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 unsigned int mtu;
1871 unsigned int min_mtu;
1872 unsigned int max_mtu;
1873 unsigned short type;
1874 unsigned short hard_header_len;
1875 unsigned char min_header_len;
1876
1877 unsigned short needed_headroom;
1878 unsigned short needed_tailroom;
1879
1880 /* Interface address info. */
1881 unsigned char perm_addr[MAX_ADDR_LEN];
1882 unsigned char addr_assign_type;
1883 unsigned char addr_len;
1884 unsigned char upper_level;
1885 unsigned char lower_level;
1886 unsigned short neigh_priv_len;
1887 unsigned short dev_id;
1888 unsigned short dev_port;
1889 spinlock_t addr_list_lock;
1890 unsigned char name_assign_type;
1891 bool uc_promisc;
1892 struct netdev_hw_addr_list uc;
1893 struct netdev_hw_addr_list mc;
1894 struct netdev_hw_addr_list dev_addrs;
1895
1896 #ifdef CONFIG_SYSFS
1897 struct kset *queues_kset;
1898 #endif
1899 unsigned int promiscuity;
1900 unsigned int allmulti;
1901
1902
1903 /* Protocol-specific pointers */
1904
1905 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1906 struct vlan_info __rcu *vlan_info;
1907 #endif
1908 #if IS_ENABLED(CONFIG_NET_DSA)
1909 struct dsa_port *dsa_ptr;
1910 #endif
1911 #if IS_ENABLED(CONFIG_TIPC)
1912 struct tipc_bearer __rcu *tipc_ptr;
1913 #endif
1914 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1915 void *atalk_ptr;
1916 #endif
1917 struct in_device __rcu *ip_ptr;
1918 #if IS_ENABLED(CONFIG_DECNET)
1919 struct dn_dev __rcu *dn_ptr;
1920 #endif
1921 struct inet6_dev __rcu *ip6_ptr;
1922 #if IS_ENABLED(CONFIG_AX25)
1923 void *ax25_ptr;
1924 #endif
1925 struct wireless_dev *ieee80211_ptr;
1926 struct wpan_dev *ieee802154_ptr;
1927 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1928 struct mpls_dev __rcu *mpls_ptr;
1929 #endif
1930
1931 /*
1932 * Cache lines mostly used on receive path (including eth_type_trans())
1933 */
1934 /* Interface address info used in eth_type_trans() */
1935 unsigned char *dev_addr;
1936
1937 struct netdev_rx_queue *_rx;
1938 unsigned int num_rx_queues;
1939 unsigned int real_num_rx_queues;
1940
1941 struct bpf_prog __rcu *xdp_prog;
1942 unsigned long gro_flush_timeout;
1943 rx_handler_func_t __rcu *rx_handler;
1944 void __rcu *rx_handler_data;
1945
1946 #ifdef CONFIG_NET_CLS_ACT
1947 struct mini_Qdisc __rcu *miniq_ingress;
1948 #endif
1949 struct netdev_queue __rcu *ingress_queue;
1950 #ifdef CONFIG_NETFILTER_INGRESS
1951 struct nf_hook_entries __rcu *nf_hooks_ingress;
1952 #endif
1953
1954 unsigned char broadcast[MAX_ADDR_LEN];
1955 #ifdef CONFIG_RFS_ACCEL
1956 struct cpu_rmap *rx_cpu_rmap;
1957 #endif
1958 struct hlist_node index_hlist;
1959
1960 /*
1961 * Cache lines mostly used on transmit path
1962 */
1963 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1964 unsigned int num_tx_queues;
1965 unsigned int real_num_tx_queues;
1966 struct Qdisc *qdisc;
1967 #ifdef CONFIG_NET_SCHED
1968 DECLARE_HASHTABLE (qdisc_hash, 4);
1969 #endif
1970 unsigned int tx_queue_len;
1971 spinlock_t tx_global_lock;
1972 int watchdog_timeo;
1973
1974 #ifdef CONFIG_XPS
1975 struct xps_dev_maps __rcu *xps_cpus_map;
1976 struct xps_dev_maps __rcu *xps_rxqs_map;
1977 #endif
1978 #ifdef CONFIG_NET_CLS_ACT
1979 struct mini_Qdisc __rcu *miniq_egress;
1980 #endif
1981
1982 /* These may be needed for future network-power-down code. */
1983 struct timer_list watchdog_timer;
1984
1985 int __percpu *pcpu_refcnt;
1986 struct list_head todo_list;
1987
1988 struct list_head link_watch_list;
1989
1990 enum { NETREG_UNINITIALIZED=0,
1991 NETREG_REGISTERED, /* completed register_netdevice */
1992 NETREG_UNREGISTERING, /* called unregister_netdevice */
1993 NETREG_UNREGISTERED, /* completed unregister todo */
1994 NETREG_RELEASED, /* called free_netdev */
1995 NETREG_DUMMY, /* dummy device for NAPI poll */
1996 } reg_state:8;
1997
1998 bool dismantle;
1999
2000 enum {
2001 RTNL_LINK_INITIALIZED,
2002 RTNL_LINK_INITIALIZING,
2003 } rtnl_link_state:16;
2004
2005 bool needs_free_netdev;
2006 void (*priv_destructor)(struct net_device *dev);
2007
2008 #ifdef CONFIG_NETPOLL
2009 struct netpoll_info __rcu *npinfo;
2010 #endif
2011
2012 possible_net_t nd_net;
2013
2014 /* mid-layer private */
2015 union {
2016 void *ml_priv;
2017 struct pcpu_lstats __percpu *lstats;
2018 struct pcpu_sw_netstats __percpu *tstats;
2019 struct pcpu_dstats __percpu *dstats;
2020 };
2021
2022 #if IS_ENABLED(CONFIG_GARP)
2023 struct garp_port __rcu *garp_port;
2024 #endif
2025 #if IS_ENABLED(CONFIG_MRP)
2026 struct mrp_port __rcu *mrp_port;
2027 #endif
2028
2029 struct device dev;
2030 const struct attribute_group *sysfs_groups[4];
2031 const struct attribute_group *sysfs_rx_queue_group;
2032
2033 const struct rtnl_link_ops *rtnl_link_ops;
2034
2035 /* for setting kernel sock attribute on TCP connection setup */
2036 #define GSO_MAX_SIZE 65536
2037 unsigned int gso_max_size;
2038 #define GSO_MAX_SEGS 65535
2039 u16 gso_max_segs;
2040
2041 #ifdef CONFIG_DCB
2042 const struct dcbnl_rtnl_ops *dcbnl_ops;
2043 #endif
2044 s16 num_tc;
2045 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2046 u8 prio_tc_map[TC_BITMASK + 1];
2047
2048 #if IS_ENABLED(CONFIG_FCOE)
2049 unsigned int fcoe_ddp_xid;
2050 #endif
2051 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2052 struct netprio_map __rcu *priomap;
2053 #endif
2054 struct phy_device *phydev;
2055 struct sfp_bus *sfp_bus;
2056 struct lock_class_key qdisc_tx_busylock_key;
2057 struct lock_class_key qdisc_running_key;
2058 struct lock_class_key qdisc_xmit_lock_key;
2059 struct lock_class_key addr_list_lock_key;
2060 bool proto_down;
2061 unsigned wol_enabled:1;
2062 };
2063 #define to_net_dev(d) container_of(d, struct net_device, dev)
2064
netif_elide_gro(const struct net_device * dev)2065 static inline bool netif_elide_gro(const struct net_device *dev)
2066 {
2067 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2068 return true;
2069 return false;
2070 }
2071
2072 #define NETDEV_ALIGN 32
2073
2074 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2075 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2076 {
2077 return dev->prio_tc_map[prio & TC_BITMASK];
2078 }
2079
2080 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2081 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2082 {
2083 if (tc >= dev->num_tc)
2084 return -EINVAL;
2085
2086 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2087 return 0;
2088 }
2089
2090 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2091 void netdev_reset_tc(struct net_device *dev);
2092 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2093 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2094
2095 static inline
netdev_get_num_tc(struct net_device * dev)2096 int netdev_get_num_tc(struct net_device *dev)
2097 {
2098 return dev->num_tc;
2099 }
2100
2101 void netdev_unbind_sb_channel(struct net_device *dev,
2102 struct net_device *sb_dev);
2103 int netdev_bind_sb_channel_queue(struct net_device *dev,
2104 struct net_device *sb_dev,
2105 u8 tc, u16 count, u16 offset);
2106 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2107 static inline int netdev_get_sb_channel(struct net_device *dev)
2108 {
2109 return max_t(int, -dev->num_tc, 0);
2110 }
2111
2112 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2113 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2114 unsigned int index)
2115 {
2116 return &dev->_tx[index];
2117 }
2118
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2119 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2120 const struct sk_buff *skb)
2121 {
2122 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2123 }
2124
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2125 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2126 void (*f)(struct net_device *,
2127 struct netdev_queue *,
2128 void *),
2129 void *arg)
2130 {
2131 unsigned int i;
2132
2133 for (i = 0; i < dev->num_tx_queues; i++)
2134 f(dev, &dev->_tx[i], arg);
2135 }
2136
2137 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2138 struct net_device *sb_dev);
2139 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2140 struct sk_buff *skb,
2141 struct net_device *sb_dev);
2142
2143 /* returns the headroom that the master device needs to take in account
2144 * when forwarding to this dev
2145 */
netdev_get_fwd_headroom(struct net_device * dev)2146 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2147 {
2148 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2149 }
2150
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2151 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2152 {
2153 if (dev->netdev_ops->ndo_set_rx_headroom)
2154 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2155 }
2156
2157 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2158 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2159 {
2160 netdev_set_rx_headroom(dev, -1);
2161 }
2162
2163 /*
2164 * Net namespace inlines
2165 */
2166 static inline
dev_net(const struct net_device * dev)2167 struct net *dev_net(const struct net_device *dev)
2168 {
2169 return read_pnet(&dev->nd_net);
2170 }
2171
2172 static inline
dev_net_set(struct net_device * dev,struct net * net)2173 void dev_net_set(struct net_device *dev, struct net *net)
2174 {
2175 write_pnet(&dev->nd_net, net);
2176 }
2177
2178 /**
2179 * netdev_priv - access network device private data
2180 * @dev: network device
2181 *
2182 * Get network device private data
2183 */
netdev_priv(const struct net_device * dev)2184 static inline void *netdev_priv(const struct net_device *dev)
2185 {
2186 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2187 }
2188
2189 /* Set the sysfs physical device reference for the network logical device
2190 * if set prior to registration will cause a symlink during initialization.
2191 */
2192 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2193
2194 /* Set the sysfs device type for the network logical device to allow
2195 * fine-grained identification of different network device types. For
2196 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2197 */
2198 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2199
2200 /* Default NAPI poll() weight
2201 * Device drivers are strongly advised to not use bigger value
2202 */
2203 #define NAPI_POLL_WEIGHT 64
2204
2205 /**
2206 * netif_napi_add - initialize a NAPI context
2207 * @dev: network device
2208 * @napi: NAPI context
2209 * @poll: polling function
2210 * @weight: default weight
2211 *
2212 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2213 * *any* of the other NAPI-related functions.
2214 */
2215 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2216 int (*poll)(struct napi_struct *, int), int weight);
2217
2218 /**
2219 * netif_tx_napi_add - initialize a NAPI context
2220 * @dev: network device
2221 * @napi: NAPI context
2222 * @poll: polling function
2223 * @weight: default weight
2224 *
2225 * This variant of netif_napi_add() should be used from drivers using NAPI
2226 * to exclusively poll a TX queue.
2227 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2228 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2229 static inline void netif_tx_napi_add(struct net_device *dev,
2230 struct napi_struct *napi,
2231 int (*poll)(struct napi_struct *, int),
2232 int weight)
2233 {
2234 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2235 netif_napi_add(dev, napi, poll, weight);
2236 }
2237
2238 /**
2239 * netif_napi_del - remove a NAPI context
2240 * @napi: NAPI context
2241 *
2242 * netif_napi_del() removes a NAPI context from the network device NAPI list
2243 */
2244 void netif_napi_del(struct napi_struct *napi);
2245
2246 struct napi_gro_cb {
2247 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2248 void *frag0;
2249
2250 /* Length of frag0. */
2251 unsigned int frag0_len;
2252
2253 /* This indicates where we are processing relative to skb->data. */
2254 int data_offset;
2255
2256 /* This is non-zero if the packet cannot be merged with the new skb. */
2257 u16 flush;
2258
2259 /* Save the IP ID here and check when we get to the transport layer */
2260 u16 flush_id;
2261
2262 /* Number of segments aggregated. */
2263 u16 count;
2264
2265 /* Start offset for remote checksum offload */
2266 u16 gro_remcsum_start;
2267
2268 /* jiffies when first packet was created/queued */
2269 unsigned long age;
2270
2271 /* Used in ipv6_gro_receive() and foo-over-udp */
2272 u16 proto;
2273
2274 /* This is non-zero if the packet may be of the same flow. */
2275 u8 same_flow:1;
2276
2277 /* Used in tunnel GRO receive */
2278 u8 encap_mark:1;
2279
2280 /* GRO checksum is valid */
2281 u8 csum_valid:1;
2282
2283 /* Number of checksums via CHECKSUM_UNNECESSARY */
2284 u8 csum_cnt:3;
2285
2286 /* Free the skb? */
2287 u8 free:2;
2288 #define NAPI_GRO_FREE 1
2289 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2290
2291 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2292 u8 is_ipv6:1;
2293
2294 /* Used in GRE, set in fou/gue_gro_receive */
2295 u8 is_fou:1;
2296
2297 /* Used to determine if flush_id can be ignored */
2298 u8 is_atomic:1;
2299
2300 /* Number of gro_receive callbacks this packet already went through */
2301 u8 recursion_counter:4;
2302
2303 /* 1 bit hole */
2304
2305 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2306 __wsum csum;
2307
2308 /* used in skb_gro_receive() slow path */
2309 struct sk_buff *last;
2310 };
2311
2312 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2313
2314 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2315 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2316 {
2317 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2318 }
2319
2320 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2321 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2322 struct list_head *head,
2323 struct sk_buff *skb)
2324 {
2325 if (unlikely(gro_recursion_inc_test(skb))) {
2326 NAPI_GRO_CB(skb)->flush |= 1;
2327 return NULL;
2328 }
2329
2330 return cb(head, skb);
2331 }
2332
2333 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2334 struct sk_buff *);
2335 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2336 struct sock *sk,
2337 struct list_head *head,
2338 struct sk_buff *skb)
2339 {
2340 if (unlikely(gro_recursion_inc_test(skb))) {
2341 NAPI_GRO_CB(skb)->flush |= 1;
2342 return NULL;
2343 }
2344
2345 return cb(sk, head, skb);
2346 }
2347
2348 struct packet_type {
2349 __be16 type; /* This is really htons(ether_type). */
2350 bool ignore_outgoing;
2351 struct net_device *dev; /* NULL is wildcarded here */
2352 int (*func) (struct sk_buff *,
2353 struct net_device *,
2354 struct packet_type *,
2355 struct net_device *);
2356 void (*list_func) (struct list_head *,
2357 struct packet_type *,
2358 struct net_device *);
2359 bool (*id_match)(struct packet_type *ptype,
2360 struct sock *sk);
2361 void *af_packet_priv;
2362 struct list_head list;
2363 };
2364
2365 struct offload_callbacks {
2366 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2367 netdev_features_t features);
2368 struct sk_buff *(*gro_receive)(struct list_head *head,
2369 struct sk_buff *skb);
2370 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2371 };
2372
2373 struct packet_offload {
2374 __be16 type; /* This is really htons(ether_type). */
2375 u16 priority;
2376 struct offload_callbacks callbacks;
2377 struct list_head list;
2378 };
2379
2380 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2381 struct pcpu_sw_netstats {
2382 u64 rx_packets;
2383 u64 rx_bytes;
2384 u64 tx_packets;
2385 u64 tx_bytes;
2386 struct u64_stats_sync syncp;
2387 } __aligned(4 * sizeof(u64));
2388
2389 struct pcpu_lstats {
2390 u64 packets;
2391 u64 bytes;
2392 struct u64_stats_sync syncp;
2393 } __aligned(2 * sizeof(u64));
2394
2395 #define __netdev_alloc_pcpu_stats(type, gfp) \
2396 ({ \
2397 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2398 if (pcpu_stats) { \
2399 int __cpu; \
2400 for_each_possible_cpu(__cpu) { \
2401 typeof(type) *stat; \
2402 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2403 u64_stats_init(&stat->syncp); \
2404 } \
2405 } \
2406 pcpu_stats; \
2407 })
2408
2409 #define netdev_alloc_pcpu_stats(type) \
2410 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2411
2412 enum netdev_lag_tx_type {
2413 NETDEV_LAG_TX_TYPE_UNKNOWN,
2414 NETDEV_LAG_TX_TYPE_RANDOM,
2415 NETDEV_LAG_TX_TYPE_BROADCAST,
2416 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2417 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2418 NETDEV_LAG_TX_TYPE_HASH,
2419 };
2420
2421 enum netdev_lag_hash {
2422 NETDEV_LAG_HASH_NONE,
2423 NETDEV_LAG_HASH_L2,
2424 NETDEV_LAG_HASH_L34,
2425 NETDEV_LAG_HASH_L23,
2426 NETDEV_LAG_HASH_E23,
2427 NETDEV_LAG_HASH_E34,
2428 NETDEV_LAG_HASH_UNKNOWN,
2429 };
2430
2431 struct netdev_lag_upper_info {
2432 enum netdev_lag_tx_type tx_type;
2433 enum netdev_lag_hash hash_type;
2434 };
2435
2436 struct netdev_lag_lower_state_info {
2437 u8 link_up : 1,
2438 tx_enabled : 1;
2439 };
2440
2441 #include <linux/notifier.h>
2442
2443 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2444 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2445 * adding new types.
2446 */
2447 enum netdev_cmd {
2448 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2449 NETDEV_DOWN,
2450 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2451 detected a hardware crash and restarted
2452 - we can use this eg to kick tcp sessions
2453 once done */
2454 NETDEV_CHANGE, /* Notify device state change */
2455 NETDEV_REGISTER,
2456 NETDEV_UNREGISTER,
2457 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2458 NETDEV_CHANGEADDR, /* notify after the address change */
2459 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2460 NETDEV_GOING_DOWN,
2461 NETDEV_CHANGENAME,
2462 NETDEV_FEAT_CHANGE,
2463 NETDEV_BONDING_FAILOVER,
2464 NETDEV_PRE_UP,
2465 NETDEV_PRE_TYPE_CHANGE,
2466 NETDEV_POST_TYPE_CHANGE,
2467 NETDEV_POST_INIT,
2468 NETDEV_RELEASE,
2469 NETDEV_NOTIFY_PEERS,
2470 NETDEV_JOIN,
2471 NETDEV_CHANGEUPPER,
2472 NETDEV_RESEND_IGMP,
2473 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2474 NETDEV_CHANGEINFODATA,
2475 NETDEV_BONDING_INFO,
2476 NETDEV_PRECHANGEUPPER,
2477 NETDEV_CHANGELOWERSTATE,
2478 NETDEV_UDP_TUNNEL_PUSH_INFO,
2479 NETDEV_UDP_TUNNEL_DROP_INFO,
2480 NETDEV_CHANGE_TX_QUEUE_LEN,
2481 NETDEV_CVLAN_FILTER_PUSH_INFO,
2482 NETDEV_CVLAN_FILTER_DROP_INFO,
2483 NETDEV_SVLAN_FILTER_PUSH_INFO,
2484 NETDEV_SVLAN_FILTER_DROP_INFO,
2485 };
2486 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2487
2488 int register_netdevice_notifier(struct notifier_block *nb);
2489 int unregister_netdevice_notifier(struct notifier_block *nb);
2490
2491 struct netdev_notifier_info {
2492 struct net_device *dev;
2493 struct netlink_ext_ack *extack;
2494 };
2495
2496 struct netdev_notifier_info_ext {
2497 struct netdev_notifier_info info; /* must be first */
2498 union {
2499 u32 mtu;
2500 } ext;
2501 };
2502
2503 struct netdev_notifier_change_info {
2504 struct netdev_notifier_info info; /* must be first */
2505 unsigned int flags_changed;
2506 };
2507
2508 struct netdev_notifier_changeupper_info {
2509 struct netdev_notifier_info info; /* must be first */
2510 struct net_device *upper_dev; /* new upper dev */
2511 bool master; /* is upper dev master */
2512 bool linking; /* is the notification for link or unlink */
2513 void *upper_info; /* upper dev info */
2514 };
2515
2516 struct netdev_notifier_changelowerstate_info {
2517 struct netdev_notifier_info info; /* must be first */
2518 void *lower_state_info; /* is lower dev state */
2519 };
2520
2521 struct netdev_notifier_pre_changeaddr_info {
2522 struct netdev_notifier_info info; /* must be first */
2523 const unsigned char *dev_addr;
2524 };
2525
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2526 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2527 struct net_device *dev)
2528 {
2529 info->dev = dev;
2530 info->extack = NULL;
2531 }
2532
2533 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2534 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2535 {
2536 return info->dev;
2537 }
2538
2539 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2540 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2541 {
2542 return info->extack;
2543 }
2544
2545 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2546
2547
2548 extern rwlock_t dev_base_lock; /* Device list lock */
2549
2550 #define for_each_netdev(net, d) \
2551 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2552 #define for_each_netdev_reverse(net, d) \
2553 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2554 #define for_each_netdev_rcu(net, d) \
2555 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2556 #define for_each_netdev_safe(net, d, n) \
2557 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2558 #define for_each_netdev_continue(net, d) \
2559 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2560 #define for_each_netdev_continue_rcu(net, d) \
2561 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2562 #define for_each_netdev_in_bond_rcu(bond, slave) \
2563 for_each_netdev_rcu(&init_net, slave) \
2564 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2565 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2566
next_net_device(struct net_device * dev)2567 static inline struct net_device *next_net_device(struct net_device *dev)
2568 {
2569 struct list_head *lh;
2570 struct net *net;
2571
2572 net = dev_net(dev);
2573 lh = dev->dev_list.next;
2574 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2575 }
2576
next_net_device_rcu(struct net_device * dev)2577 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2578 {
2579 struct list_head *lh;
2580 struct net *net;
2581
2582 net = dev_net(dev);
2583 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2584 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2585 }
2586
first_net_device(struct net * net)2587 static inline struct net_device *first_net_device(struct net *net)
2588 {
2589 return list_empty(&net->dev_base_head) ? NULL :
2590 net_device_entry(net->dev_base_head.next);
2591 }
2592
first_net_device_rcu(struct net * net)2593 static inline struct net_device *first_net_device_rcu(struct net *net)
2594 {
2595 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2596
2597 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2598 }
2599
2600 int netdev_boot_setup_check(struct net_device *dev);
2601 unsigned long netdev_boot_base(const char *prefix, int unit);
2602 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2603 const char *hwaddr);
2604 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2605 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2606 void dev_add_pack(struct packet_type *pt);
2607 void dev_remove_pack(struct packet_type *pt);
2608 void __dev_remove_pack(struct packet_type *pt);
2609 void dev_add_offload(struct packet_offload *po);
2610 void dev_remove_offload(struct packet_offload *po);
2611
2612 int dev_get_iflink(const struct net_device *dev);
2613 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2614 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2615 unsigned short mask);
2616 struct net_device *dev_get_by_name(struct net *net, const char *name);
2617 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2618 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2619 int dev_alloc_name(struct net_device *dev, const char *name);
2620 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2621 void dev_close(struct net_device *dev);
2622 void dev_close_many(struct list_head *head, bool unlink);
2623 void dev_disable_lro(struct net_device *dev);
2624 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2625 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2626 struct net_device *sb_dev);
2627 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2628 struct net_device *sb_dev);
2629 int dev_queue_xmit(struct sk_buff *skb);
2630 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2631 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2632 int register_netdevice(struct net_device *dev);
2633 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2634 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2635 static inline void unregister_netdevice(struct net_device *dev)
2636 {
2637 unregister_netdevice_queue(dev, NULL);
2638 }
2639
2640 int netdev_refcnt_read(const struct net_device *dev);
2641 void free_netdev(struct net_device *dev);
2642 void netdev_freemem(struct net_device *dev);
2643 void synchronize_net(void);
2644 int init_dummy_netdev(struct net_device *dev);
2645
2646 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2647 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2648 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2649 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2650 int netdev_get_name(struct net *net, char *name, int ifindex);
2651 int dev_restart(struct net_device *dev);
2652 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2653
skb_gro_offset(const struct sk_buff * skb)2654 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2655 {
2656 return NAPI_GRO_CB(skb)->data_offset;
2657 }
2658
skb_gro_len(const struct sk_buff * skb)2659 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2660 {
2661 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2662 }
2663
skb_gro_pull(struct sk_buff * skb,unsigned int len)2664 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2665 {
2666 NAPI_GRO_CB(skb)->data_offset += len;
2667 }
2668
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2669 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2670 unsigned int offset)
2671 {
2672 return NAPI_GRO_CB(skb)->frag0 + offset;
2673 }
2674
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2675 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2676 {
2677 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2678 }
2679
skb_gro_frag0_invalidate(struct sk_buff * skb)2680 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2681 {
2682 NAPI_GRO_CB(skb)->frag0 = NULL;
2683 NAPI_GRO_CB(skb)->frag0_len = 0;
2684 }
2685
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2686 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2687 unsigned int offset)
2688 {
2689 if (!pskb_may_pull(skb, hlen))
2690 return NULL;
2691
2692 skb_gro_frag0_invalidate(skb);
2693 return skb->data + offset;
2694 }
2695
skb_gro_network_header(struct sk_buff * skb)2696 static inline void *skb_gro_network_header(struct sk_buff *skb)
2697 {
2698 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2699 skb_network_offset(skb);
2700 }
2701
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2702 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2703 const void *start, unsigned int len)
2704 {
2705 if (NAPI_GRO_CB(skb)->csum_valid)
2706 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2707 csum_partial(start, len, 0));
2708 }
2709
2710 /* GRO checksum functions. These are logical equivalents of the normal
2711 * checksum functions (in skbuff.h) except that they operate on the GRO
2712 * offsets and fields in sk_buff.
2713 */
2714
2715 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2716
skb_at_gro_remcsum_start(struct sk_buff * skb)2717 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2718 {
2719 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2720 }
2721
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2722 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2723 bool zero_okay,
2724 __sum16 check)
2725 {
2726 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2727 skb_checksum_start_offset(skb) <
2728 skb_gro_offset(skb)) &&
2729 !skb_at_gro_remcsum_start(skb) &&
2730 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2731 (!zero_okay || check));
2732 }
2733
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2734 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2735 __wsum psum)
2736 {
2737 if (NAPI_GRO_CB(skb)->csum_valid &&
2738 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2739 return 0;
2740
2741 NAPI_GRO_CB(skb)->csum = psum;
2742
2743 return __skb_gro_checksum_complete(skb);
2744 }
2745
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)2746 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2747 {
2748 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2749 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2750 NAPI_GRO_CB(skb)->csum_cnt--;
2751 } else {
2752 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2753 * verified a new top level checksum or an encapsulated one
2754 * during GRO. This saves work if we fallback to normal path.
2755 */
2756 __skb_incr_checksum_unnecessary(skb);
2757 }
2758 }
2759
2760 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2761 compute_pseudo) \
2762 ({ \
2763 __sum16 __ret = 0; \
2764 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2765 __ret = __skb_gro_checksum_validate_complete(skb, \
2766 compute_pseudo(skb, proto)); \
2767 if (!__ret) \
2768 skb_gro_incr_csum_unnecessary(skb); \
2769 __ret; \
2770 })
2771
2772 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2773 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2774
2775 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2776 compute_pseudo) \
2777 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2778
2779 #define skb_gro_checksum_simple_validate(skb) \
2780 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2781
__skb_gro_checksum_convert_check(struct sk_buff * skb)2782 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2783 {
2784 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2785 !NAPI_GRO_CB(skb)->csum_valid);
2786 }
2787
__skb_gro_checksum_convert(struct sk_buff * skb,__sum16 check,__wsum pseudo)2788 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2789 __sum16 check, __wsum pseudo)
2790 {
2791 NAPI_GRO_CB(skb)->csum = ~pseudo;
2792 NAPI_GRO_CB(skb)->csum_valid = 1;
2793 }
2794
2795 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2796 do { \
2797 if (__skb_gro_checksum_convert_check(skb)) \
2798 __skb_gro_checksum_convert(skb, check, \
2799 compute_pseudo(skb, proto)); \
2800 } while (0)
2801
2802 struct gro_remcsum {
2803 int offset;
2804 __wsum delta;
2805 };
2806
skb_gro_remcsum_init(struct gro_remcsum * grc)2807 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2808 {
2809 grc->offset = 0;
2810 grc->delta = 0;
2811 }
2812
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)2813 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2814 unsigned int off, size_t hdrlen,
2815 int start, int offset,
2816 struct gro_remcsum *grc,
2817 bool nopartial)
2818 {
2819 __wsum delta;
2820 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2821
2822 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2823
2824 if (!nopartial) {
2825 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2826 return ptr;
2827 }
2828
2829 ptr = skb_gro_header_fast(skb, off);
2830 if (skb_gro_header_hard(skb, off + plen)) {
2831 ptr = skb_gro_header_slow(skb, off + plen, off);
2832 if (!ptr)
2833 return NULL;
2834 }
2835
2836 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2837 start, offset);
2838
2839 /* Adjust skb->csum since we changed the packet */
2840 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2841
2842 grc->offset = off + hdrlen + offset;
2843 grc->delta = delta;
2844
2845 return ptr;
2846 }
2847
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)2848 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2849 struct gro_remcsum *grc)
2850 {
2851 void *ptr;
2852 size_t plen = grc->offset + sizeof(u16);
2853
2854 if (!grc->delta)
2855 return;
2856
2857 ptr = skb_gro_header_fast(skb, grc->offset);
2858 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2859 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2860 if (!ptr)
2861 return;
2862 }
2863
2864 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2865 }
2866
2867 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)2868 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2869 {
2870 if (PTR_ERR(pp) != -EINPROGRESS)
2871 NAPI_GRO_CB(skb)->flush |= flush;
2872 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)2873 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2874 struct sk_buff *pp,
2875 int flush,
2876 struct gro_remcsum *grc)
2877 {
2878 if (PTR_ERR(pp) != -EINPROGRESS) {
2879 NAPI_GRO_CB(skb)->flush |= flush;
2880 skb_gro_remcsum_cleanup(skb, grc);
2881 skb->remcsum_offload = 0;
2882 }
2883 }
2884 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)2885 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2886 {
2887 NAPI_GRO_CB(skb)->flush |= flush;
2888 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)2889 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2890 struct sk_buff *pp,
2891 int flush,
2892 struct gro_remcsum *grc)
2893 {
2894 NAPI_GRO_CB(skb)->flush |= flush;
2895 skb_gro_remcsum_cleanup(skb, grc);
2896 skb->remcsum_offload = 0;
2897 }
2898 #endif
2899
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)2900 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2901 unsigned short type,
2902 const void *daddr, const void *saddr,
2903 unsigned int len)
2904 {
2905 if (!dev->header_ops || !dev->header_ops->create)
2906 return 0;
2907
2908 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2909 }
2910
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)2911 static inline int dev_parse_header(const struct sk_buff *skb,
2912 unsigned char *haddr)
2913 {
2914 const struct net_device *dev = skb->dev;
2915
2916 if (!dev->header_ops || !dev->header_ops->parse)
2917 return 0;
2918 return dev->header_ops->parse(skb, haddr);
2919 }
2920
dev_parse_header_protocol(const struct sk_buff * skb)2921 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2922 {
2923 const struct net_device *dev = skb->dev;
2924
2925 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2926 return 0;
2927 return dev->header_ops->parse_protocol(skb);
2928 }
2929
2930 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)2931 static inline bool dev_validate_header(const struct net_device *dev,
2932 char *ll_header, int len)
2933 {
2934 if (likely(len >= dev->hard_header_len))
2935 return true;
2936 if (len < dev->min_header_len)
2937 return false;
2938
2939 if (capable(CAP_SYS_RAWIO)) {
2940 memset(ll_header + len, 0, dev->hard_header_len - len);
2941 return true;
2942 }
2943
2944 if (dev->header_ops && dev->header_ops->validate)
2945 return dev->header_ops->validate(ll_header, len);
2946
2947 return false;
2948 }
2949
2950 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2951 int len, int size);
2952 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
unregister_gifconf(unsigned int family)2953 static inline int unregister_gifconf(unsigned int family)
2954 {
2955 return register_gifconf(family, NULL);
2956 }
2957
2958 #ifdef CONFIG_NET_FLOW_LIMIT
2959 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2960 struct sd_flow_limit {
2961 u64 count;
2962 unsigned int num_buckets;
2963 unsigned int history_head;
2964 u16 history[FLOW_LIMIT_HISTORY];
2965 u8 buckets[];
2966 };
2967
2968 extern int netdev_flow_limit_table_len;
2969 #endif /* CONFIG_NET_FLOW_LIMIT */
2970
2971 /*
2972 * Incoming packets are placed on per-CPU queues
2973 */
2974 struct softnet_data {
2975 struct list_head poll_list;
2976 struct sk_buff_head process_queue;
2977
2978 /* stats */
2979 unsigned int processed;
2980 unsigned int time_squeeze;
2981 unsigned int received_rps;
2982 #ifdef CONFIG_RPS
2983 struct softnet_data *rps_ipi_list;
2984 #endif
2985 #ifdef CONFIG_NET_FLOW_LIMIT
2986 struct sd_flow_limit __rcu *flow_limit;
2987 #endif
2988 struct Qdisc *output_queue;
2989 struct Qdisc **output_queue_tailp;
2990 struct sk_buff *completion_queue;
2991 #ifdef CONFIG_XFRM_OFFLOAD
2992 struct sk_buff_head xfrm_backlog;
2993 #endif
2994 /* written and read only by owning cpu: */
2995 struct {
2996 u16 recursion;
2997 u8 more;
2998 } xmit;
2999 #ifdef CONFIG_RPS
3000 /* input_queue_head should be written by cpu owning this struct,
3001 * and only read by other cpus. Worth using a cache line.
3002 */
3003 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3004
3005 /* Elements below can be accessed between CPUs for RPS/RFS */
3006 call_single_data_t csd ____cacheline_aligned_in_smp;
3007 struct softnet_data *rps_ipi_next;
3008 unsigned int cpu;
3009 unsigned int input_queue_tail;
3010 #endif
3011 unsigned int dropped;
3012 struct sk_buff_head input_pkt_queue;
3013 struct napi_struct backlog;
3014
3015 };
3016
input_queue_head_incr(struct softnet_data * sd)3017 static inline void input_queue_head_incr(struct softnet_data *sd)
3018 {
3019 #ifdef CONFIG_RPS
3020 sd->input_queue_head++;
3021 #endif
3022 }
3023
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3024 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3025 unsigned int *qtail)
3026 {
3027 #ifdef CONFIG_RPS
3028 *qtail = ++sd->input_queue_tail;
3029 #endif
3030 }
3031
3032 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3033
dev_recursion_level(void)3034 static inline int dev_recursion_level(void)
3035 {
3036 return this_cpu_read(softnet_data.xmit.recursion);
3037 }
3038
3039 #define XMIT_RECURSION_LIMIT 10
dev_xmit_recursion(void)3040 static inline bool dev_xmit_recursion(void)
3041 {
3042 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3043 XMIT_RECURSION_LIMIT);
3044 }
3045
dev_xmit_recursion_inc(void)3046 static inline void dev_xmit_recursion_inc(void)
3047 {
3048 __this_cpu_inc(softnet_data.xmit.recursion);
3049 }
3050
dev_xmit_recursion_dec(void)3051 static inline void dev_xmit_recursion_dec(void)
3052 {
3053 __this_cpu_dec(softnet_data.xmit.recursion);
3054 }
3055
3056 void __netif_schedule(struct Qdisc *q);
3057 void netif_schedule_queue(struct netdev_queue *txq);
3058
netif_tx_schedule_all(struct net_device * dev)3059 static inline void netif_tx_schedule_all(struct net_device *dev)
3060 {
3061 unsigned int i;
3062
3063 for (i = 0; i < dev->num_tx_queues; i++)
3064 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3065 }
3066
netif_tx_start_queue(struct netdev_queue * dev_queue)3067 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3068 {
3069 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3070 }
3071
3072 /**
3073 * netif_start_queue - allow transmit
3074 * @dev: network device
3075 *
3076 * Allow upper layers to call the device hard_start_xmit routine.
3077 */
netif_start_queue(struct net_device * dev)3078 static inline void netif_start_queue(struct net_device *dev)
3079 {
3080 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3081 }
3082
netif_tx_start_all_queues(struct net_device * dev)3083 static inline void netif_tx_start_all_queues(struct net_device *dev)
3084 {
3085 unsigned int i;
3086
3087 for (i = 0; i < dev->num_tx_queues; i++) {
3088 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3089 netif_tx_start_queue(txq);
3090 }
3091 }
3092
3093 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3094
3095 /**
3096 * netif_wake_queue - restart transmit
3097 * @dev: network device
3098 *
3099 * Allow upper layers to call the device hard_start_xmit routine.
3100 * Used for flow control when transmit resources are available.
3101 */
netif_wake_queue(struct net_device * dev)3102 static inline void netif_wake_queue(struct net_device *dev)
3103 {
3104 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3105 }
3106
netif_tx_wake_all_queues(struct net_device * dev)3107 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3108 {
3109 unsigned int i;
3110
3111 for (i = 0; i < dev->num_tx_queues; i++) {
3112 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3113 netif_tx_wake_queue(txq);
3114 }
3115 }
3116
netif_tx_stop_queue(struct netdev_queue * dev_queue)3117 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3118 {
3119 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3120 }
3121
3122 /**
3123 * netif_stop_queue - stop transmitted packets
3124 * @dev: network device
3125 *
3126 * Stop upper layers calling the device hard_start_xmit routine.
3127 * Used for flow control when transmit resources are unavailable.
3128 */
netif_stop_queue(struct net_device * dev)3129 static inline void netif_stop_queue(struct net_device *dev)
3130 {
3131 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3132 }
3133
3134 void netif_tx_stop_all_queues(struct net_device *dev);
3135 void netdev_update_lockdep_key(struct net_device *dev);
3136
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3137 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3138 {
3139 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3140 }
3141
3142 /**
3143 * netif_queue_stopped - test if transmit queue is flowblocked
3144 * @dev: network device
3145 *
3146 * Test if transmit queue on device is currently unable to send.
3147 */
netif_queue_stopped(const struct net_device * dev)3148 static inline bool netif_queue_stopped(const struct net_device *dev)
3149 {
3150 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3151 }
3152
netif_xmit_stopped(const struct netdev_queue * dev_queue)3153 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3154 {
3155 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3156 }
3157
3158 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3159 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3160 {
3161 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3162 }
3163
3164 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3165 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3166 {
3167 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3168 }
3169
3170 /**
3171 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3172 * @dev_queue: pointer to transmit queue
3173 *
3174 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3175 * to give appropriate hint to the CPU.
3176 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3177 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3178 {
3179 #ifdef CONFIG_BQL
3180 prefetchw(&dev_queue->dql.num_queued);
3181 #endif
3182 }
3183
3184 /**
3185 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3186 * @dev_queue: pointer to transmit queue
3187 *
3188 * BQL enabled drivers might use this helper in their TX completion path,
3189 * to give appropriate hint to the CPU.
3190 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3191 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3192 {
3193 #ifdef CONFIG_BQL
3194 prefetchw(&dev_queue->dql.limit);
3195 #endif
3196 }
3197
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3198 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3199 unsigned int bytes)
3200 {
3201 #ifdef CONFIG_BQL
3202 dql_queued(&dev_queue->dql, bytes);
3203
3204 if (likely(dql_avail(&dev_queue->dql) >= 0))
3205 return;
3206
3207 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3208
3209 /*
3210 * The XOFF flag must be set before checking the dql_avail below,
3211 * because in netdev_tx_completed_queue we update the dql_completed
3212 * before checking the XOFF flag.
3213 */
3214 smp_mb();
3215
3216 /* check again in case another CPU has just made room avail */
3217 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3218 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3219 #endif
3220 }
3221
3222 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3223 * that they should not test BQL status themselves.
3224 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3225 * skb of a batch.
3226 * Returns true if the doorbell must be used to kick the NIC.
3227 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3228 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3229 unsigned int bytes,
3230 bool xmit_more)
3231 {
3232 if (xmit_more) {
3233 #ifdef CONFIG_BQL
3234 dql_queued(&dev_queue->dql, bytes);
3235 #endif
3236 return netif_tx_queue_stopped(dev_queue);
3237 }
3238 netdev_tx_sent_queue(dev_queue, bytes);
3239 return true;
3240 }
3241
3242 /**
3243 * netdev_sent_queue - report the number of bytes queued to hardware
3244 * @dev: network device
3245 * @bytes: number of bytes queued to the hardware device queue
3246 *
3247 * Report the number of bytes queued for sending/completion to the network
3248 * device hardware queue. @bytes should be a good approximation and should
3249 * exactly match netdev_completed_queue() @bytes
3250 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3251 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3252 {
3253 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3254 }
3255
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3256 static inline bool __netdev_sent_queue(struct net_device *dev,
3257 unsigned int bytes,
3258 bool xmit_more)
3259 {
3260 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3261 xmit_more);
3262 }
3263
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3264 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3265 unsigned int pkts, unsigned int bytes)
3266 {
3267 #ifdef CONFIG_BQL
3268 if (unlikely(!bytes))
3269 return;
3270
3271 dql_completed(&dev_queue->dql, bytes);
3272
3273 /*
3274 * Without the memory barrier there is a small possiblity that
3275 * netdev_tx_sent_queue will miss the update and cause the queue to
3276 * be stopped forever
3277 */
3278 smp_mb();
3279
3280 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3281 return;
3282
3283 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3284 netif_schedule_queue(dev_queue);
3285 #endif
3286 }
3287
3288 /**
3289 * netdev_completed_queue - report bytes and packets completed by device
3290 * @dev: network device
3291 * @pkts: actual number of packets sent over the medium
3292 * @bytes: actual number of bytes sent over the medium
3293 *
3294 * Report the number of bytes and packets transmitted by the network device
3295 * hardware queue over the physical medium, @bytes must exactly match the
3296 * @bytes amount passed to netdev_sent_queue()
3297 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3298 static inline void netdev_completed_queue(struct net_device *dev,
3299 unsigned int pkts, unsigned int bytes)
3300 {
3301 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3302 }
3303
netdev_tx_reset_queue(struct netdev_queue * q)3304 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3305 {
3306 #ifdef CONFIG_BQL
3307 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3308 dql_reset(&q->dql);
3309 #endif
3310 }
3311
3312 /**
3313 * netdev_reset_queue - reset the packets and bytes count of a network device
3314 * @dev_queue: network device
3315 *
3316 * Reset the bytes and packet count of a network device and clear the
3317 * software flow control OFF bit for this network device
3318 */
netdev_reset_queue(struct net_device * dev_queue)3319 static inline void netdev_reset_queue(struct net_device *dev_queue)
3320 {
3321 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3322 }
3323
3324 /**
3325 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3326 * @dev: network device
3327 * @queue_index: given tx queue index
3328 *
3329 * Returns 0 if given tx queue index >= number of device tx queues,
3330 * otherwise returns the originally passed tx queue index.
3331 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3332 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3333 {
3334 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3335 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3336 dev->name, queue_index,
3337 dev->real_num_tx_queues);
3338 return 0;
3339 }
3340
3341 return queue_index;
3342 }
3343
3344 /**
3345 * netif_running - test if up
3346 * @dev: network device
3347 *
3348 * Test if the device has been brought up.
3349 */
netif_running(const struct net_device * dev)3350 static inline bool netif_running(const struct net_device *dev)
3351 {
3352 return test_bit(__LINK_STATE_START, &dev->state);
3353 }
3354
3355 /*
3356 * Routines to manage the subqueues on a device. We only need start,
3357 * stop, and a check if it's stopped. All other device management is
3358 * done at the overall netdevice level.
3359 * Also test the device if we're multiqueue.
3360 */
3361
3362 /**
3363 * netif_start_subqueue - allow sending packets on subqueue
3364 * @dev: network device
3365 * @queue_index: sub queue index
3366 *
3367 * Start individual transmit queue of a device with multiple transmit queues.
3368 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3369 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3370 {
3371 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3372
3373 netif_tx_start_queue(txq);
3374 }
3375
3376 /**
3377 * netif_stop_subqueue - stop sending packets on subqueue
3378 * @dev: network device
3379 * @queue_index: sub queue index
3380 *
3381 * Stop individual transmit queue of a device with multiple transmit queues.
3382 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3383 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3384 {
3385 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3386 netif_tx_stop_queue(txq);
3387 }
3388
3389 /**
3390 * netif_subqueue_stopped - test status of subqueue
3391 * @dev: network device
3392 * @queue_index: sub queue index
3393 *
3394 * Check individual transmit queue of a device with multiple transmit queues.
3395 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3396 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3397 u16 queue_index)
3398 {
3399 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3400
3401 return netif_tx_queue_stopped(txq);
3402 }
3403
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3404 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3405 struct sk_buff *skb)
3406 {
3407 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3408 }
3409
3410 /**
3411 * netif_wake_subqueue - allow sending packets on subqueue
3412 * @dev: network device
3413 * @queue_index: sub queue index
3414 *
3415 * Resume individual transmit queue of a device with multiple transmit queues.
3416 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3417 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3418 {
3419 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3420
3421 netif_tx_wake_queue(txq);
3422 }
3423
3424 #ifdef CONFIG_XPS
3425 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3426 u16 index);
3427 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3428 u16 index, bool is_rxqs_map);
3429
3430 /**
3431 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3432 * @j: CPU/Rx queue index
3433 * @mask: bitmask of all cpus/rx queues
3434 * @nr_bits: number of bits in the bitmask
3435 *
3436 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3437 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3438 static inline bool netif_attr_test_mask(unsigned long j,
3439 const unsigned long *mask,
3440 unsigned int nr_bits)
3441 {
3442 cpu_max_bits_warn(j, nr_bits);
3443 return test_bit(j, mask);
3444 }
3445
3446 /**
3447 * netif_attr_test_online - Test for online CPU/Rx queue
3448 * @j: CPU/Rx queue index
3449 * @online_mask: bitmask for CPUs/Rx queues that are online
3450 * @nr_bits: number of bits in the bitmask
3451 *
3452 * Returns true if a CPU/Rx queue is online.
3453 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3454 static inline bool netif_attr_test_online(unsigned long j,
3455 const unsigned long *online_mask,
3456 unsigned int nr_bits)
3457 {
3458 cpu_max_bits_warn(j, nr_bits);
3459
3460 if (online_mask)
3461 return test_bit(j, online_mask);
3462
3463 return (j < nr_bits);
3464 }
3465
3466 /**
3467 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3468 * @n: CPU/Rx queue index
3469 * @srcp: the cpumask/Rx queue mask pointer
3470 * @nr_bits: number of bits in the bitmask
3471 *
3472 * Returns >= nr_bits if no further CPUs/Rx queues set.
3473 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3474 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3475 unsigned int nr_bits)
3476 {
3477 /* -1 is a legal arg here. */
3478 if (n != -1)
3479 cpu_max_bits_warn(n, nr_bits);
3480
3481 if (srcp)
3482 return find_next_bit(srcp, nr_bits, n + 1);
3483
3484 return n + 1;
3485 }
3486
3487 /**
3488 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3489 * @n: CPU/Rx queue index
3490 * @src1p: the first CPUs/Rx queues mask pointer
3491 * @src2p: the second CPUs/Rx queues mask pointer
3492 * @nr_bits: number of bits in the bitmask
3493 *
3494 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3495 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3496 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3497 const unsigned long *src2p,
3498 unsigned int nr_bits)
3499 {
3500 /* -1 is a legal arg here. */
3501 if (n != -1)
3502 cpu_max_bits_warn(n, nr_bits);
3503
3504 if (src1p && src2p)
3505 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3506 else if (src1p)
3507 return find_next_bit(src1p, nr_bits, n + 1);
3508 else if (src2p)
3509 return find_next_bit(src2p, nr_bits, n + 1);
3510
3511 return n + 1;
3512 }
3513 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3514 static inline int netif_set_xps_queue(struct net_device *dev,
3515 const struct cpumask *mask,
3516 u16 index)
3517 {
3518 return 0;
3519 }
3520
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,bool is_rxqs_map)3521 static inline int __netif_set_xps_queue(struct net_device *dev,
3522 const unsigned long *mask,
3523 u16 index, bool is_rxqs_map)
3524 {
3525 return 0;
3526 }
3527 #endif
3528
3529 /**
3530 * netif_is_multiqueue - test if device has multiple transmit queues
3531 * @dev: network device
3532 *
3533 * Check if device has multiple transmit queues
3534 */
netif_is_multiqueue(const struct net_device * dev)3535 static inline bool netif_is_multiqueue(const struct net_device *dev)
3536 {
3537 return dev->num_tx_queues > 1;
3538 }
3539
3540 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3541
3542 #ifdef CONFIG_SYSFS
3543 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3544 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3545 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3546 unsigned int rxqs)
3547 {
3548 dev->real_num_rx_queues = rxqs;
3549 return 0;
3550 }
3551 #endif
3552
3553 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3554 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3555 {
3556 return dev->_rx + rxq;
3557 }
3558
3559 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3560 static inline unsigned int get_netdev_rx_queue_index(
3561 struct netdev_rx_queue *queue)
3562 {
3563 struct net_device *dev = queue->dev;
3564 int index = queue - dev->_rx;
3565
3566 BUG_ON(index >= dev->num_rx_queues);
3567 return index;
3568 }
3569 #endif
3570
3571 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3572 int netif_get_num_default_rss_queues(void);
3573
3574 enum skb_free_reason {
3575 SKB_REASON_CONSUMED,
3576 SKB_REASON_DROPPED,
3577 };
3578
3579 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3580 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3581
3582 /*
3583 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3584 * interrupt context or with hardware interrupts being disabled.
3585 * (in_irq() || irqs_disabled())
3586 *
3587 * We provide four helpers that can be used in following contexts :
3588 *
3589 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3590 * replacing kfree_skb(skb)
3591 *
3592 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3593 * Typically used in place of consume_skb(skb) in TX completion path
3594 *
3595 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3596 * replacing kfree_skb(skb)
3597 *
3598 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3599 * and consumed a packet. Used in place of consume_skb(skb)
3600 */
dev_kfree_skb_irq(struct sk_buff * skb)3601 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3602 {
3603 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3604 }
3605
dev_consume_skb_irq(struct sk_buff * skb)3606 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3607 {
3608 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3609 }
3610
dev_kfree_skb_any(struct sk_buff * skb)3611 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3612 {
3613 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3614 }
3615
dev_consume_skb_any(struct sk_buff * skb)3616 static inline void dev_consume_skb_any(struct sk_buff *skb)
3617 {
3618 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3619 }
3620
3621 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3622 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3623 int netif_rx(struct sk_buff *skb);
3624 int netif_rx_ni(struct sk_buff *skb);
3625 int netif_receive_skb(struct sk_buff *skb);
3626 int netif_receive_skb_core(struct sk_buff *skb);
3627 void netif_receive_skb_list(struct list_head *head);
3628 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3629 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3630 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3631 gro_result_t napi_gro_frags(struct napi_struct *napi);
3632 struct packet_offload *gro_find_receive_by_type(__be16 type);
3633 struct packet_offload *gro_find_complete_by_type(__be16 type);
3634
napi_free_frags(struct napi_struct * napi)3635 static inline void napi_free_frags(struct napi_struct *napi)
3636 {
3637 kfree_skb(napi->skb);
3638 napi->skb = NULL;
3639 }
3640
3641 bool netdev_is_rx_handler_busy(struct net_device *dev);
3642 int netdev_rx_handler_register(struct net_device *dev,
3643 rx_handler_func_t *rx_handler,
3644 void *rx_handler_data);
3645 void netdev_rx_handler_unregister(struct net_device *dev);
3646
3647 bool dev_valid_name(const char *name);
3648 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3649 bool *need_copyout);
3650 int dev_ifconf(struct net *net, struct ifconf *, int);
3651 int dev_ethtool(struct net *net, struct ifreq *);
3652 unsigned int dev_get_flags(const struct net_device *);
3653 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3654 struct netlink_ext_ack *extack);
3655 int dev_change_flags(struct net_device *dev, unsigned int flags,
3656 struct netlink_ext_ack *extack);
3657 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3658 unsigned int gchanges);
3659 int dev_change_name(struct net_device *, const char *);
3660 int dev_set_alias(struct net_device *, const char *, size_t);
3661 int dev_get_alias(const struct net_device *, char *, size_t);
3662 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3663 int __dev_set_mtu(struct net_device *, int);
3664 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3665 struct netlink_ext_ack *extack);
3666 int dev_set_mtu(struct net_device *, int);
3667 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3668 void dev_set_group(struct net_device *, int);
3669 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3670 struct netlink_ext_ack *extack);
3671 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3672 struct netlink_ext_ack *extack);
3673 int dev_change_carrier(struct net_device *, bool new_carrier);
3674 int dev_get_phys_port_id(struct net_device *dev,
3675 struct netdev_phys_item_id *ppid);
3676 int dev_get_phys_port_name(struct net_device *dev,
3677 char *name, size_t len);
3678 int dev_get_port_parent_id(struct net_device *dev,
3679 struct netdev_phys_item_id *ppid, bool recurse);
3680 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3681 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3682 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3683 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3684 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3685 struct netdev_queue *txq, int *ret);
3686
3687 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3688 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3689 int fd, u32 flags);
3690 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3691 enum bpf_netdev_command cmd);
3692 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3693
3694 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3695 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3696 bool is_skb_forwardable(const struct net_device *dev,
3697 const struct sk_buff *skb);
3698
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb)3699 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3700 struct sk_buff *skb)
3701 {
3702 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3703 unlikely(!is_skb_forwardable(dev, skb))) {
3704 atomic_long_inc(&dev->rx_dropped);
3705 kfree_skb(skb);
3706 return NET_RX_DROP;
3707 }
3708
3709 skb_scrub_packet(skb, true);
3710 skb->priority = 0;
3711 return 0;
3712 }
3713
3714 bool dev_nit_active(struct net_device *dev);
3715 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3716
3717 extern int netdev_budget;
3718 extern unsigned int netdev_budget_usecs;
3719
3720 /* Called by rtnetlink.c:rtnl_unlock() */
3721 void netdev_run_todo(void);
3722
3723 /**
3724 * dev_put - release reference to device
3725 * @dev: network device
3726 *
3727 * Release reference to device to allow it to be freed.
3728 */
dev_put(struct net_device * dev)3729 static inline void dev_put(struct net_device *dev)
3730 {
3731 this_cpu_dec(*dev->pcpu_refcnt);
3732 }
3733
3734 /**
3735 * dev_hold - get reference to device
3736 * @dev: network device
3737 *
3738 * Hold reference to device to keep it from being freed.
3739 */
dev_hold(struct net_device * dev)3740 static inline void dev_hold(struct net_device *dev)
3741 {
3742 this_cpu_inc(*dev->pcpu_refcnt);
3743 }
3744
3745 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3746 * and _off may be called from IRQ context, but it is caller
3747 * who is responsible for serialization of these calls.
3748 *
3749 * The name carrier is inappropriate, these functions should really be
3750 * called netif_lowerlayer_*() because they represent the state of any
3751 * kind of lower layer not just hardware media.
3752 */
3753
3754 void linkwatch_init_dev(struct net_device *dev);
3755 void linkwatch_fire_event(struct net_device *dev);
3756 void linkwatch_forget_dev(struct net_device *dev);
3757
3758 /**
3759 * netif_carrier_ok - test if carrier present
3760 * @dev: network device
3761 *
3762 * Check if carrier is present on device
3763 */
netif_carrier_ok(const struct net_device * dev)3764 static inline bool netif_carrier_ok(const struct net_device *dev)
3765 {
3766 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3767 }
3768
3769 unsigned long dev_trans_start(struct net_device *dev);
3770
3771 void __netdev_watchdog_up(struct net_device *dev);
3772
3773 void netif_carrier_on(struct net_device *dev);
3774
3775 void netif_carrier_off(struct net_device *dev);
3776
3777 /**
3778 * netif_dormant_on - mark device as dormant.
3779 * @dev: network device
3780 *
3781 * Mark device as dormant (as per RFC2863).
3782 *
3783 * The dormant state indicates that the relevant interface is not
3784 * actually in a condition to pass packets (i.e., it is not 'up') but is
3785 * in a "pending" state, waiting for some external event. For "on-
3786 * demand" interfaces, this new state identifies the situation where the
3787 * interface is waiting for events to place it in the up state.
3788 */
netif_dormant_on(struct net_device * dev)3789 static inline void netif_dormant_on(struct net_device *dev)
3790 {
3791 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3792 linkwatch_fire_event(dev);
3793 }
3794
3795 /**
3796 * netif_dormant_off - set device as not dormant.
3797 * @dev: network device
3798 *
3799 * Device is not in dormant state.
3800 */
netif_dormant_off(struct net_device * dev)3801 static inline void netif_dormant_off(struct net_device *dev)
3802 {
3803 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3804 linkwatch_fire_event(dev);
3805 }
3806
3807 /**
3808 * netif_dormant - test if device is dormant
3809 * @dev: network device
3810 *
3811 * Check if device is dormant.
3812 */
netif_dormant(const struct net_device * dev)3813 static inline bool netif_dormant(const struct net_device *dev)
3814 {
3815 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3816 }
3817
3818
3819 /**
3820 * netif_oper_up - test if device is operational
3821 * @dev: network device
3822 *
3823 * Check if carrier is operational
3824 */
netif_oper_up(const struct net_device * dev)3825 static inline bool netif_oper_up(const struct net_device *dev)
3826 {
3827 return (dev->operstate == IF_OPER_UP ||
3828 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3829 }
3830
3831 /**
3832 * netif_device_present - is device available or removed
3833 * @dev: network device
3834 *
3835 * Check if device has not been removed from system.
3836 */
netif_device_present(struct net_device * dev)3837 static inline bool netif_device_present(struct net_device *dev)
3838 {
3839 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3840 }
3841
3842 void netif_device_detach(struct net_device *dev);
3843
3844 void netif_device_attach(struct net_device *dev);
3845
3846 /*
3847 * Network interface message level settings
3848 */
3849
3850 enum {
3851 NETIF_MSG_DRV = 0x0001,
3852 NETIF_MSG_PROBE = 0x0002,
3853 NETIF_MSG_LINK = 0x0004,
3854 NETIF_MSG_TIMER = 0x0008,
3855 NETIF_MSG_IFDOWN = 0x0010,
3856 NETIF_MSG_IFUP = 0x0020,
3857 NETIF_MSG_RX_ERR = 0x0040,
3858 NETIF_MSG_TX_ERR = 0x0080,
3859 NETIF_MSG_TX_QUEUED = 0x0100,
3860 NETIF_MSG_INTR = 0x0200,
3861 NETIF_MSG_TX_DONE = 0x0400,
3862 NETIF_MSG_RX_STATUS = 0x0800,
3863 NETIF_MSG_PKTDATA = 0x1000,
3864 NETIF_MSG_HW = 0x2000,
3865 NETIF_MSG_WOL = 0x4000,
3866 };
3867
3868 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3869 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3870 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3871 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3872 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3873 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3874 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3875 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3876 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3877 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3878 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3879 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3880 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3881 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3882 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3883
netif_msg_init(int debug_value,int default_msg_enable_bits)3884 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3885 {
3886 /* use default */
3887 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3888 return default_msg_enable_bits;
3889 if (debug_value == 0) /* no output */
3890 return 0;
3891 /* set low N bits */
3892 return (1U << debug_value) - 1;
3893 }
3894
__netif_tx_lock(struct netdev_queue * txq,int cpu)3895 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3896 {
3897 spin_lock(&txq->_xmit_lock);
3898 txq->xmit_lock_owner = cpu;
3899 }
3900
__netif_tx_acquire(struct netdev_queue * txq)3901 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3902 {
3903 __acquire(&txq->_xmit_lock);
3904 return true;
3905 }
3906
__netif_tx_release(struct netdev_queue * txq)3907 static inline void __netif_tx_release(struct netdev_queue *txq)
3908 {
3909 __release(&txq->_xmit_lock);
3910 }
3911
__netif_tx_lock_bh(struct netdev_queue * txq)3912 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3913 {
3914 spin_lock_bh(&txq->_xmit_lock);
3915 txq->xmit_lock_owner = smp_processor_id();
3916 }
3917
__netif_tx_trylock(struct netdev_queue * txq)3918 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3919 {
3920 bool ok = spin_trylock(&txq->_xmit_lock);
3921 if (likely(ok))
3922 txq->xmit_lock_owner = smp_processor_id();
3923 return ok;
3924 }
3925
__netif_tx_unlock(struct netdev_queue * txq)3926 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3927 {
3928 txq->xmit_lock_owner = -1;
3929 spin_unlock(&txq->_xmit_lock);
3930 }
3931
__netif_tx_unlock_bh(struct netdev_queue * txq)3932 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3933 {
3934 txq->xmit_lock_owner = -1;
3935 spin_unlock_bh(&txq->_xmit_lock);
3936 }
3937
txq_trans_update(struct netdev_queue * txq)3938 static inline void txq_trans_update(struct netdev_queue *txq)
3939 {
3940 if (txq->xmit_lock_owner != -1)
3941 txq->trans_start = jiffies;
3942 }
3943
3944 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)3945 static inline void netif_trans_update(struct net_device *dev)
3946 {
3947 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3948
3949 if (txq->trans_start != jiffies)
3950 txq->trans_start = jiffies;
3951 }
3952
3953 /**
3954 * netif_tx_lock - grab network device transmit lock
3955 * @dev: network device
3956 *
3957 * Get network device transmit lock
3958 */
netif_tx_lock(struct net_device * dev)3959 static inline void netif_tx_lock(struct net_device *dev)
3960 {
3961 unsigned int i;
3962 int cpu;
3963
3964 spin_lock(&dev->tx_global_lock);
3965 cpu = smp_processor_id();
3966 for (i = 0; i < dev->num_tx_queues; i++) {
3967 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3968
3969 /* We are the only thread of execution doing a
3970 * freeze, but we have to grab the _xmit_lock in
3971 * order to synchronize with threads which are in
3972 * the ->hard_start_xmit() handler and already
3973 * checked the frozen bit.
3974 */
3975 __netif_tx_lock(txq, cpu);
3976 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3977 __netif_tx_unlock(txq);
3978 }
3979 }
3980
netif_tx_lock_bh(struct net_device * dev)3981 static inline void netif_tx_lock_bh(struct net_device *dev)
3982 {
3983 local_bh_disable();
3984 netif_tx_lock(dev);
3985 }
3986
netif_tx_unlock(struct net_device * dev)3987 static inline void netif_tx_unlock(struct net_device *dev)
3988 {
3989 unsigned int i;
3990
3991 for (i = 0; i < dev->num_tx_queues; i++) {
3992 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3993
3994 /* No need to grab the _xmit_lock here. If the
3995 * queue is not stopped for another reason, we
3996 * force a schedule.
3997 */
3998 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3999 netif_schedule_queue(txq);
4000 }
4001 spin_unlock(&dev->tx_global_lock);
4002 }
4003
netif_tx_unlock_bh(struct net_device * dev)4004 static inline void netif_tx_unlock_bh(struct net_device *dev)
4005 {
4006 netif_tx_unlock(dev);
4007 local_bh_enable();
4008 }
4009
4010 #define HARD_TX_LOCK(dev, txq, cpu) { \
4011 if ((dev->features & NETIF_F_LLTX) == 0) { \
4012 __netif_tx_lock(txq, cpu); \
4013 } else { \
4014 __netif_tx_acquire(txq); \
4015 } \
4016 }
4017
4018 #define HARD_TX_TRYLOCK(dev, txq) \
4019 (((dev->features & NETIF_F_LLTX) == 0) ? \
4020 __netif_tx_trylock(txq) : \
4021 __netif_tx_acquire(txq))
4022
4023 #define HARD_TX_UNLOCK(dev, txq) { \
4024 if ((dev->features & NETIF_F_LLTX) == 0) { \
4025 __netif_tx_unlock(txq); \
4026 } else { \
4027 __netif_tx_release(txq); \
4028 } \
4029 }
4030
netif_tx_disable(struct net_device * dev)4031 static inline void netif_tx_disable(struct net_device *dev)
4032 {
4033 unsigned int i;
4034 int cpu;
4035
4036 local_bh_disable();
4037 cpu = smp_processor_id();
4038 for (i = 0; i < dev->num_tx_queues; i++) {
4039 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4040
4041 __netif_tx_lock(txq, cpu);
4042 netif_tx_stop_queue(txq);
4043 __netif_tx_unlock(txq);
4044 }
4045 local_bh_enable();
4046 }
4047
netif_addr_lock(struct net_device * dev)4048 static inline void netif_addr_lock(struct net_device *dev)
4049 {
4050 spin_lock(&dev->addr_list_lock);
4051 }
4052
netif_addr_lock_bh(struct net_device * dev)4053 static inline void netif_addr_lock_bh(struct net_device *dev)
4054 {
4055 spin_lock_bh(&dev->addr_list_lock);
4056 }
4057
netif_addr_unlock(struct net_device * dev)4058 static inline void netif_addr_unlock(struct net_device *dev)
4059 {
4060 spin_unlock(&dev->addr_list_lock);
4061 }
4062
netif_addr_unlock_bh(struct net_device * dev)4063 static inline void netif_addr_unlock_bh(struct net_device *dev)
4064 {
4065 spin_unlock_bh(&dev->addr_list_lock);
4066 }
4067
4068 /*
4069 * dev_addrs walker. Should be used only for read access. Call with
4070 * rcu_read_lock held.
4071 */
4072 #define for_each_dev_addr(dev, ha) \
4073 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4074
4075 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4076
4077 void ether_setup(struct net_device *dev);
4078
4079 /* Support for loadable net-drivers */
4080 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4081 unsigned char name_assign_type,
4082 void (*setup)(struct net_device *),
4083 unsigned int txqs, unsigned int rxqs);
4084 int dev_get_valid_name(struct net *net, struct net_device *dev,
4085 const char *name);
4086
4087 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4088 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4089
4090 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4091 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4092 count)
4093
4094 int register_netdev(struct net_device *dev);
4095 void unregister_netdev(struct net_device *dev);
4096
4097 /* General hardware address lists handling functions */
4098 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4099 struct netdev_hw_addr_list *from_list, int addr_len);
4100 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4101 struct netdev_hw_addr_list *from_list, int addr_len);
4102 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4103 struct net_device *dev,
4104 int (*sync)(struct net_device *, const unsigned char *),
4105 int (*unsync)(struct net_device *,
4106 const unsigned char *));
4107 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4108 struct net_device *dev,
4109 int (*sync)(struct net_device *,
4110 const unsigned char *, int),
4111 int (*unsync)(struct net_device *,
4112 const unsigned char *, int));
4113 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4114 struct net_device *dev,
4115 int (*unsync)(struct net_device *,
4116 const unsigned char *, int));
4117 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4118 struct net_device *dev,
4119 int (*unsync)(struct net_device *,
4120 const unsigned char *));
4121 void __hw_addr_init(struct netdev_hw_addr_list *list);
4122
4123 /* Functions used for device addresses handling */
4124 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4125 unsigned char addr_type);
4126 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4127 unsigned char addr_type);
4128 void dev_addr_flush(struct net_device *dev);
4129 int dev_addr_init(struct net_device *dev);
4130
4131 /* Functions used for unicast addresses handling */
4132 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4133 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4134 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4135 int dev_uc_sync(struct net_device *to, struct net_device *from);
4136 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4137 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4138 void dev_uc_flush(struct net_device *dev);
4139 void dev_uc_init(struct net_device *dev);
4140
4141 /**
4142 * __dev_uc_sync - Synchonize device's unicast list
4143 * @dev: device to sync
4144 * @sync: function to call if address should be added
4145 * @unsync: function to call if address should be removed
4146 *
4147 * Add newly added addresses to the interface, and release
4148 * addresses that have been deleted.
4149 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4150 static inline int __dev_uc_sync(struct net_device *dev,
4151 int (*sync)(struct net_device *,
4152 const unsigned char *),
4153 int (*unsync)(struct net_device *,
4154 const unsigned char *))
4155 {
4156 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4157 }
4158
4159 /**
4160 * __dev_uc_unsync - Remove synchronized addresses from device
4161 * @dev: device to sync
4162 * @unsync: function to call if address should be removed
4163 *
4164 * Remove all addresses that were added to the device by dev_uc_sync().
4165 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4166 static inline void __dev_uc_unsync(struct net_device *dev,
4167 int (*unsync)(struct net_device *,
4168 const unsigned char *))
4169 {
4170 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4171 }
4172
4173 /* Functions used for multicast addresses handling */
4174 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4175 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4176 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4177 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4178 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4179 int dev_mc_sync(struct net_device *to, struct net_device *from);
4180 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4181 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4182 void dev_mc_flush(struct net_device *dev);
4183 void dev_mc_init(struct net_device *dev);
4184
4185 /**
4186 * __dev_mc_sync - Synchonize device's multicast list
4187 * @dev: device to sync
4188 * @sync: function to call if address should be added
4189 * @unsync: function to call if address should be removed
4190 *
4191 * Add newly added addresses to the interface, and release
4192 * addresses that have been deleted.
4193 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4194 static inline int __dev_mc_sync(struct net_device *dev,
4195 int (*sync)(struct net_device *,
4196 const unsigned char *),
4197 int (*unsync)(struct net_device *,
4198 const unsigned char *))
4199 {
4200 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4201 }
4202
4203 /**
4204 * __dev_mc_unsync - Remove synchronized addresses from device
4205 * @dev: device to sync
4206 * @unsync: function to call if address should be removed
4207 *
4208 * Remove all addresses that were added to the device by dev_mc_sync().
4209 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4210 static inline void __dev_mc_unsync(struct net_device *dev,
4211 int (*unsync)(struct net_device *,
4212 const unsigned char *))
4213 {
4214 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4215 }
4216
4217 /* Functions used for secondary unicast and multicast support */
4218 void dev_set_rx_mode(struct net_device *dev);
4219 void __dev_set_rx_mode(struct net_device *dev);
4220 int dev_set_promiscuity(struct net_device *dev, int inc);
4221 int dev_set_allmulti(struct net_device *dev, int inc);
4222 void netdev_state_change(struct net_device *dev);
4223 void netdev_notify_peers(struct net_device *dev);
4224 void netdev_features_change(struct net_device *dev);
4225 /* Load a device via the kmod */
4226 void dev_load(struct net *net, const char *name);
4227 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4228 struct rtnl_link_stats64 *storage);
4229 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4230 const struct net_device_stats *netdev_stats);
4231
4232 extern int netdev_max_backlog;
4233 extern int netdev_tstamp_prequeue;
4234 extern int weight_p;
4235 extern int dev_weight_rx_bias;
4236 extern int dev_weight_tx_bias;
4237 extern int dev_rx_weight;
4238 extern int dev_tx_weight;
4239 extern int gro_normal_batch;
4240
4241 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4242 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4243 struct list_head **iter);
4244 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4245 struct list_head **iter);
4246
4247 /* iterate through upper list, must be called under RCU read lock */
4248 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4249 for (iter = &(dev)->adj_list.upper, \
4250 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4251 updev; \
4252 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4253
4254 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4255 int (*fn)(struct net_device *upper_dev,
4256 void *data),
4257 void *data);
4258
4259 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4260 struct net_device *upper_dev);
4261
4262 bool netdev_has_any_upper_dev(struct net_device *dev);
4263
4264 void *netdev_lower_get_next_private(struct net_device *dev,
4265 struct list_head **iter);
4266 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4267 struct list_head **iter);
4268
4269 #define netdev_for_each_lower_private(dev, priv, iter) \
4270 for (iter = (dev)->adj_list.lower.next, \
4271 priv = netdev_lower_get_next_private(dev, &(iter)); \
4272 priv; \
4273 priv = netdev_lower_get_next_private(dev, &(iter)))
4274
4275 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4276 for (iter = &(dev)->adj_list.lower, \
4277 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4278 priv; \
4279 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4280
4281 void *netdev_lower_get_next(struct net_device *dev,
4282 struct list_head **iter);
4283
4284 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4285 for (iter = (dev)->adj_list.lower.next, \
4286 ldev = netdev_lower_get_next(dev, &(iter)); \
4287 ldev; \
4288 ldev = netdev_lower_get_next(dev, &(iter)))
4289
4290 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4291 struct list_head **iter);
4292 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4293 struct list_head **iter);
4294
4295 int netdev_walk_all_lower_dev(struct net_device *dev,
4296 int (*fn)(struct net_device *lower_dev,
4297 void *data),
4298 void *data);
4299 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4300 int (*fn)(struct net_device *lower_dev,
4301 void *data),
4302 void *data);
4303
4304 void *netdev_adjacent_get_private(struct list_head *adj_list);
4305 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4306 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4307 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4308 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4309 struct netlink_ext_ack *extack);
4310 int netdev_master_upper_dev_link(struct net_device *dev,
4311 struct net_device *upper_dev,
4312 void *upper_priv, void *upper_info,
4313 struct netlink_ext_ack *extack);
4314 void netdev_upper_dev_unlink(struct net_device *dev,
4315 struct net_device *upper_dev);
4316 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4317 struct net_device *new_dev,
4318 struct net_device *dev,
4319 struct netlink_ext_ack *extack);
4320 void netdev_adjacent_change_commit(struct net_device *old_dev,
4321 struct net_device *new_dev,
4322 struct net_device *dev);
4323 void netdev_adjacent_change_abort(struct net_device *old_dev,
4324 struct net_device *new_dev,
4325 struct net_device *dev);
4326 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4327 void *netdev_lower_dev_get_private(struct net_device *dev,
4328 struct net_device *lower_dev);
4329 void netdev_lower_state_changed(struct net_device *lower_dev,
4330 void *lower_state_info);
4331
4332 /* RSS keys are 40 or 52 bytes long */
4333 #define NETDEV_RSS_KEY_LEN 52
4334 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4335 void netdev_rss_key_fill(void *buffer, size_t len);
4336
4337 int skb_checksum_help(struct sk_buff *skb);
4338 int skb_crc32c_csum_help(struct sk_buff *skb);
4339 int skb_csum_hwoffload_help(struct sk_buff *skb,
4340 const netdev_features_t features);
4341
4342 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4343 netdev_features_t features, bool tx_path);
4344 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4345 netdev_features_t features);
4346
4347 struct netdev_bonding_info {
4348 ifslave slave;
4349 ifbond master;
4350 };
4351
4352 struct netdev_notifier_bonding_info {
4353 struct netdev_notifier_info info; /* must be first */
4354 struct netdev_bonding_info bonding_info;
4355 };
4356
4357 void netdev_bonding_info_change(struct net_device *dev,
4358 struct netdev_bonding_info *bonding_info);
4359
4360 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4361 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4362 {
4363 return __skb_gso_segment(skb, features, true);
4364 }
4365 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4366
can_checksum_protocol(netdev_features_t features,__be16 protocol)4367 static inline bool can_checksum_protocol(netdev_features_t features,
4368 __be16 protocol)
4369 {
4370 if (protocol == htons(ETH_P_FCOE))
4371 return !!(features & NETIF_F_FCOE_CRC);
4372
4373 /* Assume this is an IP checksum (not SCTP CRC) */
4374
4375 if (features & NETIF_F_HW_CSUM) {
4376 /* Can checksum everything */
4377 return true;
4378 }
4379
4380 switch (protocol) {
4381 case htons(ETH_P_IP):
4382 return !!(features & NETIF_F_IP_CSUM);
4383 case htons(ETH_P_IPV6):
4384 return !!(features & NETIF_F_IPV6_CSUM);
4385 default:
4386 return false;
4387 }
4388 }
4389
4390 #ifdef CONFIG_BUG
4391 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4392 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4393 static inline void netdev_rx_csum_fault(struct net_device *dev,
4394 struct sk_buff *skb)
4395 {
4396 }
4397 #endif
4398 /* rx skb timestamps */
4399 void net_enable_timestamp(void);
4400 void net_disable_timestamp(void);
4401
4402 #ifdef CONFIG_PROC_FS
4403 int __init dev_proc_init(void);
4404 #else
4405 #define dev_proc_init() 0
4406 #endif
4407
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4408 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4409 struct sk_buff *skb, struct net_device *dev,
4410 bool more)
4411 {
4412 __this_cpu_write(softnet_data.xmit.more, more);
4413 return ops->ndo_start_xmit(skb, dev);
4414 }
4415
netdev_xmit_more(void)4416 static inline bool netdev_xmit_more(void)
4417 {
4418 return __this_cpu_read(softnet_data.xmit.more);
4419 }
4420
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4421 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4422 struct netdev_queue *txq, bool more)
4423 {
4424 const struct net_device_ops *ops = dev->netdev_ops;
4425 netdev_tx_t rc;
4426
4427 rc = __netdev_start_xmit(ops, skb, dev, more);
4428 if (rc == NETDEV_TX_OK)
4429 txq_trans_update(txq);
4430
4431 return rc;
4432 }
4433
4434 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4435 const void *ns);
4436 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4437 const void *ns);
4438
netdev_class_create_file(const struct class_attribute * class_attr)4439 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4440 {
4441 return netdev_class_create_file_ns(class_attr, NULL);
4442 }
4443
netdev_class_remove_file(const struct class_attribute * class_attr)4444 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4445 {
4446 netdev_class_remove_file_ns(class_attr, NULL);
4447 }
4448
4449 extern const struct kobj_ns_type_operations net_ns_type_operations;
4450
4451 const char *netdev_drivername(const struct net_device *dev);
4452
4453 void linkwatch_run_queue(void);
4454
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4455 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4456 netdev_features_t f2)
4457 {
4458 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4459 if (f1 & NETIF_F_HW_CSUM)
4460 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4461 else
4462 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4463 }
4464
4465 return f1 & f2;
4466 }
4467
netdev_get_wanted_features(struct net_device * dev)4468 static inline netdev_features_t netdev_get_wanted_features(
4469 struct net_device *dev)
4470 {
4471 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4472 }
4473 netdev_features_t netdev_increment_features(netdev_features_t all,
4474 netdev_features_t one, netdev_features_t mask);
4475
4476 /* Allow TSO being used on stacked device :
4477 * Performing the GSO segmentation before last device
4478 * is a performance improvement.
4479 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4480 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4481 netdev_features_t mask)
4482 {
4483 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4484 }
4485
4486 int __netdev_update_features(struct net_device *dev);
4487 void netdev_update_features(struct net_device *dev);
4488 void netdev_change_features(struct net_device *dev);
4489
4490 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4491 struct net_device *dev);
4492
4493 netdev_features_t passthru_features_check(struct sk_buff *skb,
4494 struct net_device *dev,
4495 netdev_features_t features);
4496 netdev_features_t netif_skb_features(struct sk_buff *skb);
4497
net_gso_ok(netdev_features_t features,int gso_type)4498 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4499 {
4500 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4501
4502 /* check flags correspondence */
4503 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4504 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4505 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4506 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4507 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4508 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4509 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4510 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4511 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4512 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4513 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4514 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4515 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4516 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4517 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4518 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4519 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4520 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4521
4522 return (features & feature) == feature;
4523 }
4524
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)4525 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4526 {
4527 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4528 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4529 }
4530
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)4531 static inline bool netif_needs_gso(struct sk_buff *skb,
4532 netdev_features_t features)
4533 {
4534 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4535 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4536 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4537 }
4538
netif_set_gso_max_size(struct net_device * dev,unsigned int size)4539 static inline void netif_set_gso_max_size(struct net_device *dev,
4540 unsigned int size)
4541 {
4542 dev->gso_max_size = size;
4543 }
4544
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)4545 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4546 int pulled_hlen, u16 mac_offset,
4547 int mac_len)
4548 {
4549 skb->protocol = protocol;
4550 skb->encapsulation = 1;
4551 skb_push(skb, pulled_hlen);
4552 skb_reset_transport_header(skb);
4553 skb->mac_header = mac_offset;
4554 skb->network_header = skb->mac_header + mac_len;
4555 skb->mac_len = mac_len;
4556 }
4557
netif_is_macsec(const struct net_device * dev)4558 static inline bool netif_is_macsec(const struct net_device *dev)
4559 {
4560 return dev->priv_flags & IFF_MACSEC;
4561 }
4562
netif_is_macvlan(const struct net_device * dev)4563 static inline bool netif_is_macvlan(const struct net_device *dev)
4564 {
4565 return dev->priv_flags & IFF_MACVLAN;
4566 }
4567
netif_is_macvlan_port(const struct net_device * dev)4568 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4569 {
4570 return dev->priv_flags & IFF_MACVLAN_PORT;
4571 }
4572
netif_is_bond_master(const struct net_device * dev)4573 static inline bool netif_is_bond_master(const struct net_device *dev)
4574 {
4575 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4576 }
4577
netif_is_bond_slave(const struct net_device * dev)4578 static inline bool netif_is_bond_slave(const struct net_device *dev)
4579 {
4580 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4581 }
4582
netif_supports_nofcs(struct net_device * dev)4583 static inline bool netif_supports_nofcs(struct net_device *dev)
4584 {
4585 return dev->priv_flags & IFF_SUPP_NOFCS;
4586 }
4587
netif_has_l3_rx_handler(const struct net_device * dev)4588 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4589 {
4590 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4591 }
4592
netif_is_l3_master(const struct net_device * dev)4593 static inline bool netif_is_l3_master(const struct net_device *dev)
4594 {
4595 return dev->priv_flags & IFF_L3MDEV_MASTER;
4596 }
4597
netif_is_l3_slave(const struct net_device * dev)4598 static inline bool netif_is_l3_slave(const struct net_device *dev)
4599 {
4600 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4601 }
4602
netif_is_bridge_master(const struct net_device * dev)4603 static inline bool netif_is_bridge_master(const struct net_device *dev)
4604 {
4605 return dev->priv_flags & IFF_EBRIDGE;
4606 }
4607
netif_is_bridge_port(const struct net_device * dev)4608 static inline bool netif_is_bridge_port(const struct net_device *dev)
4609 {
4610 return dev->priv_flags & IFF_BRIDGE_PORT;
4611 }
4612
netif_is_ovs_master(const struct net_device * dev)4613 static inline bool netif_is_ovs_master(const struct net_device *dev)
4614 {
4615 return dev->priv_flags & IFF_OPENVSWITCH;
4616 }
4617
netif_is_ovs_port(const struct net_device * dev)4618 static inline bool netif_is_ovs_port(const struct net_device *dev)
4619 {
4620 return dev->priv_flags & IFF_OVS_DATAPATH;
4621 }
4622
netif_is_team_master(const struct net_device * dev)4623 static inline bool netif_is_team_master(const struct net_device *dev)
4624 {
4625 return dev->priv_flags & IFF_TEAM;
4626 }
4627
netif_is_team_port(const struct net_device * dev)4628 static inline bool netif_is_team_port(const struct net_device *dev)
4629 {
4630 return dev->priv_flags & IFF_TEAM_PORT;
4631 }
4632
netif_is_lag_master(const struct net_device * dev)4633 static inline bool netif_is_lag_master(const struct net_device *dev)
4634 {
4635 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4636 }
4637
netif_is_lag_port(const struct net_device * dev)4638 static inline bool netif_is_lag_port(const struct net_device *dev)
4639 {
4640 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4641 }
4642
netif_is_rxfh_configured(const struct net_device * dev)4643 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4644 {
4645 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4646 }
4647
netif_is_failover(const struct net_device * dev)4648 static inline bool netif_is_failover(const struct net_device *dev)
4649 {
4650 return dev->priv_flags & IFF_FAILOVER;
4651 }
4652
netif_is_failover_slave(const struct net_device * dev)4653 static inline bool netif_is_failover_slave(const struct net_device *dev)
4654 {
4655 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4656 }
4657
4658 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)4659 static inline void netif_keep_dst(struct net_device *dev)
4660 {
4661 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4662 }
4663
4664 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)4665 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4666 {
4667 /* TODO: reserve and use an additional IFF bit, if we get more users */
4668 return dev->priv_flags & IFF_MACSEC;
4669 }
4670
4671 extern struct pernet_operations __net_initdata loopback_net_ops;
4672
4673 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4674
4675 /* netdev_printk helpers, similar to dev_printk */
4676
netdev_name(const struct net_device * dev)4677 static inline const char *netdev_name(const struct net_device *dev)
4678 {
4679 if (!dev->name[0] || strchr(dev->name, '%'))
4680 return "(unnamed net_device)";
4681 return dev->name;
4682 }
4683
netdev_unregistering(const struct net_device * dev)4684 static inline bool netdev_unregistering(const struct net_device *dev)
4685 {
4686 return dev->reg_state == NETREG_UNREGISTERING;
4687 }
4688
netdev_reg_state(const struct net_device * dev)4689 static inline const char *netdev_reg_state(const struct net_device *dev)
4690 {
4691 switch (dev->reg_state) {
4692 case NETREG_UNINITIALIZED: return " (uninitialized)";
4693 case NETREG_REGISTERED: return "";
4694 case NETREG_UNREGISTERING: return " (unregistering)";
4695 case NETREG_UNREGISTERED: return " (unregistered)";
4696 case NETREG_RELEASED: return " (released)";
4697 case NETREG_DUMMY: return " (dummy)";
4698 }
4699
4700 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4701 return " (unknown)";
4702 }
4703
4704 __printf(3, 4) __cold
4705 void netdev_printk(const char *level, const struct net_device *dev,
4706 const char *format, ...);
4707 __printf(2, 3) __cold
4708 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4709 __printf(2, 3) __cold
4710 void netdev_alert(const struct net_device *dev, const char *format, ...);
4711 __printf(2, 3) __cold
4712 void netdev_crit(const struct net_device *dev, const char *format, ...);
4713 __printf(2, 3) __cold
4714 void netdev_err(const struct net_device *dev, const char *format, ...);
4715 __printf(2, 3) __cold
4716 void netdev_warn(const struct net_device *dev, const char *format, ...);
4717 __printf(2, 3) __cold
4718 void netdev_notice(const struct net_device *dev, const char *format, ...);
4719 __printf(2, 3) __cold
4720 void netdev_info(const struct net_device *dev, const char *format, ...);
4721
4722 #define netdev_level_once(level, dev, fmt, ...) \
4723 do { \
4724 static bool __print_once __read_mostly; \
4725 \
4726 if (!__print_once) { \
4727 __print_once = true; \
4728 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4729 } \
4730 } while (0)
4731
4732 #define netdev_emerg_once(dev, fmt, ...) \
4733 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4734 #define netdev_alert_once(dev, fmt, ...) \
4735 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4736 #define netdev_crit_once(dev, fmt, ...) \
4737 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4738 #define netdev_err_once(dev, fmt, ...) \
4739 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4740 #define netdev_warn_once(dev, fmt, ...) \
4741 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4742 #define netdev_notice_once(dev, fmt, ...) \
4743 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4744 #define netdev_info_once(dev, fmt, ...) \
4745 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4746
4747 #define MODULE_ALIAS_NETDEV(device) \
4748 MODULE_ALIAS("netdev-" device)
4749
4750 #if defined(CONFIG_DYNAMIC_DEBUG)
4751 #define netdev_dbg(__dev, format, args...) \
4752 do { \
4753 dynamic_netdev_dbg(__dev, format, ##args); \
4754 } while (0)
4755 #elif defined(DEBUG)
4756 #define netdev_dbg(__dev, format, args...) \
4757 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4758 #else
4759 #define netdev_dbg(__dev, format, args...) \
4760 ({ \
4761 if (0) \
4762 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4763 })
4764 #endif
4765
4766 #if defined(VERBOSE_DEBUG)
4767 #define netdev_vdbg netdev_dbg
4768 #else
4769
4770 #define netdev_vdbg(dev, format, args...) \
4771 ({ \
4772 if (0) \
4773 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4774 0; \
4775 })
4776 #endif
4777
4778 /*
4779 * netdev_WARN() acts like dev_printk(), but with the key difference
4780 * of using a WARN/WARN_ON to get the message out, including the
4781 * file/line information and a backtrace.
4782 */
4783 #define netdev_WARN(dev, format, args...) \
4784 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4785 netdev_reg_state(dev), ##args)
4786
4787 #define netdev_WARN_ONCE(dev, format, args...) \
4788 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4789 netdev_reg_state(dev), ##args)
4790
4791 /* netif printk helpers, similar to netdev_printk */
4792
4793 #define netif_printk(priv, type, level, dev, fmt, args...) \
4794 do { \
4795 if (netif_msg_##type(priv)) \
4796 netdev_printk(level, (dev), fmt, ##args); \
4797 } while (0)
4798
4799 #define netif_level(level, priv, type, dev, fmt, args...) \
4800 do { \
4801 if (netif_msg_##type(priv)) \
4802 netdev_##level(dev, fmt, ##args); \
4803 } while (0)
4804
4805 #define netif_emerg(priv, type, dev, fmt, args...) \
4806 netif_level(emerg, priv, type, dev, fmt, ##args)
4807 #define netif_alert(priv, type, dev, fmt, args...) \
4808 netif_level(alert, priv, type, dev, fmt, ##args)
4809 #define netif_crit(priv, type, dev, fmt, args...) \
4810 netif_level(crit, priv, type, dev, fmt, ##args)
4811 #define netif_err(priv, type, dev, fmt, args...) \
4812 netif_level(err, priv, type, dev, fmt, ##args)
4813 #define netif_warn(priv, type, dev, fmt, args...) \
4814 netif_level(warn, priv, type, dev, fmt, ##args)
4815 #define netif_notice(priv, type, dev, fmt, args...) \
4816 netif_level(notice, priv, type, dev, fmt, ##args)
4817 #define netif_info(priv, type, dev, fmt, args...) \
4818 netif_level(info, priv, type, dev, fmt, ##args)
4819
4820 #if defined(CONFIG_DYNAMIC_DEBUG)
4821 #define netif_dbg(priv, type, netdev, format, args...) \
4822 do { \
4823 if (netif_msg_##type(priv)) \
4824 dynamic_netdev_dbg(netdev, format, ##args); \
4825 } while (0)
4826 #elif defined(DEBUG)
4827 #define netif_dbg(priv, type, dev, format, args...) \
4828 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4829 #else
4830 #define netif_dbg(priv, type, dev, format, args...) \
4831 ({ \
4832 if (0) \
4833 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4834 0; \
4835 })
4836 #endif
4837
4838 /* if @cond then downgrade to debug, else print at @level */
4839 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4840 do { \
4841 if (cond) \
4842 netif_dbg(priv, type, netdev, fmt, ##args); \
4843 else \
4844 netif_ ## level(priv, type, netdev, fmt, ##args); \
4845 } while (0)
4846
4847 #if defined(VERBOSE_DEBUG)
4848 #define netif_vdbg netif_dbg
4849 #else
4850 #define netif_vdbg(priv, type, dev, format, args...) \
4851 ({ \
4852 if (0) \
4853 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4854 0; \
4855 })
4856 #endif
4857
4858 /*
4859 * The list of packet types we will receive (as opposed to discard)
4860 * and the routines to invoke.
4861 *
4862 * Why 16. Because with 16 the only overlap we get on a hash of the
4863 * low nibble of the protocol value is RARP/SNAP/X.25.
4864 *
4865 * 0800 IP
4866 * 0001 802.3
4867 * 0002 AX.25
4868 * 0004 802.2
4869 * 8035 RARP
4870 * 0005 SNAP
4871 * 0805 X.25
4872 * 0806 ARP
4873 * 8137 IPX
4874 * 0009 Localtalk
4875 * 86DD IPv6
4876 */
4877 #define PTYPE_HASH_SIZE (16)
4878 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4879
4880 extern struct net_device *blackhole_netdev;
4881
4882 #endif /* _LINUX_NETDEVICE_H */
4883