VLAN filtering this function is called when a
961 * VLAN id is unregistered.
962 *
963 * void (*ndo_poll_controller)(struct net_device *dev);
964 *
965 * SR-IOV management functions.
966 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
967 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
968 * u8 qos, __be16 proto);
969 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
970 * int max_tx_rate);
971 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
972 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
973 * int (*ndo_get_vf_config)(struct net_device *dev,
974 * int vf, struct ifla_vf_info *ivf);
975 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
976 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
977 * struct nlattr *port[]);
978 *
979 * Enable or disable the VF ability to query its RSS Redirection Table and
980 * Hash Key. This is needed since on some devices VF share this information
981 * with PF and querying it may introduce a theoretical security risk.
982 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
983 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
984 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
985 * void *type_data);
986 * Called to setup any 'tc' scheduler, classifier or action on @dev.
987 * This is always called from the stack with the rtnl lock held and netif
988 * tx queues stopped. This allows the netdevice to perform queue
989 * management safely.
990 *
991 * Fiber Channel over Ethernet (FCoE) offload functions.
992 * int (*ndo_fcoe_enable)(struct net_device *dev);
993 * Called when the FCoE protocol stack wants to start using LLD for FCoE
994 * so the underlying device can perform whatever needed configuration or
995 * initialization to support acceleration of FCoE traffic.
996 *
997 * int (*ndo_fcoe_disable)(struct net_device *dev);
998 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
999 * so the underlying device can perform whatever needed clean-ups to
1000 * stop supporting acceleration of FCoE traffic.
1001 *
1002 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1003 * struct scatterlist *sgl, unsigned int sgc);
1004 * Called when the FCoE Initiator wants to initialize an I/O that
1005 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1006 * perform necessary setup and returns 1 to indicate the device is set up
1007 * successfully to perform DDP on this I/O, otherwise this returns 0.
1008 *
1009 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1010 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1011 * indicated by the FC exchange id 'xid', so the underlying device can
1012 * clean up and reuse resources for later DDP requests.
1013 *
1014 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1015 * struct scatterlist *sgl, unsigned int sgc);
1016 * Called when the FCoE Target wants to initialize an I/O that
1017 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1018 * perform necessary setup and returns 1 to indicate the device is set up
1019 * successfully to perform DDP on this I/O, otherwise this returns 0.
1020 *
1021 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1022 * struct netdev_fcoe_hbainfo *hbainfo);
1023 * Called when the FCoE Protocol stack wants information on the underlying
1024 * device. This information is utilized by the FCoE protocol stack to
1025 * register attributes with Fiber Channel management service as per the
1026 * FC-GS Fabric Device Management Information(FDMI) specification.
1027 *
1028 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1029 * Called when the underlying device wants to override default World Wide
1030 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1031 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1032 * protocol stack to use.
1033 *
1034 * RFS acceleration.
1035 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1036 * u16 rxq_index, u32 flow_id);
1037 * Set hardware filter for RFS. rxq_index is the target queue index;
1038 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1039 * Return the filter ID on success, or a negative error code.
1040 *
1041 * Slave management functions (for bridge, bonding, etc).
1042 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1043 * Called to make another netdev an underling.
1044 *
1045 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1046 * Called to release previously enslaved netdev.
1047 *
1048 * Feature/offload setting functions.
1049 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1050 * netdev_features_t features);
1051 * Adjusts the requested feature flags according to device-specific
1052 * constraints, and returns the resulting flags. Must not modify
1053 * the device state.
1054 *
1055 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1056 * Called to update device configuration to new features. Passed
1057 * feature set might be less than what was returned by ndo_fix_features()).
1058 * Must return >0 or -errno if it changed dev->features itself.
1059 *
1060 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1061 * struct net_device *dev,
1062 * const unsigned char *addr, u16 vid, u16 flags)
1063 * Adds an FDB entry to dev for addr.
1064 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1065 * struct net_device *dev,
1066 * const unsigned char *addr, u16 vid)
1067 * Deletes the FDB entry from dev coresponding to addr.
1068 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1069 * struct net_device *dev, struct net_device *filter_dev,
1070 * int *idx)
1071 * Used to add FDB entries to dump requests. Implementers should add
1072 * entries to skb and update idx with the number of entries.
1073 *
1074 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1075 * u16 flags)
1076 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1077 * struct net_device *dev, u32 filter_mask,
1078 * int nlflags)
1079 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1080 * u16 flags);
1081 *
1082 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1083 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1084 * which do not represent real hardware may define this to allow their
1085 * userspace components to manage their virtual carrier state. Devices
1086 * that determine carrier state from physical hardware properties (eg
1087 * network cables) or protocol-dependent mechanisms (eg
1088 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1089 *
1090 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1091 * struct netdev_phys_item_id *ppid);
1092 * Called to get ID of physical port of this device. If driver does
1093 * not implement this, it is assumed that the hw is not able to have
1094 * multiple net devices on single physical port.
1095 *
1096 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1097 * struct udp_tunnel_info *ti);
1098 * Called by UDP tunnel to notify a driver about the UDP port and socket
1099 * address family that a UDP tunnel is listnening to. It is called only
1100 * when a new port starts listening. The operation is protected by the
1101 * RTNL.
1102 *
1103 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1104 * struct udp_tunnel_info *ti);
1105 * Called by UDP tunnel to notify the driver about a UDP port and socket
1106 * address family that the UDP tunnel is not listening to anymore. The
1107 * operation is protected by the RTNL.
1108 *
1109 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1110 * struct net_device *dev)
1111 * Called by upper layer devices to accelerate switching or other
1112 * station functionality into hardware. 'pdev is the lowerdev
1113 * to use for the offload and 'dev' is the net device that will
1114 * back the offload. Returns a pointer to the private structure
1115 * the upper layer will maintain.
1116 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1117 * Called by upper layer device to delete the station created
1118 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1119 * the station and priv is the structure returned by the add
1120 * operation.
1121 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1122 * int queue_index, u32 maxrate);
1123 * Called when a user wants to set a max-rate limitation of specific
1124 * TX queue.
1125 * int (*ndo_get_iflink)(const struct net_device *dev);
1126 * Called to get the iflink value of this device.
1127 * void (*ndo_change_proto_down)(struct net_device *dev,
1128 * bool proto_down);
1129 * This function is used to pass protocol port error state information
1130 * to the switch driver. The switch driver can react to the proto_down
1131 * by doing a phys down on the associated switch port.
1132 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1133 * This function is used to get egress tunnel information for given skb.
1134 * This is useful for retrieving outer tunnel header parameters while
1135 * sampling packet.
1136 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1137 * This function is used to specify the headroom that the skb must
1138 * consider when allocation skb during packet reception. Setting
1139 * appropriate rx headroom value allows avoiding skb head copy on
1140 * forward. Setting a negative value resets the rx headroom to the
1141 * default value.
1142 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1143 * This function is used to set or query state related to XDP on the
1144 * netdevice and manage BPF offload. See definition of
1145 * enum bpf_netdev_command for details.
1146 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1147 * This function is used to submit a XDP packet for transmit on a
1148 * netdevice.
1149 * void (*ndo_xdp_flush)(struct net_device *dev);
1150 * This function is used to inform the driver to flush a particular
1151 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1152 */
1153 struct net_device_ops {
1154 int (*ndo_init)(struct net_device *dev);
1155 void (*ndo_uninit)(struct net_device *dev);
1156 int (*ndo_open)(struct net_device *dev);
1157 int (*ndo_stop)(struct net_device *dev);
1158 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1159 struct net_device *dev);
1160 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1161 struct net_device *dev,
1162 netdev_features_t features);
1163 u16 (*ndo_select_queue)(struct net_device *dev,
1164 struct sk_buff *skb,
1165 void *accel_priv,
1166 select_queue_fallback_t fallback);
1167 void (*ndo_change_rx_flags)(struct net_device *dev,
1168 int flags);
1169 void (*ndo_set_rx_mode)(struct net_device *dev);
1170 int (*ndo_set_mac_address)(struct net_device *dev,
1171 void *addr);
1172 int (*ndo_validate_addr)(struct net_device *dev);
1173 int (*ndo_do_ioctl)(struct net_device *dev,
1174 struct ifreq *ifr, int cmd);
1175 int (*ndo_set_config)(struct net_device *dev,
1176 struct ifmap *map);
1177 int (*ndo_change_mtu)(struct net_device *dev,
1178 int new_mtu);
1179 int (*ndo_neigh_setup)(struct net_device *dev,
1180 struct neigh_parms *);
1181 void (*ndo_tx_timeout) (struct net_device *dev);
1182
1183 void (*ndo_get_stats64)(struct net_device *dev,
1184 struct rtnl_link_stats64 *storage);
1185 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1186 int (*ndo_get_offload_stats)(int attr_id,
1187 const struct net_device *dev,
1188 void *attr_data);
1189 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1190
1191 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1192 __be16 proto, u16 vid);
1193 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1194 __be16 proto, u16 vid);
1195 #ifdef CONFIG_NET_POLL_CONTROLLER
1196 void (*ndo_poll_controller)(struct net_device *dev);
1197 int (*ndo_netpoll_setup)(struct net_device *dev,
1198 struct netpoll_info *info);
1199 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1200 #endif
1201 int (*ndo_set_vf_mac)(struct net_device *dev,
1202 int queue, u8 *mac);
1203 int (*ndo_set_vf_vlan)(struct net_device *dev,
1204 int queue, u16 vlan,
1205 u8 qos, __be16 proto);
1206 int (*ndo_set_vf_rate)(struct net_device *dev,
1207 int vf, int min_tx_rate,
1208 int max_tx_rate);
1209 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1210 int vf, bool setting);
1211 int (*ndo_set_vf_trust)(struct net_device *dev,
1212 int vf, bool setting);
1213 int (*ndo_get_vf_config)(struct net_device *dev,
1214 int vf,
1215 struct ifla_vf_info *ivf);
1216 int (*ndo_set_vf_link_state)(struct net_device *dev,
1217 int vf, int link_state);
1218 int (*ndo_get_vf_stats)(struct net_device *dev,
1219 int vf,
1220 struct ifla_vf_stats
1221 *vf_stats);
1222 int (*ndo_set_vf_port)(struct net_device *dev,
1223 int vf,
1224 struct nlattr *port[]);
1225 int (*ndo_get_vf_port)(struct net_device *dev,
1226 int vf, struct sk_buff *skb);
1227 int (*ndo_set_vf_guid)(struct net_device *dev,
1228 int vf, u64 guid,
1229 int guid_type);
1230 int (*ndo_set_vf_rss_query_en)(
1231 struct net_device *dev,
1232 int vf, bool setting);
1233 int (*ndo_setup_tc)(struct net_device *dev,
1234 enum tc_setup_type type,
1235 void *type_data);
1236 #if IS_ENABLED(CONFIG_FCOE)
1237 int (*ndo_fcoe_enable)(struct net_device *dev);
1238 int (*ndo_fcoe_disable)(struct net_device *dev);
1239 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1240 u16 xid,
1241 struct scatterlist *sgl,
1242 unsigned int sgc);
1243 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1244 u16 xid);
1245 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1246 u16 xid,
1247 struct scatterlist *sgl,
1248 unsigned int sgc);
1249 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1250 struct netdev_fcoe_hbainfo *hbainfo);
1251 #endif
1252
1253 #if IS_ENABLED(CONFIG_LIBFCOE)
1254 #define NETDEV_FCOE_WWNN 0
1255 #define NETDEV_FCOE_WWPN 1
1256 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1257 u64 *wwn, int type);
1258 #endif
1259
1260 #ifdef CONFIG_RFS_ACCEL
1261 int (*ndo_rx_flow_steer)(struct net_device *dev,
1262 const struct sk_buff *skb,
1263 u16 rxq_index,
1264 u32 flow_id);
1265 #endif
1266 int (*ndo_add_slave)(struct net_device *dev,
1267 struct net_device *slave_dev,
1268 struct netlink_ext_ack *extack);
1269 int (*ndo_del_slave)(struct net_device *dev,
1270 struct net_device *slave_dev);
1271 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1272 netdev_features_t features);
1273 int (*ndo_set_features)(struct net_device *dev,
1274 netdev_features_t features);
1275 int (*ndo_neigh_construct)(struct net_device *dev,
1276 struct neighbour *n);
1277 void (*ndo_neigh_destroy)(struct net_device *dev,
1278 struct neighbour *n);
1279
1280 int (*ndo_fdb_add)(struct ndmsg *ndm,
1281 struct nlattr *tb[],
1282 struct net_device *dev,
1283 const unsigned char *addr,
1284 u16 vid,
1285 u16 flags);
1286 int (*ndo_fdb_del)(struct ndmsg *ndm,
1287 struct nlattr *tb[],
1288 struct net_device *dev,
1289 const unsigned char *addr,
1290 u16 vid);
1291 int (*ndo_fdb_dump)(struct sk_buff *skb,
1292 struct netlink_callback *cb,
1293 struct net_device *dev,
1294 struct net_device *filter_dev,
1295 int *idx);
1296
1297 int (*ndo_bridge_setlink)(struct net_device *dev,
1298 struct nlmsghdr *nlh,
1299 u16 flags);
1300 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1301 u32 pid, u32 seq,
1302 struct net_device *dev,
1303 u32 filter_mask,
1304 int nlflags);
1305 int (*ndo_bridge_dellink)(struct net_device *dev,
1306 struct nlmsghdr *nlh,
1307 u16 flags);
1308 int (*ndo_change_carrier)(struct net_device *dev,
1309 bool new_carrier);
1310 int (*ndo_get_phys_port_id)(struct net_device *dev,
1311 struct netdev_phys_item_id *ppid);
1312 int (*ndo_get_phys_port_name)(struct net_device *dev,
1313 char *name, size_t len);
1314 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1315 struct udp_tunnel_info *ti);
1316 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1317 struct udp_tunnel_info *ti);
1318 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1319 struct net_device *dev);
1320 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1321 void *priv);
1322
1323 int (*ndo_get_lock_subclass)(struct net_device *dev);
1324 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1325 int queue_index,
1326 u32 maxrate);
1327 int (*ndo_get_iflink)(const struct net_device *dev);
1328 int (*ndo_change_proto_down)(struct net_device *dev,
1329 bool proto_down);
1330 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1331 struct sk_buff *skb);
1332 void (*ndo_set_rx_headroom)(struct net_device *dev,
1333 int needed_headroom);
1334 int (*ndo_bpf)(struct net_device *dev,
1335 struct netdev_bpf *bpf);
1336 int (*ndo_xdp_xmit)(struct net_device *dev,
1337 struct xdp_buff *xdp);
1338 void (*ndo_xdp_flush)(struct net_device *dev);
1339 };
1340
1341 /**
1342 * enum net_device_priv_flags - &struct net_device priv_flags
1343 *
1344 * These are the &struct net_device, they are only set internally
1345 * by drivers and used in the kernel. These flags are invisible to
1346 * userspace; this means that the order of these flags can change
1347 * during any kernel release.
1348 *
1349 * You should have a pretty good reason to be extending these flags.
1350 *
1351 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1352 * @IFF_EBRIDGE: Ethernet bridging device
1353 * @IFF_BONDING: bonding master or slave
1354 * @IFF_ISATAP: ISATAP interface (RFC4214)
1355 * @IFF_WAN_HDLC: WAN HDLC device
1356 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1357 * release skb->dst
1358 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1359 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1360 * @IFF_MACVLAN_PORT: device used as macvlan port
1361 * @IFF_BRIDGE_PORT: device used as bridge port
1362 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1363 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1364 * @IFF_UNICAST_FLT: Supports unicast filtering
1365 * @IFF_TEAM_PORT: device used as team port
1366 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1367 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1368 * change when it's running
1369 * @IFF_MACVLAN: Macvlan device
1370 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1371 * underlying stacked devices
1372 * @IFF_IPVLAN_MASTER: IPvlan master device
1373 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1374 * @IFF_L3MDEV_MASTER: device is an L3 master device
1375 * @IFF_NO_QUEUE: device can run without qdisc attached
1376 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1377 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1378 * @IFF_TEAM: device is a team device
1379 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1380 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1381 * entity (i.e. the master device for bridged veth)
1382 * @IFF_MACSEC: device is a MACsec device
1383 */
1384 enum netdev_priv_flags {
1385 IFF_802_1Q_VLAN = 1<<0,
1386 IFF_EBRIDGE = 1<<1,
1387 IFF_BONDING = 1<<2,
1388 IFF_ISATAP = 1<<3,
1389 IFF_WAN_HDLC = 1<<4,
1390 IFF_XMIT_DST_RELEASE = 1<<5,
1391 IFF_DONT_BRIDGE = 1<<6,
1392 IFF_DISABLE_NETPOLL = 1<<7,
1393 IFF_MACVLAN_PORT = 1<<8,
1394 IFF_BRIDGE_PORT = 1<<9,
1395 IFF_OVS_DATAPATH = 1<<10,
1396 IFF_TX_SKB_SHARING = 1<<11,
1397 IFF_UNICAST_FLT = 1<<12,
1398 IFF_TEAM_PORT = 1<<13,
1399 IFF_SUPP_NOFCS = 1<<14,
1400 IFF_LIVE_ADDR_CHANGE = 1<<15,
1401 IFF_MACVLAN = 1<<16,
1402 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1403 IFF_IPVLAN_MASTER = 1<<18,
1404 IFF_IPVLAN_SLAVE = 1<<19,
1405 IFF_L3MDEV_MASTER = 1<<20,
1406 IFF_NO_QUEUE = 1<<21,
1407 IFF_OPENVSWITCH = 1<<22,
1408 IFF_L3MDEV_SLAVE = 1<<23,
1409 IFF_TEAM = 1<<24,
1410 IFF_RXFH_CONFIGURED = 1<<25,
1411 IFF_PHONY_HEADROOM = 1<<26,
1412 IFF_MACSEC = 1<<27,
1413 };
1414
1415 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1416 #define IFF_EBRIDGE IFF_EBRIDGE
1417 #define IFF_BONDING IFF_BONDING
1418 #define IFF_ISATAP IFF_ISATAP
1419 #define IFF_WAN_HDLC IFF_WAN_HDLC
1420 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1421 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1422 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1423 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1424 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1425 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1426 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1427 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1428 #define IFF_TEAM_PORT IFF_TEAM_PORT
1429 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1430 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1431 #define IFF_MACVLAN IFF_MACVLAN
1432 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1433 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1434 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1435 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1436 #define IFF_NO_QUEUE IFF_NO_QUEUE
1437 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1438 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1439 #define IFF_TEAM IFF_TEAM
1440 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1441 #define IFF_MACSEC IFF_MACSEC
1442
1443 /**
1444 * struct net_device - The DEVICE structure.
1445 *
1446 * Actually, this whole structure is a big mistake. It mixes I/O
1447 * data with strictly "high-level" data, and it has to know about
1448 * almost every data structure used in the INET module.
1449 *
1450 * @name: This is the first field of the "visible" part of this structure
1451 * (i.e. as seen by users in the "Space.c" file). It is the name
1452 * of the interface.
1453 *
1454 * @name_hlist: Device name hash chain, please keep it close to name[]
1455 * @ifalias: SNMP alias
1456 * @mem_end: Shared memory end
1457 * @mem_start: Shared memory start
1458 * @base_addr: Device I/O address
1459 * @irq: Device IRQ number
1460 *
1461 * @carrier_changes: Stats to monitor carrier on<->off transitions
1462 *
1463 * @state: Generic network queuing layer state, see netdev_state_t
1464 * @dev_list: The global list of network devices
1465 * @napi_list: List entry used for polling NAPI devices
1466 * @unreg_list: List entry when we are unregistering the
1467 * device; see the function unregister_netdev
1468 * @close_list: List entry used when we are closing the device
1469 * @ptype_all: Device-specific packet handlers for all protocols
1470 * @ptype_specific: Device-specific, protocol-specific packet handlers
1471 *
1472 * @adj_list: Directly linked devices, like slaves for bonding
1473 * @features: Currently active device features
1474 * @hw_features: User-changeable features
1475 *
1476 * @wanted_features: User-requested features
1477 * @vlan_features: Mask of features inheritable by VLAN devices
1478 *
1479 * @hw_enc_features: Mask of features inherited by encapsulating devices
1480 * This field indicates what encapsulation
1481 * offloads the hardware is capable of doing,
1482 * and drivers will need to set them appropriately.
1483 *
1484 * @mpls_features: Mask of features inheritable by MPLS
1485 *
1486 * @ifindex: interface index
1487 * @group: The group the device belongs to
1488 *
1489 * @stats: Statistics struct, which was left as a legacy, use
1490 * rtnl_link_stats64 instead
1491 *
1492 * @rx_dropped: Dropped packets by core network,
1493 * do not use this in drivers
1494 * @tx_dropped: Dropped packets by core network,
1495 * do not use this in drivers
1496 * @rx_nohandler: nohandler dropped packets by core network on
1497 * inactive devices, do not use this in drivers
1498 *
1499 * @wireless_handlers: List of functions to handle Wireless Extensions,
1500 * instead of ioctl,
1501 * see <net/iw_handler.h> for details.
1502 * @wireless_data: Instance data managed by the core of wireless extensions
1503 *
1504 * @netdev_ops: Includes several pointers to callbacks,
1505 * if one wants to override the ndo_*() functions
1506 * @ethtool_ops: Management operations
1507 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1508 * discovery handling. Necessary for e.g. 6LoWPAN.
1509 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1510 * of Layer 2 headers.
1511 *
1512 * @flags: Interface flags (a la BSD)
1513 * @priv_flags: Like 'flags' but invisible to userspace,
1514 * see if.h for the definitions
1515 * @gflags: Global flags ( kept as legacy )
1516 * @padded: How much padding added by alloc_netdev()
1517 * @operstate: RFC2863 operstate
1518 * @link_mode: Mapping policy to operstate
1519 * @if_port: Selectable AUI, TP, ...
1520 * @dma: DMA channel
1521 * @mtu: Interface MTU value
1522 * @min_mtu: Interface Minimum MTU value
1523 * @max_mtu: Interface Maximum MTU value
1524 * @type: Interface hardware type
1525 * @hard_header_len: Maximum hardware header length.
1526 * @min_header_len: Minimum hardware header length
1527 *
1528 * @needed_headroom: Extra headroom the hardware may need, but not in all
1529 * cases can this be guaranteed
1530 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1531 * cases can this be guaranteed. Some cases also use
1532 * LL_MAX_HEADER instead to allocate the skb
1533 *
1534 * interface address info:
1535 *
1536 * @perm_addr: Permanent hw address
1537 * @addr_assign_type: Hw address assignment type
1538 * @addr_len: Hardware address length
1539 * @neigh_priv_len: Used in neigh_alloc()
1540 * @dev_id: Used to differentiate devices that share
1541 * the same link layer address
1542 * @dev_port: Used to differentiate devices that share
1543 * the same function
1544 * @addr_list_lock: XXX: need comments on this one
1545 * @uc_promisc: Counter that indicates promiscuous mode
1546 * has been enabled due to the need to listen to
1547 * additional unicast addresses in a device that
1548 * does not implement ndo_set_rx_mode()
1549 * @uc: unicast mac addresses
1550 * @mc: multicast mac addresses
1551 * @dev_addrs: list of device hw addresses
1552 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1553 * @promiscuity: Number of times the NIC is told to work in
1554 * promiscuous mode; if it becomes 0 the NIC will
1555 * exit promiscuous mode
1556 * @allmulti: Counter, enables or disables allmulticast mode
1557 *
1558 * @vlan_info: VLAN info
1559 * @dsa_ptr: dsa specific data
1560 * @tipc_ptr: TIPC specific data
1561 * @atalk_ptr: AppleTalk link
1562 * @ip_ptr: IPv4 specific data
1563 * @dn_ptr: DECnet specific data
1564 * @ip6_ptr: IPv6 specific data
1565 * @ax25_ptr: AX.25 specific data
1566 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1567 *
1568 * @dev_addr: Hw address (before bcast,
1569 * because most packets are unicast)
1570 *
1571 * @_rx: Array of RX queues
1572 * @num_rx_queues: Number of RX queues
1573 * allocated at register_netdev() time
1574 * @real_num_rx_queues: Number of RX queues currently active in device
1575 *
1576 * @rx_handler: handler for received packets
1577 * @rx_handler_data: XXX: need comments on this one
1578 * @miniq_ingress: ingress/clsact qdisc specific data for
1579 * ingress processing
1580 * @ingress_queue: XXX: need comments on this one
1581 * @broadcast: hw bcast address
1582 *
1583 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1584 * indexed by RX queue number. Assigned by driver.
1585 * This must only be set if the ndo_rx_flow_steer
1586 * operation is defined
1587 * @index_hlist: Device index hash chain
1588 *
1589 * @_tx: Array of TX queues
1590 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1591 * @real_num_tx_queues: Number of TX queues currently active in device
1592 * @qdisc: Root qdisc from userspace point of view
1593 * @tx_queue_len: Max frames per queue allowed
1594 * @tx_global_lock: XXX: need comments on this one
1595 *
1596 * @xps_maps: XXX: need comments on this one
1597 * @miniq_egress: clsact qdisc specific data for
1598 * egress processing
1599 * @watchdog_timeo: Represents the timeout that is used by
1600 * the watchdog (see dev_watchdog())
1601 * @watchdog_timer: List of timers
1602 *
1603 * @pcpu_refcnt: Number of references to this device
1604 * @todo_list: Delayed register/unregister
1605 * @link_watch_list: XXX: need comments on this one
1606 *
1607 * @reg_state: Register/unregister state machine
1608 * @dismantle: Device is going to be freed
1609 * @rtnl_link_state: This enum represents the phases of creating
1610 * a new link
1611 *
1612 * @needs_free_netdev: Should unregister perform free_netdev?
1613 * @priv_destructor: Called from unregister
1614 * @npinfo: XXX: need comments on this one
1615 * @nd_net: Network namespace this network device is inside
1616 *
1617 * @ml_priv: Mid-layer private
1618 * @lstats: Loopback statistics
1619 * @tstats: Tunnel statistics
1620 * @dstats: Dummy statistics
1621 * @vstats: Virtual ethernet statistics
1622 *
1623 * @garp_port: GARP
1624 * @mrp_port: MRP
1625 *
1626 * @dev: Class/net/name entry
1627 * @sysfs_groups: Space for optional device, statistics and wireless
1628 * sysfs groups
1629 *
1630 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1631 * @rtnl_link_ops: Rtnl_link_ops
1632 *
1633 * @gso_max_size: Maximum size of generic segmentation offload
1634 * @gso_max_segs: Maximum number of segments that can be passed to the
1635 * NIC for GSO
1636 *
1637 * @dcbnl_ops: Data Center Bridging netlink ops
1638 * @num_tc: Number of traffic classes in the net device
1639 * @tc_to_txq: XXX: need comments on this one
1640 * @prio_tc_map: XXX: need comments on this one
1641 *
1642 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1643 *
1644 * @priomap: XXX: need comments on this one
1645 * @phydev: Physical device may attach itself
1646 * for hardware timestamping
1647 *
1648 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1649 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1650 *
1651 * @proto_down: protocol port state information can be sent to the
1652 * switch driver and used to set the phys state of the
1653 * switch port.
1654 *
1655 * FIXME: cleanup struct net_device such that network protocol info
1656 * moves out.
1657 */
1658
1659 struct net_device {
1660 char name[IFNAMSIZ];
1661 struct hlist_node name_hlist;
1662 struct dev_ifalias __rcu *ifalias;
1663 /*
1664 * I/O specific fields
1665 * FIXME: Merge these and struct ifmap into one
1666 */
1667 unsigned long mem_end;
1668 unsigned long mem_start;
1669 unsigned long base_addr;
1670 int irq;
1671
1672 /*
1673 * Some hardware also needs these fields (state,dev_list,
1674 * napi_list,unreg_list,close_list) but they are not
1675 * part of the usual set specified in Space.c.
1676 */
1677
1678 unsigned long state;
1679
1680 struct list_head dev_list;
1681 struct list_head napi_list;
1682 struct list_head unreg_list;
1683 struct list_head close_list;
1684 struct list_head ptype_all;
1685 struct list_head ptype_specific;
1686
1687 struct {
1688 struct list_head upper;
1689 struct list_head lower;
1690 } adj_list;
1691
1692 netdev_features_t features;
1693 netdev_features_t hw_features;
1694 netdev_features_t wanted_features;
1695 netdev_features_t vlan_features;
1696 netdev_features_t hw_enc_features;
1697 netdev_features_t mpls_features;
1698 netdev_features_t gso_partial_features;
1699
1700 int ifindex;
1701 int group;
1702
1703 struct net_device_stats stats;
1704
1705 atomic_long_t rx_dropped;
1706 atomic_long_t tx_dropped;
1707 atomic_long_t rx_nohandler;
1708
1709 /* Stats to monitor link on/off, flapping */
1710 atomic_t carrier_up_count;
1711 atomic_t carrier_down_count;
1712
1713 #ifdef CONFIG_WIRELESS_EXT
1714 const struct iw_handler_def *wireless_handlers;
1715 struct iw_public_data *wireless_data;
1716 #endif
1717 const struct net_device_ops *netdev_ops;
1718 const struct ethtool_ops *ethtool_ops;
1719 #ifdef CONFIG_NET_SWITCHDEV
1720 const struct switchdev_ops *switchdev_ops;
1721 #endif
1722 #ifdef CONFIG_NET_L3_MASTER_DEV
1723 const struct l3mdev_ops *l3mdev_ops;
1724 #endif
1725 #if IS_ENABLED(CONFIG_IPV6)
1726 const struct ndisc_ops *ndisc_ops;
1727 #endif
1728
1729 #ifdef CONFIG_XFRM
1730 const struct xfrmdev_ops *xfrmdev_ops;
1731 #endif
1732
1733 const struct header_ops *header_ops;
1734
1735 unsigned int flags;
1736 unsigned int priv_flags;
1737
1738 unsigned short gflags;
1739 unsigned short padded;
1740
1741 unsigned char operstate;
1742 unsigned char link_mode;
1743
1744 unsigned char if_port;
1745 unsigned char dma;
1746
1747 unsigned int mtu;
1748 unsigned int min_mtu;
1749 unsigned int max_mtu;
1750 unsigned short type;
1751 unsigned short hard_header_len;
1752 unsigned char min_header_len;
1753
1754 unsigned short needed_headroom;
1755 unsigned short needed_tailroom;
1756
1757 /* Interface address info. */
1758 unsigned char perm_addr[MAX_ADDR_LEN];
1759 unsigned char addr_assign_type;
1760 unsigned char addr_len;
1761 unsigned short neigh_priv_len;
1762 unsigned short dev_id;
1763 unsigned short dev_port;
1764 spinlock_t addr_list_lock;
1765 unsigned char name_assign_type;
1766 bool uc_promisc;
1767 struct netdev_hw_addr_list uc;
1768 struct netdev_hw_addr_list mc;
1769 struct netdev_hw_addr_list dev_addrs;
1770
1771 #ifdef CONFIG_SYSFS
1772 struct kset *queues_kset;
1773 #endif
1774 unsigned int promiscuity;
1775 unsigned int allmulti;
1776
1777
1778 /* Protocol-specific pointers */
1779
1780 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1781 struct vlan_info __rcu *vlan_info;
1782 #endif
1783 #if IS_ENABLED(CONFIG_NET_DSA)
1784 struct dsa_port *dsa_ptr;
1785 #endif
1786 #if IS_ENABLED(CONFIG_TIPC)
1787 struct tipc_bearer __rcu *tipc_ptr;
1788 #endif
1789 void *atalk_ptr;
1790 struct in_device __rcu *ip_ptr;
1791 struct dn_dev __rcu *dn_ptr;
1792 struct inet6_dev __rcu *ip6_ptr;
1793 void *ax25_ptr;
1794 struct wireless_dev *ieee80211_ptr;
1795 struct wpan_dev *ieee802154_ptr;
1796 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1797 struct mpls_dev __rcu *mpls_ptr;
1798 #endif
1799
1800 /*
1801 * Cache lines mostly used on receive path (including eth_type_trans())
1802 */
1803 /* Interface address info used in eth_type_trans() */
1804 unsigned char *dev_addr;
1805
1806 #ifdef CONFIG_SYSFS
1807 struct netdev_rx_queue *_rx;
1808
1809 unsigned int num_rx_queues;
1810 unsigned int real_num_rx_queues;
1811 #endif
1812
1813 struct bpf_prog __rcu *xdp_prog;
1814 unsigned long gro_flush_timeout;
1815 rx_handler_func_t __rcu *rx_handler;
1816 void __rcu *rx_handler_data;
1817
1818 #ifdef CONFIG_NET_CLS_ACT
1819 struct mini_Qdisc __rcu *miniq_ingress;
1820 #endif
1821 struct netdev_queue __rcu *ingress_queue;
1822 #ifdef CONFIG_NETFILTER_INGRESS
1823 struct nf_hook_entries __rcu *nf_hooks_ingress;
1824 #endif
1825
1826 unsigned char broadcast[MAX_ADDR_LEN];
1827 #ifdef CONFIG_RFS_ACCEL
1828 struct cpu_rmap *rx_cpu_rmap;
1829 #endif
1830 struct hlist_node index_hlist;
1831
1832 /*
1833 * Cache lines mostly used on transmit path
1834 */
1835 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1836 unsigned int num_tx_queues;
1837 unsigned int real_num_tx_queues;
1838 struct Qdisc *qdisc;
1839 #ifdef CONFIG_NET_SCHED
1840 DECLARE_HASHTABLE (qdisc_hash, 4);
1841 #endif
1842 unsigned int tx_queue_len;
1843 spinlock_t tx_global_lock;
1844 int watchdog_timeo;
1845
1846 #ifdef CONFIG_XPS
1847 struct xps_dev_maps __rcu *xps_maps;
1848 #endif
1849 #ifdef CONFIG_NET_CLS_ACT
1850 struct mini_Qdisc __rcu *miniq_egress;
1851 #endif
1852
1853 /* These may be needed for future network-power-down code. */
1854 struct timer_list watchdog_timer;
1855
1856 int __percpu *pcpu_refcnt;
1857 struct list_head todo_list;
1858
1859 struct list_head link_watch_list;
1860
1861 enum { NETREG_UNINITIALIZED=0,
1862 NETREG_REGISTERED, /* completed register_netdevice */
1863 NETREG_UNREGISTERING, /* called unregister_netdevice */
1864 NETREG_UNREGISTERED, /* completed unregister todo */
1865 NETREG_RELEASED, /* called free_netdev */
1866 NETREG_DUMMY, /* dummy device for NAPI poll */
1867 } reg_state:8;
1868
1869 bool dismantle;
1870
1871 enum {
1872 RTNL_LINK_INITIALIZED,
1873 RTNL_LINK_INITIALIZING,
1874 } rtnl_link_state:16;
1875
1876 bool needs_free_netdev;
1877 void (*priv_destructor)(struct net_device *dev);
1878
1879 #ifdef CONFIG_NETPOLL
1880 struct netpoll_info __rcu *npinfo;
1881 #endif
1882
1883 possible_net_t nd_net;
1884
1885 /* mid-layer private */
1886 union {
1887 void *ml_priv;
1888 struct pcpu_lstats __percpu *lstats;
1889 struct pcpu_sw_netstats __percpu *tstats;
1890 struct pcpu_dstats __percpu *dstats;
1891 struct pcpu_vstats __percpu *vstats;
1892 };
1893
1894 #if IS_ENABLED(CONFIG_GARP)
1895 struct garp_port __rcu *garp_port;
1896 #endif
1897 #if IS_ENABLED(CONFIG_MRP)
1898 struct mrp_port __rcu *mrp_port;
1899 #endif
1900
1901 struct device dev;
1902 const struct attribute_group *sysfs_groups[4];
1903 const struct attribute_group *sysfs_rx_queue_group;
1904
1905 const struct rtnl_link_ops *rtnl_link_ops;
1906
1907 /* for setting kernel sock attribute on TCP connection setup */
1908 #define GSO_MAX_SIZE 65536
1909 unsigned int gso_max_size;
1910 #define GSO_MAX_SEGS 65535
1911 u16 gso_max_segs;
1912
1913 #ifdef CONFIG_DCB
1914 const struct dcbnl_rtnl_ops *dcbnl_ops;
1915 #endif
1916 u8 num_tc;
1917 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1918 u8 prio_tc_map[TC_BITMASK + 1];
1919
1920 #if IS_ENABLED(CONFIG_FCOE)
1921 unsigned int fcoe_ddp_xid;
1922 #endif
1923 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1924 struct netprio_map __rcu *priomap;
1925 #endif
1926 struct phy_device *phydev;
1927 struct lock_class_key *qdisc_tx_busylock;
1928 struct lock_class_key *qdisc_running_key;
1929 bool proto_down;
1930 };
1931 #define to_net_dev(d) container_of(d, struct net_device, dev)
1932
1933 static inline bool netif_elide_gro(const struct net_device *dev)
1934 {
1935 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1936 return true;
1937 return false;
1938 }
1939
1940 #define NETDEV_ALIGN 32
1941
1942 static inline
1943 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1944 {
1945 return dev->prio_tc_map[prio & TC_BITMASK];
1946 }
1947
1948 static inline
1949 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1950 {
1951 if (tc >= dev->num_tc)
1952 return -EINVAL;
1953
1954 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1955 return 0;
1956 }
1957
1958 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1959 void netdev_reset_tc(struct net_device *dev);
1960 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1961 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1962
1963 static inline
1964 int netdev_get_num_tc(struct net_device *dev)
1965 {
1966 return dev->num_tc;
1967 }
1968
1969 static inline
1970 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1971 unsigned int index)
1972 {
1973 return &dev->_tx[index];
1974 }
1975
1976 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1977 const struct sk_buff *skb)
1978 {
1979 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1980 }
1981
1982 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1983 void (*f)(struct net_device *,
1984 struct netdev_queue *,
1985 void *),
1986 void *arg)
1987 {
1988 unsigned int i;
1989
1990 for (i = 0; i < dev->num_tx_queues; i++)
1991 f(dev, &dev->_tx[i], arg);
1992 }
1993
1994 #define netdev_lockdep_set_classes(dev) \
1995 { \
1996 static struct lock_class_key qdisc_tx_busylock_key; \
1997 static struct lock_class_key qdisc_running_key; \
1998 static struct lock_class_key qdisc_xmit_lock_key; \
1999 static struct lock_class_key dev_addr_list_lock_key; \
2000 unsigned int i; \
2001 \
2002 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2003 (dev)->qdisc_running_key = &qdisc_running_key; \
2004 lockdep_set_class(&(dev)->addr_list_lock, \
2005 &dev_addr_list_lock_key); \
2006 for (i = 0; i < (dev)->num_tx_queues; i++) \
2007 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2008 &qdisc_xmit_lock_key); \
2009 }
2010
2011 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2012 struct sk_buff *skb,
2013 void *accel_priv);
2014
2015 /* returns the headroom that the master device needs to take in account
2016 * when forwarding to this dev
2017 */
2018 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2019 {
2020 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2021 }
2022
2023 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2024 {
2025 if (dev->netdev_ops->ndo_set_rx_headroom)
2026 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2027 }
2028
2029 /* set the device rx headroom to the dev's default */
2030 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2031 {
2032 netdev_set_rx_headroom(dev, -1);
2033 }
2034
2035 /*
2036 * Net namespace inlines
2037 */
2038 static inline
2039 struct net *dev_net(const struct net_device *dev)
2040 {
2041 return read_pnet(&dev->nd_net);
2042 }
2043
2044 static inline
2045 void dev_net_set(struct net_device *dev, struct net *net)
2046 {
2047 write_pnet(&dev->nd_net, net);
2048 }
2049
2050 /**
2051 * netdev_priv - access network device private data
2052 * @dev: network device
2053 *
2054 * Get network device private data
2055 */
2056 static inline void *netdev_priv(const struct net_device *dev)
2057 {
2058 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2059 }
2060
2061 /* Set the sysfs physical device reference for the network logical device
2062 * if set prior to registration will cause a symlink during initialization.
2063 */
2064 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2065
2066 /* Set the sysfs device type for the network logical device to allow
2067 * fine-grained identification of different network device types. For
2068 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2069 */
2070 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2071
2072 /* Default NAPI poll() weight
2073 * Device drivers are strongly advised to not use bigger value
2074 */
2075 #define NAPI_POLL_WEIGHT 64
2076
2077 /**
2078 * netif_napi_add - initialize a NAPI context
2079 * @dev: network device
2080 * @napi: NAPI context
2081 * @poll: polling function
2082 * @weight: default weight
2083 *
2084 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2085 * *any* of the other NAPI-related functions.
2086 */
2087 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2088 int (*poll)(struct napi_struct *, int), int weight);
2089
2090 /**
2091 * netif_tx_napi_add - initialize a NAPI context
2092 * @dev: network device
2093 * @napi: NAPI context
2094 * @poll: polling function
2095 * @weight: default weight
2096 *
2097 * This variant of netif_napi_add() should be used from drivers using NAPI
2098 * to exclusively poll a TX queue.
2099 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2100 */
2101 static inline void netif_tx_napi_add(struct net_device *dev,
2102 struct napi_struct *napi,
2103 int (*poll)(struct napi_struct *, int),
2104 int weight)
2105 {
2106 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2107 netif_napi_add(dev, napi, poll, weight);
2108 }
2109
2110 /**
2111 * netif_napi_del - remove a NAPI context
2112 * @napi: NAPI context
2113 *
2114 * netif_napi_del() removes a NAPI context from the network device NAPI list
2115 */
2116 void netif_napi_del(struct napi_struct *napi);
2117
2118 struct napi_gro_cb {
2119 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2120 void *frag0;
2121
2122 /* Length of frag0. */
2123 unsigned int frag0_len;
2124
2125 /* This indicates where we are processing relative to skb->data. */
2126 int data_offset;
2127
2128 /* This is non-zero if the packet cannot be merged with the new skb. */
2129 u16 flush;
2130
2131 /* Save the IP ID here and check when we get to the transport layer */
2132 u16 flush_id;
2133
2134 /* Number of segments aggregated. */
2135 u16 count;
2136
2137 /* Start offset for remote checksum offload */
2138 u16 gro_remcsum_start;
2139
2140 /* jiffies when first packet was created/queued */
2141 unsigned long age;
2142
2143 /* Used in ipv6_gro_receive() and foo-over-udp */
2144 u16 proto;
2145
2146 /* This is non-zero if the packet may be of the same flow. */
2147 u8 same_flow:1;
2148
2149 /* Used in tunnel GRO receive */
2150 u8 encap_mark:1;
2151
2152 /* GRO checksum is valid */
2153 u8 csum_valid:1;
2154
2155 /* Number of checksums via CHECKSUM_UNNECESSARY */
2156 u8 csum_cnt:3;
2157
2158 /* Free the skb? */
2159 u8 free:2;
2160 #define NAPI_GRO_FREE 1
2161 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2162
2163 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2164 u8 is_ipv6:1;
2165
2166 /* Used in GRE, set in fou/gue_gro_receive */
2167 u8 is_fou:1;
2168
2169 /* Used to determine if flush_id can be ignored */
2170 u8 is_atomic:1;
2171
2172 /* Number of gro_receive callbacks this packet already went through */
2173 u8 recursion_counter:4;
2174
2175 /* 1 bit hole */
2176
2177 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2178 __wsum csum;
2179
2180 /* used in skb_gro_receive() slow path */
2181 struct sk_buff *last;
2182 };
2183
2184 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2185
2186 #define GRO_RECURSION_LIMIT 15
2187 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2188 {
2189 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2190 }
2191
2192 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2193 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2194 struct sk_buff **head,
2195 struct sk_buff *skb)
2196 {
2197 if (unlikely(gro_recursion_inc_test(skb))) {
2198 NAPI_GRO_CB(skb)->flush |= 1;
2199 return NULL;
2200 }
2201
2202 return cb(head, skb);
2203 }
2204
2205 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2206 struct sk_buff *);
2207 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2208 struct sock *sk,
2209 struct sk_buff **head,
2210 struct sk_buff *skb)
2211 {
2212 if (unlikely(gro_recursion_inc_test(skb))) {
2213 NAPI_GRO_CB(skb)->flush |= 1;
2214 return NULL;
2215 }
2216
2217 return cb(sk, head, skb);
2218 }
2219
2220 struct packet_type {
2221 __be16 type; /* This is really htons(ether_type). */
2222 struct net_device *dev; /* NULL is wildcarded here */
2223 int (*func) (struct sk_buff *,
2224 struct net_device *,
2225 struct packet_type *,
2226 struct net_device *);
2227 bool (*id_match)(struct packet_type *ptype,
2228 struct sock *sk);
2229 void *af_packet_priv;
2230 struct list_head list;
2231 };
2232
2233 struct offload_callbacks {
2234 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2235 netdev_features_t features);
2236 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2237 struct sk_buff *skb);
2238 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2239 };
2240
2241 struct packet_offload {
2242 __be16 type; /* This is really htons(ether_type). */
2243 u16 priority;
2244 struct offload_callbacks callbacks;
2245 struct list_head list;
2246 };
2247
2248 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2249 struct pcpu_sw_netstats {
2250 u64 rx_packets;
2251 u64 rx_bytes;
2252 u64 tx_packets;
2253 u64 tx_bytes;
2254 struct u64_stats_sync syncp;
2255 };
2256
2257 #define __netdev_alloc_pcpu_stats(type, gfp) \
2258 ({ \
2259 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2260 if (pcpu_stats) { \
2261 int __cpu; \
2262 for_each_possible_cpu(__cpu) { \
2263 typeof(type) *stat; \
2264 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2265 u64_stats_init(&stat->syncp); \
2266 } \
2267 } \
2268 pcpu_stats; \
2269 })
2270
2271 #define netdev_alloc_pcpu_stats(type) \
2272 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2273
2274 enum netdev_lag_tx_type {
2275 NETDEV_LAG_TX_TYPE_UNKNOWN,
2276 NETDEV_LAG_TX_TYPE_RANDOM,
2277 NETDEV_LAG_TX_TYPE_BROADCAST,
2278 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2279 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2280 NETDEV_LAG_TX_TYPE_HASH,
2281 };
2282
2283 struct netdev_lag_upper_info {
2284 enum netdev_lag_tx_type tx_type;
2285 };
2286
2287 struct netdev_lag_lower_state_info {
2288 u8 link_up : 1,
2289 tx_enabled : 1;
2290 };
2291
2292 #include <linux/notifier.h>
2293
2294 /* netdevice notifier chain. Please remember to update the rtnetlink
2295 * notification exclusion list in rtnetlink_event() when adding new
2296 * types.
2297 */
2298 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2299 #define NETDEV_DOWN 0x0002
2300 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2301 detected a hardware crash and restarted
2302 - we can use this eg to kick tcp sessions
2303 once done */
2304 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2305 #define NETDEV_REGISTER 0x0005
2306 #define NETDEV_UNREGISTER 0x0006
2307 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2308 #define NETDEV_CHANGEADDR 0x0008
2309 #define NETDEV_GOING_DOWN 0x0009
2310 #define NETDEV_CHANGENAME 0x000A
2311 #define NETDEV_FEAT_CHANGE 0x000B
2312 #define NETDEV_BONDING_FAILOVER 0x000C
2313 #define NETDEV_PRE_UP 0x000D
2314 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2315 #define NETDEV_POST_TYPE_CHANGE 0x000F
2316 #define NETDEV_POST_INIT 0x0010
2317 #define NETDEV_UNREGISTER_FINAL 0x0011
2318 #define NETDEV_RELEASE 0x0012
2319 #define NETDEV_NOTIFY_PEERS 0x0013
2320 #define NETDEV_JOIN 0x0014
2321 #define NETDEV_CHANGEUPPER 0x0015
2322 #define NETDEV_RESEND_IGMP 0x0016
2323 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2324 #define NETDEV_CHANGEINFODATA 0x0018
2325 #define NETDEV_BONDING_INFO 0x0019
2326 #define NETDEV_PRECHANGEUPPER 0x001A
2327 #define NETDEV_CHANGELOWERSTATE 0x001B
2328 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2329 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2330 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2331
2332 int register_netdevice_notifier(struct notifier_block *nb);
2333 int unregister_netdevice_notifier(struct notifier_block *nb);
2334
2335 struct netdev_notifier_info {
2336 struct net_device *dev;
2337 struct netlink_ext_ack *extack;
2338 };
2339
2340 struct netdev_notifier_change_info {
2341 struct netdev_notifier_info info; /* must be first */
2342 unsigned int flags_changed;
2343 };
2344
2345 struct netdev_notifier_changeupper_info {
2346 struct netdev_notifier_info info; /* must be first */
2347 struct net_device *upper_dev; /* new upper dev */
2348 bool master; /* is upper dev master */
2349 bool linking; /* is the notification for link or unlink */
2350 void *upper_info; /* upper dev info */
2351 };
2352
2353 struct netdev_notifier_changelowerstate_info {
2354 struct netdev_notifier_info info; /* must be first */
2355 void *lower_state_info; /* is lower dev state */
2356 };
2357
2358 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2359 struct net_device *dev)
2360 {
2361 info->dev = dev;
2362 info->extack = NULL;
2363 }
2364
2365 static inline struct net_device *
2366 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2367 {
2368 return info->dev;
2369 }
2370
2371 static inline struct netlink_ext_ack *
2372 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2373 {
2374 return info->extack;
2375 }
2376
2377 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2378
2379
2380 extern rwlock_t dev_base_lock; /* Device list lock */
2381
2382 #define for_each_netdev(net, d) \
2383 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2384 #define for_each_netdev_reverse(net, d) \
2385 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2386 #define for_each_netdev_rcu(net, d) \
2387 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2388 #define for_each_netdev_safe(net, d, n) \
2389 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2390 #define for_each_netdev_continue(net, d) \
2391 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2392 #define for_each_netdev_continue_rcu(net, d) \
2393 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2394 #define for_each_netdev_in_bond_rcu(bond, slave) \
2395 for_each_netdev_rcu(&init_net, slave) \
2396 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2397 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2398
2399 static inline struct net_device *next_net_device(struct net_device *dev)
2400 {
2401 struct list_head *lh;
2402 struct net *net;
2403
2404 net = dev_net(dev);
2405 lh = dev->dev_list.next;
2406 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2407 }
2408
2409 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2410 {
2411 struct list_head *lh;
2412 struct net *net;
2413
2414 net = dev_net(dev);
2415 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2416 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2417 }
2418
2419 static inline struct net_device *first_net_device(struct net *net)
2420 {
2421 return list_empty(&net->dev_base_head) ? NULL :
2422 net_device_entry(net->dev_base_head.next);
2423 }
2424
2425 static inline struct net_device *first_net_device_rcu(struct net *net)
2426 {
2427 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2428
2429 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2430 }
2431
2432 int netdev_boot_setup_check(struct net_device *dev);
2433 unsigned long netdev_boot_base(const char *prefix, int unit);
2434 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2435 const char *hwaddr);
2436 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2437 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2438 void dev_add_pack(struct packet_type *pt);
2439 void dev_remove_pack(struct packet_type *pt);
2440 void __dev_remove_pack(struct packet_type *pt);
2441 void dev_add_offload(struct packet_offload *po);
2442 void dev_remove_offload(struct packet_offload *po);
2443
2444 int dev_get_iflink(const struct net_device *dev);
2445 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2446 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2447 unsigned short mask);
2448 struct net_device *dev_get_by_name(struct net *net, const char *name);
2449 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2450 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2451 int dev_alloc_name(struct net_device *dev, const char *name);
2452 int dev_open(struct net_device *dev);
2453 void dev_close(struct net_device *dev);
2454 void dev_close_many(struct list_head *head, bool unlink);
2455 void dev_disable_lro(struct net_device *dev);
2456 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2457 int dev_queue_xmit(struct sk_buff *skb);
2458 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2459 int register_netdevice(struct net_device *dev);
2460 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2461 void unregister_netdevice_many(struct list_head *head);
2462 static inline void unregister_netdevice(struct net_device *dev)
2463 {
2464 unregister_netdevice_queue(dev, NULL);
2465 }
2466
2467 int netdev_refcnt_read(const struct net_device *dev);
2468 void free_netdev(struct net_device *dev);
2469 void netdev_freemem(struct net_device *dev);
2470 void synchronize_net(void);
2471 int init_dummy_netdev(struct net_device *dev);
2472
2473 DECLARE_PER_CPU(int, xmit_recursion);
2474 #define XMIT_RECURSION_LIMIT 10
2475
2476 static inline int dev_recursion_level(void)
2477 {
2478 return this_cpu_read(xmit_recursion);
2479 }
2480
2481 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2482 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2483 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2484 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2485 int netdev_get_name(struct net *net, char *name, int ifindex);
2486 int dev_restart(struct net_device *dev);
2487 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2488
2489 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2490 {
2491 return NAPI_GRO_CB(skb)->data_offset;
2492 }
2493
2494 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2495 {
2496 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2497 }
2498
2499 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2500 {
2501 NAPI_GRO_CB(skb)->data_offset += len;
2502 }
2503
2504 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2505 unsigned int offset)
2506 {
2507 return NAPI_GRO_CB(skb)->frag0 + offset;
2508 }
2509
2510 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2511 {
2512 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2513 }
2514
2515 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2516 {
2517 NAPI_GRO_CB(skb)->frag0 = NULL;
2518 NAPI_GRO_CB(skb)->frag0_len = 0;
2519 }
2520
2521 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2522 unsigned int offset)
2523 {
2524 if (!pskb_may_pull(skb, hlen))
2525 return NULL;
2526
2527 skb_gro_frag0_invalidate(skb);
2528 return skb->data + offset;
2529 }
2530
2531 static inline void *skb_gro_network_header(struct sk_buff *skb)
2532 {
2533 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2534 skb_network_offset(skb);
2535 }
2536
2537 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2538 const void *start, unsigned int len)
2539 {
2540 if (NAPI_GRO_CB(skb)->csum_valid)
2541 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2542 csum_partial(start, len, 0));
2543 }
2544
2545 /* GRO checksum functions. These are logical equivalents of the normal
2546 * checksum functions (in skbuff.h) except that they operate on the GRO
2547 * offsets and fields in sk_buff.
2548 */
2549
2550 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2551
2552 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2553 {
2554 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2555 }
2556
2557 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2558 bool zero_okay,
2559 __sum16 check)
2560 {
2561 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2562 skb_checksum_start_offset(skb) <
2563 skb_gro_offset(skb)) &&
2564 !skb_at_gro_remcsum_start(skb) &&
2565 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2566 (!zero_okay || check));
2567 }
2568
2569 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2570 __wsum psum)
2571 {
2572 if (NAPI_GRO_CB(skb)->csum_valid &&
2573 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2574 return 0;
2575
2576 NAPI_GRO_CB(skb)->csum = psum;
2577
2578 return __skb_gro_checksum_complete(skb);
2579 }
2580
2581 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2582 {
2583 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2584 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2585 NAPI_GRO_CB(skb)->csum_cnt--;
2586 } else {
2587 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2588 * verified a new top level checksum or an encapsulated one
2589 * during GRO. This saves work if we fallback to normal path.
2590 */
2591 __skb_incr_checksum_unnecessary(skb);
2592 }
2593 }
2594
2595 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2596 compute_pseudo) \
2597 ({ \
2598 __sum16 __ret = 0; \
2599 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2600 __ret = __skb_gro_checksum_validate_complete(skb, \
2601 compute_pseudo(skb, proto)); \
2602 if (!__ret) \
2603 skb_gro_incr_csum_unnecessary(skb); \
2604 __ret; \
2605 })
2606
2607 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2608 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2609
2610 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2611 compute_pseudo) \
2612 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2613
2614 #define skb_gro_checksum_simple_validate(skb) \
2615 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2616
2617 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2618 {
2619 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2620 !NAPI_GRO_CB(skb)->csum_valid);
2621 }
2622
2623 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2624 __sum16 check, __wsum pseudo)
2625 {
2626 NAPI_GRO_CB(skb)->csum = ~pseudo;
2627 NAPI_GRO_CB(skb)->csum_valid = 1;
2628 }
2629
2630 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2631 do { \
2632 if (__skb_gro_checksum_convert_check(skb)) \
2633 __skb_gro_checksum_convert(skb, check, \
2634 compute_pseudo(skb, proto)); \
2635 } while (0)
2636
2637 struct gro_remcsum {
2638 int offset;
2639 __wsum delta;
2640 };
2641
2642 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2643 {
2644 grc->offset = 0;
2645 grc->delta = 0;
2646 }
2647
2648 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2649 unsigned int off, size_t hdrlen,
2650 int start, int offset,
2651 struct gro_remcsum *grc,
2652 bool nopartial)
2653 {
2654 __wsum delta;
2655 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2656
2657 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2658
2659 if (!nopartial) {
2660 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2661 return ptr;
2662 }
2663
2664 ptr = skb_gro_header_fast(skb, off);
2665 if (skb_gro_header_hard(skb, off + plen)) {
2666 ptr = skb_gro_header_slow(skb, off + plen, off);
2667 if (!ptr)
2668 return NULL;
2669 }
2670
2671 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2672 start, offset);
2673
2674 /* Adjust skb->csum since we changed the packet */
2675 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2676
2677 grc->offset = off + hdrlen + offset;
2678 grc->delta = delta;
2679
2680 return ptr;
2681 }
2682
2683 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2684 struct gro_remcsum *grc)
2685 {
2686 void *ptr;
2687 size_t plen = grc->offset + sizeof(u16);
2688
2689 if (!grc->delta)
2690 return;
2691
2692 ptr = skb_gro_header_fast(skb, grc->offset);
2693 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2694 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2695 if (!ptr)
2696 return;
2697 }
2698
2699 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2700 }
2701
2702 #ifdef CONFIG_XFRM_OFFLOAD
2703 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2704 {
2705 if (PTR_ERR(pp) != -EINPROGRESS)
2706 NAPI_GRO_CB(skb)->flush |= flush;
2707 }
2708 #else
2709 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2710 {
2711 NAPI_GRO_CB(skb)->flush |= flush;
2712 }
2713 #endif
2714
2715 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2716 unsigned short type,
2717 const void *daddr, const void *saddr,
2718 unsigned int len)
2719 {
2720 if (!dev->header_ops || !dev->header_ops->create)
2721 return 0;
2722
2723 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2724 }
2725
2726 static inline int dev_parse_header(const struct sk_buff *skb,
2727 unsigned char *haddr)
2728 {
2729 const struct net_device *dev = skb->dev;
2730
2731 if (!dev->header_ops || !dev->header_ops->parse)
2732 return 0;
2733 return dev->header_ops->parse(skb, haddr);
2734 }
2735
2736 /* ll_header must have at least hard_header_len allocated */
2737 static inline bool dev_validate_header(const struct net_device *dev,
2738 char *ll_header, int len)
2739 {
2740 if (likely(len >= dev->hard_header_len))
2741 return true;
2742 if (len < dev->min_header_len)
2743 return false;
2744
2745 if (capable(CAP_SYS_RAWIO)) {
2746 memset(ll_header + len, 0, dev->hard_header_len - len);
2747 return true;
2748 }
2749
2750 if (dev->header_ops && dev->header_ops->validate)
2751 return dev->header_ops->validate(ll_header, len);
2752
2753 return false;
2754 }
2755
2756 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2757 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2758 static inline int unregister_gifconf(unsigned int family)
2759 {
2760 return register_gifconf(family, NULL);
2761 }
2762
2763 #ifdef CONFIG_NET_FLOW_LIMIT
2764 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2765 struct sd_flow_limit {
2766 u64 count;
2767 unsigned int num_buckets;
2768 unsigned int history_head;
2769 u16 history[FLOW_LIMIT_HISTORY];
2770 u8 buckets[];
2771 };
2772
2773 extern int netdev_flow_limit_table_len;
2774 #endif /* CONFIG_NET_FLOW_LIMIT */
2775
2776 /*
2777 * Incoming packets are placed on per-CPU queues
2778 */
2779 struct softnet_data {
2780 struct list_head poll_list;
2781 struct sk_buff_head process_queue;
2782
2783 /* stats */
2784 unsigned int processed;
2785 unsigned int time_squeeze;
2786 unsigned int received_rps;
2787 #ifdef CONFIG_RPS
2788 struct softnet_data *rps_ipi_list;
2789 #endif
2790 #ifdef CONFIG_NET_FLOW_LIMIT
2791 struct sd_flow_limit __rcu *flow_limit;
2792 #endif
2793 struct Qdisc *output_queue;
2794 struct Qdisc **output_queue_tailp;
2795 struct sk_buff *completion_queue;
2796
2797 #ifdef CONFIG_RPS
2798 /* input_queue_head should be written by cpu owning this struct,
2799 * and only read by other cpus. Worth using a cache line.
2800 */
2801 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2802
2803 /* Elements below can be accessed between CPUs for RPS/RFS */
2804 call_single_data_t csd ____cacheline_aligned_in_smp;
2805 struct softnet_data *rps_ipi_next;
2806 unsigned int cpu;
2807 unsigned int input_queue_tail;
2808 #endif
2809 unsigned int dropped;
2810 struct sk_buff_head input_pkt_queue;
2811 struct napi_struct backlog;
2812
2813 };
2814
2815 static inline void input_queue_head_incr(struct softnet_data *sd)
2816 {
2817 #ifdef CONFIG_RPS
2818 sd->input_queue_head++;
2819 #endif
2820 }
2821
2822 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2823 unsigned int *qtail)
2824 {
2825 #ifdef CONFIG_RPS
2826 *qtail = ++sd->input_queue_tail;
2827 #endif
2828 }
2829
2830 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2831
2832 void __netif_schedule(struct Qdisc *q);
2833 void netif_schedule_queue(struct netdev_queue *txq);
2834
2835 static inline void netif_tx_schedule_all(struct net_device *dev)
2836 {
2837 unsigned int i;
2838
2839 for (i = 0; i < dev->num_tx_queues; i++)
2840 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2841 }
2842
2843 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2844 {
2845 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2846 }
2847
2848 /**
2849 * netif_start_queue - allow transmit
2850 * @dev: network device
2851 *
2852 * Allow upper layers to call the device hard_start_xmit routine.
2853 */
2854 static inline void netif_start_queue(struct net_device *dev)
2855 {
2856 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2857 }
2858
2859 static inline void netif_tx_start_all_queues(struct net_device *dev)
2860 {
2861 unsigned int i;
2862
2863 for (i = 0; i < dev->num_tx_queues; i++) {
2864 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2865 netif_tx_start_queue(txq);
2866 }
2867 }
2868
2869 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2870
2871 /**
2872 * netif_wake_queue - restart transmit
2873 * @dev: network device
2874 *
2875 * Allow upper layers to call the device hard_start_xmit routine.
2876 * Used for flow control when transmit resources are available.
2877 */
2878 static inline void netif_wake_queue(struct net_device *dev)
2879 {
2880 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2881 }
2882
2883 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2884 {
2885 unsigned int i;
2886
2887 for (i = 0; i < dev->num_tx_queues; i++) {
2888 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2889 netif_tx_wake_queue(txq);
2890 }
2891 }
2892
2893 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2894 {
2895 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2896 }
2897
2898 /**
2899 * netif_stop_queue - stop transmitted packets
2900 * @dev: network device
2901 *
2902 * Stop upper layers calling the device hard_start_xmit routine.
2903 * Used for flow control when transmit resources are unavailable.
2904 */
2905 static inline void netif_stop_queue(struct net_device *dev)
2906 {
2907 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2908 }
2909
2910 void netif_tx_stop_all_queues(struct net_device *dev);
2911
2912 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2913 {
2914 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2915 }
2916
2917 /**
2918 * netif_queue_stopped - test if transmit queue is flowblocked
2919 * @dev: network device
2920 *
2921 * Test if transmit queue on device is currently unable to send.
2922 */
2923 static inline bool netif_queue_stopped(const struct net_device *dev)
2924 {
2925 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2926 }
2927
2928 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2929 {
2930 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2931 }
2932
2933 static inline bool
2934 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2935 {
2936 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2937 }
2938
2939 static inline bool
2940 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2941 {
2942 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2943 }
2944
2945 /**
2946 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2947 * @dev_queue: pointer to transmit queue
2948 *
2949 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2950 * to give appropriate hint to the CPU.
2951 */
2952 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2953 {
2954 #ifdef CONFIG_BQL
2955 prefetchw(&dev_queue->dql.num_queued);
2956 #endif
2957 }
2958
2959 /**
2960 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2961 * @dev_queue: pointer to transmit queue
2962 *
2963 * BQL enabled drivers might use this helper in their TX completion path,
2964 * to give appropriate hint to the CPU.
2965 */
2966 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2967 {
2968 #ifdef CONFIG_BQL
2969 prefetchw(&dev_queue->dql.limit);
2970 #endif
2971 }
2972
2973 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2974 unsigned int bytes)
2975 {
2976 #ifdef CONFIG_BQL
2977 dql_queued(&dev_queue->dql, bytes);
2978
2979 if (likely(dql_avail(&dev_queue->dql) >= 0))
2980 return;
2981
2982 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2983
2984 /*
2985 * The XOFF flag must be set before checking the dql_avail below,
2986 * because in netdev_tx_completed_queue we update the dql_completed
2987 * before checking the XOFF flag.
2988 */
2989 smp_mb();
2990
2991 /* check again in case another CPU has just made room avail */
2992 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2993 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2994 #endif
2995 }
2996
2997 /**
2998 * netdev_sent_queue - report the number of bytes queued to hardware
2999 * @dev: network device
3000 * @bytes: number of bytes queued to the hardware device queue
3001 *
3002 * Report the number of bytes queued for sending/completion to the network
3003 * device hardware queue. @bytes should be a good approximation and should
3004 * exactly match netdev_completed_queue() @bytes
3005 */
3006 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3007 {
3008 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3009 }
3010
3011 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3012 unsigned int pkts, unsigned int bytes)
3013 {
3014 #ifdef CONFIG_BQL
3015 if (unlikely(!bytes))
3016 return;
3017
3018 dql_completed(&dev_queue->dql, bytes);
3019
3020 /*
3021 * Without the memory barrier there is a small possiblity that
3022 * netdev_tx_sent_queue will miss the update and cause the queue to
3023 * be stopped forever
3024 */
3025 smp_mb();
3026
3027 if (dql_avail(&dev_queue->dql) < 0)
3028 return;
3029
3030 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3031 netif_schedule_queue(dev_queue);
3032 #endif
3033 }
3034
3035 /**
3036 * netdev_completed_queue - report bytes and packets completed by device
3037 * @dev: network device
3038 * @pkts: actual number of packets sent over the medium
3039 * @bytes: actual number of bytes sent over the medium
3040 *
3041 * Report the number of bytes and packets transmitted by the network device
3042 * hardware queue over the physical medium, @bytes must exactly match the
3043 * @bytes amount passed to netdev_sent_queue()
3044 */
3045 static inline void netdev_completed_queue(struct net_device *dev,
3046 unsigned int pkts, unsigned int bytes)
3047 {
3048 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3049 }
3050
3051 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3052 {
3053 #ifdef CONFIG_BQL
3054 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3055 dql_reset(&q->dql);
3056 #endif
3057 }
3058
3059 /**
3060 * netdev_reset_queue - reset the packets and bytes count of a network device
3061 * @dev_queue: network device
3062 *
3063 * Reset the bytes and packet count of a network device and clear the
3064 * software flow control OFF bit for this network device
3065 */
3066 static inline void netdev_reset_queue(struct net_device *dev_queue)
3067 {
3068 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3069 }
3070
3071 /**
3072 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3073 * @dev: network device
3074 * @queue_index: given tx queue index
3075 *
3076 * Returns 0 if given tx queue index >= number of device tx queues,
3077 * otherwise returns the originally passed tx queue index.
3078 */
3079 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3080 {
3081 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3082 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3083 dev->name, queue_index,
3084 dev->real_num_tx_queues);
3085 return 0;
3086 }
3087
3088 return queue_index;
3089 }
3090
3091 /**
3092 * netif_running - test if up
3093 * @dev: network device
3094 *
3095 * Test if the device has been brought up.
3096 */
3097 static inline bool netif_running(const struct net_device *dev)
3098 {
3099 return test_bit(__LINK_STATE_START, &dev->state);
3100 }
3101
3102 /*
3103 * Routines to manage the subqueues on a device. We only need start,
3104 * stop, and a check if it's stopped. All other device management is
3105 * done at the overall netdevice level.
3106 * Also test the device if we're multiqueue.
3107 */
3108
3109 /**
3110 * netif_start_subqueue - allow sending packets on subqueue
3111 * @dev: network device
3112 * @queue_index: sub queue index
3113 *
3114 * Start individual transmit queue of a device with multiple transmit queues.
3115 */
3116 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3117 {
3118 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3119
3120 netif_tx_start_queue(txq);
3121 }
3122
3123 /**
3124 * netif_stop_subqueue - stop sending packets on subqueue
3125 * @dev: network device
3126 * @queue_index: sub queue index
3127 *
3128 * Stop individual transmit queue of a device with multiple transmit queues.
3129 */
3130 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3131 {
3132 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3133 netif_tx_stop_queue(txq);
3134 }
3135
3136 /**
3137 * netif_subqueue_stopped - test status of subqueue
3138 * @dev: network device
3139 * @queue_index: sub queue index
3140 *
3141 * Check individual transmit queue of a device with multiple transmit queues.
3142 */
3143 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3144 u16 queue_index)
3145 {
3146 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3147
3148 return netif_tx_queue_stopped(txq);
3149 }
3150
3151 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3152 struct sk_buff *skb)
3153 {
3154 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3155 }
3156
3157 /**
3158 * netif_wake_subqueue - allow sending packets on subqueue
3159 * @dev: network device
3160 * @queue_index: sub queue index
3161 *
3162 * Resume individual transmit queue of a device with multiple transmit queues.
3163 */
3164 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3165 {
3166 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3167
3168 netif_tx_wake_queue(txq);
3169 }
3170
3171 #ifdef CONFIG_XPS
3172 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3173 u16 index);
3174 #else
3175 static inline int netif_set_xps_queue(struct net_device *dev,
3176 const struct cpumask *mask,
3177 u16 index)
3178 {
3179 return 0;
3180 }
3181 #endif
3182
3183 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3184 unsigned int num_tx_queues);
3185
3186 /*
3187 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3188 * as a distribution range limit for the returned value.
3189 */
3190 static inline u16 skb_tx_hash(const struct net_device *dev,
3191 struct sk_buff *skb)
3192 {
3193 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3194 }
3195
3196 /**
3197 * netif_is_multiqueue - test if device has multiple transmit queues
3198 * @dev: network device
3199 *
3200 * Check if device has multiple transmit queues
3201 */
3202 static inline bool netif_is_multiqueue(const struct net_device *dev)
3203 {
3204 return dev->num_tx_queues > 1;
3205 }
3206
3207 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3208
3209 #ifdef CONFIG_SYSFS
3210 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3211 #else
3212 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3213 unsigned int rxq)
3214 {
3215 return 0;
3216 }
3217 #endif
3218
3219 #ifdef CONFIG_SYSFS
3220 static inline unsigned int get_netdev_rx_queue_index(
3221 struct netdev_rx_queue *queue)
3222 {
3223 struct net_device *dev = queue->dev;
3224 int index = queue - dev->_rx;
3225
3226 BUG_ON(index >= dev->num_rx_queues);
3227 return index;
3228 }
3229 #endif
3230
3231 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3232 int netif_get_num_default_rss_queues(void);
3233
3234 enum skb_free_reason {
3235 SKB_REASON_CONSUMED,
3236 SKB_REASON_DROPPED,
3237 };
3238
3239 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3240 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3241
3242 /*
3243 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3244 * interrupt context or with hardware interrupts being disabled.
3245 * (in_irq() || irqs_disabled())
3246 *
3247 * We provide four helpers that can be used in following contexts :
3248 *
3249 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3250 * replacing kfree_skb(skb)
3251 *
3252 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3253 * Typically used in place of consume_skb(skb) in TX completion path
3254 *
3255 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3256 * replacing kfree_skb(skb)
3257 *
3258 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3259 * and consumed a packet. Used in place of consume_skb(skb)
3260 */
3261 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3262 {
3263 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3264 }
3265
3266 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3267 {
3268 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3269 }
3270
3271 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3272 {
3273 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3274 }
3275
3276 static inline void dev_consume_skb_any(struct sk_buff *skb)
3277 {
3278 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3279 }
3280
3281 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3282 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3283 int netif_rx(struct sk_buff *skb);
3284 int netif_rx_ni(struct sk_buff *skb);
3285 int netif_receive_skb(struct sk_buff *skb);
3286 int netif_receive_skb_core(struct sk_buff *skb);
3287 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3288 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3289 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3290 gro_result_t napi_gro_frags(struct napi_struct *napi);
3291 struct packet_offload *gro_find_receive_by_type(__be16 type);
3292 struct packet_offload *gro_find_complete_by_type(__be16 type);
3293
3294 static inline void napi_free_frags(struct napi_struct *napi)
3295 {
3296 kfree_skb(napi->skb);
3297 napi->skb = NULL;
3298 }
3299
3300 bool netdev_is_rx_handler_busy(struct net_device *dev);
3301 int netdev_rx_handler_register(struct net_device *dev,
3302 rx_handler_func_t *rx_handler,
3303 void *rx_handler_data);
3304 void netdev_rx_handler_unregister(struct net_device *dev);
3305
3306 bool dev_valid_name(const char *name);
3307 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3308 int dev_ethtool(struct net *net, struct ifreq *);
3309 unsigned int dev_get_flags(const struct net_device *);
3310 int __dev_change_flags(struct net_device *, unsigned int flags);
3311 int dev_change_flags(struct net_device *, unsigned int);
3312 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3313 unsigned int gchanges);
3314 int dev_change_name(struct net_device *, const char *);
3315 int dev_set_alias(struct net_device *, const char *, size_t);
3316 int dev_get_alias(const struct net_device *, char *, size_t);
3317 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3318 int __dev_set_mtu(struct net_device *, int);
3319 int dev_set_mtu(struct net_device *, int);
3320 void dev_set_group(struct net_device *, int);
3321 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3322 int dev_change_carrier(struct net_device *, bool new_carrier);
3323 int dev_get_phys_port_id(struct net_device *dev,
3324 struct netdev_phys_item_id *ppid);
3325 int dev_get_phys_port_name(struct net_device *dev,
3326 char *name, size_t len);
3327 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3328 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3329 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3330 struct netdev_queue *txq, int *ret);
3331
3332 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3333 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3334 int fd, u32 flags);
3335 u8 __dev_xdp_attached(struct net_device *dev, bpf_op_t xdp_op, u32 *prog_id);
3336
3337 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3338 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3339 bool is_skb_forwardable(const struct net_device *dev,
3340 const struct sk_buff *skb);
3341
3342 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3343 struct sk_buff *skb)
3344 {
3345 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3346 unlikely(!is_skb_forwardable(dev, skb))) {
3347 atomic_long_inc(&dev->rx_dropped);
3348 kfree_skb(skb);
3349 return NET_RX_DROP;
3350 }
3351
3352 skb_scrub_packet(skb, true);
3353 skb->priority = 0;
3354 return 0;
3355 }
3356
3357 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3358
3359 extern int netdev_budget;
3360 extern unsigned int netdev_budget_usecs;
3361
3362 /* Called by rtnetlink.c:rtnl_unlock() */
3363 void netdev_run_todo(void);
3364
3365 /**
3366 * dev_put - release reference to device
3367 * @dev: network device
3368 *
3369 * Release reference to device to allow it to be freed.
3370 */
3371 static inline void dev_put(struct net_device *dev)
3372 {
3373 this_cpu_dec(*dev->pcpu_refcnt);
3374 }
3375
3376 /**
3377 * dev_hold - get reference to device
3378 * @dev: network device
3379 *
3380 * Hold reference to device to keep it from being freed.
3381 */
3382 static inline void dev_hold(struct net_device *dev)
3383 {
3384 this_cpu_inc(*dev->pcpu_refcnt);
3385 }
3386
3387 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3388 * and _off may be called from IRQ context, but it is caller
3389 * who is responsible for serialization of these calls.
3390 *
3391 * The name carrier is inappropriate, these functions should really be
3392 * called netif_lowerlayer_*() because they represent the state of any
3393 * kind of lower layer not just hardware media.
3394 */
3395
3396 void linkwatch_init_dev(struct net_device *dev);
3397 void linkwatch_fire_event(struct net_device *dev);
3398 void linkwatch_forget_dev(struct net_device *dev);
3399
3400 /**
3401 * netif_carrier_ok - test if carrier present
3402 * @dev: network device
3403 *
3404 * Check if carrier is present on device
3405 */
3406 static inline bool netif_carrier_ok(const struct net_device *dev)
3407 {
3408 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3409 }
3410
3411 unsigned long dev_trans_start(struct net_device *dev);
3412
3413 void __netdev_watchdog_up(struct net_device *dev);
3414
3415 void netif_carrier_on(struct net_device *dev);
3416
3417 void netif_carrier_off(struct net_device *dev);
3418
3419 /**
3420 * netif_dormant_on - mark device as dormant.
3421 * @dev: network device
3422 *
3423 * Mark device as dormant (as per RFC2863).
3424 *
3425 * The dormant state indicates that the relevant interface is not
3426 * actually in a condition to pass packets (i.e., it is not 'up') but is
3427 * in a "pending" state, waiting for some external event. For "on-
3428 * demand" interfaces, this new state identifies the situation where the
3429 * interface is waiting for events to place it in the up state.
3430 */
3431 static inline void netif_dormant_on(struct net_device *dev)
3432 {
3433 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3434 linkwatch_fire_event(dev);
3435 }
3436
3437 /**
3438 * netif_dormant_off - set device as not dormant.
3439 * @dev: network device
3440 *
3441 * Device is not in dormant state.
3442 */
3443 static inline void netif_dormant_off(struct net_device *dev)
3444 {
3445 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3446 linkwatch_fire_event(dev);
3447 }
3448
3449 /**
3450 * netif_dormant - test if device is dormant
3451 * @dev: network device
3452 *
3453 * Check if device is dormant.
3454 */
3455 static inline bool netif_dormant(const struct net_device *dev)
3456 {
3457 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3458 }
3459
3460
3461 /**
3462 * netif_oper_up - test if device is operational
3463 * @dev: network device
3464 *
3465 * Check if carrier is operational
3466 */
3467 static inline bool netif_oper_up(const struct net_device *dev)
3468 {
3469 return (dev->operstate == IF_OPER_UP ||
3470 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3471 }
3472
3473 /**
3474 * netif_device_present - is device available or removed
3475 * @dev: network device
3476 *
3477 * Check if device has not been removed from system.
3478 */
3479 static inline bool netif_device_present(struct net_device *dev)
3480 {
3481 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3482 }
3483
3484 void netif_device_detach(struct net_device *dev);
3485
3486 void netif_device_attach(struct net_device *dev);
3487
3488 /*
3489 * Network interface message level settings
3490 */
3491
3492 enum {
3493 NETIF_MSG_DRV = 0x0001,
3494 NETIF_MSG_PROBE = 0x0002,
3495 NETIF_MSG_LINK = 0x0004,
3496 NETIF_MSG_TIMER = 0x0008,
3497 NETIF_MSG_IFDOWN = 0x0010,
3498 NETIF_MSG_IFUP = 0x0020,
3499 NETIF_MSG_RX_ERR = 0x0040,
3500 NETIF_MSG_TX_ERR = 0x0080,
3501 NETIF_MSG_TX_QUEUED = 0x0100,
3502 NETIF_MSG_INTR = 0x0200,
3503 NETIF_MSG_TX_DONE = 0x0400,
3504 NETIF_MSG_RX_STATUS = 0x0800,
3505 NETIF_MSG_PKTDATA = 0x1000,
3506 NETIF_MSG_HW = 0x2000,
3507 NETIF_MSG_WOL = 0x4000,
3508 };
3509
3510 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3511 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3512 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3513 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3514 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3515 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3516 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3517 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3518 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3519 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3520 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3521 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3522 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3523 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3524 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3525
3526 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3527 {
3528 /* use default */
3529 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3530 return default_msg_enable_bits;
3531 if (debug_value == 0) /* no output */
3532 return 0;
3533 /* set low N bits */
3534 return (1 << debug_value) - 1;
3535 }
3536
3537 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3538 {
3539 spin_lock(&txq->_xmit_lock);
3540 txq->xmit_lock_owner = cpu;
3541 }
3542
3543 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3544 {
3545 __acquire(&txq->_xmit_lock);
3546 return true;
3547 }
3548
3549 static inline void __netif_tx_release(struct netdev_queue *txq)
3550 {
3551 __release(&txq->_xmit_lock);
3552 }
3553
3554 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3555 {
3556 spin_lock_bh(&txq->_xmit_lock);
3557 txq->xmit_lock_owner = smp_processor_id();
3558 }
3559
3560 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3561 {
3562 bool ok = spin_trylock(&txq->_xmit_lock);
3563 if (likely(ok))
3564 txq->xmit_lock_owner = smp_processor_id();
3565 return ok;
3566 }
3567
3568 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3569 {
3570 txq->xmit_lock_owner = -1;
3571 spin_unlock(&txq->_xmit_lock);
3572 }
3573
3574 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3575 {
3576 txq->xmit_lock_owner = -1;
3577 spin_unlock_bh(&txq->_xmit_lock);
3578 }
3579
3580 static inline void txq_trans_update(struct netdev_queue *txq)
3581 {
3582 if (txq->xmit_lock_owner != -1)
3583 txq->trans_start = jiffies;
3584 }
3585
3586 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3587 static inline void netif_trans_update(struct net_device *dev)
3588 {
3589 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3590
3591 if (txq->trans_start != jiffies)
3592 txq->trans_start = jiffies;
3593 }
3594
3595 /**
3596 * netif_tx_lock - grab network device transmit lock
3597 * @dev: network device
3598 *
3599 * Get network device transmit lock
3600 */
3601 static inline void netif_tx_lock(struct net_device *dev)
3602 {
3603 unsigned int i;
3604 int cpu;
3605
3606 spin_lock(&dev->tx_global_lock);
3607 cpu = smp_processor_id();
3608 for (i = 0; i < dev->num_tx_queues; i++) {
3609 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3610
3611 /* We are the only thread of execution doing a
3612 * freeze, but we have to grab the _xmit_lock in
3613 * order to synchronize with threads which are in
3614 * the ->hard_start_xmit() handler and already
3615 * checked the frozen bit.
3616 */
3617 __netif_tx_lock(txq, cpu);
3618 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3619 __netif_tx_unlock(txq);
3620 }
3621 }
3622
3623 static inline void netif_tx_lock_bh(struct net_device *dev)
3624 {
3625 local_bh_disable();
3626 netif_tx_lock(dev);
3627 }
3628
3629 static inline void netif_tx_unlock(struct net_device *dev)
3630 {
3631 unsigned int i;
3632
3633 for (i = 0; i < dev->num_tx_queues; i++) {
3634 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3635
3636 /* No need to grab the _xmit_lock here. If the
3637 * queue is not stopped for another reason, we
3638 * force a schedule.
3639 */
3640 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3641 netif_schedule_queue(txq);
3642 }
3643 spin_unlock(&dev->tx_global_lock);
3644 }
3645
3646 static inline void netif_tx_unlock_bh(struct net_device *dev)
3647 {
3648 netif_tx_unlock(dev);
3649 local_bh_enable();
3650 }
3651
3652 #define HARD_TX_LOCK(dev, txq, cpu) { \
3653 if ((dev->features & NETIF_F_LLTX) == 0) { \
3654 __netif_tx_lock(txq, cpu); \
3655 } else { \
3656 __netif_tx_acquire(txq); \
3657 } \
3658 }
3659
3660 #define HARD_TX_TRYLOCK(dev, txq) \
3661 (((dev->features & NETIF_F_LLTX) == 0) ? \
3662 __netif_tx_trylock(txq) : \
3663 __netif_tx_acquire(txq))
3664
3665 #define HARD_TX_UNLOCK(dev, txq) { \
3666 if ((dev->features & NETIF_F_LLTX) == 0) { \
3667 __netif_tx_unlock(txq); \
3668 } else { \
3669 __netif_tx_release(txq); \
3670 } \
3671 }
3672
3673 static inline void netif_tx_disable(struct net_device *dev)
3674 {
3675 unsigned int i;
3676 int cpu;
3677
3678 local_bh_disable();
3679 cpu = smp_processor_id();
3680 for (i = 0; i < dev->num_tx_queues; i++) {
3681 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3682
3683 __netif_tx_lock(txq, cpu);
3684 netif_tx_stop_queue(txq);
3685 __netif_tx_unlock(txq);
3686 }
3687 local_bh_enable();
3688 }
3689
3690 static inline void netif_addr_lock(struct net_device *dev)
3691 {
3692 spin_lock(&dev->addr_list_lock);
3693 }
3694
3695 static inline void netif_addr_lock_nested(struct net_device *dev)
3696 {
3697 int subclass = SINGLE_DEPTH_NESTING;
3698
3699 if (dev->netdev_ops->ndo_get_lock_subclass)
3700 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3701
3702 spin_lock_nested(&dev->addr_list_lock, subclass);
3703 }
3704
3705 static inline void netif_addr_lock_bh(struct net_device *dev)
3706 {
3707 spin_lock_bh(&dev->addr_list_lock);
3708 }
3709
3710 static inline void netif_addr_unlock(struct net_device *dev)
3711 {
3712 spin_unlock(&dev->addr_list_lock);
3713 }
3714
3715 static inline void netif_addr_unlock_bh(struct net_device *dev)
3716 {
3717 spin_unlock_bh(&dev->addr_list_lock);
3718 }
3719
3720 /*
3721 * dev_addrs walker. Should be used only for read access. Call with
3722 * rcu_read_lock held.
3723 */
3724 #define for_each_dev_addr(dev, ha) \
3725 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3726
3727 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3728
3729 void ether_setup(struct net_device *dev);
3730
3731 /* Support for loadable net-drivers */
3732 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3733 unsigned char name_assign_type,
3734 void (*setup)(struct net_device *),
3735 unsigned int txqs, unsigned int rxqs);
3736 int dev_get_valid_name(struct net *net, struct net_device *dev,
3737 const char *name);
3738
3739 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3740 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3741
3742 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3743 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3744 count)
3745
3746 int register_netdev(struct net_device *dev);
3747 void unregister_netdev(struct net_device *dev);
3748
3749 /* General hardware address lists handling functions */
3750 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3751 struct netdev_hw_addr_list *from_list, int addr_len);
3752 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3753 struct netdev_hw_addr_list *from_list, int addr_len);
3754 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3755 struct net_device *dev,
3756 int (*sync)(struct net_device *, const unsigned char *),
3757 int (*unsync)(struct net_device *,
3758 const unsigned char *));
3759 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3760 struct net_device *dev,
3761 int (*unsync)(struct net_device *,
3762 const unsigned char *));
3763 void __hw_addr_init(struct netdev_hw_addr_list *list);
3764
3765 /* Functions used for device addresses handling */
3766 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3767 unsigned char addr_type);
3768 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3769 unsigned char addr_type);
3770 void dev_addr_flush(struct net_device *dev);
3771 int dev_addr_init(struct net_device *dev);
3772
3773 /* Functions used for unicast addresses handling */
3774 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3775 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3776 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3777 int dev_uc_sync(struct net_device *to, struct net_device *from);
3778 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3779 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3780 void dev_uc_flush(struct net_device *dev);
3781 void dev_uc_init(struct net_device *dev);
3782
3783 /**
3784 * __dev_uc_sync - Synchonize device's unicast list
3785 * @dev: device to sync
3786 * @sync: function to call if address should be added
3787 * @unsync: function to call if address should be removed
3788 *
3789 * Add newly added addresses to the interface, and release
3790 * addresses that have been deleted.
3791 */
3792 static inline int __dev_uc_sync(struct net_device *dev,
3793 int (*sync)(struct net_device *,
3794 const unsigned char *),
3795 int (*unsync)(struct net_device *,
3796 const unsigned char *))
3797 {
3798 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3799 }
3800
3801 /**
3802 * __dev_uc_unsync - Remove synchronized addresses from device
3803 * @dev: device to sync
3804 * @unsync: function to call if address should be removed
3805 *
3806 * Remove all addresses that were added to the device by dev_uc_sync().
3807 */
3808 static inline void __dev_uc_unsync(struct net_device *dev,
3809 int (*unsync)(struct net_device *,
3810 const unsigned char *))
3811 {
3812 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3813 }
3814
3815 /* Functions used for multicast addresses handling */
3816 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3817 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3818 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3819 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3820 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3821 int dev_mc_sync(struct net_device *to, struct net_device *from);
3822 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3823 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3824 void dev_mc_flush(struct net_device *dev);
3825 void dev_mc_init(struct net_device *dev);
3826
3827 /**
3828 * __dev_mc_sync - Synchonize device's multicast list
3829 * @dev: device to sync
3830 * @sync: function to call if address should be added
3831 * @unsync: function to call if address should be removed
3832 *
3833 * Add newly added addresses to the interface, and release
3834 * addresses that have been deleted.
3835 */
3836 static inline int __dev_mc_sync(struct net_device *dev,
3837 int (*sync)(struct net_device *,
3838 const unsigned char *),
3839 int (*unsync)(struct net_device *,
3840 const unsigned char *))
3841 {
3842 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3843 }
3844
3845 /**
3846 * __dev_mc_unsync - Remove synchronized addresses from device
3847 * @dev: device to sync
3848 * @unsync: function to call if address should be removed
3849 *
3850 * Remove all addresses that were added to the device by dev_mc_sync().
3851 */
3852 static inline void __dev_mc_unsync(struct net_device *dev,
3853 int (*unsync)(struct net_device *,
3854 const unsigned char *))
3855 {
3856 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3857 }
3858
3859 /* Functions used for secondary unicast and multicast support */
3860 void dev_set_rx_mode(struct net_device *dev);
3861 void __dev_set_rx_mode(struct net_device *dev);
3862 int dev_set_promiscuity(struct net_device *dev, int inc);
3863 int dev_set_allmulti(struct net_device *dev, int inc);
3864 void netdev_state_change(struct net_device *dev);
3865 void netdev_notify_peers(struct net_device *dev);
3866 void netdev_features_change(struct net_device *dev);
3867 /* Load a device via the kmod */
3868 void dev_load(struct net *net, const char *name);
3869 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3870 struct rtnl_link_stats64 *storage);
3871 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3872 const struct net_device_stats *netdev_stats);
3873
3874 extern int netdev_max_backlog;
3875 extern int netdev_tstamp_prequeue;
3876 extern int weight_p;
3877 extern int dev_weight_rx_bias;
3878 extern int dev_weight_tx_bias;
3879 extern int dev_rx_weight;
3880 extern int dev_tx_weight;
3881
3882 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3883 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3884 struct list_head **iter);
3885 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3886 struct list_head **iter);
3887
3888 /* iterate through upper list, must be called under RCU read lock */
3889 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3890 for (iter = &(dev)->adj_list.upper, \
3891 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3892 updev; \
3893 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3894
3895 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3896 int (*fn)(struct net_device *upper_dev,
3897 void *data),
3898 void *data);
3899
3900 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3901 struct net_device *upper_dev);
3902
3903 bool netdev_has_any_upper_dev(struct net_device *dev);
3904
3905 void *netdev_lower_get_next_private(struct net_device *dev,
3906 struct list_head **iter);
3907 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3908 struct list_head **iter);
3909
3910 #define netdev_for_each_lower_private(dev, priv, iter) \
3911 for (iter = (dev)->adj_list.lower.next, \
3912 priv = netdev_lower_get_next_private(dev, &(iter)); \
3913 priv; \
3914 priv = netdev_lower_get_next_private(dev, &(iter)))
3915
3916 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3917 for (iter = &(dev)->adj_list.lower, \
3918 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3919 priv; \
3920 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3921
3922 void *netdev_lower_get_next(struct net_device *dev,
3923 struct list_head **iter);
3924
3925 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3926 for (iter = (dev)->adj_list.lower.next, \
3927 ldev = netdev_lower_get_next(dev, &(iter)); \
3928 ldev; \
3929 ldev = netdev_lower_get_next(dev, &(iter)))
3930
3931 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3932 struct list_head **iter);
3933 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3934 struct list_head **iter);
3935
3936 int netdev_walk_all_lower_dev(struct net_device *dev,
3937 int (*fn)(struct net_device *lower_dev,
3938 void *data),
3939 void *data);
3940 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3941 int (*fn)(struct net_device *lower_dev,
3942 void *data),
3943 void *data);
3944
3945 void *netdev_adjacent_get_private(struct list_head *adj_list);
3946 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3947 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3948 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3949 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
3950 struct netlink_ext_ack *extack);
3951 int netdev_master_upper_dev_link(struct net_device *dev,
3952 struct net_device *upper_dev,
3953 void *upper_priv, void *upper_info,
3954 struct netlink_ext_ack *extack);
3955 void netdev_upper_dev_unlink(struct net_device *dev,
3956 struct net_device *upper_dev);
3957 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3958 void *netdev_lower_dev_get_private(struct net_device *dev,
3959 struct net_device *lower_dev);
3960 void netdev_lower_state_changed(struct net_device *lower_dev,
3961 void *lower_state_info);
3962
3963 /* RSS keys are 40 or 52 bytes long */
3964 #define NETDEV_RSS_KEY_LEN 52
3965 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3966 void netdev_rss_key_fill(void *buffer, size_t len);
3967
3968 int dev_get_nest_level(struct net_device *dev);
3969 int skb_checksum_help(struct sk_buff *skb);
3970 int skb_crc32c_csum_help(struct sk_buff *skb);
3971 int skb_csum_hwoffload_help(struct sk_buff *skb,
3972 const netdev_features_t features);
3973
3974 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3975 netdev_features_t features, bool tx_path);
3976 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3977 netdev_features_t features);
3978
3979 struct netdev_bonding_info {
3980 ifslave slave;
3981 ifbond master;
3982 };
3983
3984 struct netdev_notifier_bonding_info {
3985 struct netdev_notifier_info info; /* must be first */
3986 struct netdev_bonding_info bonding_info;
3987 };
3988
3989 void netdev_bonding_info_change(struct net_device *dev,
3990 struct netdev_bonding_info *bonding_info);
3991
3992 static inline
3993 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3994 {
3995 return __skb_gso_segment(skb, features, true);
3996 }
3997 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3998
3999 static inline bool can_checksum_protocol(netdev_features_t features,
4000 __be16 protocol)
4001 {
4002 if (protocol == htons(ETH_P_FCOE))
4003 return !!(features & NETIF_F_FCOE_CRC);
4004
4005 /* Assume this is an IP checksum (not SCTP CRC) */
4006
4007 if (features & NETIF_F_HW_CSUM) {
4008 /* Can checksum everything */
4009 return true;
4010 }
4011
4012 switch (protocol) {
4013 case htons(ETH_P_IP):
4014 return !!(features & NETIF_F_IP_CSUM);
4015 case htons(ETH_P_IPV6):
4016 return !!(features & NETIF_F_IPV6_CSUM);
4017 default:
4018 return false;
4019 }
4020 }
4021
4022 #ifdef CONFIG_BUG
4023 void netdev_rx_csum_fault(struct net_device *dev);
4024 #else
4025 static inline void netdev_rx_csum_fault(struct net_device *dev)
4026 {
4027 }
4028 #endif
4029 /* rx skb timestamps */
4030 void net_enable_timestamp(void);
4031 void net_disable_timestamp(void);
4032
4033 #ifdef CONFIG_PROC_FS
4034 int __init dev_proc_init(void);
4035 #else
4036 #define dev_proc_init() 0
4037 #endif
4038
4039 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4040 struct sk_buff *skb, struct net_device *dev,
4041 bool more)
4042 {
4043 skb->xmit_more = more ? 1 : 0;
4044 return ops->ndo_start_xmit(skb, dev);
4045 }
4046
4047 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4048 struct netdev_queue *txq, bool more)
4049 {
4050 const struct net_device_ops *ops = dev->netdev_ops;
4051 int rc;
4052
4053 rc = __netdev_start_xmit(ops, skb, dev, more);
4054 if (rc == NETDEV_TX_OK)
4055 txq_trans_update(txq);
4056
4057 return rc;
4058 }
4059
4060 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4061 const void *ns);
4062 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4063 const void *ns);
4064
4065 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4066 {
4067 return netdev_class_create_file_ns(class_attr, NULL);
4068 }
4069
4070 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4071 {
4072 netdev_class_remove_file_ns(class_attr, NULL);
4073 }
4074
4075 extern const struct kobj_ns_type_operations net_ns_type_operations;
4076
4077 const char *netdev_drivername(const struct net_device *dev);
4078
4079 void linkwatch_run_queue(void);
4080
4081 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4082 netdev_features_t f2)
4083 {
4084 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4085 if (f1 & NETIF_F_HW_CSUM)
4086 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4087 else
4088 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4089 }
4090
4091 return f1 & f2;
4092 }
4093
4094 static inline netdev_features_t netdev_get_wanted_features(
4095 struct net_device *dev)
4096 {
4097 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4098 }
4099 netdev_features_t netdev_increment_features(netdev_features_t all,
4100 netdev_features_t one, netdev_features_t mask);
4101
4102 /* Allow TSO being used on stacked device :
4103 * Performing the GSO segmentation before last device
4104 * is a performance improvement.
4105 */
4106 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4107 netdev_features_t mask)
4108 {
4109 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4110 }
4111
4112 int __netdev_update_features(struct net_device *dev);
4113 void netdev_update_features(struct net_device *dev);
4114 void netdev_change_features(struct net_device *dev);
4115
4116 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4117 struct net_device *dev);
4118
4119 netdev_features_t passthru_features_check(struct sk_buff *skb,
4120 struct net_device *dev,
4121 netdev_features_t features);
4122 netdev_features_t netif_skb_features(struct sk_buff *skb);
4123
4124 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4125 {
4126 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4127
4128 /* check flags correspondence */
4129 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4130 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4131 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4132 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4133 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4134 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4135 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4136 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4137 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4138 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4139 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4140 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4141 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4142 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4143 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4144 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4145 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4146
4147 return (features & feature) == feature;
4148 }
4149
4150 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4151 {
4152 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4153 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4154 }
4155
4156 static inline bool netif_needs_gso(struct sk_buff *skb,
4157 netdev_features_t features)
4158 {
4159 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4160 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4161 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4162 }
4163
4164 static inline void netif_set_gso_max_size(struct net_device *dev,
4165 unsigned int size)
4166 {
4167 dev->gso_max_size = size;
4168 }
4169
4170 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4171 int pulled_hlen, u16 mac_offset,
4172 int mac_len)
4173 {
4174 skb->protocol = protocol;
4175 skb->encapsulation = 1;
4176 skb_push(skb, pulled_hlen);
4177 skb_reset_transport_header(skb);
4178 skb->mac_header = mac_offset;
4179 skb->network_header = skb->mac_header + mac_len;
4180 skb->mac_len = mac_len;
4181 }
4182
4183 static inline bool netif_is_macsec(const struct net_device *dev)
4184 {
4185 return dev->priv_flags & IFF_MACSEC;
4186 }
4187
4188 static inline bool netif_is_macvlan(const struct net_device *dev)
4189 {
4190 return dev->priv_flags & IFF_MACVLAN;
4191 }
4192
4193 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4194 {
4195 return dev->priv_flags & IFF_MACVLAN_PORT;
4196 }
4197
4198 static inline bool netif_is_ipvlan(const struct net_device *dev)
4199 {
4200 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4201 }
4202
4203 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4204 {
4205 return dev->priv_flags & IFF_IPVLAN_MASTER;
4206 }
4207
4208 static inline bool netif_is_bond_master(const struct net_device *dev)
4209 {
4210 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4211 }
4212
4213 static inline bool netif_is_bond_slave(const struct net_device *dev)
4214 {
4215 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4216 }
4217
4218 static inline bool netif_supports_nofcs(struct net_device *dev)
4219 {
4220 return dev->priv_flags & IFF_SUPP_NOFCS;
4221 }
4222
4223 static inline bool netif_is_l3_master(const struct net_device *dev)
4224 {
4225 return dev->priv_flags & IFF_L3MDEV_MASTER;
4226 }
4227
4228 static inline bool netif_is_l3_slave(const struct net_device *dev)
4229 {
4230 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4231 }
4232
4233 static inline bool netif_is_bridge_master(const struct net_device *dev)
4234 {
4235 return dev->priv_flags & IFF_EBRIDGE;
4236 }
4237
4238 static inline bool netif_is_bridge_port(const struct net_device *dev)
4239 {
4240 return dev->priv_flags & IFF_BRIDGE_PORT;
4241 }
4242
4243 static inline bool netif_is_ovs_master(const struct net_device *dev)
4244 {
4245 return dev->priv_flags & IFF_OPENVSWITCH;
4246 }
4247
4248 static inline bool netif_is_ovs_port(const struct net_device *dev)
4249 {
4250 return dev->priv_flags & IFF_OVS_DATAPATH;
4251 }
4252
4253 static inline bool netif_is_team_master(const struct net_device *dev)
4254 {
4255 return dev->priv_flags & IFF_TEAM;
4256 }
4257
4258 static inline bool netif_is_team_port(const struct net_device *dev)
4259 {
4260 return dev->priv_flags & IFF_TEAM_PORT;
4261 }
4262
4263 static inline bool netif_is_lag_master(const struct net_device *dev)
4264 {
4265 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4266 }
4267
4268 static inline bool netif_is_lag_port(const struct net_device *dev)
4269 {
4270 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4271 }
4272
4273 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4274 {
4275 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4276 }
4277
4278 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4279 static inline void netif_keep_dst(struct net_device *dev)
4280 {
4281 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4282 }
4283
4284 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4285 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4286 {
4287 /* TODO: reserve and use an additional IFF bit, if we get more users */
4288 return dev->priv_flags & IFF_MACSEC;
4289 }
4290
4291 extern struct pernet_operations __net_initdata loopback_net_ops;
4292
4293 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4294
4295 /* netdev_printk helpers, similar to dev_printk */
4296
4297 static inline const char *netdev_name(const struct net_device *dev)
4298 {
4299 if (!dev->name[0] || strchr(dev->name, '%'))
4300 return "(unnamed net_device)";
4301 return dev->name;
4302 }
4303
4304 static inline bool netdev_unregistering(const struct net_device *dev)
4305 {
4306 return dev->reg_state == NETREG_UNREGISTERING;
4307 }
4308
4309 static inline const char *netdev_reg_state(const struct net_device *dev)
4310 {
4311 switch (dev->reg_state) {
4312 case NETREG_UNINITIALIZED: return " (uninitialized)";
4313 case NETREG_REGISTERED: return "";
4314 case NETREG_UNREGISTERING: return " (unregistering)";
4315 case NETREG_UNREGISTERED: return " (unregistered)";
4316 case NETREG_RELEASED: return " (released)";
4317 case NETREG_DUMMY: return " (dummy)";
4318 }
4319
4320 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4321 return " (unknown)";
4322 }
4323
4324 __printf(3, 4)
4325 void netdev_printk(const char *level, const struct net_device *dev,
4326 const char *format, ...);
4327 __printf(2, 3)
4328 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4329 __printf(2, 3)
4330 void netdev_alert(const struct net_device *dev, const char *format, ...);
4331 __printf(2, 3)
4332 void netdev_crit(const struct net_device *dev, const char *format, ...);
4333 __printf(2, 3)
4334 void netdev_err(const struct net_device *dev, const char *format, ...);
4335 __printf(2, 3)
4336 void netdev_warn(const struct net_device *dev, const char *format, ...);
4337 __printf(2, 3)
4338 void netdev_notice(const struct net_device *dev, const char *format, ...);
4339 __printf(2, 3)
4340 void netdev_info(const struct net_device *dev, const char *format, ...);
4341
4342 #define netdev_level_once(level, dev, fmt, ...) \
4343 do { \
4344 static bool __print_once __read_mostly; \
4345 \
4346 if (!__print_once) { \
4347 __print_once = true; \
4348 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4349 } \
4350 } while (0)
4351
4352 #define netdev_emerg_once(dev, fmt, ...) \
4353 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4354 #define netdev_alert_once(dev, fmt, ...) \
4355 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4356 #define netdev_crit_once(dev, fmt, ...) \
4357 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4358 #define netdev_err_once(dev, fmt, ...) \
4359 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4360 #define netdev_warn_once(dev, fmt, ...) \
4361 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4362 #define netdev_notice_once(dev, fmt, ...) \
4363 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4364 #define netdev_info_once(dev, fmt, ...) \
4365 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4366
4367 #define MODULE_ALIAS_NETDEV(device) \
4368 MODULE_ALIAS("netdev-" device)
4369
4370 #if defined(CONFIG_DYNAMIC_DEBUG)
4371 #define netdev_dbg(__dev, format, args...) \
4372 do { \
4373 dynamic_netdev_dbg(__dev, format, ##args); \
4374 } while (0)
4375 #elif defined(DEBUG)
4376 #define netdev_dbg(__dev, format, args...) \
4377 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4378 #else
4379 #define netdev_dbg(__dev, format, args...) \
4380 ({ \
4381 if (0) \
4382 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4383 })
4384 #endif
4385
4386 #if defined(VERBOSE_DEBUG)
4387 #define netdev_vdbg netdev_dbg
4388 #else
4389
4390 #define netdev_vdbg(dev, format, args...) \
4391 ({ \
4392 if (0) \
4393 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4394 0; \
4395 })
4396 #endif
4397
4398 /*
4399 * netdev_WARN() acts like dev_printk(), but with the key difference
4400 * of using a WARN/WARN_ON to get the message out, including the
4401 * file/line information and a backtrace.
4402 */
4403 #define netdev_WARN(dev, format, args...) \
4404 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4405 netdev_reg_state(dev), ##args)
4406
4407 #define netdev_WARN_ONCE(dev, format, args...) \
4408 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4409 netdev_reg_state(dev), ##args)
4410
4411 /* netif printk helpers, similar to netdev_printk */
4412
4413 #define netif_printk(priv, type, level, dev, fmt, args...) \
4414 do { \
4415 if (netif_msg_##type(priv)) \
4416 netdev_printk(level, (dev), fmt, ##args); \
4417 } while (0)
4418
4419 #define netif_level(level, priv, type, dev, fmt, args...) \
4420 do { \
4421 if (netif_msg_##type(priv)) \
4422 netdev_##level(dev, fmt, ##args); \
4423 } while (0)
4424
4425 #define netif_emerg(priv, type, dev, fmt, args...) \
4426 netif_level(emerg, priv, type, dev, fmt, ##args)
4427 #define netif_alert(priv, type, dev, fmt, args...) \
4428 netif_level(alert, priv, type, dev, fmt, ##args)
4429 #define netif_crit(priv, type, dev, fmt, args...) \
4430 netif_level(crit, priv, type, dev, fmt, ##args)
4431 #define netif_err(priv, type, dev, fmt, args...) \
4432 netif_level(err, priv, type, dev, fmt, ##args)
4433 #define netif_warn(priv, type, dev, fmt, args...) \
4434 netif_level(warn, priv, type, dev, fmt, ##args)
4435 #define netif_notice(priv, type, dev, fmt, args...) \
4436 netif_level(notice, priv, type, dev, fmt, ##args)
4437 #define netif_info(priv, type, dev, fmt, args...) \
4438 netif_level(info, priv, type, dev, fmt, ##args)
4439
4440 #if defined(CONFIG_DYNAMIC_DEBUG)
4441 #define netif_dbg(priv, type, netdev, format, args...) \
4442 do { \
4443 if (netif_msg_##type(priv)) \
4444 dynamic_netdev_dbg(netdev, format, ##args); \
4445 } while (0)
4446 #elif defined(DEBUG)
4447 #define netif_dbg(priv, type, dev, format, args...) \
4448 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4449 #else
4450 #define netif_dbg(priv, type, dev, format, args...) \
4451 ({ \
4452 if (0) \
4453 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4454 0; \
4455 })
4456 #endif
4457
4458 /* if @cond then downgrade to debug, else print at @level */
4459 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4460 do { \
4461 if (cond) \
4462 netif_dbg(priv, type, netdev, fmt, ##args); \
4463 else \
4464 netif_ ## level(priv, type, netdev, fmt, ##args); \
4465 } while (0)
4466
4467 #if defined(VERBOSE_DEBUG)
4468 #define netif_vdbg netif_dbg
4469 #else
4470 #define netif_vdbg(priv, type, dev, format, args...) \
4471 ({ \
4472 if (0) \
4473 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4474 0; \
4475 })
4476 #endif
4477
4478 /*
4479 * The list of packet types we will receive (as opposed to discard)
4480 * and the routines to invoke.
4481 *
4482 * Why 16. Because with 16 the only overlap we get on a hash of the
4483 * low nibble of the protocol value is RARP/SNAP/X.25.
4484 *
4485 * 0800 IP
4486 * 0001 802.3
4487 * 0002 AX.25
4488 * 0004 802.2
4489 * 8035 RARP
4490 * 0005 SNAP
4491 * 0805 X.25
4492 * 0806 ARP
4493 * 8137 IPX
4494 * 0009 Localtalk
4495 * 86DD IPv6
4496 */
4497 #define PTYPE_HASH_SIZE (16)
4498 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4499
4500 #endif /* _LINUX_NETDEVICE_H */