1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
36
37 #include <linux/ethtool.h>
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 #include <net/xdp.h>
44
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
51
52 struct netpoll_info;
53 struct device;
54 struct phy_device;
55 struct dsa_port;
56
57 struct sfp_bus;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63 /* UDP Tunnel offloads */
64 struct udp_tunnel_info;
65 struct bpf_prog;
66 struct xdp_buff;
67
68 void netdev_set_default_ethtool_ops(struct net_device *dev,
69 const struct ethtool_ops *ops);
70
71 /* Backlog congestion levels */
72 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
73 #define NET_RX_DROP 1 /* packet dropped */
74
75 /*
76 * Transmit return codes: transmit return codes originate from three different
77 * namespaces:
78 *
79 * - qdisc return codes
80 * - driver transmit return codes
81 * - errno values
82 *
83 * Drivers are allowed to return any one of those in their hard_start_xmit()
84 * function. Real network devices commonly used with qdiscs should only return
85 * the driver transmit return codes though - when qdiscs are used, the actual
86 * transmission happens asynchronously, so the value is not propagated to
87 * higher layers. Virtual network devices transmit synchronously; in this case
88 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
89 * others are propagated to higher layers.
90 */
91
92 /* qdisc ->enqueue() return codes. */
93 #define NET_XMIT_SUCCESS 0x00
94 #define NET_XMIT_DROP 0x01 /* skb dropped */
95 #define NET_XMIT_CN 0x02 /* congestion notification */
96 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
97
98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99 * indicates that the device will soon be dropping packets, or already drops
100 * some packets of the same priority; prompting us to send less aggressively. */
101 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
102 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103
104 /* Driver transmit return codes */
105 #define NETDEV_TX_MASK 0xf0
106
107 enum netdev_tx {
108 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
109 NETDEV_TX_OK = 0x00, /* driver took care of packet */
110 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
111 };
112 typedef enum netdev_tx netdev_tx_t;
113
114 /*
115 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 */
dev_xmit_complete(int rc)118 static inline bool dev_xmit_complete(int rc)
119 {
120 /*
121 * Positive cases with an skb consumed by a driver:
122 * - successful transmission (rc == NETDEV_TX_OK)
123 * - error while transmitting (rc < 0)
124 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 */
126 if (likely(rc < NET_XMIT_MASK))
127 return true;
128
129 return false;
130 }
131
132 /*
133 * Compute the worst-case header length according to the protocols
134 * used.
135 */
136
137 #if defined(CONFIG_HYPERV_NET)
138 # define LL_MAX_HEADER 128
139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140 # if defined(CONFIG_MAC80211_MESH)
141 # define LL_MAX_HEADER 128
142 # else
143 # define LL_MAX_HEADER 96
144 # endif
145 #else
146 # define LL_MAX_HEADER 32
147 #endif
148
149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
151 #define MAX_HEADER LL_MAX_HEADER
152 #else
153 #define MAX_HEADER (LL_MAX_HEADER + 48)
154 #endif
155
156 /*
157 * Old network device statistics. Fields are native words
158 * (unsigned long) so they can be read and written atomically.
159 */
160
161 struct net_device_stats {
162 unsigned long rx_packets;
163 unsigned long tx_packets;
164 unsigned long rx_bytes;
165 unsigned long tx_bytes;
166 unsigned long rx_errors;
167 unsigned long tx_errors;
168 unsigned long rx_dropped;
169 unsigned long tx_dropped;
170 unsigned long multicast;
171 unsigned long collisions;
172 unsigned long rx_length_errors;
173 unsigned long rx_over_errors;
174 unsigned long rx_crc_errors;
175 unsigned long rx_frame_errors;
176 unsigned long rx_fifo_errors;
177 unsigned long rx_missed_errors;
178 unsigned long tx_aborted_errors;
179 unsigned long tx_carrier_errors;
180 unsigned long tx_fifo_errors;
181 unsigned long tx_heartbeat_errors;
182 unsigned long tx_window_errors;
183 unsigned long rx_compressed;
184 unsigned long tx_compressed;
185 };
186
187
188 #include <linux/cache.h>
189 #include <linux/skbuff.h>
190
191 #ifdef CONFIG_RPS
192 #include <linux/static_key.h>
193 extern struct static_key_false rps_needed;
194 extern struct static_key_false rfs_needed;
195 #endif
196
197 struct neighbour;
198 struct neigh_parms;
199 struct sk_buff;
200
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
204 unsigned char type;
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
210 bool global_use;
211 int sync_cnt;
212 int refcount;
213 int synced;
214 struct rcu_head rcu_head;
215 };
216
217 struct netdev_hw_addr_list {
218 struct list_head list;
219 int count;
220 };
221
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
226
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236
237 struct hh_cache {
238 unsigned int hh_len;
239 seqlock_t hh_lock;
240
241 /* cached hardware header; allow for machine alignment needs. */
242 #define HH_DATA_MOD 16
243 #define HH_DATA_OFF(__len) \
244 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
245 #define HH_DATA_ALIGN(__len) \
246 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248 };
249
250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
251 * Alternative is:
252 * dev->hard_header_len ? (dev->hard_header_len +
253 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 *
255 * We could use other alignment values, but we must maintain the
256 * relationship HH alignment <= LL alignment.
257 */
258 #define LL_RESERVED_SPACE(dev) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262
263 struct header_ops {
264 int (*create) (struct sk_buff *skb, struct net_device *dev,
265 unsigned short type, const void *daddr,
266 const void *saddr, unsigned int len);
267 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
268 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
269 void (*cache_update)(struct hh_cache *hh,
270 const struct net_device *dev,
271 const unsigned char *haddr);
272 bool (*validate)(const char *ll_header, unsigned int len);
273 __be16 (*parse_protocol)(const struct sk_buff *skb);
274 };
275
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
278 * code.
279 */
280
281 enum netdev_state_t {
282 __LINK_STATE_START,
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
287 };
288
289
290 /*
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
293 */
294 struct netdev_boot_setup {
295 char name[IFNAMSIZ];
296 struct ifmap map;
297 };
298 #define NETDEV_BOOT_SETUP_MAX 8
299
300 int __init netdev_boot_setup(char *str);
301
302 struct gro_list {
303 struct list_head list;
304 int count;
305 };
306
307 /*
308 * size of gro hash buckets, must less than bit number of
309 * napi_struct::gro_bitmask
310 */
311 #define GRO_HASH_BUCKETS 8
312
313 /*
314 * Structure for NAPI scheduling similar to tasklet but with weighting
315 */
316 struct napi_struct {
317 /* The poll_list must only be managed by the entity which
318 * changes the state of the NAPI_STATE_SCHED bit. This means
319 * whoever atomically sets that bit can add this napi_struct
320 * to the per-CPU poll_list, and whoever clears that bit
321 * can remove from the list right before clearing the bit.
322 */
323 struct list_head poll_list;
324
325 unsigned long state;
326 int weight;
327 unsigned long gro_bitmask;
328 int (*poll)(struct napi_struct *, int);
329 #ifdef CONFIG_NETPOLL
330 int poll_owner;
331 #endif
332 struct net_device *dev;
333 struct gro_list gro_hash[GRO_HASH_BUCKETS];
334 struct sk_buff *skb;
335 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
336 int rx_count; /* length of rx_list */
337 struct hrtimer timer;
338 struct list_head dev_list;
339 struct hlist_node napi_hash_node;
340 unsigned int napi_id;
341 };
342
343 enum {
344 NAPI_STATE_SCHED, /* Poll is scheduled */
345 NAPI_STATE_MISSED, /* reschedule a napi */
346 NAPI_STATE_DISABLE, /* Disable pending */
347 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
348 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
349 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
350 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
351 };
352
353 enum {
354 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
355 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
356 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
357 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
358 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
359 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
360 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
361 };
362
363 enum gro_result {
364 GRO_MERGED,
365 GRO_MERGED_FREE,
366 GRO_HELD,
367 GRO_NORMAL,
368 GRO_DROP,
369 GRO_CONSUMED,
370 };
371 typedef enum gro_result gro_result_t;
372
373 /*
374 * enum rx_handler_result - Possible return values for rx_handlers.
375 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
376 * further.
377 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
378 * case skb->dev was changed by rx_handler.
379 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
380 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
381 *
382 * rx_handlers are functions called from inside __netif_receive_skb(), to do
383 * special processing of the skb, prior to delivery to protocol handlers.
384 *
385 * Currently, a net_device can only have a single rx_handler registered. Trying
386 * to register a second rx_handler will return -EBUSY.
387 *
388 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
389 * To unregister a rx_handler on a net_device, use
390 * netdev_rx_handler_unregister().
391 *
392 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
393 * do with the skb.
394 *
395 * If the rx_handler consumed the skb in some way, it should return
396 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
397 * the skb to be delivered in some other way.
398 *
399 * If the rx_handler changed skb->dev, to divert the skb to another
400 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
401 * new device will be called if it exists.
402 *
403 * If the rx_handler decides the skb should be ignored, it should return
404 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
405 * are registered on exact device (ptype->dev == skb->dev).
406 *
407 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
408 * delivered, it should return RX_HANDLER_PASS.
409 *
410 * A device without a registered rx_handler will behave as if rx_handler
411 * returned RX_HANDLER_PASS.
412 */
413
414 enum rx_handler_result {
415 RX_HANDLER_CONSUMED,
416 RX_HANDLER_ANOTHER,
417 RX_HANDLER_EXACT,
418 RX_HANDLER_PASS,
419 };
420 typedef enum rx_handler_result rx_handler_result_t;
421 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
422
423 void __napi_schedule(struct napi_struct *n);
424 void __napi_schedule_irqoff(struct napi_struct *n);
425
napi_disable_pending(struct napi_struct * n)426 static inline bool napi_disable_pending(struct napi_struct *n)
427 {
428 return test_bit(NAPI_STATE_DISABLE, &n->state);
429 }
430
431 bool napi_schedule_prep(struct napi_struct *n);
432
433 /**
434 * napi_schedule - schedule NAPI poll
435 * @n: NAPI context
436 *
437 * Schedule NAPI poll routine to be called if it is not already
438 * running.
439 */
napi_schedule(struct napi_struct * n)440 static inline void napi_schedule(struct napi_struct *n)
441 {
442 if (napi_schedule_prep(n))
443 __napi_schedule(n);
444 }
445
446 /**
447 * napi_schedule_irqoff - schedule NAPI poll
448 * @n: NAPI context
449 *
450 * Variant of napi_schedule(), assuming hard irqs are masked.
451 */
napi_schedule_irqoff(struct napi_struct * n)452 static inline void napi_schedule_irqoff(struct napi_struct *n)
453 {
454 if (napi_schedule_prep(n))
455 __napi_schedule_irqoff(n);
456 }
457
458 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)459 static inline bool napi_reschedule(struct napi_struct *napi)
460 {
461 if (napi_schedule_prep(napi)) {
462 __napi_schedule(napi);
463 return true;
464 }
465 return false;
466 }
467
468 bool napi_complete_done(struct napi_struct *n, int work_done);
469 /**
470 * napi_complete - NAPI processing complete
471 * @n: NAPI context
472 *
473 * Mark NAPI processing as complete.
474 * Consider using napi_complete_done() instead.
475 * Return false if device should avoid rearming interrupts.
476 */
napi_complete(struct napi_struct * n)477 static inline bool napi_complete(struct napi_struct *n)
478 {
479 return napi_complete_done(n, 0);
480 }
481
482 /**
483 * napi_hash_del - remove a NAPI from global table
484 * @napi: NAPI context
485 *
486 * Warning: caller must observe RCU grace period
487 * before freeing memory containing @napi, if
488 * this function returns true.
489 * Note: core networking stack automatically calls it
490 * from netif_napi_del().
491 * Drivers might want to call this helper to combine all
492 * the needed RCU grace periods into a single one.
493 */
494 bool napi_hash_del(struct napi_struct *napi);
495
496 /**
497 * napi_disable - prevent NAPI from scheduling
498 * @n: NAPI context
499 *
500 * Stop NAPI from being scheduled on this context.
501 * Waits till any outstanding processing completes.
502 */
503 void napi_disable(struct napi_struct *n);
504
505 /**
506 * napi_enable - enable NAPI scheduling
507 * @n: NAPI context
508 *
509 * Resume NAPI from being scheduled on this context.
510 * Must be paired with napi_disable.
511 */
napi_enable(struct napi_struct * n)512 static inline void napi_enable(struct napi_struct *n)
513 {
514 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
515 smp_mb__before_atomic();
516 clear_bit(NAPI_STATE_SCHED, &n->state);
517 clear_bit(NAPI_STATE_NPSVC, &n->state);
518 }
519
520 /**
521 * napi_synchronize - wait until NAPI is not running
522 * @n: NAPI context
523 *
524 * Wait until NAPI is done being scheduled on this context.
525 * Waits till any outstanding processing completes but
526 * does not disable future activations.
527 */
napi_synchronize(const struct napi_struct * n)528 static inline void napi_synchronize(const struct napi_struct *n)
529 {
530 if (IS_ENABLED(CONFIG_SMP))
531 while (test_bit(NAPI_STATE_SCHED, &n->state))
532 msleep(1);
533 else
534 barrier();
535 }
536
537 /**
538 * napi_if_scheduled_mark_missed - if napi is running, set the
539 * NAPIF_STATE_MISSED
540 * @n: NAPI context
541 *
542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
543 * NAPI is scheduled.
544 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
546 {
547 unsigned long val, new;
548
549 do {
550 val = READ_ONCE(n->state);
551 if (val & NAPIF_STATE_DISABLE)
552 return true;
553
554 if (!(val & NAPIF_STATE_SCHED))
555 return false;
556
557 new = val | NAPIF_STATE_MISSED;
558 } while (cmpxchg(&n->state, val, new) != val);
559
560 return true;
561 }
562
563 enum netdev_queue_state_t {
564 __QUEUE_STATE_DRV_XOFF,
565 __QUEUE_STATE_STACK_XOFF,
566 __QUEUE_STATE_FROZEN,
567 };
568
569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
572
573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
575 QUEUE_STATE_FROZEN)
576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
577 QUEUE_STATE_FROZEN)
578
579 /*
580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
581 * netif_tx_* functions below are used to manipulate this flag. The
582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
583 * queue independently. The netif_xmit_*stopped functions below are called
584 * to check if the queue has been stopped by the driver or stack (either
585 * of the XOFF bits are set in the state). Drivers should not need to call
586 * netif_xmit*stopped functions, they should only be using netif_tx_*.
587 */
588
589 struct netdev_queue {
590 /*
591 * read-mostly part
592 */
593 struct net_device *dev;
594 struct Qdisc __rcu *qdisc;
595 struct Qdisc *qdisc_sleeping;
596 #ifdef CONFIG_SYSFS
597 struct kobject kobj;
598 #endif
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 int numa_node;
601 #endif
602 unsigned long tx_maxrate;
603 /*
604 * Number of TX timeouts for this queue
605 * (/sys/class/net/DEV/Q/trans_timeout)
606 */
607 unsigned long trans_timeout;
608
609 /* Subordinate device that the queue has been assigned to */
610 struct net_device *sb_dev;
611 #ifdef CONFIG_XDP_SOCKETS
612 struct xdp_umem *umem;
613 #endif
614 /*
615 * write-mostly part
616 */
617 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
618 int xmit_lock_owner;
619 /*
620 * Time (in jiffies) of last Tx
621 */
622 unsigned long trans_start;
623
624 unsigned long state;
625
626 #ifdef CONFIG_BQL
627 struct dql dql;
628 #endif
629 } ____cacheline_aligned_in_smp;
630
631 extern int sysctl_fb_tunnels_only_for_init_net;
632 extern int sysctl_devconf_inherit_init_net;
633
net_has_fallback_tunnels(const struct net * net)634 static inline bool net_has_fallback_tunnels(const struct net *net)
635 {
636 return net == &init_net ||
637 !IS_ENABLED(CONFIG_SYSCTL) ||
638 !sysctl_fb_tunnels_only_for_init_net;
639 }
640
netdev_queue_numa_node_read(const struct netdev_queue * q)641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
642 {
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
644 return q->numa_node;
645 #else
646 return NUMA_NO_NODE;
647 #endif
648 }
649
netdev_queue_numa_node_write(struct netdev_queue * q,int node)650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
651 {
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
653 q->numa_node = node;
654 #endif
655 }
656
657 #ifdef CONFIG_RPS
658 /*
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
661 */
662 struct rps_map {
663 unsigned int len;
664 struct rcu_head rcu;
665 u16 cpus[0];
666 };
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668
669 /*
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
673 */
674 struct rps_dev_flow {
675 u16 cpu;
676 u16 filter;
677 unsigned int last_qtail;
678 };
679 #define RPS_NO_FILTER 0xffff
680
681 /*
682 * The rps_dev_flow_table structure contains a table of flow mappings.
683 */
684 struct rps_dev_flow_table {
685 unsigned int mask;
686 struct rcu_head rcu;
687 struct rps_dev_flow flows[0];
688 };
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
691
692 /*
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
701 */
702 struct rps_sock_flow_table {
703 u32 mask;
704
705 u32 ents[0] ____cacheline_aligned_in_smp;
706 };
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
708
709 #define RPS_NO_CPU 0xffff
710
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
713
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
715 u32 hash)
716 {
717 if (table && hash) {
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
720
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
723
724 if (table->ents[index] != val)
725 table->ents[index] = val;
726 }
727 }
728
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 u16 filter_id);
732 #endif
733 #endif /* CONFIG_RPS */
734
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
737 #ifdef CONFIG_RPS
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
740 #endif
741 struct kobject kobj;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xdp_umem *umem;
746 #endif
747 } ____cacheline_aligned_in_smp;
748
749 /*
750 * RX queue sysfs structures and functions.
751 */
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
757 };
758
759 #ifdef CONFIG_XPS
760 /*
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
763 */
764 struct xps_map {
765 unsigned int len;
766 unsigned int alloc_len;
767 struct rcu_head rcu;
768 u16 queues[0];
769 };
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
773
774 /*
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
776 */
777 struct xps_dev_maps {
778 struct rcu_head rcu;
779 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
780 };
781
782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
784
785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
787
788 #endif /* CONFIG_XPS */
789
790 #define TC_MAX_QUEUE 16
791 #define TC_BITMASK 15
792 /* HW offloaded queuing disciplines txq count and offset maps */
793 struct netdev_tc_txq {
794 u16 count;
795 u16 offset;
796 };
797
798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
799 /*
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
802 */
803 struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
810 char model[256];
811 char model_description[256];
812 };
813 #endif
814
815 #define MAX_PHYS_ITEM_ID_LEN 32
816
817 /* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
819 */
820 struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 unsigned char id_len;
823 };
824
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
827 {
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
830 }
831
832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
833 struct sk_buff *skb,
834 struct net_device *sb_dev);
835
836 enum tc_setup_type {
837 TC_SETUP_QDISC_MQPRIO,
838 TC_SETUP_CLSU32,
839 TC_SETUP_CLSFLOWER,
840 TC_SETUP_CLSMATCHALL,
841 TC_SETUP_CLSBPF,
842 TC_SETUP_BLOCK,
843 TC_SETUP_QDISC_CBS,
844 TC_SETUP_QDISC_RED,
845 TC_SETUP_QDISC_PRIO,
846 TC_SETUP_QDISC_MQ,
847 TC_SETUP_QDISC_ETF,
848 TC_SETUP_ROOT_QDISC,
849 TC_SETUP_QDISC_GRED,
850 TC_SETUP_QDISC_TAPRIO,
851 };
852
853 /* These structures hold the attributes of bpf state that are being passed
854 * to the netdevice through the bpf op.
855 */
856 enum bpf_netdev_command {
857 /* Set or clear a bpf program used in the earliest stages of packet
858 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
859 * is responsible for calling bpf_prog_put on any old progs that are
860 * stored. In case of error, the callee need not release the new prog
861 * reference, but on success it takes ownership and must bpf_prog_put
862 * when it is no longer used.
863 */
864 XDP_SETUP_PROG,
865 XDP_SETUP_PROG_HW,
866 XDP_QUERY_PROG,
867 XDP_QUERY_PROG_HW,
868 /* BPF program for offload callbacks, invoked at program load time. */
869 BPF_OFFLOAD_MAP_ALLOC,
870 BPF_OFFLOAD_MAP_FREE,
871 XDP_SETUP_XSK_UMEM,
872 };
873
874 struct bpf_prog_offload_ops;
875 struct netlink_ext_ack;
876 struct xdp_umem;
877
878 struct netdev_bpf {
879 enum bpf_netdev_command command;
880 union {
881 /* XDP_SETUP_PROG */
882 struct {
883 u32 flags;
884 struct bpf_prog *prog;
885 struct netlink_ext_ack *extack;
886 };
887 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
888 struct {
889 u32 prog_id;
890 /* flags with which program was installed */
891 u32 prog_flags;
892 };
893 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
894 struct {
895 struct bpf_offloaded_map *offmap;
896 };
897 /* XDP_SETUP_XSK_UMEM */
898 struct {
899 struct xdp_umem *umem;
900 u16 queue_id;
901 } xsk;
902 };
903 };
904
905 /* Flags for ndo_xsk_wakeup. */
906 #define XDP_WAKEUP_RX (1 << 0)
907 #define XDP_WAKEUP_TX (1 << 1)
908
909 #ifdef CONFIG_XFRM_OFFLOAD
910 struct xfrmdev_ops {
911 int (*xdo_dev_state_add) (struct xfrm_state *x);
912 void (*xdo_dev_state_delete) (struct xfrm_state *x);
913 void (*xdo_dev_state_free) (struct xfrm_state *x);
914 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
915 struct xfrm_state *x);
916 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
917 };
918 #endif
919
920 struct dev_ifalias {
921 struct rcu_head rcuhead;
922 char ifalias[];
923 };
924
925 struct devlink;
926 struct tlsdev_ops;
927
928
929 /*
930 * This structure defines the management hooks for network devices.
931 * The following hooks can be defined; unless noted otherwise, they are
932 * optional and can be filled with a null pointer.
933 *
934 * int (*ndo_init)(struct net_device *dev);
935 * This function is called once when a network device is registered.
936 * The network device can use this for any late stage initialization
937 * or semantic validation. It can fail with an error code which will
938 * be propagated back to register_netdev.
939 *
940 * void (*ndo_uninit)(struct net_device *dev);
941 * This function is called when device is unregistered or when registration
942 * fails. It is not called if init fails.
943 *
944 * int (*ndo_open)(struct net_device *dev);
945 * This function is called when a network device transitions to the up
946 * state.
947 *
948 * int (*ndo_stop)(struct net_device *dev);
949 * This function is called when a network device transitions to the down
950 * state.
951 *
952 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
953 * struct net_device *dev);
954 * Called when a packet needs to be transmitted.
955 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
956 * the queue before that can happen; it's for obsolete devices and weird
957 * corner cases, but the stack really does a non-trivial amount
958 * of useless work if you return NETDEV_TX_BUSY.
959 * Required; cannot be NULL.
960 *
961 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
962 * struct net_device *dev
963 * netdev_features_t features);
964 * Called by core transmit path to determine if device is capable of
965 * performing offload operations on a given packet. This is to give
966 * the device an opportunity to implement any restrictions that cannot
967 * be otherwise expressed by feature flags. The check is called with
968 * the set of features that the stack has calculated and it returns
969 * those the driver believes to be appropriate.
970 *
971 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
972 * struct net_device *sb_dev);
973 * Called to decide which queue to use when device supports multiple
974 * transmit queues.
975 *
976 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
977 * This function is called to allow device receiver to make
978 * changes to configuration when multicast or promiscuous is enabled.
979 *
980 * void (*ndo_set_rx_mode)(struct net_device *dev);
981 * This function is called device changes address list filtering.
982 * If driver handles unicast address filtering, it should set
983 * IFF_UNICAST_FLT in its priv_flags.
984 *
985 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
986 * This function is called when the Media Access Control address
987 * needs to be changed. If this interface is not defined, the
988 * MAC address can not be changed.
989 *
990 * int (*ndo_validate_addr)(struct net_device *dev);
991 * Test if Media Access Control address is valid for the device.
992 *
993 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
994 * Called when a user requests an ioctl which can't be handled by
995 * the generic interface code. If not defined ioctls return
996 * not supported error code.
997 *
998 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
999 * Used to set network devices bus interface parameters. This interface
1000 * is retained for legacy reasons; new devices should use the bus
1001 * interface (PCI) for low level management.
1002 *
1003 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1004 * Called when a user wants to change the Maximum Transfer Unit
1005 * of a device.
1006 *
1007 * void (*ndo_tx_timeout)(struct net_device *dev);
1008 * Callback used when the transmitter has not made any progress
1009 * for dev->watchdog ticks.
1010 *
1011 * void (*ndo_get_stats64)(struct net_device *dev,
1012 * struct rtnl_link_stats64 *storage);
1013 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1014 * Called when a user wants to get the network device usage
1015 * statistics. Drivers must do one of the following:
1016 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1017 * rtnl_link_stats64 structure passed by the caller.
1018 * 2. Define @ndo_get_stats to update a net_device_stats structure
1019 * (which should normally be dev->stats) and return a pointer to
1020 * it. The structure may be changed asynchronously only if each
1021 * field is written atomically.
1022 * 3. Update dev->stats asynchronously and atomically, and define
1023 * neither operation.
1024 *
1025 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1026 * Return true if this device supports offload stats of this attr_id.
1027 *
1028 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1029 * void *attr_data)
1030 * Get statistics for offload operations by attr_id. Write it into the
1031 * attr_data pointer.
1032 *
1033 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1034 * If device supports VLAN filtering this function is called when a
1035 * VLAN id is registered.
1036 *
1037 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1038 * If device supports VLAN filtering this function is called when a
1039 * VLAN id is unregistered.
1040 *
1041 * void (*ndo_poll_controller)(struct net_device *dev);
1042 *
1043 * SR-IOV management functions.
1044 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1045 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1046 * u8 qos, __be16 proto);
1047 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1048 * int max_tx_rate);
1049 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1050 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1051 * int (*ndo_get_vf_config)(struct net_device *dev,
1052 * int vf, struct ifla_vf_info *ivf);
1053 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1054 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1055 * struct nlattr *port[]);
1056 *
1057 * Enable or disable the VF ability to query its RSS Redirection Table and
1058 * Hash Key. This is needed since on some devices VF share this information
1059 * with PF and querying it may introduce a theoretical security risk.
1060 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1061 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1062 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1063 * void *type_data);
1064 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1065 * This is always called from the stack with the rtnl lock held and netif
1066 * tx queues stopped. This allows the netdevice to perform queue
1067 * management safely.
1068 *
1069 * Fiber Channel over Ethernet (FCoE) offload functions.
1070 * int (*ndo_fcoe_enable)(struct net_device *dev);
1071 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1072 * so the underlying device can perform whatever needed configuration or
1073 * initialization to support acceleration of FCoE traffic.
1074 *
1075 * int (*ndo_fcoe_disable)(struct net_device *dev);
1076 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1077 * so the underlying device can perform whatever needed clean-ups to
1078 * stop supporting acceleration of FCoE traffic.
1079 *
1080 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1081 * struct scatterlist *sgl, unsigned int sgc);
1082 * Called when the FCoE Initiator wants to initialize an I/O that
1083 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1084 * perform necessary setup and returns 1 to indicate the device is set up
1085 * successfully to perform DDP on this I/O, otherwise this returns 0.
1086 *
1087 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1088 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1089 * indicated by the FC exchange id 'xid', so the underlying device can
1090 * clean up and reuse resources for later DDP requests.
1091 *
1092 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1093 * struct scatterlist *sgl, unsigned int sgc);
1094 * Called when the FCoE Target wants to initialize an I/O that
1095 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1096 * perform necessary setup and returns 1 to indicate the device is set up
1097 * successfully to perform DDP on this I/O, otherwise this returns 0.
1098 *
1099 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1100 * struct netdev_fcoe_hbainfo *hbainfo);
1101 * Called when the FCoE Protocol stack wants information on the underlying
1102 * device. This information is utilized by the FCoE protocol stack to
1103 * register attributes with Fiber Channel management service as per the
1104 * FC-GS Fabric Device Management Information(FDMI) specification.
1105 *
1106 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1107 * Called when the underlying device wants to override default World Wide
1108 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1109 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1110 * protocol stack to use.
1111 *
1112 * RFS acceleration.
1113 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1114 * u16 rxq_index, u32 flow_id);
1115 * Set hardware filter for RFS. rxq_index is the target queue index;
1116 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1117 * Return the filter ID on success, or a negative error code.
1118 *
1119 * Slave management functions (for bridge, bonding, etc).
1120 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1121 * Called to make another netdev an underling.
1122 *
1123 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1124 * Called to release previously enslaved netdev.
1125 *
1126 * Feature/offload setting functions.
1127 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1128 * netdev_features_t features);
1129 * Adjusts the requested feature flags according to device-specific
1130 * constraints, and returns the resulting flags. Must not modify
1131 * the device state.
1132 *
1133 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1134 * Called to update device configuration to new features. Passed
1135 * feature set might be less than what was returned by ndo_fix_features()).
1136 * Must return >0 or -errno if it changed dev->features itself.
1137 *
1138 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1139 * struct net_device *dev,
1140 * const unsigned char *addr, u16 vid, u16 flags,
1141 * struct netlink_ext_ack *extack);
1142 * Adds an FDB entry to dev for addr.
1143 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1144 * struct net_device *dev,
1145 * const unsigned char *addr, u16 vid)
1146 * Deletes the FDB entry from dev coresponding to addr.
1147 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1148 * struct net_device *dev, struct net_device *filter_dev,
1149 * int *idx)
1150 * Used to add FDB entries to dump requests. Implementers should add
1151 * entries to skb and update idx with the number of entries.
1152 *
1153 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1154 * u16 flags, struct netlink_ext_ack *extack)
1155 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1156 * struct net_device *dev, u32 filter_mask,
1157 * int nlflags)
1158 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1159 * u16 flags);
1160 *
1161 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1162 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1163 * which do not represent real hardware may define this to allow their
1164 * userspace components to manage their virtual carrier state. Devices
1165 * that determine carrier state from physical hardware properties (eg
1166 * network cables) or protocol-dependent mechanisms (eg
1167 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1168 *
1169 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1170 * struct netdev_phys_item_id *ppid);
1171 * Called to get ID of physical port of this device. If driver does
1172 * not implement this, it is assumed that the hw is not able to have
1173 * multiple net devices on single physical port.
1174 *
1175 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1176 * struct netdev_phys_item_id *ppid)
1177 * Called to get the parent ID of the physical port of this device.
1178 *
1179 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1180 * struct udp_tunnel_info *ti);
1181 * Called by UDP tunnel to notify a driver about the UDP port and socket
1182 * address family that a UDP tunnel is listnening to. It is called only
1183 * when a new port starts listening. The operation is protected by the
1184 * RTNL.
1185 *
1186 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1187 * struct udp_tunnel_info *ti);
1188 * Called by UDP tunnel to notify the driver about a UDP port and socket
1189 * address family that the UDP tunnel is not listening to anymore. The
1190 * operation is protected by the RTNL.
1191 *
1192 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1193 * struct net_device *dev)
1194 * Called by upper layer devices to accelerate switching or other
1195 * station functionality into hardware. 'pdev is the lowerdev
1196 * to use for the offload and 'dev' is the net device that will
1197 * back the offload. Returns a pointer to the private structure
1198 * the upper layer will maintain.
1199 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1200 * Called by upper layer device to delete the station created
1201 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1202 * the station and priv is the structure returned by the add
1203 * operation.
1204 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1205 * int queue_index, u32 maxrate);
1206 * Called when a user wants to set a max-rate limitation of specific
1207 * TX queue.
1208 * int (*ndo_get_iflink)(const struct net_device *dev);
1209 * Called to get the iflink value of this device.
1210 * void (*ndo_change_proto_down)(struct net_device *dev,
1211 * bool proto_down);
1212 * This function is used to pass protocol port error state information
1213 * to the switch driver. The switch driver can react to the proto_down
1214 * by doing a phys down on the associated switch port.
1215 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1216 * This function is used to get egress tunnel information for given skb.
1217 * This is useful for retrieving outer tunnel header parameters while
1218 * sampling packet.
1219 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1220 * This function is used to specify the headroom that the skb must
1221 * consider when allocation skb during packet reception. Setting
1222 * appropriate rx headroom value allows avoiding skb head copy on
1223 * forward. Setting a negative value resets the rx headroom to the
1224 * default value.
1225 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1226 * This function is used to set or query state related to XDP on the
1227 * netdevice and manage BPF offload. See definition of
1228 * enum bpf_netdev_command for details.
1229 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1230 * u32 flags);
1231 * This function is used to submit @n XDP packets for transmit on a
1232 * netdevice. Returns number of frames successfully transmitted, frames
1233 * that got dropped are freed/returned via xdp_return_frame().
1234 * Returns negative number, means general error invoking ndo, meaning
1235 * no frames were xmit'ed and core-caller will free all frames.
1236 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1237 * This function is used to wake up the softirq, ksoftirqd or kthread
1238 * responsible for sending and/or receiving packets on a specific
1239 * queue id bound to an AF_XDP socket. The flags field specifies if
1240 * only RX, only Tx, or both should be woken up using the flags
1241 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1242 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1243 * Get devlink port instance associated with a given netdev.
1244 * Called with a reference on the netdevice and devlink locks only,
1245 * rtnl_lock is not held.
1246 */
1247 struct net_device_ops {
1248 int (*ndo_init)(struct net_device *dev);
1249 void (*ndo_uninit)(struct net_device *dev);
1250 int (*ndo_open)(struct net_device *dev);
1251 int (*ndo_stop)(struct net_device *dev);
1252 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1253 struct net_device *dev);
1254 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1255 struct net_device *dev,
1256 netdev_features_t features);
1257 u16 (*ndo_select_queue)(struct net_device *dev,
1258 struct sk_buff *skb,
1259 struct net_device *sb_dev);
1260 void (*ndo_change_rx_flags)(struct net_device *dev,
1261 int flags);
1262 void (*ndo_set_rx_mode)(struct net_device *dev);
1263 int (*ndo_set_mac_address)(struct net_device *dev,
1264 void *addr);
1265 int (*ndo_validate_addr)(struct net_device *dev);
1266 int (*ndo_do_ioctl)(struct net_device *dev,
1267 struct ifreq *ifr, int cmd);
1268 int (*ndo_set_config)(struct net_device *dev,
1269 struct ifmap *map);
1270 int (*ndo_change_mtu)(struct net_device *dev,
1271 int new_mtu);
1272 int (*ndo_neigh_setup)(struct net_device *dev,
1273 struct neigh_parms *);
1274 void (*ndo_tx_timeout) (struct net_device *dev);
1275
1276 void (*ndo_get_stats64)(struct net_device *dev,
1277 struct rtnl_link_stats64 *storage);
1278 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1279 int (*ndo_get_offload_stats)(int attr_id,
1280 const struct net_device *dev,
1281 void *attr_data);
1282 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1283
1284 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1285 __be16 proto, u16 vid);
1286 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1287 __be16 proto, u16 vid);
1288 #ifdef CONFIG_NET_POLL_CONTROLLER
1289 void (*ndo_poll_controller)(struct net_device *dev);
1290 int (*ndo_netpoll_setup)(struct net_device *dev,
1291 struct netpoll_info *info);
1292 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1293 #endif
1294 int (*ndo_set_vf_mac)(struct net_device *dev,
1295 int queue, u8 *mac);
1296 int (*ndo_set_vf_vlan)(struct net_device *dev,
1297 int queue, u16 vlan,
1298 u8 qos, __be16 proto);
1299 int (*ndo_set_vf_rate)(struct net_device *dev,
1300 int vf, int min_tx_rate,
1301 int max_tx_rate);
1302 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1303 int vf, bool setting);
1304 int (*ndo_set_vf_trust)(struct net_device *dev,
1305 int vf, bool setting);
1306 int (*ndo_get_vf_config)(struct net_device *dev,
1307 int vf,
1308 struct ifla_vf_info *ivf);
1309 int (*ndo_set_vf_link_state)(struct net_device *dev,
1310 int vf, int link_state);
1311 int (*ndo_get_vf_stats)(struct net_device *dev,
1312 int vf,
1313 struct ifla_vf_stats
1314 *vf_stats);
1315 int (*ndo_set_vf_port)(struct net_device *dev,
1316 int vf,
1317 struct nlattr *port[]);
1318 int (*ndo_get_vf_port)(struct net_device *dev,
1319 int vf, struct sk_buff *skb);
1320 int (*ndo_set_vf_guid)(struct net_device *dev,
1321 int vf, u64 guid,
1322 int guid_type);
1323 int (*ndo_set_vf_rss_query_en)(
1324 struct net_device *dev,
1325 int vf, bool setting);
1326 int (*ndo_setup_tc)(struct net_device *dev,
1327 enum tc_setup_type type,
1328 void *type_data);
1329 #if IS_ENABLED(CONFIG_FCOE)
1330 int (*ndo_fcoe_enable)(struct net_device *dev);
1331 int (*ndo_fcoe_disable)(struct net_device *dev);
1332 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1333 u16 xid,
1334 struct scatterlist *sgl,
1335 unsigned int sgc);
1336 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1337 u16 xid);
1338 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1339 u16 xid,
1340 struct scatterlist *sgl,
1341 unsigned int sgc);
1342 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1343 struct netdev_fcoe_hbainfo *hbainfo);
1344 #endif
1345
1346 #if IS_ENABLED(CONFIG_LIBFCOE)
1347 #define NETDEV_FCOE_WWNN 0
1348 #define NETDEV_FCOE_WWPN 1
1349 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1350 u64 *wwn, int type);
1351 #endif
1352
1353 #ifdef CONFIG_RFS_ACCEL
1354 int (*ndo_rx_flow_steer)(struct net_device *dev,
1355 const struct sk_buff *skb,
1356 u16 rxq_index,
1357 u32 flow_id);
1358 #endif
1359 int (*ndo_add_slave)(struct net_device *dev,
1360 struct net_device *slave_dev,
1361 struct netlink_ext_ack *extack);
1362 int (*ndo_del_slave)(struct net_device *dev,
1363 struct net_device *slave_dev);
1364 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1365 netdev_features_t features);
1366 int (*ndo_set_features)(struct net_device *dev,
1367 netdev_features_t features);
1368 int (*ndo_neigh_construct)(struct net_device *dev,
1369 struct neighbour *n);
1370 void (*ndo_neigh_destroy)(struct net_device *dev,
1371 struct neighbour *n);
1372
1373 int (*ndo_fdb_add)(struct ndmsg *ndm,
1374 struct nlattr *tb[],
1375 struct net_device *dev,
1376 const unsigned char *addr,
1377 u16 vid,
1378 u16 flags,
1379 struct netlink_ext_ack *extack);
1380 int (*ndo_fdb_del)(struct ndmsg *ndm,
1381 struct nlattr *tb[],
1382 struct net_device *dev,
1383 const unsigned char *addr,
1384 u16 vid);
1385 int (*ndo_fdb_dump)(struct sk_buff *skb,
1386 struct netlink_callback *cb,
1387 struct net_device *dev,
1388 struct net_device *filter_dev,
1389 int *idx);
1390 int (*ndo_fdb_get)(struct sk_buff *skb,
1391 struct nlattr *tb[],
1392 struct net_device *dev,
1393 const unsigned char *addr,
1394 u16 vid, u32 portid, u32 seq,
1395 struct netlink_ext_ack *extack);
1396 int (*ndo_bridge_setlink)(struct net_device *dev,
1397 struct nlmsghdr *nlh,
1398 u16 flags,
1399 struct netlink_ext_ack *extack);
1400 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1401 u32 pid, u32 seq,
1402 struct net_device *dev,
1403 u32 filter_mask,
1404 int nlflags);
1405 int (*ndo_bridge_dellink)(struct net_device *dev,
1406 struct nlmsghdr *nlh,
1407 u16 flags);
1408 int (*ndo_change_carrier)(struct net_device *dev,
1409 bool new_carrier);
1410 int (*ndo_get_phys_port_id)(struct net_device *dev,
1411 struct netdev_phys_item_id *ppid);
1412 int (*ndo_get_port_parent_id)(struct net_device *dev,
1413 struct netdev_phys_item_id *ppid);
1414 int (*ndo_get_phys_port_name)(struct net_device *dev,
1415 char *name, size_t len);
1416 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1417 struct udp_tunnel_info *ti);
1418 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1419 struct udp_tunnel_info *ti);
1420 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1421 struct net_device *dev);
1422 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1423 void *priv);
1424
1425 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1426 int queue_index,
1427 u32 maxrate);
1428 int (*ndo_get_iflink)(const struct net_device *dev);
1429 int (*ndo_change_proto_down)(struct net_device *dev,
1430 bool proto_down);
1431 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1432 struct sk_buff *skb);
1433 void (*ndo_set_rx_headroom)(struct net_device *dev,
1434 int needed_headroom);
1435 int (*ndo_bpf)(struct net_device *dev,
1436 struct netdev_bpf *bpf);
1437 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1438 struct xdp_frame **xdp,
1439 u32 flags);
1440 int (*ndo_xsk_wakeup)(struct net_device *dev,
1441 u32 queue_id, u32 flags);
1442 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1443 };
1444
1445 /**
1446 * enum net_device_priv_flags - &struct net_device priv_flags
1447 *
1448 * These are the &struct net_device, they are only set internally
1449 * by drivers and used in the kernel. These flags are invisible to
1450 * userspace; this means that the order of these flags can change
1451 * during any kernel release.
1452 *
1453 * You should have a pretty good reason to be extending these flags.
1454 *
1455 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1456 * @IFF_EBRIDGE: Ethernet bridging device
1457 * @IFF_BONDING: bonding master or slave
1458 * @IFF_ISATAP: ISATAP interface (RFC4214)
1459 * @IFF_WAN_HDLC: WAN HDLC device
1460 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1461 * release skb->dst
1462 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1463 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1464 * @IFF_MACVLAN_PORT: device used as macvlan port
1465 * @IFF_BRIDGE_PORT: device used as bridge port
1466 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1467 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1468 * @IFF_UNICAST_FLT: Supports unicast filtering
1469 * @IFF_TEAM_PORT: device used as team port
1470 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1471 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1472 * change when it's running
1473 * @IFF_MACVLAN: Macvlan device
1474 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1475 * underlying stacked devices
1476 * @IFF_L3MDEV_MASTER: device is an L3 master device
1477 * @IFF_NO_QUEUE: device can run without qdisc attached
1478 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1479 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1480 * @IFF_TEAM: device is a team device
1481 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1482 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1483 * entity (i.e. the master device for bridged veth)
1484 * @IFF_MACSEC: device is a MACsec device
1485 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1486 * @IFF_FAILOVER: device is a failover master device
1487 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1488 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1489 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1490 */
1491 enum netdev_priv_flags {
1492 IFF_802_1Q_VLAN = 1<<0,
1493 IFF_EBRIDGE = 1<<1,
1494 IFF_BONDING = 1<<2,
1495 IFF_ISATAP = 1<<3,
1496 IFF_WAN_HDLC = 1<<4,
1497 IFF_XMIT_DST_RELEASE = 1<<5,
1498 IFF_DONT_BRIDGE = 1<<6,
1499 IFF_DISABLE_NETPOLL = 1<<7,
1500 IFF_MACVLAN_PORT = 1<<8,
1501 IFF_BRIDGE_PORT = 1<<9,
1502 IFF_OVS_DATAPATH = 1<<10,
1503 IFF_TX_SKB_SHARING = 1<<11,
1504 IFF_UNICAST_FLT = 1<<12,
1505 IFF_TEAM_PORT = 1<<13,
1506 IFF_SUPP_NOFCS = 1<<14,
1507 IFF_LIVE_ADDR_CHANGE = 1<<15,
1508 IFF_MACVLAN = 1<<16,
1509 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1510 IFF_L3MDEV_MASTER = 1<<18,
1511 IFF_NO_QUEUE = 1<<19,
1512 IFF_OPENVSWITCH = 1<<20,
1513 IFF_L3MDEV_SLAVE = 1<<21,
1514 IFF_TEAM = 1<<22,
1515 IFF_RXFH_CONFIGURED = 1<<23,
1516 IFF_PHONY_HEADROOM = 1<<24,
1517 IFF_MACSEC = 1<<25,
1518 IFF_NO_RX_HANDLER = 1<<26,
1519 IFF_FAILOVER = 1<<27,
1520 IFF_FAILOVER_SLAVE = 1<<28,
1521 IFF_L3MDEV_RX_HANDLER = 1<<29,
1522 IFF_LIVE_RENAME_OK = 1<<30,
1523 };
1524
1525 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1526 #define IFF_EBRIDGE IFF_EBRIDGE
1527 #define IFF_BONDING IFF_BONDING
1528 #define IFF_ISATAP IFF_ISATAP
1529 #define IFF_WAN_HDLC IFF_WAN_HDLC
1530 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1531 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1532 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1533 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1534 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1535 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1536 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1537 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1538 #define IFF_TEAM_PORT IFF_TEAM_PORT
1539 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1540 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1541 #define IFF_MACVLAN IFF_MACVLAN
1542 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1543 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1544 #define IFF_NO_QUEUE IFF_NO_QUEUE
1545 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1546 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1547 #define IFF_TEAM IFF_TEAM
1548 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1549 #define IFF_MACSEC IFF_MACSEC
1550 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1551 #define IFF_FAILOVER IFF_FAILOVER
1552 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1553 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1554 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1555
1556 /**
1557 * struct net_device - The DEVICE structure.
1558 *
1559 * Actually, this whole structure is a big mistake. It mixes I/O
1560 * data with strictly "high-level" data, and it has to know about
1561 * almost every data structure used in the INET module.
1562 *
1563 * @name: This is the first field of the "visible" part of this structure
1564 * (i.e. as seen by users in the "Space.c" file). It is the name
1565 * of the interface.
1566 *
1567 * @name_hlist: Device name hash chain, please keep it close to name[]
1568 * @ifalias: SNMP alias
1569 * @mem_end: Shared memory end
1570 * @mem_start: Shared memory start
1571 * @base_addr: Device I/O address
1572 * @irq: Device IRQ number
1573 *
1574 * @state: Generic network queuing layer state, see netdev_state_t
1575 * @dev_list: The global list of network devices
1576 * @napi_list: List entry used for polling NAPI devices
1577 * @unreg_list: List entry when we are unregistering the
1578 * device; see the function unregister_netdev
1579 * @close_list: List entry used when we are closing the device
1580 * @ptype_all: Device-specific packet handlers for all protocols
1581 * @ptype_specific: Device-specific, protocol-specific packet handlers
1582 *
1583 * @adj_list: Directly linked devices, like slaves for bonding
1584 * @features: Currently active device features
1585 * @hw_features: User-changeable features
1586 *
1587 * @wanted_features: User-requested features
1588 * @vlan_features: Mask of features inheritable by VLAN devices
1589 *
1590 * @hw_enc_features: Mask of features inherited by encapsulating devices
1591 * This field indicates what encapsulation
1592 * offloads the hardware is capable of doing,
1593 * and drivers will need to set them appropriately.
1594 *
1595 * @mpls_features: Mask of features inheritable by MPLS
1596 *
1597 * @ifindex: interface index
1598 * @group: The group the device belongs to
1599 *
1600 * @stats: Statistics struct, which was left as a legacy, use
1601 * rtnl_link_stats64 instead
1602 *
1603 * @rx_dropped: Dropped packets by core network,
1604 * do not use this in drivers
1605 * @tx_dropped: Dropped packets by core network,
1606 * do not use this in drivers
1607 * @rx_nohandler: nohandler dropped packets by core network on
1608 * inactive devices, do not use this in drivers
1609 * @carrier_up_count: Number of times the carrier has been up
1610 * @carrier_down_count: Number of times the carrier has been down
1611 *
1612 * @wireless_handlers: List of functions to handle Wireless Extensions,
1613 * instead of ioctl,
1614 * see <net/iw_handler.h> for details.
1615 * @wireless_data: Instance data managed by the core of wireless extensions
1616 *
1617 * @netdev_ops: Includes several pointers to callbacks,
1618 * if one wants to override the ndo_*() functions
1619 * @ethtool_ops: Management operations
1620 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1621 * discovery handling. Necessary for e.g. 6LoWPAN.
1622 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1623 * of Layer 2 headers.
1624 *
1625 * @flags: Interface flags (a la BSD)
1626 * @priv_flags: Like 'flags' but invisible to userspace,
1627 * see if.h for the definitions
1628 * @gflags: Global flags ( kept as legacy )
1629 * @padded: How much padding added by alloc_netdev()
1630 * @operstate: RFC2863 operstate
1631 * @link_mode: Mapping policy to operstate
1632 * @if_port: Selectable AUI, TP, ...
1633 * @dma: DMA channel
1634 * @mtu: Interface MTU value
1635 * @min_mtu: Interface Minimum MTU value
1636 * @max_mtu: Interface Maximum MTU value
1637 * @type: Interface hardware type
1638 * @hard_header_len: Maximum hardware header length.
1639 * @min_header_len: Minimum hardware header length
1640 *
1641 * @needed_headroom: Extra headroom the hardware may need, but not in all
1642 * cases can this be guaranteed
1643 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1644 * cases can this be guaranteed. Some cases also use
1645 * LL_MAX_HEADER instead to allocate the skb
1646 *
1647 * interface address info:
1648 *
1649 * @perm_addr: Permanent hw address
1650 * @addr_assign_type: Hw address assignment type
1651 * @addr_len: Hardware address length
1652 * @upper_level: Maximum depth level of upper devices.
1653 * @lower_level: Maximum depth level of lower devices.
1654 * @neigh_priv_len: Used in neigh_alloc()
1655 * @dev_id: Used to differentiate devices that share
1656 * the same link layer address
1657 * @dev_port: Used to differentiate devices that share
1658 * the same function
1659 * @addr_list_lock: XXX: need comments on this one
1660 * @uc_promisc: Counter that indicates promiscuous mode
1661 * has been enabled due to the need to listen to
1662 * additional unicast addresses in a device that
1663 * does not implement ndo_set_rx_mode()
1664 * @uc: unicast mac addresses
1665 * @mc: multicast mac addresses
1666 * @dev_addrs: list of device hw addresses
1667 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1668 * @promiscuity: Number of times the NIC is told to work in
1669 * promiscuous mode; if it becomes 0 the NIC will
1670 * exit promiscuous mode
1671 * @allmulti: Counter, enables or disables allmulticast mode
1672 *
1673 * @vlan_info: VLAN info
1674 * @dsa_ptr: dsa specific data
1675 * @tipc_ptr: TIPC specific data
1676 * @atalk_ptr: AppleTalk link
1677 * @ip_ptr: IPv4 specific data
1678 * @dn_ptr: DECnet specific data
1679 * @ip6_ptr: IPv6 specific data
1680 * @ax25_ptr: AX.25 specific data
1681 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1682 *
1683 * @dev_addr: Hw address (before bcast,
1684 * because most packets are unicast)
1685 *
1686 * @_rx: Array of RX queues
1687 * @num_rx_queues: Number of RX queues
1688 * allocated at register_netdev() time
1689 * @real_num_rx_queues: Number of RX queues currently active in device
1690 *
1691 * @rx_handler: handler for received packets
1692 * @rx_handler_data: XXX: need comments on this one
1693 * @miniq_ingress: ingress/clsact qdisc specific data for
1694 * ingress processing
1695 * @ingress_queue: XXX: need comments on this one
1696 * @broadcast: hw bcast address
1697 *
1698 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1699 * indexed by RX queue number. Assigned by driver.
1700 * This must only be set if the ndo_rx_flow_steer
1701 * operation is defined
1702 * @index_hlist: Device index hash chain
1703 *
1704 * @_tx: Array of TX queues
1705 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1706 * @real_num_tx_queues: Number of TX queues currently active in device
1707 * @qdisc: Root qdisc from userspace point of view
1708 * @tx_queue_len: Max frames per queue allowed
1709 * @tx_global_lock: XXX: need comments on this one
1710 *
1711 * @xps_maps: XXX: need comments on this one
1712 * @miniq_egress: clsact qdisc specific data for
1713 * egress processing
1714 * @watchdog_timeo: Represents the timeout that is used by
1715 * the watchdog (see dev_watchdog())
1716 * @watchdog_timer: List of timers
1717 *
1718 * @pcpu_refcnt: Number of references to this device
1719 * @todo_list: Delayed register/unregister
1720 * @link_watch_list: XXX: need comments on this one
1721 *
1722 * @reg_state: Register/unregister state machine
1723 * @dismantle: Device is going to be freed
1724 * @rtnl_link_state: This enum represents the phases of creating
1725 * a new link
1726 *
1727 * @needs_free_netdev: Should unregister perform free_netdev?
1728 * @priv_destructor: Called from unregister
1729 * @npinfo: XXX: need comments on this one
1730 * @nd_net: Network namespace this network device is inside
1731 *
1732 * @ml_priv: Mid-layer private
1733 * @lstats: Loopback statistics
1734 * @tstats: Tunnel statistics
1735 * @dstats: Dummy statistics
1736 * @vstats: Virtual ethernet statistics
1737 *
1738 * @garp_port: GARP
1739 * @mrp_port: MRP
1740 *
1741 * @dev: Class/net/name entry
1742 * @sysfs_groups: Space for optional device, statistics and wireless
1743 * sysfs groups
1744 *
1745 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1746 * @rtnl_link_ops: Rtnl_link_ops
1747 *
1748 * @gso_max_size: Maximum size of generic segmentation offload
1749 * @gso_max_segs: Maximum number of segments that can be passed to the
1750 * NIC for GSO
1751 *
1752 * @dcbnl_ops: Data Center Bridging netlink ops
1753 * @num_tc: Number of traffic classes in the net device
1754 * @tc_to_txq: XXX: need comments on this one
1755 * @prio_tc_map: XXX: need comments on this one
1756 *
1757 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1758 *
1759 * @priomap: XXX: need comments on this one
1760 * @phydev: Physical device may attach itself
1761 * for hardware timestamping
1762 * @sfp_bus: attached &struct sfp_bus structure.
1763 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1764 * spinlock
1765 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1766 * @qdisc_xmit_lock_key: lockdep class annotating
1767 * netdev_queue->_xmit_lock spinlock
1768 * @addr_list_lock_key: lockdep class annotating
1769 * net_device->addr_list_lock spinlock
1770 *
1771 * @proto_down: protocol port state information can be sent to the
1772 * switch driver and used to set the phys state of the
1773 * switch port.
1774 *
1775 * @wol_enabled: Wake-on-LAN is enabled
1776 *
1777 * FIXME: cleanup struct net_device such that network protocol info
1778 * moves out.
1779 */
1780
1781 struct net_device {
1782 char name[IFNAMSIZ];
1783 struct hlist_node name_hlist;
1784 struct dev_ifalias __rcu *ifalias;
1785 /*
1786 * I/O specific fields
1787 * FIXME: Merge these and struct ifmap into one
1788 */
1789 unsigned long mem_end;
1790 unsigned long mem_start;
1791 unsigned long base_addr;
1792 int irq;
1793
1794 /*
1795 * Some hardware also needs these fields (state,dev_list,
1796 * napi_list,unreg_list,close_list) but they are not
1797 * part of the usual set specified in Space.c.
1798 */
1799
1800 unsigned long state;
1801
1802 struct list_head dev_list;
1803 struct list_head napi_list;
1804 struct list_head unreg_list;
1805 struct list_head close_list;
1806 struct list_head ptype_all;
1807 struct list_head ptype_specific;
1808
1809 struct {
1810 struct list_head upper;
1811 struct list_head lower;
1812 } adj_list;
1813
1814 netdev_features_t features;
1815 netdev_features_t hw_features;
1816 netdev_features_t wanted_features;
1817 netdev_features_t vlan_features;
1818 netdev_features_t hw_enc_features;
1819 netdev_features_t mpls_features;
1820 netdev_features_t gso_partial_features;
1821
1822 int ifindex;
1823 int group;
1824
1825 struct net_device_stats stats;
1826
1827 atomic_long_t rx_dropped;
1828 atomic_long_t tx_dropped;
1829 atomic_long_t rx_nohandler;
1830
1831 /* Stats to monitor link on/off, flapping */
1832 atomic_t carrier_up_count;
1833 atomic_t carrier_down_count;
1834
1835 #ifdef CONFIG_WIRELESS_EXT
1836 const struct iw_handler_def *wireless_handlers;
1837 struct iw_public_data *wireless_data;
1838 #endif
1839 const struct net_device_ops *netdev_ops;
1840 const struct ethtool_ops *ethtool_ops;
1841 #ifdef CONFIG_NET_L3_MASTER_DEV
1842 const struct l3mdev_ops *l3mdev_ops;
1843 #endif
1844 #if IS_ENABLED(CONFIG_IPV6)
1845 const struct ndisc_ops *ndisc_ops;
1846 #endif
1847
1848 #ifdef CONFIG_XFRM_OFFLOAD
1849 const struct xfrmdev_ops *xfrmdev_ops;
1850 #endif
1851
1852 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1853 const struct tlsdev_ops *tlsdev_ops;
1854 #endif
1855
1856 const struct header_ops *header_ops;
1857
1858 unsigned int flags;
1859 unsigned int priv_flags;
1860
1861 unsigned short gflags;
1862 unsigned short padded;
1863
1864 unsigned char operstate;
1865 unsigned char link_mode;
1866
1867 unsigned char if_port;
1868 unsigned char dma;
1869
1870 /* Note : dev->mtu is often read without holding a lock.
1871 * Writers usually hold RTNL.
1872 * It is recommended to use READ_ONCE() to annotate the reads,
1873 * and to use WRITE_ONCE() to annotate the writes.
1874 */
1875 unsigned int mtu;
1876 unsigned int min_mtu;
1877 unsigned int max_mtu;
1878 unsigned short type;
1879 unsigned short hard_header_len;
1880 unsigned char min_header_len;
1881
1882 unsigned short needed_headroom;
1883 unsigned short needed_tailroom;
1884
1885 /* Interface address info. */
1886 unsigned char perm_addr[MAX_ADDR_LEN];
1887 unsigned char addr_assign_type;
1888 unsigned char addr_len;
1889 unsigned char upper_level;
1890 unsigned char lower_level;
1891 unsigned short neigh_priv_len;
1892 unsigned short dev_id;
1893 unsigned short dev_port;
1894 spinlock_t addr_list_lock;
1895 unsigned char name_assign_type;
1896 bool uc_promisc;
1897 struct netdev_hw_addr_list uc;
1898 struct netdev_hw_addr_list mc;
1899 struct netdev_hw_addr_list dev_addrs;
1900
1901 #ifdef CONFIG_SYSFS
1902 struct kset *queues_kset;
1903 #endif
1904 unsigned int promiscuity;
1905 unsigned int allmulti;
1906
1907
1908 /* Protocol-specific pointers */
1909
1910 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1911 struct vlan_info __rcu *vlan_info;
1912 #endif
1913 #if IS_ENABLED(CONFIG_NET_DSA)
1914 struct dsa_port *dsa_ptr;
1915 #endif
1916 #if IS_ENABLED(CONFIG_TIPC)
1917 struct tipc_bearer __rcu *tipc_ptr;
1918 #endif
1919 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1920 void *atalk_ptr;
1921 #endif
1922 struct in_device __rcu *ip_ptr;
1923 #if IS_ENABLED(CONFIG_DECNET)
1924 struct dn_dev __rcu *dn_ptr;
1925 #endif
1926 struct inet6_dev __rcu *ip6_ptr;
1927 #if IS_ENABLED(CONFIG_AX25)
1928 void *ax25_ptr;
1929 #endif
1930 struct wireless_dev *ieee80211_ptr;
1931 struct wpan_dev *ieee802154_ptr;
1932 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1933 struct mpls_dev __rcu *mpls_ptr;
1934 #endif
1935
1936 /*
1937 * Cache lines mostly used on receive path (including eth_type_trans())
1938 */
1939 /* Interface address info used in eth_type_trans() */
1940 unsigned char *dev_addr;
1941
1942 struct netdev_rx_queue *_rx;
1943 unsigned int num_rx_queues;
1944 unsigned int real_num_rx_queues;
1945
1946 struct bpf_prog __rcu *xdp_prog;
1947 unsigned long gro_flush_timeout;
1948 rx_handler_func_t __rcu *rx_handler;
1949 void __rcu *rx_handler_data;
1950
1951 #ifdef CONFIG_NET_CLS_ACT
1952 struct mini_Qdisc __rcu *miniq_ingress;
1953 #endif
1954 struct netdev_queue __rcu *ingress_queue;
1955 #ifdef CONFIG_NETFILTER_INGRESS
1956 struct nf_hook_entries __rcu *nf_hooks_ingress;
1957 #endif
1958
1959 unsigned char broadcast[MAX_ADDR_LEN];
1960 #ifdef CONFIG_RFS_ACCEL
1961 struct cpu_rmap *rx_cpu_rmap;
1962 #endif
1963 struct hlist_node index_hlist;
1964
1965 /*
1966 * Cache lines mostly used on transmit path
1967 */
1968 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1969 unsigned int num_tx_queues;
1970 unsigned int real_num_tx_queues;
1971 struct Qdisc *qdisc;
1972 #ifdef CONFIG_NET_SCHED
1973 DECLARE_HASHTABLE (qdisc_hash, 4);
1974 #endif
1975 unsigned int tx_queue_len;
1976 spinlock_t tx_global_lock;
1977 int watchdog_timeo;
1978
1979 #ifdef CONFIG_XPS
1980 struct xps_dev_maps __rcu *xps_cpus_map;
1981 struct xps_dev_maps __rcu *xps_rxqs_map;
1982 #endif
1983 #ifdef CONFIG_NET_CLS_ACT
1984 struct mini_Qdisc __rcu *miniq_egress;
1985 #endif
1986
1987 /* These may be needed for future network-power-down code. */
1988 struct timer_list watchdog_timer;
1989
1990 int __percpu *pcpu_refcnt;
1991 struct list_head todo_list;
1992
1993 struct list_head link_watch_list;
1994
1995 enum { NETREG_UNINITIALIZED=0,
1996 NETREG_REGISTERED, /* completed register_netdevice */
1997 NETREG_UNREGISTERING, /* called unregister_netdevice */
1998 NETREG_UNREGISTERED, /* completed unregister todo */
1999 NETREG_RELEASED, /* called free_netdev */
2000 NETREG_DUMMY, /* dummy device for NAPI poll */
2001 } reg_state:8;
2002
2003 bool dismantle;
2004
2005 enum {
2006 RTNL_LINK_INITIALIZED,
2007 RTNL_LINK_INITIALIZING,
2008 } rtnl_link_state:16;
2009
2010 bool needs_free_netdev;
2011 void (*priv_destructor)(struct net_device *dev);
2012
2013 #ifdef CONFIG_NETPOLL
2014 struct netpoll_info __rcu *npinfo;
2015 #endif
2016
2017 possible_net_t nd_net;
2018
2019 /* mid-layer private */
2020 union {
2021 void *ml_priv;
2022 struct pcpu_lstats __percpu *lstats;
2023 struct pcpu_sw_netstats __percpu *tstats;
2024 struct pcpu_dstats __percpu *dstats;
2025 };
2026
2027 #if IS_ENABLED(CONFIG_GARP)
2028 struct garp_port __rcu *garp_port;
2029 #endif
2030 #if IS_ENABLED(CONFIG_MRP)
2031 struct mrp_port __rcu *mrp_port;
2032 #endif
2033
2034 struct device dev;
2035 const struct attribute_group *sysfs_groups[4];
2036 const struct attribute_group *sysfs_rx_queue_group;
2037
2038 const struct rtnl_link_ops *rtnl_link_ops;
2039
2040 /* for setting kernel sock attribute on TCP connection setup */
2041 #define GSO_MAX_SIZE 65536
2042 unsigned int gso_max_size;
2043 #define GSO_MAX_SEGS 65535
2044 u16 gso_max_segs;
2045
2046 #ifdef CONFIG_DCB
2047 const struct dcbnl_rtnl_ops *dcbnl_ops;
2048 #endif
2049 s16 num_tc;
2050 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2051 u8 prio_tc_map[TC_BITMASK + 1];
2052
2053 #if IS_ENABLED(CONFIG_FCOE)
2054 unsigned int fcoe_ddp_xid;
2055 #endif
2056 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2057 struct netprio_map __rcu *priomap;
2058 #endif
2059 struct phy_device *phydev;
2060 struct sfp_bus *sfp_bus;
2061 struct lock_class_key qdisc_tx_busylock_key;
2062 struct lock_class_key qdisc_running_key;
2063 struct lock_class_key qdisc_xmit_lock_key;
2064 struct lock_class_key addr_list_lock_key;
2065 bool proto_down;
2066 unsigned wol_enabled:1;
2067 };
2068 #define to_net_dev(d) container_of(d, struct net_device, dev)
2069
netif_elide_gro(const struct net_device * dev)2070 static inline bool netif_elide_gro(const struct net_device *dev)
2071 {
2072 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2073 return true;
2074 return false;
2075 }
2076
2077 #define NETDEV_ALIGN 32
2078
2079 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2080 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2081 {
2082 return dev->prio_tc_map[prio & TC_BITMASK];
2083 }
2084
2085 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2086 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2087 {
2088 if (tc >= dev->num_tc)
2089 return -EINVAL;
2090
2091 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2092 return 0;
2093 }
2094
2095 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2096 void netdev_reset_tc(struct net_device *dev);
2097 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2098 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2099
2100 static inline
netdev_get_num_tc(struct net_device * dev)2101 int netdev_get_num_tc(struct net_device *dev)
2102 {
2103 return dev->num_tc;
2104 }
2105
2106 void netdev_unbind_sb_channel(struct net_device *dev,
2107 struct net_device *sb_dev);
2108 int netdev_bind_sb_channel_queue(struct net_device *dev,
2109 struct net_device *sb_dev,
2110 u8 tc, u16 count, u16 offset);
2111 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2112 static inline int netdev_get_sb_channel(struct net_device *dev)
2113 {
2114 return max_t(int, -dev->num_tc, 0);
2115 }
2116
2117 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2118 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2119 unsigned int index)
2120 {
2121 return &dev->_tx[index];
2122 }
2123
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2124 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2125 const struct sk_buff *skb)
2126 {
2127 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2128 }
2129
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2130 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2131 void (*f)(struct net_device *,
2132 struct netdev_queue *,
2133 void *),
2134 void *arg)
2135 {
2136 unsigned int i;
2137
2138 for (i = 0; i < dev->num_tx_queues; i++)
2139 f(dev, &dev->_tx[i], arg);
2140 }
2141
2142 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2143 struct net_device *sb_dev);
2144 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2145 struct sk_buff *skb,
2146 struct net_device *sb_dev);
2147
2148 /* returns the headroom that the master device needs to take in account
2149 * when forwarding to this dev
2150 */
netdev_get_fwd_headroom(struct net_device * dev)2151 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2152 {
2153 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2154 }
2155
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2156 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2157 {
2158 if (dev->netdev_ops->ndo_set_rx_headroom)
2159 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2160 }
2161
2162 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2163 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2164 {
2165 netdev_set_rx_headroom(dev, -1);
2166 }
2167
2168 /*
2169 * Net namespace inlines
2170 */
2171 static inline
dev_net(const struct net_device * dev)2172 struct net *dev_net(const struct net_device *dev)
2173 {
2174 return read_pnet(&dev->nd_net);
2175 }
2176
2177 static inline
dev_net_set(struct net_device * dev,struct net * net)2178 void dev_net_set(struct net_device *dev, struct net *net)
2179 {
2180 write_pnet(&dev->nd_net, net);
2181 }
2182
2183 /**
2184 * netdev_priv - access network device private data
2185 * @dev: network device
2186 *
2187 * Get network device private data
2188 */
netdev_priv(const struct net_device * dev)2189 static inline void *netdev_priv(const struct net_device *dev)
2190 {
2191 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2192 }
2193
2194 /* Set the sysfs physical device reference for the network logical device
2195 * if set prior to registration will cause a symlink during initialization.
2196 */
2197 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2198
2199 /* Set the sysfs device type for the network logical device to allow
2200 * fine-grained identification of different network device types. For
2201 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2202 */
2203 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2204
2205 /* Default NAPI poll() weight
2206 * Device drivers are strongly advised to not use bigger value
2207 */
2208 #define NAPI_POLL_WEIGHT 64
2209
2210 /**
2211 * netif_napi_add - initialize a NAPI context
2212 * @dev: network device
2213 * @napi: NAPI context
2214 * @poll: polling function
2215 * @weight: default weight
2216 *
2217 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2218 * *any* of the other NAPI-related functions.
2219 */
2220 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2221 int (*poll)(struct napi_struct *, int), int weight);
2222
2223 /**
2224 * netif_tx_napi_add - initialize a NAPI context
2225 * @dev: network device
2226 * @napi: NAPI context
2227 * @poll: polling function
2228 * @weight: default weight
2229 *
2230 * This variant of netif_napi_add() should be used from drivers using NAPI
2231 * to exclusively poll a TX queue.
2232 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2233 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2234 static inline void netif_tx_napi_add(struct net_device *dev,
2235 struct napi_struct *napi,
2236 int (*poll)(struct napi_struct *, int),
2237 int weight)
2238 {
2239 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2240 netif_napi_add(dev, napi, poll, weight);
2241 }
2242
2243 /**
2244 * netif_napi_del - remove a NAPI context
2245 * @napi: NAPI context
2246 *
2247 * netif_napi_del() removes a NAPI context from the network device NAPI list
2248 */
2249 void netif_napi_del(struct napi_struct *napi);
2250
2251 struct napi_gro_cb {
2252 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2253 void *frag0;
2254
2255 /* Length of frag0. */
2256 unsigned int frag0_len;
2257
2258 /* This indicates where we are processing relative to skb->data. */
2259 int data_offset;
2260
2261 /* This is non-zero if the packet cannot be merged with the new skb. */
2262 u16 flush;
2263
2264 /* Save the IP ID here and check when we get to the transport layer */
2265 u16 flush_id;
2266
2267 /* Number of segments aggregated. */
2268 u16 count;
2269
2270 /* Start offset for remote checksum offload */
2271 u16 gro_remcsum_start;
2272
2273 /* jiffies when first packet was created/queued */
2274 unsigned long age;
2275
2276 /* Used in ipv6_gro_receive() and foo-over-udp */
2277 u16 proto;
2278
2279 /* This is non-zero if the packet may be of the same flow. */
2280 u8 same_flow:1;
2281
2282 /* Used in tunnel GRO receive */
2283 u8 encap_mark:1;
2284
2285 /* GRO checksum is valid */
2286 u8 csum_valid:1;
2287
2288 /* Number of checksums via CHECKSUM_UNNECESSARY */
2289 u8 csum_cnt:3;
2290
2291 /* Free the skb? */
2292 u8 free:2;
2293 #define NAPI_GRO_FREE 1
2294 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2295
2296 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2297 u8 is_ipv6:1;
2298
2299 /* Used in GRE, set in fou/gue_gro_receive */
2300 u8 is_fou:1;
2301
2302 /* Used to determine if flush_id can be ignored */
2303 u8 is_atomic:1;
2304
2305 /* Number of gro_receive callbacks this packet already went through */
2306 u8 recursion_counter:4;
2307
2308 /* 1 bit hole */
2309
2310 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2311 __wsum csum;
2312
2313 /* used in skb_gro_receive() slow path */
2314 struct sk_buff *last;
2315 };
2316
2317 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2318
2319 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2320 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2321 {
2322 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2323 }
2324
2325 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
call_gro_receive(gro_receive_t cb,struct list_head * head,struct sk_buff * skb)2326 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2327 struct list_head *head,
2328 struct sk_buff *skb)
2329 {
2330 if (unlikely(gro_recursion_inc_test(skb))) {
2331 NAPI_GRO_CB(skb)->flush |= 1;
2332 return NULL;
2333 }
2334
2335 return cb(head, skb);
2336 }
2337
2338 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2339 struct sk_buff *);
call_gro_receive_sk(gro_receive_sk_t cb,struct sock * sk,struct list_head * head,struct sk_buff * skb)2340 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2341 struct sock *sk,
2342 struct list_head *head,
2343 struct sk_buff *skb)
2344 {
2345 if (unlikely(gro_recursion_inc_test(skb))) {
2346 NAPI_GRO_CB(skb)->flush |= 1;
2347 return NULL;
2348 }
2349
2350 return cb(sk, head, skb);
2351 }
2352
2353 struct packet_type {
2354 __be16 type; /* This is really htons(ether_type). */
2355 bool ignore_outgoing;
2356 struct net_device *dev; /* NULL is wildcarded here */
2357 int (*func) (struct sk_buff *,
2358 struct net_device *,
2359 struct packet_type *,
2360 struct net_device *);
2361 void (*list_func) (struct list_head *,
2362 struct packet_type *,
2363 struct net_device *);
2364 bool (*id_match)(struct packet_type *ptype,
2365 struct sock *sk);
2366 void *af_packet_priv;
2367 struct list_head list;
2368 };
2369
2370 struct offload_callbacks {
2371 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2372 netdev_features_t features);
2373 struct sk_buff *(*gro_receive)(struct list_head *head,
2374 struct sk_buff *skb);
2375 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2376 };
2377
2378 struct packet_offload {
2379 __be16 type; /* This is really htons(ether_type). */
2380 u16 priority;
2381 struct offload_callbacks callbacks;
2382 struct list_head list;
2383 };
2384
2385 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2386 struct pcpu_sw_netstats {
2387 u64 rx_packets;
2388 u64 rx_bytes;
2389 u64 tx_packets;
2390 u64 tx_bytes;
2391 struct u64_stats_sync syncp;
2392 } __aligned(4 * sizeof(u64));
2393
2394 struct pcpu_lstats {
2395 u64 packets;
2396 u64 bytes;
2397 struct u64_stats_sync syncp;
2398 } __aligned(2 * sizeof(u64));
2399
2400 #define __netdev_alloc_pcpu_stats(type, gfp) \
2401 ({ \
2402 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2403 if (pcpu_stats) { \
2404 int __cpu; \
2405 for_each_possible_cpu(__cpu) { \
2406 typeof(type) *stat; \
2407 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2408 u64_stats_init(&stat->syncp); \
2409 } \
2410 } \
2411 pcpu_stats; \
2412 })
2413
2414 #define netdev_alloc_pcpu_stats(type) \
2415 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2416
2417 enum netdev_lag_tx_type {
2418 NETDEV_LAG_TX_TYPE_UNKNOWN,
2419 NETDEV_LAG_TX_TYPE_RANDOM,
2420 NETDEV_LAG_TX_TYPE_BROADCAST,
2421 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2422 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2423 NETDEV_LAG_TX_TYPE_HASH,
2424 };
2425
2426 enum netdev_lag_hash {
2427 NETDEV_LAG_HASH_NONE,
2428 NETDEV_LAG_HASH_L2,
2429 NETDEV_LAG_HASH_L34,
2430 NETDEV_LAG_HASH_L23,
2431 NETDEV_LAG_HASH_E23,
2432 NETDEV_LAG_HASH_E34,
2433 NETDEV_LAG_HASH_UNKNOWN,
2434 };
2435
2436 struct netdev_lag_upper_info {
2437 enum netdev_lag_tx_type tx_type;
2438 enum netdev_lag_hash hash_type;
2439 };
2440
2441 struct netdev_lag_lower_state_info {
2442 u8 link_up : 1,
2443 tx_enabled : 1;
2444 };
2445
2446 #include <linux/notifier.h>
2447
2448 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2449 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2450 * adding new types.
2451 */
2452 enum netdev_cmd {
2453 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2454 NETDEV_DOWN,
2455 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2456 detected a hardware crash and restarted
2457 - we can use this eg to kick tcp sessions
2458 once done */
2459 NETDEV_CHANGE, /* Notify device state change */
2460 NETDEV_REGISTER,
2461 NETDEV_UNREGISTER,
2462 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2463 NETDEV_CHANGEADDR, /* notify after the address change */
2464 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2465 NETDEV_GOING_DOWN,
2466 NETDEV_CHANGENAME,
2467 NETDEV_FEAT_CHANGE,
2468 NETDEV_BONDING_FAILOVER,
2469 NETDEV_PRE_UP,
2470 NETDEV_PRE_TYPE_CHANGE,
2471 NETDEV_POST_TYPE_CHANGE,
2472 NETDEV_POST_INIT,
2473 NETDEV_RELEASE,
2474 NETDEV_NOTIFY_PEERS,
2475 NETDEV_JOIN,
2476 NETDEV_CHANGEUPPER,
2477 NETDEV_RESEND_IGMP,
2478 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2479 NETDEV_CHANGEINFODATA,
2480 NETDEV_BONDING_INFO,
2481 NETDEV_PRECHANGEUPPER,
2482 NETDEV_CHANGELOWERSTATE,
2483 NETDEV_UDP_TUNNEL_PUSH_INFO,
2484 NETDEV_UDP_TUNNEL_DROP_INFO,
2485 NETDEV_CHANGE_TX_QUEUE_LEN,
2486 NETDEV_CVLAN_FILTER_PUSH_INFO,
2487 NETDEV_CVLAN_FILTER_DROP_INFO,
2488 NETDEV_SVLAN_FILTER_PUSH_INFO,
2489 NETDEV_SVLAN_FILTER_DROP_INFO,
2490 };
2491 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2492
2493 int register_netdevice_notifier(struct notifier_block *nb);
2494 int unregister_netdevice_notifier(struct notifier_block *nb);
2495
2496 struct netdev_notifier_info {
2497 struct net_device *dev;
2498 struct netlink_ext_ack *extack;
2499 };
2500
2501 struct netdev_notifier_info_ext {
2502 struct netdev_notifier_info info; /* must be first */
2503 union {
2504 u32 mtu;
2505 } ext;
2506 };
2507
2508 struct netdev_notifier_change_info {
2509 struct netdev_notifier_info info; /* must be first */
2510 unsigned int flags_changed;
2511 };
2512
2513 struct netdev_notifier_changeupper_info {
2514 struct netdev_notifier_info info; /* must be first */
2515 struct net_device *upper_dev; /* new upper dev */
2516 bool master; /* is upper dev master */
2517 bool linking; /* is the notification for link or unlink */
2518 void *upper_info; /* upper dev info */
2519 };
2520
2521 struct netdev_notifier_changelowerstate_info {
2522 struct netdev_notifier_info info; /* must be first */
2523 void *lower_state_info; /* is lower dev state */
2524 };
2525
2526 struct netdev_notifier_pre_changeaddr_info {
2527 struct netdev_notifier_info info; /* must be first */
2528 const unsigned char *dev_addr;
2529 };
2530
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2531 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2532 struct net_device *dev)
2533 {
2534 info->dev = dev;
2535 info->extack = NULL;
2536 }
2537
2538 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2539 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2540 {
2541 return info->dev;
2542 }
2543
2544 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2545 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2546 {
2547 return info->extack;
2548 }
2549
2550 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2551
2552
2553 extern rwlock_t dev_base_lock; /* Device list lock */
2554
2555 #define for_each_netdev(net, d) \
2556 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2557 #define for_each_netdev_reverse(net, d) \
2558 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2559 #define for_each_netdev_rcu(net, d) \
2560 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2561 #define for_each_netdev_safe(net, d, n) \
2562 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2563 #define for_each_netdev_continue(net, d) \
2564 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2565 #define for_each_netdev_continue_rcu(net, d) \
2566 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2567 #define for_each_netdev_in_bond_rcu(bond, slave) \
2568 for_each_netdev_rcu(&init_net, slave) \
2569 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2570 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2571
next_net_device(struct net_device * dev)2572 static inline struct net_device *next_net_device(struct net_device *dev)
2573 {
2574 struct list_head *lh;
2575 struct net *net;
2576
2577 net = dev_net(dev);
2578 lh = dev->dev_list.next;
2579 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2580 }
2581
next_net_device_rcu(struct net_device * dev)2582 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2583 {
2584 struct list_head *lh;
2585 struct net *net;
2586
2587 net = dev_net(dev);
2588 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2589 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2590 }
2591
first_net_device(struct net * net)2592 static inline struct net_device *first_net_device(struct net *net)
2593 {
2594 return list_empty(&net->dev_base_head) ? NULL :
2595 net_device_entry(net->dev_base_head.next);
2596 }
2597
first_net_device_rcu(struct net * net)2598 static inline struct net_device *first_net_device_rcu(struct net *net)
2599 {
2600 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2601
2602 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2603 }
2604
2605 int netdev_boot_setup_check(struct net_device *dev);
2606 unsigned long netdev_boot_base(const char *prefix, int unit);
2607 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2608 const char *hwaddr);
2609 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2610 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2611 void dev_add_pack(struct packet_type *pt);
2612 void dev_remove_pack(struct packet_type *pt);
2613 void __dev_remove_pack(struct packet_type *pt);
2614 void dev_add_offload(struct packet_offload *po);
2615 void dev_remove_offload(struct packet_offload *po);
2616
2617 int dev_get_iflink(const struct net_device *dev);
2618 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2619 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2620 unsigned short mask);
2621 struct net_device *dev_get_by_name(struct net *net, const char *name);
2622 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2623 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2624 int dev_alloc_name(struct net_device *dev, const char *name);
2625 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2626 void dev_close(struct net_device *dev);
2627 void dev_close_many(struct list_head *head, bool unlink);
2628 void dev_disable_lro(struct net_device *dev);
2629 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2630 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2631 struct net_device *sb_dev);
2632 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2633 struct net_device *sb_dev);
2634 int dev_queue_xmit(struct sk_buff *skb);
2635 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2636 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2637 int register_netdevice(struct net_device *dev);
2638 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2639 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2640 static inline void unregister_netdevice(struct net_device *dev)
2641 {
2642 unregister_netdevice_queue(dev, NULL);
2643 }
2644
2645 int netdev_refcnt_read(const struct net_device *dev);
2646 void free_netdev(struct net_device *dev);
2647 void netdev_freemem(struct net_device *dev);
2648 void synchronize_net(void);
2649 int init_dummy_netdev(struct net_device *dev);
2650
2651 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2652 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2653 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2654 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2655 int netdev_get_name(struct net *net, char *name, int ifindex);
2656 int dev_restart(struct net_device *dev);
2657 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2658
skb_gro_offset(const struct sk_buff * skb)2659 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2660 {
2661 return NAPI_GRO_CB(skb)->data_offset;
2662 }
2663
skb_gro_len(const struct sk_buff * skb)2664 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2665 {
2666 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2667 }
2668
skb_gro_pull(struct sk_buff * skb,unsigned int len)2669 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2670 {
2671 NAPI_GRO_CB(skb)->data_offset += len;
2672 }
2673
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2674 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2675 unsigned int offset)
2676 {
2677 return NAPI_GRO_CB(skb)->frag0 + offset;
2678 }
2679
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2680 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2681 {
2682 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2683 }
2684
skb_gro_frag0_invalidate(struct sk_buff * skb)2685 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2686 {
2687 NAPI_GRO_CB(skb)->frag0 = NULL;
2688 NAPI_GRO_CB(skb)->frag0_len = 0;
2689 }
2690
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2691 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2692 unsigned int offset)
2693 {
2694 if (!pskb_may_pull(skb, hlen))
2695 return NULL;
2696
2697 skb_gro_frag0_invalidate(skb);
2698 return skb->data + offset;
2699 }
2700
skb_gro_network_header(struct sk_buff * skb)2701 static inline void *skb_gro_network_header(struct sk_buff *skb)
2702 {
2703 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2704 skb_network_offset(skb);
2705 }
2706
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2707 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2708 const void *start, unsigned int len)
2709 {
2710 if (NAPI_GRO_CB(skb)->csum_valid)
2711 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2712 csum_partial(start, len, 0));
2713 }
2714
2715 /* GRO checksum functions. These are logical equivalents of the normal
2716 * checksum functions (in skbuff.h) except that they operate on the GRO
2717 * offsets and fields in sk_buff.
2718 */
2719
2720 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2721
skb_at_gro_remcsum_start(struct sk_buff * skb)2722 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2723 {
2724 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2725 }
2726
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2727 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2728 bool zero_okay,
2729 __sum16 check)
2730 {
2731 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2732 skb_checksum_start_offset(skb) <
2733 skb_gro_offset(skb)) &&
2734 !skb_at_gro_remcsum_start(skb) &&
2735 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2736 (!zero_okay || check));
2737 }
2738
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2739 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2740 __wsum psum)
2741 {
2742 if (NAPI_GRO_CB(skb)->csum_valid &&
2743 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2744 return 0;
2745
2746 NAPI_GRO_CB(skb)->csum = psum;
2747
2748 return __skb_gro_checksum_complete(skb);
2749 }
2750
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)2751 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2752 {
2753 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2754 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2755 NAPI_GRO_CB(skb)->csum_cnt--;
2756 } else {
2757 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2758 * verified a new top level checksum or an encapsulated one
2759 * during GRO. This saves work if we fallback to normal path.
2760 */
2761 __skb_incr_checksum_unnecessary(skb);
2762 }
2763 }
2764
2765 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2766 compute_pseudo) \
2767 ({ \
2768 __sum16 __ret = 0; \
2769 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2770 __ret = __skb_gro_checksum_validate_complete(skb, \
2771 compute_pseudo(skb, proto)); \
2772 if (!__ret) \
2773 skb_gro_incr_csum_unnecessary(skb); \
2774 __ret; \
2775 })
2776
2777 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2778 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2779
2780 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2781 compute_pseudo) \
2782 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2783
2784 #define skb_gro_checksum_simple_validate(skb) \
2785 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2786
__skb_gro_checksum_convert_check(struct sk_buff * skb)2787 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2788 {
2789 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2790 !NAPI_GRO_CB(skb)->csum_valid);
2791 }
2792
__skb_gro_checksum_convert(struct sk_buff * skb,__sum16 check,__wsum pseudo)2793 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2794 __sum16 check, __wsum pseudo)
2795 {
2796 NAPI_GRO_CB(skb)->csum = ~pseudo;
2797 NAPI_GRO_CB(skb)->csum_valid = 1;
2798 }
2799
2800 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2801 do { \
2802 if (__skb_gro_checksum_convert_check(skb)) \
2803 __skb_gro_checksum_convert(skb, check, \
2804 compute_pseudo(skb, proto)); \
2805 } while (0)
2806
2807 struct gro_remcsum {
2808 int offset;
2809 __wsum delta;
2810 };
2811
skb_gro_remcsum_init(struct gro_remcsum * grc)2812 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2813 {
2814 grc->offset = 0;
2815 grc->delta = 0;
2816 }
2817
skb_gro_remcsum_process(struct sk_buff * skb,void * ptr,unsigned int off,size_t hdrlen,int start,int offset,struct gro_remcsum * grc,bool nopartial)2818 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2819 unsigned int off, size_t hdrlen,
2820 int start, int offset,
2821 struct gro_remcsum *grc,
2822 bool nopartial)
2823 {
2824 __wsum delta;
2825 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2826
2827 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2828
2829 if (!nopartial) {
2830 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2831 return ptr;
2832 }
2833
2834 ptr = skb_gro_header_fast(skb, off);
2835 if (skb_gro_header_hard(skb, off + plen)) {
2836 ptr = skb_gro_header_slow(skb, off + plen, off);
2837 if (!ptr)
2838 return NULL;
2839 }
2840
2841 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2842 start, offset);
2843
2844 /* Adjust skb->csum since we changed the packet */
2845 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2846
2847 grc->offset = off + hdrlen + offset;
2848 grc->delta = delta;
2849
2850 return ptr;
2851 }
2852
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)2853 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2854 struct gro_remcsum *grc)
2855 {
2856 void *ptr;
2857 size_t plen = grc->offset + sizeof(u16);
2858
2859 if (!grc->delta)
2860 return;
2861
2862 ptr = skb_gro_header_fast(skb, grc->offset);
2863 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2864 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2865 if (!ptr)
2866 return;
2867 }
2868
2869 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2870 }
2871
2872 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)2873 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2874 {
2875 if (PTR_ERR(pp) != -EINPROGRESS)
2876 NAPI_GRO_CB(skb)->flush |= flush;
2877 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)2878 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2879 struct sk_buff *pp,
2880 int flush,
2881 struct gro_remcsum *grc)
2882 {
2883 if (PTR_ERR(pp) != -EINPROGRESS) {
2884 NAPI_GRO_CB(skb)->flush |= flush;
2885 skb_gro_remcsum_cleanup(skb, grc);
2886 skb->remcsum_offload = 0;
2887 }
2888 }
2889 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)2890 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2891 {
2892 NAPI_GRO_CB(skb)->flush |= flush;
2893 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)2894 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2895 struct sk_buff *pp,
2896 int flush,
2897 struct gro_remcsum *grc)
2898 {
2899 NAPI_GRO_CB(skb)->flush |= flush;
2900 skb_gro_remcsum_cleanup(skb, grc);
2901 skb->remcsum_offload = 0;
2902 }
2903 #endif
2904
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)2905 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2906 unsigned short type,
2907 const void *daddr, const void *saddr,
2908 unsigned int len)
2909 {
2910 if (!dev->header_ops || !dev->header_ops->create)
2911 return 0;
2912
2913 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2914 }
2915
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)2916 static inline int dev_parse_header(const struct sk_buff *skb,
2917 unsigned char *haddr)
2918 {
2919 const struct net_device *dev = skb->dev;
2920
2921 if (!dev->header_ops || !dev->header_ops->parse)
2922 return 0;
2923 return dev->header_ops->parse(skb, haddr);
2924 }
2925
dev_parse_header_protocol(const struct sk_buff * skb)2926 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2927 {
2928 const struct net_device *dev = skb->dev;
2929
2930 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2931 return 0;
2932 return dev->header_ops->parse_protocol(skb);
2933 }
2934
2935 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)2936 static inline bool dev_validate_header(const struct net_device *dev,
2937 char *ll_header, int len)
2938 {
2939 if (likely(len >= dev->hard_header_len))
2940 return true;
2941 if (len < dev->min_header_len)
2942 return false;
2943
2944 if (capable(CAP_SYS_RAWIO)) {
2945 memset(ll_header + len, 0, dev->hard_header_len - len);
2946 return true;
2947 }
2948
2949 if (dev->header_ops && dev->header_ops->validate)
2950 return dev->header_ops->validate(ll_header, len);
2951
2952 return false;
2953 }
2954
2955 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2956 int len, int size);
2957 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
unregister_gifconf(unsigned int family)2958 static inline int unregister_gifconf(unsigned int family)
2959 {
2960 return register_gifconf(family, NULL);
2961 }
2962
2963 #ifdef CONFIG_NET_FLOW_LIMIT
2964 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2965 struct sd_flow_limit {
2966 u64 count;
2967 unsigned int num_buckets;
2968 unsigned int history_head;
2969 u16 history[FLOW_LIMIT_HISTORY];
2970 u8 buckets[];
2971 };
2972
2973 extern int netdev_flow_limit_table_len;
2974 #endif /* CONFIG_NET_FLOW_LIMIT */
2975
2976 /*
2977 * Incoming packets are placed on per-CPU queues
2978 */
2979 struct softnet_data {
2980 struct list_head poll_list;
2981 struct sk_buff_head process_queue;
2982
2983 /* stats */
2984 unsigned int processed;
2985 unsigned int time_squeeze;
2986 unsigned int received_rps;
2987 #ifdef CONFIG_RPS
2988 struct softnet_data *rps_ipi_list;
2989 #endif
2990 #ifdef CONFIG_NET_FLOW_LIMIT
2991 struct sd_flow_limit __rcu *flow_limit;
2992 #endif
2993 struct Qdisc *output_queue;
2994 struct Qdisc **output_queue_tailp;
2995 struct sk_buff *completion_queue;
2996 #ifdef CONFIG_XFRM_OFFLOAD
2997 struct sk_buff_head xfrm_backlog;
2998 #endif
2999 /* written and read only by owning cpu: */
3000 struct {
3001 u16 recursion;
3002 u8 more;
3003 } xmit;
3004 #ifdef CONFIG_RPS
3005 /* input_queue_head should be written by cpu owning this struct,
3006 * and only read by other cpus. Worth using a cache line.
3007 */
3008 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3009
3010 /* Elements below can be accessed between CPUs for RPS/RFS */
3011 call_single_data_t csd ____cacheline_aligned_in_smp;
3012 struct softnet_data *rps_ipi_next;
3013 unsigned int cpu;
3014 unsigned int input_queue_tail;
3015 #endif
3016 unsigned int dropped;
3017 struct sk_buff_head input_pkt_queue;
3018 struct napi_struct backlog;
3019
3020 };
3021
input_queue_head_incr(struct softnet_data * sd)3022 static inline void input_queue_head_incr(struct softnet_data *sd)
3023 {
3024 #ifdef CONFIG_RPS
3025 sd->input_queue_head++;
3026 #endif
3027 }
3028
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3029 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3030 unsigned int *qtail)
3031 {
3032 #ifdef CONFIG_RPS
3033 *qtail = ++sd->input_queue_tail;
3034 #endif
3035 }
3036
3037 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3038
dev_recursion_level(void)3039 static inline int dev_recursion_level(void)
3040 {
3041 return this_cpu_read(softnet_data.xmit.recursion);
3042 }
3043
3044 #define XMIT_RECURSION_LIMIT 10
dev_xmit_recursion(void)3045 static inline bool dev_xmit_recursion(void)
3046 {
3047 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3048 XMIT_RECURSION_LIMIT);
3049 }
3050
dev_xmit_recursion_inc(void)3051 static inline void dev_xmit_recursion_inc(void)
3052 {
3053 __this_cpu_inc(softnet_data.xmit.recursion);
3054 }
3055
dev_xmit_recursion_dec(void)3056 static inline void dev_xmit_recursion_dec(void)
3057 {
3058 __this_cpu_dec(softnet_data.xmit.recursion);
3059 }
3060
3061 void __netif_schedule(struct Qdisc *q);
3062 void netif_schedule_queue(struct netdev_queue *txq);
3063
netif_tx_schedule_all(struct net_device * dev)3064 static inline void netif_tx_schedule_all(struct net_device *dev)
3065 {
3066 unsigned int i;
3067
3068 for (i = 0; i < dev->num_tx_queues; i++)
3069 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3070 }
3071
netif_tx_start_queue(struct netdev_queue * dev_queue)3072 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3073 {
3074 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3075 }
3076
3077 /**
3078 * netif_start_queue - allow transmit
3079 * @dev: network device
3080 *
3081 * Allow upper layers to call the device hard_start_xmit routine.
3082 */
netif_start_queue(struct net_device * dev)3083 static inline void netif_start_queue(struct net_device *dev)
3084 {
3085 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3086 }
3087
netif_tx_start_all_queues(struct net_device * dev)3088 static inline void netif_tx_start_all_queues(struct net_device *dev)
3089 {
3090 unsigned int i;
3091
3092 for (i = 0; i < dev->num_tx_queues; i++) {
3093 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3094 netif_tx_start_queue(txq);
3095 }
3096 }
3097
3098 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3099
3100 /**
3101 * netif_wake_queue - restart transmit
3102 * @dev: network device
3103 *
3104 * Allow upper layers to call the device hard_start_xmit routine.
3105 * Used for flow control when transmit resources are available.
3106 */
netif_wake_queue(struct net_device * dev)3107 static inline void netif_wake_queue(struct net_device *dev)
3108 {
3109 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3110 }
3111
netif_tx_wake_all_queues(struct net_device * dev)3112 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3113 {
3114 unsigned int i;
3115
3116 for (i = 0; i < dev->num_tx_queues; i++) {
3117 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3118 netif_tx_wake_queue(txq);
3119 }
3120 }
3121
netif_tx_stop_queue(struct netdev_queue * dev_queue)3122 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3123 {
3124 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3125 }
3126
3127 /**
3128 * netif_stop_queue - stop transmitted packets
3129 * @dev: network device
3130 *
3131 * Stop upper layers calling the device hard_start_xmit routine.
3132 * Used for flow control when transmit resources are unavailable.
3133 */
netif_stop_queue(struct net_device * dev)3134 static inline void netif_stop_queue(struct net_device *dev)
3135 {
3136 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3137 }
3138
3139 void netif_tx_stop_all_queues(struct net_device *dev);
3140 void netdev_update_lockdep_key(struct net_device *dev);
3141
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3142 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3143 {
3144 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3145 }
3146
3147 /**
3148 * netif_queue_stopped - test if transmit queue is flowblocked
3149 * @dev: network device
3150 *
3151 * Test if transmit queue on device is currently unable to send.
3152 */
netif_queue_stopped(const struct net_device * dev)3153 static inline bool netif_queue_stopped(const struct net_device *dev)
3154 {
3155 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3156 }
3157
netif_xmit_stopped(const struct netdev_queue * dev_queue)3158 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3159 {
3160 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3161 }
3162
3163 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3164 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3165 {
3166 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3167 }
3168
3169 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3170 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3171 {
3172 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3173 }
3174
3175 /**
3176 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3177 * @dev_queue: pointer to transmit queue
3178 *
3179 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3180 * to give appropriate hint to the CPU.
3181 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3182 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3183 {
3184 #ifdef CONFIG_BQL
3185 prefetchw(&dev_queue->dql.num_queued);
3186 #endif
3187 }
3188
3189 /**
3190 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3191 * @dev_queue: pointer to transmit queue
3192 *
3193 * BQL enabled drivers might use this helper in their TX completion path,
3194 * to give appropriate hint to the CPU.
3195 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3196 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3197 {
3198 #ifdef CONFIG_BQL
3199 prefetchw(&dev_queue->dql.limit);
3200 #endif
3201 }
3202
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3203 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3204 unsigned int bytes)
3205 {
3206 #ifdef CONFIG_BQL
3207 dql_queued(&dev_queue->dql, bytes);
3208
3209 if (likely(dql_avail(&dev_queue->dql) >= 0))
3210 return;
3211
3212 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3213
3214 /*
3215 * The XOFF flag must be set before checking the dql_avail below,
3216 * because in netdev_tx_completed_queue we update the dql_completed
3217 * before checking the XOFF flag.
3218 */
3219 smp_mb();
3220
3221 /* check again in case another CPU has just made room avail */
3222 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3223 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3224 #endif
3225 }
3226
3227 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3228 * that they should not test BQL status themselves.
3229 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3230 * skb of a batch.
3231 * Returns true if the doorbell must be used to kick the NIC.
3232 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3233 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3234 unsigned int bytes,
3235 bool xmit_more)
3236 {
3237 if (xmit_more) {
3238 #ifdef CONFIG_BQL
3239 dql_queued(&dev_queue->dql, bytes);
3240 #endif
3241 return netif_tx_queue_stopped(dev_queue);
3242 }
3243 netdev_tx_sent_queue(dev_queue, bytes);
3244 return true;
3245 }
3246
3247 /**
3248 * netdev_sent_queue - report the number of bytes queued to hardware
3249 * @dev: network device
3250 * @bytes: number of bytes queued to the hardware device queue
3251 *
3252 * Report the number of bytes queued for sending/completion to the network
3253 * device hardware queue. @bytes should be a good approximation and should
3254 * exactly match netdev_completed_queue() @bytes
3255 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3256 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3257 {
3258 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3259 }
3260
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3261 static inline bool __netdev_sent_queue(struct net_device *dev,
3262 unsigned int bytes,
3263 bool xmit_more)
3264 {
3265 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3266 xmit_more);
3267 }
3268
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3269 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3270 unsigned int pkts, unsigned int bytes)
3271 {
3272 #ifdef CONFIG_BQL
3273 if (unlikely(!bytes))
3274 return;
3275
3276 dql_completed(&dev_queue->dql, bytes);
3277
3278 /*
3279 * Without the memory barrier there is a small possiblity that
3280 * netdev_tx_sent_queue will miss the update and cause the queue to
3281 * be stopped forever
3282 */
3283 smp_mb();
3284
3285 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3286 return;
3287
3288 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3289 netif_schedule_queue(dev_queue);
3290 #endif
3291 }
3292
3293 /**
3294 * netdev_completed_queue - report bytes and packets completed by device
3295 * @dev: network device
3296 * @pkts: actual number of packets sent over the medium
3297 * @bytes: actual number of bytes sent over the medium
3298 *
3299 * Report the number of bytes and packets transmitted by the network device
3300 * hardware queue over the physical medium, @bytes must exactly match the
3301 * @bytes amount passed to netdev_sent_queue()
3302 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3303 static inline void netdev_completed_queue(struct net_device *dev,
3304 unsigned int pkts, unsigned int bytes)
3305 {
3306 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3307 }
3308
netdev_tx_reset_queue(struct netdev_queue * q)3309 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3310 {
3311 #ifdef CONFIG_BQL
3312 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3313 dql_reset(&q->dql);
3314 #endif
3315 }
3316
3317 /**
3318 * netdev_reset_queue - reset the packets and bytes count of a network device
3319 * @dev_queue: network device
3320 *
3321 * Reset the bytes and packet count of a network device and clear the
3322 * software flow control OFF bit for this network device
3323 */
netdev_reset_queue(struct net_device * dev_queue)3324 static inline void netdev_reset_queue(struct net_device *dev_queue)
3325 {
3326 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3327 }
3328
3329 /**
3330 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3331 * @dev: network device
3332 * @queue_index: given tx queue index
3333 *
3334 * Returns 0 if given tx queue index >= number of device tx queues,
3335 * otherwise returns the originally passed tx queue index.
3336 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3337 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3338 {
3339 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3340 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3341 dev->name, queue_index,
3342 dev->real_num_tx_queues);
3343 return 0;
3344 }
3345
3346 return queue_index;
3347 }
3348
3349 /**
3350 * netif_running - test if up
3351 * @dev: network device
3352 *
3353 * Test if the device has been brought up.
3354 */
netif_running(const struct net_device * dev)3355 static inline bool netif_running(const struct net_device *dev)
3356 {
3357 return test_bit(__LINK_STATE_START, &dev->state);
3358 }
3359
3360 /*
3361 * Routines to manage the subqueues on a device. We only need start,
3362 * stop, and a check if it's stopped. All other device management is
3363 * done at the overall netdevice level.
3364 * Also test the device if we're multiqueue.
3365 */
3366
3367 /**
3368 * netif_start_subqueue - allow sending packets on subqueue
3369 * @dev: network device
3370 * @queue_index: sub queue index
3371 *
3372 * Start individual transmit queue of a device with multiple transmit queues.
3373 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3374 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3375 {
3376 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3377
3378 netif_tx_start_queue(txq);
3379 }
3380
3381 /**
3382 * netif_stop_subqueue - stop sending packets on subqueue
3383 * @dev: network device
3384 * @queue_index: sub queue index
3385 *
3386 * Stop individual transmit queue of a device with multiple transmit queues.
3387 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3388 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3389 {
3390 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3391 netif_tx_stop_queue(txq);
3392 }
3393
3394 /**
3395 * netif_subqueue_stopped - test status of subqueue
3396 * @dev: network device
3397 * @queue_index: sub queue index
3398 *
3399 * Check individual transmit queue of a device with multiple transmit queues.
3400 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3401 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3402 u16 queue_index)
3403 {
3404 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3405
3406 return netif_tx_queue_stopped(txq);
3407 }
3408
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3409 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3410 struct sk_buff *skb)
3411 {
3412 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3413 }
3414
3415 /**
3416 * netif_wake_subqueue - allow sending packets on subqueue
3417 * @dev: network device
3418 * @queue_index: sub queue index
3419 *
3420 * Resume individual transmit queue of a device with multiple transmit queues.
3421 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3422 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3423 {
3424 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3425
3426 netif_tx_wake_queue(txq);
3427 }
3428
3429 #ifdef CONFIG_XPS
3430 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3431 u16 index);
3432 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3433 u16 index, bool is_rxqs_map);
3434
3435 /**
3436 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3437 * @j: CPU/Rx queue index
3438 * @mask: bitmask of all cpus/rx queues
3439 * @nr_bits: number of bits in the bitmask
3440 *
3441 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3442 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3443 static inline bool netif_attr_test_mask(unsigned long j,
3444 const unsigned long *mask,
3445 unsigned int nr_bits)
3446 {
3447 cpu_max_bits_warn(j, nr_bits);
3448 return test_bit(j, mask);
3449 }
3450
3451 /**
3452 * netif_attr_test_online - Test for online CPU/Rx queue
3453 * @j: CPU/Rx queue index
3454 * @online_mask: bitmask for CPUs/Rx queues that are online
3455 * @nr_bits: number of bits in the bitmask
3456 *
3457 * Returns true if a CPU/Rx queue is online.
3458 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3459 static inline bool netif_attr_test_online(unsigned long j,
3460 const unsigned long *online_mask,
3461 unsigned int nr_bits)
3462 {
3463 cpu_max_bits_warn(j, nr_bits);
3464
3465 if (online_mask)
3466 return test_bit(j, online_mask);
3467
3468 return (j < nr_bits);
3469 }
3470
3471 /**
3472 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3473 * @n: CPU/Rx queue index
3474 * @srcp: the cpumask/Rx queue mask pointer
3475 * @nr_bits: number of bits in the bitmask
3476 *
3477 * Returns >= nr_bits if no further CPUs/Rx queues set.
3478 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3479 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3480 unsigned int nr_bits)
3481 {
3482 /* -1 is a legal arg here. */
3483 if (n != -1)
3484 cpu_max_bits_warn(n, nr_bits);
3485
3486 if (srcp)
3487 return find_next_bit(srcp, nr_bits, n + 1);
3488
3489 return n + 1;
3490 }
3491
3492 /**
3493 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3494 * @n: CPU/Rx queue index
3495 * @src1p: the first CPUs/Rx queues mask pointer
3496 * @src2p: the second CPUs/Rx queues mask pointer
3497 * @nr_bits: number of bits in the bitmask
3498 *
3499 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3500 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3501 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3502 const unsigned long *src2p,
3503 unsigned int nr_bits)
3504 {
3505 /* -1 is a legal arg here. */
3506 if (n != -1)
3507 cpu_max_bits_warn(n, nr_bits);
3508
3509 if (src1p && src2p)
3510 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3511 else if (src1p)
3512 return find_next_bit(src1p, nr_bits, n + 1);
3513 else if (src2p)
3514 return find_next_bit(src2p, nr_bits, n + 1);
3515
3516 return n + 1;
3517 }
3518 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3519 static inline int netif_set_xps_queue(struct net_device *dev,
3520 const struct cpumask *mask,
3521 u16 index)
3522 {
3523 return 0;
3524 }
3525
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,bool is_rxqs_map)3526 static inline int __netif_set_xps_queue(struct net_device *dev,
3527 const unsigned long *mask,
3528 u16 index, bool is_rxqs_map)
3529 {
3530 return 0;
3531 }
3532 #endif
3533
3534 /**
3535 * netif_is_multiqueue - test if device has multiple transmit queues
3536 * @dev: network device
3537 *
3538 * Check if device has multiple transmit queues
3539 */
netif_is_multiqueue(const struct net_device * dev)3540 static inline bool netif_is_multiqueue(const struct net_device *dev)
3541 {
3542 return dev->num_tx_queues > 1;
3543 }
3544
3545 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3546
3547 #ifdef CONFIG_SYSFS
3548 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3549 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3550 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3551 unsigned int rxqs)
3552 {
3553 dev->real_num_rx_queues = rxqs;
3554 return 0;
3555 }
3556 #endif
3557
3558 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3559 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3560 {
3561 return dev->_rx + rxq;
3562 }
3563
3564 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3565 static inline unsigned int get_netdev_rx_queue_index(
3566 struct netdev_rx_queue *queue)
3567 {
3568 struct net_device *dev = queue->dev;
3569 int index = queue - dev->_rx;
3570
3571 BUG_ON(index >= dev->num_rx_queues);
3572 return index;
3573 }
3574 #endif
3575
3576 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3577 int netif_get_num_default_rss_queues(void);
3578
3579 enum skb_free_reason {
3580 SKB_REASON_CONSUMED,
3581 SKB_REASON_DROPPED,
3582 };
3583
3584 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3585 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3586
3587 /*
3588 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3589 * interrupt context or with hardware interrupts being disabled.
3590 * (in_irq() || irqs_disabled())
3591 *
3592 * We provide four helpers that can be used in following contexts :
3593 *
3594 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3595 * replacing kfree_skb(skb)
3596 *
3597 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3598 * Typically used in place of consume_skb(skb) in TX completion path
3599 *
3600 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3601 * replacing kfree_skb(skb)
3602 *
3603 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3604 * and consumed a packet. Used in place of consume_skb(skb)
3605 */
dev_kfree_skb_irq(struct sk_buff * skb)3606 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3607 {
3608 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3609 }
3610
dev_consume_skb_irq(struct sk_buff * skb)3611 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3612 {
3613 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3614 }
3615
dev_kfree_skb_any(struct sk_buff * skb)3616 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3617 {
3618 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3619 }
3620
dev_consume_skb_any(struct sk_buff * skb)3621 static inline void dev_consume_skb_any(struct sk_buff *skb)
3622 {
3623 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3624 }
3625
3626 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3627 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3628 int netif_rx(struct sk_buff *skb);
3629 int netif_rx_ni(struct sk_buff *skb);
3630 int netif_receive_skb(struct sk_buff *skb);
3631 int netif_receive_skb_core(struct sk_buff *skb);
3632 void netif_receive_skb_list(struct list_head *head);
3633 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3634 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3635 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3636 gro_result_t napi_gro_frags(struct napi_struct *napi);
3637 struct packet_offload *gro_find_receive_by_type(__be16 type);
3638 struct packet_offload *gro_find_complete_by_type(__be16 type);
3639
napi_free_frags(struct napi_struct * napi)3640 static inline void napi_free_frags(struct napi_struct *napi)
3641 {
3642 kfree_skb(napi->skb);
3643 napi->skb = NULL;
3644 }
3645
3646 bool netdev_is_rx_handler_busy(struct net_device *dev);
3647 int netdev_rx_handler_register(struct net_device *dev,
3648 rx_handler_func_t *rx_handler,
3649 void *rx_handler_data);
3650 void netdev_rx_handler_unregister(struct net_device *dev);
3651
3652 bool dev_valid_name(const char *name);
3653 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3654 bool *need_copyout);
3655 int dev_ifconf(struct net *net, struct ifconf *, int);
3656 int dev_ethtool(struct net *net, struct ifreq *);
3657 unsigned int dev_get_flags(const struct net_device *);
3658 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3659 struct netlink_ext_ack *extack);
3660 int dev_change_flags(struct net_device *dev, unsigned int flags,
3661 struct netlink_ext_ack *extack);
3662 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3663 unsigned int gchanges);
3664 int dev_change_name(struct net_device *, const char *);
3665 int dev_set_alias(struct net_device *, const char *, size_t);
3666 int dev_get_alias(const struct net_device *, char *, size_t);
3667 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3668 int __dev_set_mtu(struct net_device *, int);
3669 int dev_validate_mtu(struct net_device *dev, int mtu,
3670 struct netlink_ext_ack *extack);
3671 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3672 struct netlink_ext_ack *extack);
3673 int dev_set_mtu(struct net_device *, int);
3674 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3675 void dev_set_group(struct net_device *, int);
3676 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3677 struct netlink_ext_ack *extack);
3678 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3679 struct netlink_ext_ack *extack);
3680 int dev_change_carrier(struct net_device *, bool new_carrier);
3681 int dev_get_phys_port_id(struct net_device *dev,
3682 struct netdev_phys_item_id *ppid);
3683 int dev_get_phys_port_name(struct net_device *dev,
3684 char *name, size_t len);
3685 int dev_get_port_parent_id(struct net_device *dev,
3686 struct netdev_phys_item_id *ppid, bool recurse);
3687 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3688 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3689 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3690 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3691 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3692 struct netdev_queue *txq, int *ret);
3693
3694 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3695 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3696 int fd, u32 flags);
3697 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3698 enum bpf_netdev_command cmd);
3699 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3700
3701 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3702 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3703 bool is_skb_forwardable(const struct net_device *dev,
3704 const struct sk_buff *skb);
3705
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb)3706 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3707 struct sk_buff *skb)
3708 {
3709 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3710 unlikely(!is_skb_forwardable(dev, skb))) {
3711 atomic_long_inc(&dev->rx_dropped);
3712 kfree_skb(skb);
3713 return NET_RX_DROP;
3714 }
3715
3716 skb_scrub_packet(skb, true);
3717 skb->priority = 0;
3718 return 0;
3719 }
3720
3721 bool dev_nit_active(struct net_device *dev);
3722 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3723
3724 extern int netdev_budget;
3725 extern unsigned int netdev_budget_usecs;
3726
3727 /* Called by rtnetlink.c:rtnl_unlock() */
3728 void netdev_run_todo(void);
3729
3730 /**
3731 * dev_put - release reference to device
3732 * @dev: network device
3733 *
3734 * Release reference to device to allow it to be freed.
3735 */
dev_put(struct net_device * dev)3736 static inline void dev_put(struct net_device *dev)
3737 {
3738 this_cpu_dec(*dev->pcpu_refcnt);
3739 }
3740
3741 /**
3742 * dev_hold - get reference to device
3743 * @dev: network device
3744 *
3745 * Hold reference to device to keep it from being freed.
3746 */
dev_hold(struct net_device * dev)3747 static inline void dev_hold(struct net_device *dev)
3748 {
3749 this_cpu_inc(*dev->pcpu_refcnt);
3750 }
3751
3752 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3753 * and _off may be called from IRQ context, but it is caller
3754 * who is responsible for serialization of these calls.
3755 *
3756 * The name carrier is inappropriate, these functions should really be
3757 * called netif_lowerlayer_*() because they represent the state of any
3758 * kind of lower layer not just hardware media.
3759 */
3760
3761 void linkwatch_init_dev(struct net_device *dev);
3762 void linkwatch_fire_event(struct net_device *dev);
3763 void linkwatch_forget_dev(struct net_device *dev);
3764
3765 /**
3766 * netif_carrier_ok - test if carrier present
3767 * @dev: network device
3768 *
3769 * Check if carrier is present on device
3770 */
netif_carrier_ok(const struct net_device * dev)3771 static inline bool netif_carrier_ok(const struct net_device *dev)
3772 {
3773 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3774 }
3775
3776 unsigned long dev_trans_start(struct net_device *dev);
3777
3778 void __netdev_watchdog_up(struct net_device *dev);
3779
3780 void netif_carrier_on(struct net_device *dev);
3781
3782 void netif_carrier_off(struct net_device *dev);
3783
3784 /**
3785 * netif_dormant_on - mark device as dormant.
3786 * @dev: network device
3787 *
3788 * Mark device as dormant (as per RFC2863).
3789 *
3790 * The dormant state indicates that the relevant interface is not
3791 * actually in a condition to pass packets (i.e., it is not 'up') but is
3792 * in a "pending" state, waiting for some external event. For "on-
3793 * demand" interfaces, this new state identifies the situation where the
3794 * interface is waiting for events to place it in the up state.
3795 */
netif_dormant_on(struct net_device * dev)3796 static inline void netif_dormant_on(struct net_device *dev)
3797 {
3798 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3799 linkwatch_fire_event(dev);
3800 }
3801
3802 /**
3803 * netif_dormant_off - set device as not dormant.
3804 * @dev: network device
3805 *
3806 * Device is not in dormant state.
3807 */
netif_dormant_off(struct net_device * dev)3808 static inline void netif_dormant_off(struct net_device *dev)
3809 {
3810 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3811 linkwatch_fire_event(dev);
3812 }
3813
3814 /**
3815 * netif_dormant - test if device is dormant
3816 * @dev: network device
3817 *
3818 * Check if device is dormant.
3819 */
netif_dormant(const struct net_device * dev)3820 static inline bool netif_dormant(const struct net_device *dev)
3821 {
3822 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3823 }
3824
3825
3826 /**
3827 * netif_oper_up - test if device is operational
3828 * @dev: network device
3829 *
3830 * Check if carrier is operational
3831 */
netif_oper_up(const struct net_device * dev)3832 static inline bool netif_oper_up(const struct net_device *dev)
3833 {
3834 return (dev->operstate == IF_OPER_UP ||
3835 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3836 }
3837
3838 /**
3839 * netif_device_present - is device available or removed
3840 * @dev: network device
3841 *
3842 * Check if device has not been removed from system.
3843 */
netif_device_present(struct net_device * dev)3844 static inline bool netif_device_present(struct net_device *dev)
3845 {
3846 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3847 }
3848
3849 void netif_device_detach(struct net_device *dev);
3850
3851 void netif_device_attach(struct net_device *dev);
3852
3853 /*
3854 * Network interface message level settings
3855 */
3856
3857 enum {
3858 NETIF_MSG_DRV = 0x0001,
3859 NETIF_MSG_PROBE = 0x0002,
3860 NETIF_MSG_LINK = 0x0004,
3861 NETIF_MSG_TIMER = 0x0008,
3862 NETIF_MSG_IFDOWN = 0x0010,
3863 NETIF_MSG_IFUP = 0x0020,
3864 NETIF_MSG_RX_ERR = 0x0040,
3865 NETIF_MSG_TX_ERR = 0x0080,
3866 NETIF_MSG_TX_QUEUED = 0x0100,
3867 NETIF_MSG_INTR = 0x0200,
3868 NETIF_MSG_TX_DONE = 0x0400,
3869 NETIF_MSG_RX_STATUS = 0x0800,
3870 NETIF_MSG_PKTDATA = 0x1000,
3871 NETIF_MSG_HW = 0x2000,
3872 NETIF_MSG_WOL = 0x4000,
3873 };
3874
3875 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3876 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3877 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3878 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3879 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3880 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3881 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3882 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3883 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3884 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3885 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3886 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3887 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3888 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3889 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3890
netif_msg_init(int debug_value,int default_msg_enable_bits)3891 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3892 {
3893 /* use default */
3894 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3895 return default_msg_enable_bits;
3896 if (debug_value == 0) /* no output */
3897 return 0;
3898 /* set low N bits */
3899 return (1U << debug_value) - 1;
3900 }
3901
__netif_tx_lock(struct netdev_queue * txq,int cpu)3902 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3903 {
3904 spin_lock(&txq->_xmit_lock);
3905 txq->xmit_lock_owner = cpu;
3906 }
3907
__netif_tx_acquire(struct netdev_queue * txq)3908 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3909 {
3910 __acquire(&txq->_xmit_lock);
3911 return true;
3912 }
3913
__netif_tx_release(struct netdev_queue * txq)3914 static inline void __netif_tx_release(struct netdev_queue *txq)
3915 {
3916 __release(&txq->_xmit_lock);
3917 }
3918
__netif_tx_lock_bh(struct netdev_queue * txq)3919 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3920 {
3921 spin_lock_bh(&txq->_xmit_lock);
3922 txq->xmit_lock_owner = smp_processor_id();
3923 }
3924
__netif_tx_trylock(struct netdev_queue * txq)3925 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3926 {
3927 bool ok = spin_trylock(&txq->_xmit_lock);
3928 if (likely(ok))
3929 txq->xmit_lock_owner = smp_processor_id();
3930 return ok;
3931 }
3932
__netif_tx_unlock(struct netdev_queue * txq)3933 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3934 {
3935 txq->xmit_lock_owner = -1;
3936 spin_unlock(&txq->_xmit_lock);
3937 }
3938
__netif_tx_unlock_bh(struct netdev_queue * txq)3939 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3940 {
3941 txq->xmit_lock_owner = -1;
3942 spin_unlock_bh(&txq->_xmit_lock);
3943 }
3944
txq_trans_update(struct netdev_queue * txq)3945 static inline void txq_trans_update(struct netdev_queue *txq)
3946 {
3947 if (txq->xmit_lock_owner != -1)
3948 txq->trans_start = jiffies;
3949 }
3950
3951 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)3952 static inline void netif_trans_update(struct net_device *dev)
3953 {
3954 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3955
3956 if (txq->trans_start != jiffies)
3957 txq->trans_start = jiffies;
3958 }
3959
3960 /**
3961 * netif_tx_lock - grab network device transmit lock
3962 * @dev: network device
3963 *
3964 * Get network device transmit lock
3965 */
netif_tx_lock(struct net_device * dev)3966 static inline void netif_tx_lock(struct net_device *dev)
3967 {
3968 unsigned int i;
3969 int cpu;
3970
3971 spin_lock(&dev->tx_global_lock);
3972 cpu = smp_processor_id();
3973 for (i = 0; i < dev->num_tx_queues; i++) {
3974 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3975
3976 /* We are the only thread of execution doing a
3977 * freeze, but we have to grab the _xmit_lock in
3978 * order to synchronize with threads which are in
3979 * the ->hard_start_xmit() handler and already
3980 * checked the frozen bit.
3981 */
3982 __netif_tx_lock(txq, cpu);
3983 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3984 __netif_tx_unlock(txq);
3985 }
3986 }
3987
netif_tx_lock_bh(struct net_device * dev)3988 static inline void netif_tx_lock_bh(struct net_device *dev)
3989 {
3990 local_bh_disable();
3991 netif_tx_lock(dev);
3992 }
3993
netif_tx_unlock(struct net_device * dev)3994 static inline void netif_tx_unlock(struct net_device *dev)
3995 {
3996 unsigned int i;
3997
3998 for (i = 0; i < dev->num_tx_queues; i++) {
3999 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4000
4001 /* No need to grab the _xmit_lock here. If the
4002 * queue is not stopped for another reason, we
4003 * force a schedule.
4004 */
4005 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4006 netif_schedule_queue(txq);
4007 }
4008 spin_unlock(&dev->tx_global_lock);
4009 }
4010
netif_tx_unlock_bh(struct net_device * dev)4011 static inline void netif_tx_unlock_bh(struct net_device *dev)
4012 {
4013 netif_tx_unlock(dev);
4014 local_bh_enable();
4015 }
4016
4017 #define HARD_TX_LOCK(dev, txq, cpu) { \
4018 if ((dev->features & NETIF_F_LLTX) == 0) { \
4019 __netif_tx_lock(txq, cpu); \
4020 } else { \
4021 __netif_tx_acquire(txq); \
4022 } \
4023 }
4024
4025 #define HARD_TX_TRYLOCK(dev, txq) \
4026 (((dev->features & NETIF_F_LLTX) == 0) ? \
4027 __netif_tx_trylock(txq) : \
4028 __netif_tx_acquire(txq))
4029
4030 #define HARD_TX_UNLOCK(dev, txq) { \
4031 if ((dev->features & NETIF_F_LLTX) == 0) { \
4032 __netif_tx_unlock(txq); \
4033 } else { \
4034 __netif_tx_release(txq); \
4035 } \
4036 }
4037
netif_tx_disable(struct net_device * dev)4038 static inline void netif_tx_disable(struct net_device *dev)
4039 {
4040 unsigned int i;
4041 int cpu;
4042
4043 local_bh_disable();
4044 cpu = smp_processor_id();
4045 for (i = 0; i < dev->num_tx_queues; i++) {
4046 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4047
4048 __netif_tx_lock(txq, cpu);
4049 netif_tx_stop_queue(txq);
4050 __netif_tx_unlock(txq);
4051 }
4052 local_bh_enable();
4053 }
4054
netif_addr_lock(struct net_device * dev)4055 static inline void netif_addr_lock(struct net_device *dev)
4056 {
4057 spin_lock(&dev->addr_list_lock);
4058 }
4059
netif_addr_lock_bh(struct net_device * dev)4060 static inline void netif_addr_lock_bh(struct net_device *dev)
4061 {
4062 spin_lock_bh(&dev->addr_list_lock);
4063 }
4064
netif_addr_unlock(struct net_device * dev)4065 static inline void netif_addr_unlock(struct net_device *dev)
4066 {
4067 spin_unlock(&dev->addr_list_lock);
4068 }
4069
netif_addr_unlock_bh(struct net_device * dev)4070 static inline void netif_addr_unlock_bh(struct net_device *dev)
4071 {
4072 spin_unlock_bh(&dev->addr_list_lock);
4073 }
4074
4075 /*
4076 * dev_addrs walker. Should be used only for read access. Call with
4077 * rcu_read_lock held.
4078 */
4079 #define for_each_dev_addr(dev, ha) \
4080 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4081
4082 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4083
4084 void ether_setup(struct net_device *dev);
4085
4086 /* Support for loadable net-drivers */
4087 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4088 unsigned char name_assign_type,
4089 void (*setup)(struct net_device *),
4090 unsigned int txqs, unsigned int rxqs);
4091 int dev_get_valid_name(struct net *net, struct net_device *dev,
4092 const char *name);
4093
4094 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4095 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4096
4097 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4098 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4099 count)
4100
4101 int register_netdev(struct net_device *dev);
4102 void unregister_netdev(struct net_device *dev);
4103
4104 /* General hardware address lists handling functions */
4105 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4106 struct netdev_hw_addr_list *from_list, int addr_len);
4107 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4108 struct netdev_hw_addr_list *from_list, int addr_len);
4109 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4110 struct net_device *dev,
4111 int (*sync)(struct net_device *, const unsigned char *),
4112 int (*unsync)(struct net_device *,
4113 const unsigned char *));
4114 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4115 struct net_device *dev,
4116 int (*sync)(struct net_device *,
4117 const unsigned char *, int),
4118 int (*unsync)(struct net_device *,
4119 const unsigned char *, int));
4120 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4121 struct net_device *dev,
4122 int (*unsync)(struct net_device *,
4123 const unsigned char *, int));
4124 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4125 struct net_device *dev,
4126 int (*unsync)(struct net_device *,
4127 const unsigned char *));
4128 void __hw_addr_init(struct netdev_hw_addr_list *list);
4129
4130 /* Functions used for device addresses handling */
4131 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4132 unsigned char addr_type);
4133 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4134 unsigned char addr_type);
4135 void dev_addr_flush(struct net_device *dev);
4136 int dev_addr_init(struct net_device *dev);
4137
4138 /* Functions used for unicast addresses handling */
4139 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4140 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4141 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4142 int dev_uc_sync(struct net_device *to, struct net_device *from);
4143 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4144 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4145 void dev_uc_flush(struct net_device *dev);
4146 void dev_uc_init(struct net_device *dev);
4147
4148 /**
4149 * __dev_uc_sync - Synchonize device's unicast list
4150 * @dev: device to sync
4151 * @sync: function to call if address should be added
4152 * @unsync: function to call if address should be removed
4153 *
4154 * Add newly added addresses to the interface, and release
4155 * addresses that have been deleted.
4156 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4157 static inline int __dev_uc_sync(struct net_device *dev,
4158 int (*sync)(struct net_device *,
4159 const unsigned char *),
4160 int (*unsync)(struct net_device *,
4161 const unsigned char *))
4162 {
4163 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4164 }
4165
4166 /**
4167 * __dev_uc_unsync - Remove synchronized addresses from device
4168 * @dev: device to sync
4169 * @unsync: function to call if address should be removed
4170 *
4171 * Remove all addresses that were added to the device by dev_uc_sync().
4172 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4173 static inline void __dev_uc_unsync(struct net_device *dev,
4174 int (*unsync)(struct net_device *,
4175 const unsigned char *))
4176 {
4177 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4178 }
4179
4180 /* Functions used for multicast addresses handling */
4181 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4182 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4183 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4184 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4185 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4186 int dev_mc_sync(struct net_device *to, struct net_device *from);
4187 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4188 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4189 void dev_mc_flush(struct net_device *dev);
4190 void dev_mc_init(struct net_device *dev);
4191
4192 /**
4193 * __dev_mc_sync - Synchonize device's multicast list
4194 * @dev: device to sync
4195 * @sync: function to call if address should be added
4196 * @unsync: function to call if address should be removed
4197 *
4198 * Add newly added addresses to the interface, and release
4199 * addresses that have been deleted.
4200 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4201 static inline int __dev_mc_sync(struct net_device *dev,
4202 int (*sync)(struct net_device *,
4203 const unsigned char *),
4204 int (*unsync)(struct net_device *,
4205 const unsigned char *))
4206 {
4207 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4208 }
4209
4210 /**
4211 * __dev_mc_unsync - Remove synchronized addresses from device
4212 * @dev: device to sync
4213 * @unsync: function to call if address should be removed
4214 *
4215 * Remove all addresses that were added to the device by dev_mc_sync().
4216 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4217 static inline void __dev_mc_unsync(struct net_device *dev,
4218 int (*unsync)(struct net_device *,
4219 const unsigned char *))
4220 {
4221 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4222 }
4223
4224 /* Functions used for secondary unicast and multicast support */
4225 void dev_set_rx_mode(struct net_device *dev);
4226 void __dev_set_rx_mode(struct net_device *dev);
4227 int dev_set_promiscuity(struct net_device *dev, int inc);
4228 int dev_set_allmulti(struct net_device *dev, int inc);
4229 void netdev_state_change(struct net_device *dev);
4230 void netdev_notify_peers(struct net_device *dev);
4231 void netdev_features_change(struct net_device *dev);
4232 /* Load a device via the kmod */
4233 void dev_load(struct net *net, const char *name);
4234 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4235 struct rtnl_link_stats64 *storage);
4236 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4237 const struct net_device_stats *netdev_stats);
4238
4239 extern int netdev_max_backlog;
4240 extern int netdev_tstamp_prequeue;
4241 extern int weight_p;
4242 extern int dev_weight_rx_bias;
4243 extern int dev_weight_tx_bias;
4244 extern int dev_rx_weight;
4245 extern int dev_tx_weight;
4246 extern int gro_normal_batch;
4247
4248 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4249 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4250 struct list_head **iter);
4251 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4252 struct list_head **iter);
4253
4254 /* iterate through upper list, must be called under RCU read lock */
4255 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4256 for (iter = &(dev)->adj_list.upper, \
4257 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4258 updev; \
4259 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4260
4261 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4262 int (*fn)(struct net_device *upper_dev,
4263 void *data),
4264 void *data);
4265
4266 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4267 struct net_device *upper_dev);
4268
4269 bool netdev_has_any_upper_dev(struct net_device *dev);
4270
4271 void *netdev_lower_get_next_private(struct net_device *dev,
4272 struct list_head **iter);
4273 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4274 struct list_head **iter);
4275
4276 #define netdev_for_each_lower_private(dev, priv, iter) \
4277 for (iter = (dev)->adj_list.lower.next, \
4278 priv = netdev_lower_get_next_private(dev, &(iter)); \
4279 priv; \
4280 priv = netdev_lower_get_next_private(dev, &(iter)))
4281
4282 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4283 for (iter = &(dev)->adj_list.lower, \
4284 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4285 priv; \
4286 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4287
4288 void *netdev_lower_get_next(struct net_device *dev,
4289 struct list_head **iter);
4290
4291 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4292 for (iter = (dev)->adj_list.lower.next, \
4293 ldev = netdev_lower_get_next(dev, &(iter)); \
4294 ldev; \
4295 ldev = netdev_lower_get_next(dev, &(iter)))
4296
4297 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4298 struct list_head **iter);
4299 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4300 struct list_head **iter);
4301
4302 int netdev_walk_all_lower_dev(struct net_device *dev,
4303 int (*fn)(struct net_device *lower_dev,
4304 void *data),
4305 void *data);
4306 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4307 int (*fn)(struct net_device *lower_dev,
4308 void *data),
4309 void *data);
4310
4311 void *netdev_adjacent_get_private(struct list_head *adj_list);
4312 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4313 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4314 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4315 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4316 struct netlink_ext_ack *extack);
4317 int netdev_master_upper_dev_link(struct net_device *dev,
4318 struct net_device *upper_dev,
4319 void *upper_priv, void *upper_info,
4320 struct netlink_ext_ack *extack);
4321 void netdev_upper_dev_unlink(struct net_device *dev,
4322 struct net_device *upper_dev);
4323 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4324 struct net_device *new_dev,
4325 struct net_device *dev,
4326 struct netlink_ext_ack *extack);
4327 void netdev_adjacent_change_commit(struct net_device *old_dev,
4328 struct net_device *new_dev,
4329 struct net_device *dev);
4330 void netdev_adjacent_change_abort(struct net_device *old_dev,
4331 struct net_device *new_dev,
4332 struct net_device *dev);
4333 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4334 void *netdev_lower_dev_get_private(struct net_device *dev,
4335 struct net_device *lower_dev);
4336 void netdev_lower_state_changed(struct net_device *lower_dev,
4337 void *lower_state_info);
4338
4339 /* RSS keys are 40 or 52 bytes long */
4340 #define NETDEV_RSS_KEY_LEN 52
4341 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4342 void netdev_rss_key_fill(void *buffer, size_t len);
4343
4344 int skb_checksum_help(struct sk_buff *skb);
4345 int skb_crc32c_csum_help(struct sk_buff *skb);
4346 int skb_csum_hwoffload_help(struct sk_buff *skb,
4347 const netdev_features_t features);
4348
4349 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4350 netdev_features_t features, bool tx_path);
4351 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4352 netdev_features_t features);
4353
4354 struct netdev_bonding_info {
4355 ifslave slave;
4356 ifbond master;
4357 };
4358
4359 struct netdev_notifier_bonding_info {
4360 struct netdev_notifier_info info; /* must be first */
4361 struct netdev_bonding_info bonding_info;
4362 };
4363
4364 void netdev_bonding_info_change(struct net_device *dev,
4365 struct netdev_bonding_info *bonding_info);
4366
4367 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4368 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4369 {
4370 return __skb_gso_segment(skb, features, true);
4371 }
4372 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4373
can_checksum_protocol(netdev_features_t features,__be16 protocol)4374 static inline bool can_checksum_protocol(netdev_features_t features,
4375 __be16 protocol)
4376 {
4377 if (protocol == htons(ETH_P_FCOE))
4378 return !!(features & NETIF_F_FCOE_CRC);
4379
4380 /* Assume this is an IP checksum (not SCTP CRC) */
4381
4382 if (features & NETIF_F_HW_CSUM) {
4383 /* Can checksum everything */
4384 return true;
4385 }
4386
4387 switch (protocol) {
4388 case htons(ETH_P_IP):
4389 return !!(features & NETIF_F_IP_CSUM);
4390 case htons(ETH_P_IPV6):
4391 return !!(features & NETIF_F_IPV6_CSUM);
4392 default:
4393 return false;
4394 }
4395 }
4396
4397 #ifdef CONFIG_BUG
4398 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4399 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4400 static inline void netdev_rx_csum_fault(struct net_device *dev,
4401 struct sk_buff *skb)
4402 {
4403 }
4404 #endif
4405 /* rx skb timestamps */
4406 void net_enable_timestamp(void);
4407 void net_disable_timestamp(void);
4408
4409 #ifdef CONFIG_PROC_FS
4410 int __init dev_proc_init(void);
4411 #else
4412 #define dev_proc_init() 0
4413 #endif
4414
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4415 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4416 struct sk_buff *skb, struct net_device *dev,
4417 bool more)
4418 {
4419 __this_cpu_write(softnet_data.xmit.more, more);
4420 return ops->ndo_start_xmit(skb, dev);
4421 }
4422
netdev_xmit_more(void)4423 static inline bool netdev_xmit_more(void)
4424 {
4425 return __this_cpu_read(softnet_data.xmit.more);
4426 }
4427
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4428 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4429 struct netdev_queue *txq, bool more)
4430 {
4431 const struct net_device_ops *ops = dev->netdev_ops;
4432 netdev_tx_t rc;
4433
4434 rc = __netdev_start_xmit(ops, skb, dev, more);
4435 if (rc == NETDEV_TX_OK)
4436 txq_trans_update(txq);
4437
4438 return rc;
4439 }
4440
4441 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4442 const void *ns);
4443 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4444 const void *ns);
4445
netdev_class_create_file(const struct class_attribute * class_attr)4446 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4447 {
4448 return netdev_class_create_file_ns(class_attr, NULL);
4449 }
4450
netdev_class_remove_file(const struct class_attribute * class_attr)4451 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4452 {
4453 netdev_class_remove_file_ns(class_attr, NULL);
4454 }
4455
4456 extern const struct kobj_ns_type_operations net_ns_type_operations;
4457
4458 const char *netdev_drivername(const struct net_device *dev);
4459
4460 void linkwatch_run_queue(void);
4461
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4462 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4463 netdev_features_t f2)
4464 {
4465 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4466 if (f1 & NETIF_F_HW_CSUM)
4467 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4468 else
4469 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4470 }
4471
4472 return f1 & f2;
4473 }
4474
netdev_get_wanted_features(struct net_device * dev)4475 static inline netdev_features_t netdev_get_wanted_features(
4476 struct net_device *dev)
4477 {
4478 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4479 }
4480 netdev_features_t netdev_increment_features(netdev_features_t all,
4481 netdev_features_t one, netdev_features_t mask);
4482
4483 /* Allow TSO being used on stacked device :
4484 * Performing the GSO segmentation before last device
4485 * is a performance improvement.
4486 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4487 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4488 netdev_features_t mask)
4489 {
4490 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4491 }
4492
4493 int __netdev_update_features(struct net_device *dev);
4494 void netdev_update_features(struct net_device *dev);
4495 void netdev_change_features(struct net_device *dev);
4496
4497 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4498 struct net_device *dev);
4499
4500 netdev_features_t passthru_features_check(struct sk_buff *skb,
4501 struct net_device *dev,
4502 netdev_features_t features);
4503 netdev_features_t netif_skb_features(struct sk_buff *skb);
4504
net_gso_ok(netdev_features_t features,int gso_type)4505 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4506 {
4507 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4508
4509 /* check flags correspondence */
4510 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4511 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4512 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4513 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4514 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4515 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4516 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4517 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4518 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4519 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4520 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4521 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4522 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4523 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4524 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4525 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4526 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4527 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4528
4529 return (features & feature) == feature;
4530 }
4531
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)4532 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4533 {
4534 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4535 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4536 }
4537
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)4538 static inline bool netif_needs_gso(struct sk_buff *skb,
4539 netdev_features_t features)
4540 {
4541 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4542 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4543 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4544 }
4545
netif_set_gso_max_size(struct net_device * dev,unsigned int size)4546 static inline void netif_set_gso_max_size(struct net_device *dev,
4547 unsigned int size)
4548 {
4549 dev->gso_max_size = size;
4550 }
4551
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)4552 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4553 int pulled_hlen, u16 mac_offset,
4554 int mac_len)
4555 {
4556 skb->protocol = protocol;
4557 skb->encapsulation = 1;
4558 skb_push(skb, pulled_hlen);
4559 skb_reset_transport_header(skb);
4560 skb->mac_header = mac_offset;
4561 skb->network_header = skb->mac_header + mac_len;
4562 skb->mac_len = mac_len;
4563 }
4564
netif_is_macsec(const struct net_device * dev)4565 static inline bool netif_is_macsec(const struct net_device *dev)
4566 {
4567 return dev->priv_flags & IFF_MACSEC;
4568 }
4569
netif_is_macvlan(const struct net_device * dev)4570 static inline bool netif_is_macvlan(const struct net_device *dev)
4571 {
4572 return dev->priv_flags & IFF_MACVLAN;
4573 }
4574
netif_is_macvlan_port(const struct net_device * dev)4575 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4576 {
4577 return dev->priv_flags & IFF_MACVLAN_PORT;
4578 }
4579
netif_is_bond_master(const struct net_device * dev)4580 static inline bool netif_is_bond_master(const struct net_device *dev)
4581 {
4582 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4583 }
4584
netif_is_bond_slave(const struct net_device * dev)4585 static inline bool netif_is_bond_slave(const struct net_device *dev)
4586 {
4587 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4588 }
4589
netif_supports_nofcs(struct net_device * dev)4590 static inline bool netif_supports_nofcs(struct net_device *dev)
4591 {
4592 return dev->priv_flags & IFF_SUPP_NOFCS;
4593 }
4594
netif_has_l3_rx_handler(const struct net_device * dev)4595 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4596 {
4597 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4598 }
4599
netif_is_l3_master(const struct net_device * dev)4600 static inline bool netif_is_l3_master(const struct net_device *dev)
4601 {
4602 return dev->priv_flags & IFF_L3MDEV_MASTER;
4603 }
4604
netif_is_l3_slave(const struct net_device * dev)4605 static inline bool netif_is_l3_slave(const struct net_device *dev)
4606 {
4607 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4608 }
4609
netif_is_bridge_master(const struct net_device * dev)4610 static inline bool netif_is_bridge_master(const struct net_device *dev)
4611 {
4612 return dev->priv_flags & IFF_EBRIDGE;
4613 }
4614
netif_is_bridge_port(const struct net_device * dev)4615 static inline bool netif_is_bridge_port(const struct net_device *dev)
4616 {
4617 return dev->priv_flags & IFF_BRIDGE_PORT;
4618 }
4619
netif_is_ovs_master(const struct net_device * dev)4620 static inline bool netif_is_ovs_master(const struct net_device *dev)
4621 {
4622 return dev->priv_flags & IFF_OPENVSWITCH;
4623 }
4624
netif_is_ovs_port(const struct net_device * dev)4625 static inline bool netif_is_ovs_port(const struct net_device *dev)
4626 {
4627 return dev->priv_flags & IFF_OVS_DATAPATH;
4628 }
4629
netif_is_team_master(const struct net_device * dev)4630 static inline bool netif_is_team_master(const struct net_device *dev)
4631 {
4632 return dev->priv_flags & IFF_TEAM;
4633 }
4634
netif_is_team_port(const struct net_device * dev)4635 static inline bool netif_is_team_port(const struct net_device *dev)
4636 {
4637 return dev->priv_flags & IFF_TEAM_PORT;
4638 }
4639
netif_is_lag_master(const struct net_device * dev)4640 static inline bool netif_is_lag_master(const struct net_device *dev)
4641 {
4642 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4643 }
4644
netif_is_lag_port(const struct net_device * dev)4645 static inline bool netif_is_lag_port(const struct net_device *dev)
4646 {
4647 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4648 }
4649
netif_is_rxfh_configured(const struct net_device * dev)4650 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4651 {
4652 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4653 }
4654
netif_is_failover(const struct net_device * dev)4655 static inline bool netif_is_failover(const struct net_device *dev)
4656 {
4657 return dev->priv_flags & IFF_FAILOVER;
4658 }
4659
netif_is_failover_slave(const struct net_device * dev)4660 static inline bool netif_is_failover_slave(const struct net_device *dev)
4661 {
4662 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4663 }
4664
4665 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)4666 static inline void netif_keep_dst(struct net_device *dev)
4667 {
4668 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4669 }
4670
4671 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)4672 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4673 {
4674 /* TODO: reserve and use an additional IFF bit, if we get more users */
4675 return dev->priv_flags & IFF_MACSEC;
4676 }
4677
4678 extern struct pernet_operations __net_initdata loopback_net_ops;
4679
4680 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4681
4682 /* netdev_printk helpers, similar to dev_printk */
4683
netdev_name(const struct net_device * dev)4684 static inline const char *netdev_name(const struct net_device *dev)
4685 {
4686 if (!dev->name[0] || strchr(dev->name, '%'))
4687 return "(unnamed net_device)";
4688 return dev->name;
4689 }
4690
netdev_unregistering(const struct net_device * dev)4691 static inline bool netdev_unregistering(const struct net_device *dev)
4692 {
4693 return dev->reg_state == NETREG_UNREGISTERING;
4694 }
4695
netdev_reg_state(const struct net_device * dev)4696 static inline const char *netdev_reg_state(const struct net_device *dev)
4697 {
4698 switch (dev->reg_state) {
4699 case NETREG_UNINITIALIZED: return " (uninitialized)";
4700 case NETREG_REGISTERED: return "";
4701 case NETREG_UNREGISTERING: return " (unregistering)";
4702 case NETREG_UNREGISTERED: return " (unregistered)";
4703 case NETREG_RELEASED: return " (released)";
4704 case NETREG_DUMMY: return " (dummy)";
4705 }
4706
4707 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4708 return " (unknown)";
4709 }
4710
4711 __printf(3, 4) __cold
4712 void netdev_printk(const char *level, const struct net_device *dev,
4713 const char *format, ...);
4714 __printf(2, 3) __cold
4715 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4716 __printf(2, 3) __cold
4717 void netdev_alert(const struct net_device *dev, const char *format, ...);
4718 __printf(2, 3) __cold
4719 void netdev_crit(const struct net_device *dev, const char *format, ...);
4720 __printf(2, 3) __cold
4721 void netdev_err(const struct net_device *dev, const char *format, ...);
4722 __printf(2, 3) __cold
4723 void netdev_warn(const struct net_device *dev, const char *format, ...);
4724 __printf(2, 3) __cold
4725 void netdev_notice(const struct net_device *dev, const char *format, ...);
4726 __printf(2, 3) __cold
4727 void netdev_info(const struct net_device *dev, const char *format, ...);
4728
4729 #define netdev_level_once(level, dev, fmt, ...) \
4730 do { \
4731 static bool __print_once __read_mostly; \
4732 \
4733 if (!__print_once) { \
4734 __print_once = true; \
4735 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4736 } \
4737 } while (0)
4738
4739 #define netdev_emerg_once(dev, fmt, ...) \
4740 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4741 #define netdev_alert_once(dev, fmt, ...) \
4742 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4743 #define netdev_crit_once(dev, fmt, ...) \
4744 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4745 #define netdev_err_once(dev, fmt, ...) \
4746 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4747 #define netdev_warn_once(dev, fmt, ...) \
4748 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4749 #define netdev_notice_once(dev, fmt, ...) \
4750 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4751 #define netdev_info_once(dev, fmt, ...) \
4752 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4753
4754 #define MODULE_ALIAS_NETDEV(device) \
4755 MODULE_ALIAS("netdev-" device)
4756
4757 #if defined(CONFIG_DYNAMIC_DEBUG)
4758 #define netdev_dbg(__dev, format, args...) \
4759 do { \
4760 dynamic_netdev_dbg(__dev, format, ##args); \
4761 } while (0)
4762 #elif defined(DEBUG)
4763 #define netdev_dbg(__dev, format, args...) \
4764 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4765 #else
4766 #define netdev_dbg(__dev, format, args...) \
4767 ({ \
4768 if (0) \
4769 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4770 })
4771 #endif
4772
4773 #if defined(VERBOSE_DEBUG)
4774 #define netdev_vdbg netdev_dbg
4775 #else
4776
4777 #define netdev_vdbg(dev, format, args...) \
4778 ({ \
4779 if (0) \
4780 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4781 0; \
4782 })
4783 #endif
4784
4785 /*
4786 * netdev_WARN() acts like dev_printk(), but with the key difference
4787 * of using a WARN/WARN_ON to get the message out, including the
4788 * file/line information and a backtrace.
4789 */
4790 #define netdev_WARN(dev, format, args...) \
4791 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4792 netdev_reg_state(dev), ##args)
4793
4794 #define netdev_WARN_ONCE(dev, format, args...) \
4795 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4796 netdev_reg_state(dev), ##args)
4797
4798 /* netif printk helpers, similar to netdev_printk */
4799
4800 #define netif_printk(priv, type, level, dev, fmt, args...) \
4801 do { \
4802 if (netif_msg_##type(priv)) \
4803 netdev_printk(level, (dev), fmt, ##args); \
4804 } while (0)
4805
4806 #define netif_level(level, priv, type, dev, fmt, args...) \
4807 do { \
4808 if (netif_msg_##type(priv)) \
4809 netdev_##level(dev, fmt, ##args); \
4810 } while (0)
4811
4812 #define netif_emerg(priv, type, dev, fmt, args...) \
4813 netif_level(emerg, priv, type, dev, fmt, ##args)
4814 #define netif_alert(priv, type, dev, fmt, args...) \
4815 netif_level(alert, priv, type, dev, fmt, ##args)
4816 #define netif_crit(priv, type, dev, fmt, args...) \
4817 netif_level(crit, priv, type, dev, fmt, ##args)
4818 #define netif_err(priv, type, dev, fmt, args...) \
4819 netif_level(err, priv, type, dev, fmt, ##args)
4820 #define netif_warn(priv, type, dev, fmt, args...) \
4821 netif_level(warn, priv, type, dev, fmt, ##args)
4822 #define netif_notice(priv, type, dev, fmt, args...) \
4823 netif_level(notice, priv, type, dev, fmt, ##args)
4824 #define netif_info(priv, type, dev, fmt, args...) \
4825 netif_level(info, priv, type, dev, fmt, ##args)
4826
4827 #if defined(CONFIG_DYNAMIC_DEBUG)
4828 #define netif_dbg(priv, type, netdev, format, args...) \
4829 do { \
4830 if (netif_msg_##type(priv)) \
4831 dynamic_netdev_dbg(netdev, format, ##args); \
4832 } while (0)
4833 #elif defined(DEBUG)
4834 #define netif_dbg(priv, type, dev, format, args...) \
4835 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4836 #else
4837 #define netif_dbg(priv, type, dev, format, args...) \
4838 ({ \
4839 if (0) \
4840 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4841 0; \
4842 })
4843 #endif
4844
4845 /* if @cond then downgrade to debug, else print at @level */
4846 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4847 do { \
4848 if (cond) \
4849 netif_dbg(priv, type, netdev, fmt, ##args); \
4850 else \
4851 netif_ ## level(priv, type, netdev, fmt, ##args); \
4852 } while (0)
4853
4854 #if defined(VERBOSE_DEBUG)
4855 #define netif_vdbg netif_dbg
4856 #else
4857 #define netif_vdbg(priv, type, dev, format, args...) \
4858 ({ \
4859 if (0) \
4860 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4861 0; \
4862 })
4863 #endif
4864
4865 /*
4866 * The list of packet types we will receive (as opposed to discard)
4867 * and the routines to invoke.
4868 *
4869 * Why 16. Because with 16 the only overlap we get on a hash of the
4870 * low nibble of the protocol value is RARP/SNAP/X.25.
4871 *
4872 * 0800 IP
4873 * 0001 802.3
4874 * 0002 AX.25
4875 * 0004 802.2
4876 * 8035 RARP
4877 * 0005 SNAP
4878 * 0805 X.25
4879 * 0806 ARP
4880 * 8137 IPX
4881 * 0009 Localtalk
4882 * 86DD IPv6
4883 */
4884 #define PTYPE_HASH_SIZE (16)
4885 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4886
4887 extern struct net_device *blackhole_netdev;
4888
4889 #endif /* _LINUX_NETDEVICE_H */
4890