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 #include <linux/android_kabi.h>
52
53 struct netpoll_info;
54 struct device;
55 struct phy_device;
56 struct dsa_port;
57
58 struct sfp_bus;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_dev;
63 struct mpls_dev;
64 /* UDP Tunnel offloads */
65 struct udp_tunnel_info;
66 struct bpf_prog;
67 struct xdp_buff;
68
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
71
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
75
76 #define MAX_NEST_DEV 8
77
78 /*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
90 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
92 * others are propagated to higher layers.
93 */
94
95 /* qdisc ->enqueue() return codes. */
96 #define NET_XMIT_SUCCESS 0x00
97 #define NET_XMIT_DROP 0x01 /* skb dropped */
98 #define NET_XMIT_CN 0x02 /* congestion notification */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
109
110 enum netdev_tx {
111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
114 };
115 typedef enum netdev_tx netdev_tx_t;
116
117 /*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
dev_xmit_complete(int rc)121 static inline bool dev_xmit_complete(int rc)
122 {
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133 }
134
135 /*
136 * Compute the worst-case header length according to the protocols
137 * used.
138 */
139
140 #if defined(CONFIG_HYPERV_NET)
141 # define LL_MAX_HEADER 128
142 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
143 # if defined(CONFIG_MAC80211_MESH)
144 # define LL_MAX_HEADER 128
145 # else
146 # define LL_MAX_HEADER 96
147 # endif
148 #else
149 # define LL_MAX_HEADER 32
150 #endif
151
152 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
154 #define MAX_HEADER LL_MAX_HEADER
155 #else
156 #define MAX_HEADER (LL_MAX_HEADER + 48)
157 #endif
158
159 /*
160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
162 */
163
164 struct net_device_stats {
165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
174 unsigned long collisions;
175 unsigned long rx_length_errors;
176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188 };
189
190
191 #include <linux/cache.h>
192 #include <linux/skbuff.h>
193
194 #ifdef CONFIG_RPS
195 #include <linux/static_key.h>
196 extern struct static_key_false rps_needed;
197 extern struct static_key_false rfs_needed;
198 #endif
199
200 struct neighbour;
201 struct neigh_parms;
202 struct sk_buff;
203
204 struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
208 #define NETDEV_HW_ADDR_T_LAN 1
209 #define NETDEV_HW_ADDR_T_SAN 2
210 #define NETDEV_HW_ADDR_T_SLAVE 3
211 #define NETDEV_HW_ADDR_T_UNICAST 4
212 #define NETDEV_HW_ADDR_T_MULTICAST 5
213 bool global_use;
214 int sync_cnt;
215 int refcount;
216 int synced;
217 struct rcu_head rcu_head;
218 };
219
220 struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223 };
224
225 #define netdev_hw_addr_list_count(l) ((l)->count)
226 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227 #define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
229
230 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
232 #define netdev_for_each_uc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234
235 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
237 #define netdev_for_each_mc_addr(ha, dev) \
238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
239
240 struct hh_cache {
241 unsigned int hh_len;
242 seqlock_t hh_lock;
243
244 /* cached hardware header; allow for machine alignment needs. */
245 #define HH_DATA_MOD 16
246 #define HH_DATA_OFF(__len) \
247 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
248 #define HH_DATA_ALIGN(__len) \
249 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
250 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 };
252
253 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * Alternative is:
255 * dev->hard_header_len ? (dev->hard_header_len +
256 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 *
258 * We could use other alignment values, but we must maintain the
259 * relationship HH alignment <= LL alignment.
260 */
261 #define LL_RESERVED_SPACE(dev) \
262 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
263 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
266 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267
268 struct header_ops {
269 int (*create) (struct sk_buff *skb, struct net_device *dev,
270 unsigned short type, const void *daddr,
271 const void *saddr, unsigned int len);
272 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
273 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
274 void (*cache_update)(struct hh_cache *hh,
275 const struct net_device *dev,
276 const unsigned char *haddr);
277 bool (*validate)(const char *ll_header, unsigned int len);
278 __be16 (*parse_protocol)(const struct sk_buff *skb);
279
280 ANDROID_KABI_RESERVE(1);
281 ANDROID_KABI_RESERVE(2);
282 };
283
284 /* These flag bits are private to the generic network queueing
285 * layer; they may not be explicitly referenced by any other
286 * code.
287 */
288
289 enum netdev_state_t {
290 __LINK_STATE_START,
291 __LINK_STATE_PRESENT,
292 __LINK_STATE_NOCARRIER,
293 __LINK_STATE_LINKWATCH_PENDING,
294 __LINK_STATE_DORMANT,
295 };
296
297
298 /*
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
301 */
302 struct netdev_boot_setup {
303 char name[IFNAMSIZ];
304 struct ifmap map;
305 };
306 #define NETDEV_BOOT_SETUP_MAX 8
307
308 int __init netdev_boot_setup(char *str);
309
310 struct gro_list {
311 struct list_head list;
312 int count;
313 };
314
315 /*
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
318 */
319 #define GRO_HASH_BUCKETS 8
320
321 /*
322 * Structure for NAPI scheduling similar to tasklet but with weighting
323 */
324 struct napi_struct {
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
330 */
331 struct list_head poll_list;
332
333 unsigned long state;
334 int weight;
335 unsigned long gro_bitmask;
336 int (*poll)(struct napi_struct *, int);
337 #ifdef CONFIG_NETPOLL
338 int poll_owner;
339 #endif
340 struct net_device *dev;
341 struct gro_list gro_hash[GRO_HASH_BUCKETS];
342 struct sk_buff *skb;
343 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
344 int rx_count; /* length of rx_list */
345 struct hrtimer timer;
346 struct list_head dev_list;
347 struct hlist_node napi_hash_node;
348 unsigned int napi_id;
349
350 ANDROID_KABI_RESERVE(1);
351 ANDROID_KABI_RESERVE(2);
352 ANDROID_KABI_RESERVE(3);
353 ANDROID_KABI_RESERVE(4);
354 };
355
356 enum {
357 NAPI_STATE_SCHED, /* Poll is scheduled */
358 NAPI_STATE_MISSED, /* reschedule a napi */
359 NAPI_STATE_DISABLE, /* Disable pending */
360 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
361 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
362 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
363 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
364 };
365
366 enum {
367 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
368 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
369 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
370 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
371 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
372 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
373 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
374 };
375
376 enum gro_result {
377 GRO_MERGED,
378 GRO_MERGED_FREE,
379 GRO_HELD,
380 GRO_NORMAL,
381 GRO_DROP,
382 GRO_CONSUMED,
383 };
384 typedef enum gro_result gro_result_t;
385
386 /*
387 * enum rx_handler_result - Possible return values for rx_handlers.
388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
389 * further.
390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
391 * case skb->dev was changed by rx_handler.
392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
394 *
395 * rx_handlers are functions called from inside __netif_receive_skb(), to do
396 * special processing of the skb, prior to delivery to protocol handlers.
397 *
398 * Currently, a net_device can only have a single rx_handler registered. Trying
399 * to register a second rx_handler will return -EBUSY.
400 *
401 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
402 * To unregister a rx_handler on a net_device, use
403 * netdev_rx_handler_unregister().
404 *
405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
406 * do with the skb.
407 *
408 * If the rx_handler consumed the skb in some way, it should return
409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
410 * the skb to be delivered in some other way.
411 *
412 * If the rx_handler changed skb->dev, to divert the skb to another
413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
414 * new device will be called if it exists.
415 *
416 * If the rx_handler decides the skb should be ignored, it should return
417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
418 * are registered on exact device (ptype->dev == skb->dev).
419 *
420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
421 * delivered, it should return RX_HANDLER_PASS.
422 *
423 * A device without a registered rx_handler will behave as if rx_handler
424 * returned RX_HANDLER_PASS.
425 */
426
427 enum rx_handler_result {
428 RX_HANDLER_CONSUMED,
429 RX_HANDLER_ANOTHER,
430 RX_HANDLER_EXACT,
431 RX_HANDLER_PASS,
432 };
433 typedef enum rx_handler_result rx_handler_result_t;
434 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
435
436 void __napi_schedule(struct napi_struct *n);
437 void __napi_schedule_irqoff(struct napi_struct *n);
438
napi_disable_pending(struct napi_struct * n)439 static inline bool napi_disable_pending(struct napi_struct *n)
440 {
441 return test_bit(NAPI_STATE_DISABLE, &n->state);
442 }
443
444 bool napi_schedule_prep(struct napi_struct *n);
445
446 /**
447 * napi_schedule - schedule NAPI poll
448 * @n: NAPI context
449 *
450 * Schedule NAPI poll routine to be called if it is not already
451 * running.
452 */
napi_schedule(struct napi_struct * n)453 static inline void napi_schedule(struct napi_struct *n)
454 {
455 if (napi_schedule_prep(n))
456 __napi_schedule(n);
457 }
458
459 /**
460 * napi_schedule_irqoff - schedule NAPI poll
461 * @n: NAPI context
462 *
463 * Variant of napi_schedule(), assuming hard irqs are masked.
464 */
napi_schedule_irqoff(struct napi_struct * n)465 static inline void napi_schedule_irqoff(struct napi_struct *n)
466 {
467 if (napi_schedule_prep(n))
468 __napi_schedule_irqoff(n);
469 }
470
471 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
napi_reschedule(struct napi_struct * napi)472 static inline bool napi_reschedule(struct napi_struct *napi)
473 {
474 if (napi_schedule_prep(napi)) {
475 __napi_schedule(napi);
476 return true;
477 }
478 return false;
479 }
480
481 bool napi_complete_done(struct napi_struct *n, int work_done);
482 /**
483 * napi_complete - NAPI processing complete
484 * @n: NAPI context
485 *
486 * Mark NAPI processing as complete.
487 * Consider using napi_complete_done() instead.
488 * Return false if device should avoid rearming interrupts.
489 */
napi_complete(struct napi_struct * n)490 static inline bool napi_complete(struct napi_struct *n)
491 {
492 return napi_complete_done(n, 0);
493 }
494
495 /**
496 * napi_hash_del - remove a NAPI from global table
497 * @napi: NAPI context
498 *
499 * Warning: caller must observe RCU grace period
500 * before freeing memory containing @napi, if
501 * this function returns true.
502 * Note: core networking stack automatically calls it
503 * from netif_napi_del().
504 * Drivers might want to call this helper to combine all
505 * the needed RCU grace periods into a single one.
506 */
507 bool napi_hash_del(struct napi_struct *napi);
508
509 /**
510 * napi_disable - prevent NAPI from scheduling
511 * @n: NAPI context
512 *
513 * Stop NAPI from being scheduled on this context.
514 * Waits till any outstanding processing completes.
515 */
516 void napi_disable(struct napi_struct *n);
517
518 /**
519 * napi_enable - enable NAPI scheduling
520 * @n: NAPI context
521 *
522 * Resume NAPI from being scheduled on this context.
523 * Must be paired with napi_disable.
524 */
napi_enable(struct napi_struct * n)525 static inline void napi_enable(struct napi_struct *n)
526 {
527 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
528 smp_mb__before_atomic();
529 clear_bit(NAPI_STATE_SCHED, &n->state);
530 clear_bit(NAPI_STATE_NPSVC, &n->state);
531 }
532
533 /**
534 * napi_synchronize - wait until NAPI is not running
535 * @n: NAPI context
536 *
537 * Wait until NAPI is done being scheduled on this context.
538 * Waits till any outstanding processing completes but
539 * does not disable future activations.
540 */
napi_synchronize(const struct napi_struct * n)541 static inline void napi_synchronize(const struct napi_struct *n)
542 {
543 if (IS_ENABLED(CONFIG_SMP))
544 while (test_bit(NAPI_STATE_SCHED, &n->state))
545 msleep(1);
546 else
547 barrier();
548 }
549
550 /**
551 * napi_if_scheduled_mark_missed - if napi is running, set the
552 * NAPIF_STATE_MISSED
553 * @n: NAPI context
554 *
555 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
556 * NAPI is scheduled.
557 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)558 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
559 {
560 unsigned long val, new;
561
562 do {
563 val = READ_ONCE(n->state);
564 if (val & NAPIF_STATE_DISABLE)
565 return true;
566
567 if (!(val & NAPIF_STATE_SCHED))
568 return false;
569
570 new = val | NAPIF_STATE_MISSED;
571 } while (cmpxchg(&n->state, val, new) != val);
572
573 return true;
574 }
575
576 enum netdev_queue_state_t {
577 __QUEUE_STATE_DRV_XOFF,
578 __QUEUE_STATE_STACK_XOFF,
579 __QUEUE_STATE_FROZEN,
580 };
581
582 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
583 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
584 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
585
586 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
587 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
588 QUEUE_STATE_FROZEN)
589 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
590 QUEUE_STATE_FROZEN)
591
592 /*
593 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
594 * netif_tx_* functions below are used to manipulate this flag. The
595 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
596 * queue independently. The netif_xmit_*stopped functions below are called
597 * to check if the queue has been stopped by the driver or stack (either
598 * of the XOFF bits are set in the state). Drivers should not need to call
599 * netif_xmit*stopped functions, they should only be using netif_tx_*.
600 */
601
602 struct netdev_queue {
603 /*
604 * read-mostly part
605 */
606 struct net_device *dev;
607 struct Qdisc __rcu *qdisc;
608 struct Qdisc *qdisc_sleeping;
609 #ifdef CONFIG_SYSFS
610 struct kobject kobj;
611 #endif
612 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
613 int numa_node;
614 #endif
615 unsigned long tx_maxrate;
616 /*
617 * Number of TX timeouts for this queue
618 * (/sys/class/net/DEV/Q/trans_timeout)
619 */
620 unsigned long trans_timeout;
621
622 /* Subordinate device that the queue has been assigned to */
623 struct net_device *sb_dev;
624 #ifdef CONFIG_XDP_SOCKETS
625 struct xdp_umem *umem;
626 #endif
627 /*
628 * write-mostly part
629 */
630 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
631 int xmit_lock_owner;
632 /*
633 * Time (in jiffies) of last Tx
634 */
635 unsigned long trans_start;
636
637 unsigned long state;
638
639 #ifdef CONFIG_BQL
640 struct dql dql;
641 #endif
642
643 ANDROID_KABI_RESERVE(1);
644 ANDROID_KABI_RESERVE(2);
645 ANDROID_KABI_RESERVE(3);
646 ANDROID_KABI_RESERVE(4);
647 } ____cacheline_aligned_in_smp;
648
649 extern int sysctl_fb_tunnels_only_for_init_net;
650 extern int sysctl_devconf_inherit_init_net;
651
net_has_fallback_tunnels(const struct net * net)652 static inline bool net_has_fallback_tunnels(const struct net *net)
653 {
654 return net == &init_net ||
655 !IS_ENABLED(CONFIG_SYSCTL) ||
656 !sysctl_fb_tunnels_only_for_init_net;
657 }
658
netdev_queue_numa_node_read(const struct netdev_queue * q)659 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
660 {
661 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
662 return q->numa_node;
663 #else
664 return NUMA_NO_NODE;
665 #endif
666 }
667
netdev_queue_numa_node_write(struct netdev_queue * q,int node)668 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
669 {
670 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
671 q->numa_node = node;
672 #endif
673 }
674
675 #ifdef CONFIG_RPS
676 /*
677 * This structure holds an RPS map which can be of variable length. The
678 * map is an array of CPUs.
679 */
680 struct rps_map {
681 unsigned int len;
682 struct rcu_head rcu;
683 u16 cpus[0];
684 };
685 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
686
687 /*
688 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
689 * tail pointer for that CPU's input queue at the time of last enqueue, and
690 * a hardware filter index.
691 */
692 struct rps_dev_flow {
693 u16 cpu;
694 u16 filter;
695 unsigned int last_qtail;
696 };
697 #define RPS_NO_FILTER 0xffff
698
699 /*
700 * The rps_dev_flow_table structure contains a table of flow mappings.
701 */
702 struct rps_dev_flow_table {
703 unsigned int mask;
704 struct rcu_head rcu;
705 struct rps_dev_flow flows[0];
706 };
707 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
708 ((_num) * sizeof(struct rps_dev_flow)))
709
710 /*
711 * The rps_sock_flow_table contains mappings of flows to the last CPU
712 * on which they were processed by the application (set in recvmsg).
713 * Each entry is a 32bit value. Upper part is the high-order bits
714 * of flow hash, lower part is CPU number.
715 * rps_cpu_mask is used to partition the space, depending on number of
716 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
717 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
718 * meaning we use 32-6=26 bits for the hash.
719 */
720 struct rps_sock_flow_table {
721 u32 mask;
722
723 u32 ents[0] ____cacheline_aligned_in_smp;
724 };
725 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
726
727 #define RPS_NO_CPU 0xffff
728
729 extern u32 rps_cpu_mask;
730 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
731
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)732 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
733 u32 hash)
734 {
735 if (table && hash) {
736 unsigned int index = hash & table->mask;
737 u32 val = hash & ~rps_cpu_mask;
738
739 /* We only give a hint, preemption can change CPU under us */
740 val |= raw_smp_processor_id();
741
742 /* The following WRITE_ONCE() is paired with the READ_ONCE()
743 * here, and another one in get_rps_cpu().
744 */
745 if (READ_ONCE(table->ents[index]) != val)
746 WRITE_ONCE(table->ents[index], val);
747 }
748 }
749
750 #ifdef CONFIG_RFS_ACCEL
751 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
752 u16 filter_id);
753 #endif
754 #endif /* CONFIG_RPS */
755
756 /* This structure contains an instance of an RX queue. */
757 struct netdev_rx_queue {
758 #ifdef CONFIG_RPS
759 struct rps_map __rcu *rps_map;
760 struct rps_dev_flow_table __rcu *rps_flow_table;
761 #endif
762 struct kobject kobj;
763 struct net_device *dev;
764 struct xdp_rxq_info xdp_rxq;
765 #ifdef CONFIG_XDP_SOCKETS
766 struct xdp_umem *umem;
767 #endif
768
769 ANDROID_KABI_RESERVE(1);
770 ANDROID_KABI_RESERVE(2);
771 ANDROID_KABI_RESERVE(3);
772 ANDROID_KABI_RESERVE(4);
773 } ____cacheline_aligned_in_smp;
774
775 /*
776 * RX queue sysfs structures and functions.
777 */
778 struct rx_queue_attribute {
779 struct attribute attr;
780 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
781 ssize_t (*store)(struct netdev_rx_queue *queue,
782 const char *buf, size_t len);
783 };
784
785 #ifdef CONFIG_XPS
786 /*
787 * This structure holds an XPS map which can be of variable length. The
788 * map is an array of queues.
789 */
790 struct xps_map {
791 unsigned int len;
792 unsigned int alloc_len;
793 struct rcu_head rcu;
794 u16 queues[0];
795 };
796 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
797 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
798 - sizeof(struct xps_map)) / sizeof(u16))
799
800 /*
801 * This structure holds all XPS maps for device. Maps are indexed by CPU.
802 */
803 struct xps_dev_maps {
804 struct rcu_head rcu;
805 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
806 };
807
808 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
809 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
810
811 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
812 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
813
814 #endif /* CONFIG_XPS */
815
816 #define TC_MAX_QUEUE 16
817 #define TC_BITMASK 15
818 /* HW offloaded queuing disciplines txq count and offset maps */
819 struct netdev_tc_txq {
820 u16 count;
821 u16 offset;
822 };
823
824 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
825 /*
826 * This structure is to hold information about the device
827 * configured to run FCoE protocol stack.
828 */
829 struct netdev_fcoe_hbainfo {
830 char manufacturer[64];
831 char serial_number[64];
832 char hardware_version[64];
833 char driver_version[64];
834 char optionrom_version[64];
835 char firmware_version[64];
836 char model[256];
837 char model_description[256];
838 };
839 #endif
840
841 #define MAX_PHYS_ITEM_ID_LEN 32
842
843 /* This structure holds a unique identifier to identify some
844 * physical item (port for example) used by a netdevice.
845 */
846 struct netdev_phys_item_id {
847 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
848 unsigned char id_len;
849 };
850
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)851 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
852 struct netdev_phys_item_id *b)
853 {
854 return a->id_len == b->id_len &&
855 memcmp(a->id, b->id, a->id_len) == 0;
856 }
857
858 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
859 struct sk_buff *skb,
860 struct net_device *sb_dev);
861
862 enum tc_setup_type {
863 TC_SETUP_QDISC_MQPRIO,
864 TC_SETUP_CLSU32,
865 TC_SETUP_CLSFLOWER,
866 TC_SETUP_CLSMATCHALL,
867 TC_SETUP_CLSBPF,
868 TC_SETUP_BLOCK,
869 TC_SETUP_QDISC_CBS,
870 TC_SETUP_QDISC_RED,
871 TC_SETUP_QDISC_PRIO,
872 TC_SETUP_QDISC_MQ,
873 TC_SETUP_QDISC_ETF,
874 TC_SETUP_ROOT_QDISC,
875 TC_SETUP_QDISC_GRED,
876 TC_SETUP_QDISC_TAPRIO,
877 };
878
879 /* These structures hold the attributes of bpf state that are being passed
880 * to the netdevice through the bpf op.
881 */
882 enum bpf_netdev_command {
883 /* Set or clear a bpf program used in the earliest stages of packet
884 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
885 * is responsible for calling bpf_prog_put on any old progs that are
886 * stored. In case of error, the callee need not release the new prog
887 * reference, but on success it takes ownership and must bpf_prog_put
888 * when it is no longer used.
889 */
890 XDP_SETUP_PROG,
891 XDP_SETUP_PROG_HW,
892 XDP_QUERY_PROG,
893 XDP_QUERY_PROG_HW,
894 /* BPF program for offload callbacks, invoked at program load time. */
895 BPF_OFFLOAD_MAP_ALLOC,
896 BPF_OFFLOAD_MAP_FREE,
897 XDP_SETUP_XSK_UMEM,
898 };
899
900 struct bpf_prog_offload_ops;
901 struct netlink_ext_ack;
902 struct xdp_umem;
903
904 struct netdev_bpf {
905 enum bpf_netdev_command command;
906 union {
907 /* XDP_SETUP_PROG */
908 struct {
909 u32 flags;
910 struct bpf_prog *prog;
911 struct netlink_ext_ack *extack;
912 };
913 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
914 struct {
915 u32 prog_id;
916 /* flags with which program was installed */
917 u32 prog_flags;
918 };
919 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
920 struct {
921 struct bpf_offloaded_map *offmap;
922 };
923 /* XDP_SETUP_XSK_UMEM */
924 struct {
925 struct xdp_umem *umem;
926 u16 queue_id;
927 } xsk;
928 };
929 };
930
931 /* Flags for ndo_xsk_wakeup. */
932 #define XDP_WAKEUP_RX (1 << 0)
933 #define XDP_WAKEUP_TX (1 << 1)
934
935 #ifdef CONFIG_XFRM_OFFLOAD
936 struct xfrmdev_ops {
937 int (*xdo_dev_state_add) (struct xfrm_state *x);
938 void (*xdo_dev_state_delete) (struct xfrm_state *x);
939 void (*xdo_dev_state_free) (struct xfrm_state *x);
940 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
941 struct xfrm_state *x);
942 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
943
944 ANDROID_KABI_RESERVE(1);
945 ANDROID_KABI_RESERVE(2);
946 ANDROID_KABI_RESERVE(3);
947 ANDROID_KABI_RESERVE(4);
948 };
949 #endif
950
951 struct dev_ifalias {
952 struct rcu_head rcuhead;
953 char ifalias[];
954 };
955
956 struct devlink;
957 struct tlsdev_ops;
958
959
960 /*
961 * This structure defines the management hooks for network devices.
962 * The following hooks can be defined; unless noted otherwise, they are
963 * optional and can be filled with a null pointer.
964 *
965 * int (*ndo_init)(struct net_device *dev);
966 * This function is called once when a network device is registered.
967 * The network device can use this for any late stage initialization
968 * or semantic validation. It can fail with an error code which will
969 * be propagated back to register_netdev.
970 *
971 * void (*ndo_uninit)(struct net_device *dev);
972 * This function is called when device is unregistered or when registration
973 * fails. It is not called if init fails.
974 *
975 * int (*ndo_open)(struct net_device *dev);
976 * This function is called when a network device transitions to the up
977 * state.
978 *
979 * int (*ndo_stop)(struct net_device *dev);
980 * This function is called when a network device transitions to the down
981 * state.
982 *
983 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
984 * struct net_device *dev);
985 * Called when a packet needs to be transmitted.
986 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
987 * the queue before that can happen; it's for obsolete devices and weird
988 * corner cases, but the stack really does a non-trivial amount
989 * of useless work if you return NETDEV_TX_BUSY.
990 * Required; cannot be NULL.
991 *
992 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
993 * struct net_device *dev
994 * netdev_features_t features);
995 * Called by core transmit path to determine if device is capable of
996 * performing offload operations on a given packet. This is to give
997 * the device an opportunity to implement any restrictions that cannot
998 * be otherwise expressed by feature flags. The check is called with
999 * the set of features that the stack has calculated and it returns
1000 * those the driver believes to be appropriate.
1001 *
1002 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1003 * struct net_device *sb_dev);
1004 * Called to decide which queue to use when device supports multiple
1005 * transmit queues.
1006 *
1007 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1008 * This function is called to allow device receiver to make
1009 * changes to configuration when multicast or promiscuous is enabled.
1010 *
1011 * void (*ndo_set_rx_mode)(struct net_device *dev);
1012 * This function is called device changes address list filtering.
1013 * If driver handles unicast address filtering, it should set
1014 * IFF_UNICAST_FLT in its priv_flags.
1015 *
1016 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1017 * This function is called when the Media Access Control address
1018 * needs to be changed. If this interface is not defined, the
1019 * MAC address can not be changed.
1020 *
1021 * int (*ndo_validate_addr)(struct net_device *dev);
1022 * Test if Media Access Control address is valid for the device.
1023 *
1024 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1025 * Called when a user requests an ioctl which can't be handled by
1026 * the generic interface code. If not defined ioctls return
1027 * not supported error code.
1028 *
1029 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1030 * Used to set network devices bus interface parameters. This interface
1031 * is retained for legacy reasons; new devices should use the bus
1032 * interface (PCI) for low level management.
1033 *
1034 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1035 * Called when a user wants to change the Maximum Transfer Unit
1036 * of a device.
1037 *
1038 * void (*ndo_tx_timeout)(struct net_device *dev);
1039 * Callback used when the transmitter has not made any progress
1040 * for dev->watchdog ticks.
1041 *
1042 * void (*ndo_get_stats64)(struct net_device *dev,
1043 * struct rtnl_link_stats64 *storage);
1044 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1045 * Called when a user wants to get the network device usage
1046 * statistics. Drivers must do one of the following:
1047 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1048 * rtnl_link_stats64 structure passed by the caller.
1049 * 2. Define @ndo_get_stats to update a net_device_stats structure
1050 * (which should normally be dev->stats) and return a pointer to
1051 * it. The structure may be changed asynchronously only if each
1052 * field is written atomically.
1053 * 3. Update dev->stats asynchronously and atomically, and define
1054 * neither operation.
1055 *
1056 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1057 * Return true if this device supports offload stats of this attr_id.
1058 *
1059 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1060 * void *attr_data)
1061 * Get statistics for offload operations by attr_id. Write it into the
1062 * attr_data pointer.
1063 *
1064 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1065 * If device supports VLAN filtering this function is called when a
1066 * VLAN id is registered.
1067 *
1068 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1069 * If device supports VLAN filtering this function is called when a
1070 * VLAN id is unregistered.
1071 *
1072 * void (*ndo_poll_controller)(struct net_device *dev);
1073 *
1074 * SR-IOV management functions.
1075 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1076 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1077 * u8 qos, __be16 proto);
1078 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1079 * int max_tx_rate);
1080 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1081 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1082 * int (*ndo_get_vf_config)(struct net_device *dev,
1083 * int vf, struct ifla_vf_info *ivf);
1084 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1085 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1086 * struct nlattr *port[]);
1087 *
1088 * Enable or disable the VF ability to query its RSS Redirection Table and
1089 * Hash Key. This is needed since on some devices VF share this information
1090 * with PF and querying it may introduce a theoretical security risk.
1091 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1092 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1093 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1094 * void *type_data);
1095 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1096 * This is always called from the stack with the rtnl lock held and netif
1097 * tx queues stopped. This allows the netdevice to perform queue
1098 * management safely.
1099 *
1100 * Fiber Channel over Ethernet (FCoE) offload functions.
1101 * int (*ndo_fcoe_enable)(struct net_device *dev);
1102 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1103 * so the underlying device can perform whatever needed configuration or
1104 * initialization to support acceleration of FCoE traffic.
1105 *
1106 * int (*ndo_fcoe_disable)(struct net_device *dev);
1107 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1108 * so the underlying device can perform whatever needed clean-ups to
1109 * stop supporting acceleration of FCoE traffic.
1110 *
1111 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1112 * struct scatterlist *sgl, unsigned int sgc);
1113 * Called when the FCoE Initiator wants to initialize an I/O that
1114 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1115 * perform necessary setup and returns 1 to indicate the device is set up
1116 * successfully to perform DDP on this I/O, otherwise this returns 0.
1117 *
1118 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1119 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1120 * indicated by the FC exchange id 'xid', so the underlying device can
1121 * clean up and reuse resources for later DDP requests.
1122 *
1123 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1124 * struct scatterlist *sgl, unsigned int sgc);
1125 * Called when the FCoE Target wants to initialize an I/O that
1126 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1127 * perform necessary setup and returns 1 to indicate the device is set up
1128 * successfully to perform DDP on this I/O, otherwise this returns 0.
1129 *
1130 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1131 * struct netdev_fcoe_hbainfo *hbainfo);
1132 * Called when the FCoE Protocol stack wants information on the underlying
1133 * device. This information is utilized by the FCoE protocol stack to
1134 * register attributes with Fiber Channel management service as per the
1135 * FC-GS Fabric Device Management Information(FDMI) specification.
1136 *
1137 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1138 * Called when the underlying device wants to override default World Wide
1139 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1140 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1141 * protocol stack to use.
1142 *
1143 * RFS acceleration.
1144 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1145 * u16 rxq_index, u32 flow_id);
1146 * Set hardware filter for RFS. rxq_index is the target queue index;
1147 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1148 * Return the filter ID on success, or a negative error code.
1149 *
1150 * Slave management functions (for bridge, bonding, etc).
1151 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1152 * Called to make another netdev an underling.
1153 *
1154 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1155 * Called to release previously enslaved netdev.
1156 *
1157 * Feature/offload setting functions.
1158 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1159 * netdev_features_t features);
1160 * Adjusts the requested feature flags according to device-specific
1161 * constraints, and returns the resulting flags. Must not modify
1162 * the device state.
1163 *
1164 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1165 * Called to update device configuration to new features. Passed
1166 * feature set might be less than what was returned by ndo_fix_features()).
1167 * Must return >0 or -errno if it changed dev->features itself.
1168 *
1169 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1170 * struct net_device *dev,
1171 * const unsigned char *addr, u16 vid, u16 flags,
1172 * struct netlink_ext_ack *extack);
1173 * Adds an FDB entry to dev for addr.
1174 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1175 * struct net_device *dev,
1176 * const unsigned char *addr, u16 vid)
1177 * Deletes the FDB entry from dev coresponding to addr.
1178 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1179 * struct net_device *dev, struct net_device *filter_dev,
1180 * int *idx)
1181 * Used to add FDB entries to dump requests. Implementers should add
1182 * entries to skb and update idx with the number of entries.
1183 *
1184 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1185 * u16 flags, struct netlink_ext_ack *extack)
1186 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1187 * struct net_device *dev, u32 filter_mask,
1188 * int nlflags)
1189 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1190 * u16 flags);
1191 *
1192 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1193 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1194 * which do not represent real hardware may define this to allow their
1195 * userspace components to manage their virtual carrier state. Devices
1196 * that determine carrier state from physical hardware properties (eg
1197 * network cables) or protocol-dependent mechanisms (eg
1198 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1199 *
1200 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1201 * struct netdev_phys_item_id *ppid);
1202 * Called to get ID of physical port of this device. If driver does
1203 * not implement this, it is assumed that the hw is not able to have
1204 * multiple net devices on single physical port.
1205 *
1206 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1207 * struct netdev_phys_item_id *ppid)
1208 * Called to get the parent ID of the physical port of this device.
1209 *
1210 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1211 * struct udp_tunnel_info *ti);
1212 * Called by UDP tunnel to notify a driver about the UDP port and socket
1213 * address family that a UDP tunnel is listnening to. It is called only
1214 * when a new port starts listening. The operation is protected by the
1215 * RTNL.
1216 *
1217 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1218 * struct udp_tunnel_info *ti);
1219 * Called by UDP tunnel to notify the driver about a UDP port and socket
1220 * address family that the UDP tunnel is not listening to anymore. The
1221 * operation is protected by the RTNL.
1222 *
1223 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1224 * struct net_device *dev)
1225 * Called by upper layer devices to accelerate switching or other
1226 * station functionality into hardware. 'pdev is the lowerdev
1227 * to use for the offload and 'dev' is the net device that will
1228 * back the offload. Returns a pointer to the private structure
1229 * the upper layer will maintain.
1230 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1231 * Called by upper layer device to delete the station created
1232 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1233 * the station and priv is the structure returned by the add
1234 * operation.
1235 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1236 * int queue_index, u32 maxrate);
1237 * Called when a user wants to set a max-rate limitation of specific
1238 * TX queue.
1239 * int (*ndo_get_iflink)(const struct net_device *dev);
1240 * Called to get the iflink value of this device.
1241 * void (*ndo_change_proto_down)(struct net_device *dev,
1242 * bool proto_down);
1243 * This function is used to pass protocol port error state information
1244 * to the switch driver. The switch driver can react to the proto_down
1245 * by doing a phys down on the associated switch port.
1246 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1247 * This function is used to get egress tunnel information for given skb.
1248 * This is useful for retrieving outer tunnel header parameters while
1249 * sampling packet.
1250 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1251 * This function is used to specify the headroom that the skb must
1252 * consider when allocation skb during packet reception. Setting
1253 * appropriate rx headroom value allows avoiding skb head copy on
1254 * forward. Setting a negative value resets the rx headroom to the
1255 * default value.
1256 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1257 * This function is used to set or query state related to XDP on the
1258 * netdevice and manage BPF offload. See definition of
1259 * enum bpf_netdev_command for details.
1260 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1261 * u32 flags);
1262 * This function is used to submit @n XDP packets for transmit on a
1263 * netdevice. Returns number of frames successfully transmitted, frames
1264 * that got dropped are freed/returned via xdp_return_frame().
1265 * Returns negative number, means general error invoking ndo, meaning
1266 * no frames were xmit'ed and core-caller will free all frames.
1267 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1268 * This function is used to wake up the softirq, ksoftirqd or kthread
1269 * responsible for sending and/or receiving packets on a specific
1270 * queue id bound to an AF_XDP socket. The flags field specifies if
1271 * only RX, only Tx, or both should be woken up using the flags
1272 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1273 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1274 * Get devlink port instance associated with a given netdev.
1275 * Called with a reference on the netdevice and devlink locks only,
1276 * rtnl_lock is not held.
1277 */
1278 struct net_device_ops {
1279 int (*ndo_init)(struct net_device *dev);
1280 void (*ndo_uninit)(struct net_device *dev);
1281 int (*ndo_open)(struct net_device *dev);
1282 int (*ndo_stop)(struct net_device *dev);
1283 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1284 struct net_device *dev);
1285 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1286 struct net_device *dev,
1287 netdev_features_t features);
1288 u16 (*ndo_select_queue)(struct net_device *dev,
1289 struct sk_buff *skb,
1290 struct net_device *sb_dev);
1291 void (*ndo_change_rx_flags)(struct net_device *dev,
1292 int flags);
1293 void (*ndo_set_rx_mode)(struct net_device *dev);
1294 int (*ndo_set_mac_address)(struct net_device *dev,
1295 void *addr);
1296 int (*ndo_validate_addr)(struct net_device *dev);
1297 int (*ndo_do_ioctl)(struct net_device *dev,
1298 struct ifreq *ifr, int cmd);
1299 int (*ndo_set_config)(struct net_device *dev,
1300 struct ifmap *map);
1301 int (*ndo_change_mtu)(struct net_device *dev,
1302 int new_mtu);
1303 int (*ndo_neigh_setup)(struct net_device *dev,
1304 struct neigh_parms *);
1305 void (*ndo_tx_timeout) (struct net_device *dev);
1306
1307 void (*ndo_get_stats64)(struct net_device *dev,
1308 struct rtnl_link_stats64 *storage);
1309 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1310 int (*ndo_get_offload_stats)(int attr_id,
1311 const struct net_device *dev,
1312 void *attr_data);
1313 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1314
1315 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1316 __be16 proto, u16 vid);
1317 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1318 __be16 proto, u16 vid);
1319 #ifdef CONFIG_NET_POLL_CONTROLLER
1320 void (*ndo_poll_controller)(struct net_device *dev);
1321 int (*ndo_netpoll_setup)(struct net_device *dev,
1322 struct netpoll_info *info);
1323 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1324 #endif
1325 int (*ndo_set_vf_mac)(struct net_device *dev,
1326 int queue, u8 *mac);
1327 int (*ndo_set_vf_vlan)(struct net_device *dev,
1328 int queue, u16 vlan,
1329 u8 qos, __be16 proto);
1330 int (*ndo_set_vf_rate)(struct net_device *dev,
1331 int vf, int min_tx_rate,
1332 int max_tx_rate);
1333 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1334 int vf, bool setting);
1335 int (*ndo_set_vf_trust)(struct net_device *dev,
1336 int vf, bool setting);
1337 int (*ndo_get_vf_config)(struct net_device *dev,
1338 int vf,
1339 struct ifla_vf_info *ivf);
1340 int (*ndo_set_vf_link_state)(struct net_device *dev,
1341 int vf, int link_state);
1342 int (*ndo_get_vf_stats)(struct net_device *dev,
1343 int vf,
1344 struct ifla_vf_stats
1345 *vf_stats);
1346 int (*ndo_set_vf_port)(struct net_device *dev,
1347 int vf,
1348 struct nlattr *port[]);
1349 int (*ndo_get_vf_port)(struct net_device *dev,
1350 int vf, struct sk_buff *skb);
1351 int (*ndo_set_vf_guid)(struct net_device *dev,
1352 int vf, u64 guid,
1353 int guid_type);
1354 int (*ndo_set_vf_rss_query_en)(
1355 struct net_device *dev,
1356 int vf, bool setting);
1357 int (*ndo_setup_tc)(struct net_device *dev,
1358 enum tc_setup_type type,
1359 void *type_data);
1360 #if IS_ENABLED(CONFIG_FCOE)
1361 int (*ndo_fcoe_enable)(struct net_device *dev);
1362 int (*ndo_fcoe_disable)(struct net_device *dev);
1363 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1364 u16 xid,
1365 struct scatterlist *sgl,
1366 unsigned int sgc);
1367 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1368 u16 xid);
1369 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1370 u16 xid,
1371 struct scatterlist *sgl,
1372 unsigned int sgc);
1373 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1374 struct netdev_fcoe_hbainfo *hbainfo);
1375 #endif
1376
1377 #if IS_ENABLED(CONFIG_LIBFCOE)
1378 #define NETDEV_FCOE_WWNN 0
1379 #define NETDEV_FCOE_WWPN 1
1380 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1381 u64 *wwn, int type);
1382 #endif
1383
1384 #ifdef CONFIG_RFS_ACCEL
1385 int (*ndo_rx_flow_steer)(struct net_device *dev,
1386 const struct sk_buff *skb,
1387 u16 rxq_index,
1388 u32 flow_id);
1389 #endif
1390 int (*ndo_add_slave)(struct net_device *dev,
1391 struct net_device *slave_dev,
1392 struct netlink_ext_ack *extack);
1393 int (*ndo_del_slave)(struct net_device *dev,
1394 struct net_device *slave_dev);
1395 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1396 netdev_features_t features);
1397 int (*ndo_set_features)(struct net_device *dev,
1398 netdev_features_t features);
1399 int (*ndo_neigh_construct)(struct net_device *dev,
1400 struct neighbour *n);
1401 void (*ndo_neigh_destroy)(struct net_device *dev,
1402 struct neighbour *n);
1403
1404 int (*ndo_fdb_add)(struct ndmsg *ndm,
1405 struct nlattr *tb[],
1406 struct net_device *dev,
1407 const unsigned char *addr,
1408 u16 vid,
1409 u16 flags,
1410 struct netlink_ext_ack *extack);
1411 int (*ndo_fdb_del)(struct ndmsg *ndm,
1412 struct nlattr *tb[],
1413 struct net_device *dev,
1414 const unsigned char *addr,
1415 u16 vid);
1416 int (*ndo_fdb_dump)(struct sk_buff *skb,
1417 struct netlink_callback *cb,
1418 struct net_device *dev,
1419 struct net_device *filter_dev,
1420 int *idx);
1421 int (*ndo_fdb_get)(struct sk_buff *skb,
1422 struct nlattr *tb[],
1423 struct net_device *dev,
1424 const unsigned char *addr,
1425 u16 vid, u32 portid, u32 seq,
1426 struct netlink_ext_ack *extack);
1427 int (*ndo_bridge_setlink)(struct net_device *dev,
1428 struct nlmsghdr *nlh,
1429 u16 flags,
1430 struct netlink_ext_ack *extack);
1431 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1432 u32 pid, u32 seq,
1433 struct net_device *dev,
1434 u32 filter_mask,
1435 int nlflags);
1436 int (*ndo_bridge_dellink)(struct net_device *dev,
1437 struct nlmsghdr *nlh,
1438 u16 flags);
1439 int (*ndo_change_carrier)(struct net_device *dev,
1440 bool new_carrier);
1441 int (*ndo_get_phys_port_id)(struct net_device *dev,
1442 struct netdev_phys_item_id *ppid);
1443 int (*ndo_get_port_parent_id)(struct net_device *dev,
1444 struct netdev_phys_item_id *ppid);
1445 int (*ndo_get_phys_port_name)(struct net_device *dev,
1446 char *name, size_t len);
1447 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1448 struct udp_tunnel_info *ti);
1449 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1450 struct udp_tunnel_info *ti);
1451 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1452 struct net_device *dev);
1453 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1454 void *priv);
1455
1456 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1457 int queue_index,
1458 u32 maxrate);
1459 int (*ndo_get_iflink)(const struct net_device *dev);
1460 int (*ndo_change_proto_down)(struct net_device *dev,
1461 bool proto_down);
1462 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1463 struct sk_buff *skb);
1464 void (*ndo_set_rx_headroom)(struct net_device *dev,
1465 int needed_headroom);
1466 int (*ndo_bpf)(struct net_device *dev,
1467 struct netdev_bpf *bpf);
1468 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1469 struct xdp_frame **xdp,
1470 u32 flags);
1471 int (*ndo_xsk_wakeup)(struct net_device *dev,
1472 u32 queue_id, u32 flags);
1473 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1474
1475 ANDROID_KABI_RESERVE(1);
1476 ANDROID_KABI_RESERVE(2);
1477 ANDROID_KABI_RESERVE(3);
1478 ANDROID_KABI_RESERVE(4);
1479 ANDROID_KABI_RESERVE(5);
1480 ANDROID_KABI_RESERVE(6);
1481 ANDROID_KABI_RESERVE(7);
1482 ANDROID_KABI_RESERVE(8);
1483 };
1484
1485 /**
1486 * enum net_device_priv_flags - &struct net_device priv_flags
1487 *
1488 * These are the &struct net_device, they are only set internally
1489 * by drivers and used in the kernel. These flags are invisible to
1490 * userspace; this means that the order of these flags can change
1491 * during any kernel release.
1492 *
1493 * You should have a pretty good reason to be extending these flags.
1494 *
1495 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1496 * @IFF_EBRIDGE: Ethernet bridging device
1497 * @IFF_BONDING: bonding master or slave
1498 * @IFF_ISATAP: ISATAP interface (RFC4214)
1499 * @IFF_WAN_HDLC: WAN HDLC device
1500 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1501 * release skb->dst
1502 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1503 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1504 * @IFF_MACVLAN_PORT: device used as macvlan port
1505 * @IFF_BRIDGE_PORT: device used as bridge port
1506 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1507 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1508 * @IFF_UNICAST_FLT: Supports unicast filtering
1509 * @IFF_TEAM_PORT: device used as team port
1510 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1511 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1512 * change when it's running
1513 * @IFF_MACVLAN: Macvlan device
1514 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1515 * underlying stacked devices
1516 * @IFF_L3MDEV_MASTER: device is an L3 master device
1517 * @IFF_NO_QUEUE: device can run without qdisc attached
1518 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1519 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1520 * @IFF_TEAM: device is a team device
1521 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1522 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1523 * entity (i.e. the master device for bridged veth)
1524 * @IFF_MACSEC: device is a MACsec device
1525 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1526 * @IFF_FAILOVER: device is a failover master device
1527 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1528 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1529 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1530 */
1531 enum netdev_priv_flags {
1532 IFF_802_1Q_VLAN = 1<<0,
1533 IFF_EBRIDGE = 1<<1,
1534 IFF_BONDING = 1<<2,
1535 IFF_ISATAP = 1<<3,
1536 IFF_WAN_HDLC = 1<<4,
1537 IFF_XMIT_DST_RELEASE = 1<<5,
1538 IFF_DONT_BRIDGE = 1<<6,
1539 IFF_DISABLE_NETPOLL = 1<<7,
1540 IFF_MACVLAN_PORT = 1<<8,
1541 IFF_BRIDGE_PORT = 1<<9,
1542 IFF_OVS_DATAPATH = 1<<10,
1543 IFF_TX_SKB_SHARING = 1<<11,
1544 IFF_UNICAST_FLT = 1<<12,
1545 IFF_TEAM_PORT = 1<<13,
1546 IFF_SUPP_NOFCS = 1<<14,
1547 IFF_LIVE_ADDR_CHANGE = 1<<15,
1548 IFF_MACVLAN = 1<<16,
1549 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1550 IFF_L3MDEV_MASTER = 1<<18,
1551 IFF_NO_QUEUE = 1<<19,
1552 IFF_OPENVSWITCH = 1<<20,
1553 IFF_L3MDEV_SLAVE = 1<<21,
1554 IFF_TEAM = 1<<22,
1555 IFF_RXFH_CONFIGURED = 1<<23,
1556 IFF_PHONY_HEADROOM = 1<<24,
1557 IFF_MACSEC = 1<<25,
1558 IFF_NO_RX_HANDLER = 1<<26,
1559 IFF_FAILOVER = 1<<27,
1560 IFF_FAILOVER_SLAVE = 1<<28,
1561 IFF_L3MDEV_RX_HANDLER = 1<<29,
1562 IFF_LIVE_RENAME_OK = 1<<30,
1563 };
1564
1565 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1566 #define IFF_EBRIDGE IFF_EBRIDGE
1567 #define IFF_BONDING IFF_BONDING
1568 #define IFF_ISATAP IFF_ISATAP
1569 #define IFF_WAN_HDLC IFF_WAN_HDLC
1570 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1571 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1572 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1573 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1574 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1575 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1576 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1577 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1578 #define IFF_TEAM_PORT IFF_TEAM_PORT
1579 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1580 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1581 #define IFF_MACVLAN IFF_MACVLAN
1582 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1583 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1584 #define IFF_NO_QUEUE IFF_NO_QUEUE
1585 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1586 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1587 #define IFF_TEAM IFF_TEAM
1588 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1589 #define IFF_MACSEC IFF_MACSEC
1590 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1591 #define IFF_FAILOVER IFF_FAILOVER
1592 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1593 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1594 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1595
1596 /**
1597 * struct net_device - The DEVICE structure.
1598 *
1599 * Actually, this whole structure is a big mistake. It mixes I/O
1600 * data with strictly "high-level" data, and it has to know about
1601 * almost every data structure used in the INET module.
1602 *
1603 * @name: This is the first field of the "visible" part of this structure
1604 * (i.e. as seen by users in the "Space.c" file). It is the name
1605 * of the interface.
1606 *
1607 * @name_hlist: Device name hash chain, please keep it close to name[]
1608 * @ifalias: SNMP alias
1609 * @mem_end: Shared memory end
1610 * @mem_start: Shared memory start
1611 * @base_addr: Device I/O address
1612 * @irq: Device IRQ number
1613 *
1614 * @state: Generic network queuing layer state, see netdev_state_t
1615 * @dev_list: The global list of network devices
1616 * @napi_list: List entry used for polling NAPI devices
1617 * @unreg_list: List entry when we are unregistering the
1618 * device; see the function unregister_netdev
1619 * @close_list: List entry used when we are closing the device
1620 * @ptype_all: Device-specific packet handlers for all protocols
1621 * @ptype_specific: Device-specific, protocol-specific packet handlers
1622 *
1623 * @adj_list: Directly linked devices, like slaves for bonding
1624 * @features: Currently active device features
1625 * @hw_features: User-changeable features
1626 *
1627 * @wanted_features: User-requested features
1628 * @vlan_features: Mask of features inheritable by VLAN devices
1629 *
1630 * @hw_enc_features: Mask of features inherited by encapsulating devices
1631 * This field indicates what encapsulation
1632 * offloads the hardware is capable of doing,
1633 * and drivers will need to set them appropriately.
1634 *
1635 * @mpls_features: Mask of features inheritable by MPLS
1636 *
1637 * @ifindex: interface index
1638 * @group: The group the device belongs to
1639 *
1640 * @stats: Statistics struct, which was left as a legacy, use
1641 * rtnl_link_stats64 instead
1642 *
1643 * @rx_dropped: Dropped packets by core network,
1644 * do not use this in drivers
1645 * @tx_dropped: Dropped packets by core network,
1646 * do not use this in drivers
1647 * @rx_nohandler: nohandler dropped packets by core network on
1648 * inactive devices, do not use this in drivers
1649 * @carrier_up_count: Number of times the carrier has been up
1650 * @carrier_down_count: Number of times the carrier has been down
1651 *
1652 * @wireless_handlers: List of functions to handle Wireless Extensions,
1653 * instead of ioctl,
1654 * see <net/iw_handler.h> for details.
1655 * @wireless_data: Instance data managed by the core of wireless extensions
1656 *
1657 * @netdev_ops: Includes several pointers to callbacks,
1658 * if one wants to override the ndo_*() functions
1659 * @ethtool_ops: Management operations
1660 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1661 * discovery handling. Necessary for e.g. 6LoWPAN.
1662 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1663 * of Layer 2 headers.
1664 *
1665 * @flags: Interface flags (a la BSD)
1666 * @priv_flags: Like 'flags' but invisible to userspace,
1667 * see if.h for the definitions
1668 * @gflags: Global flags ( kept as legacy )
1669 * @padded: How much padding added by alloc_netdev()
1670 * @operstate: RFC2863 operstate
1671 * @link_mode: Mapping policy to operstate
1672 * @if_port: Selectable AUI, TP, ...
1673 * @dma: DMA channel
1674 * @mtu: Interface MTU value
1675 * @min_mtu: Interface Minimum MTU value
1676 * @max_mtu: Interface Maximum MTU value
1677 * @type: Interface hardware type
1678 * @hard_header_len: Maximum hardware header length.
1679 * @min_header_len: Minimum hardware header length
1680 *
1681 * @needed_headroom: Extra headroom the hardware may need, but not in all
1682 * cases can this be guaranteed
1683 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1684 * cases can this be guaranteed. Some cases also use
1685 * LL_MAX_HEADER instead to allocate the skb
1686 *
1687 * interface address info:
1688 *
1689 * @perm_addr: Permanent hw address
1690 * @addr_assign_type: Hw address assignment type
1691 * @addr_len: Hardware address length
1692 * @upper_level: Maximum depth level of upper devices.
1693 * @lower_level: Maximum depth level of lower devices.
1694 * @neigh_priv_len: Used in neigh_alloc()
1695 * @dev_id: Used to differentiate devices that share
1696 * the same link layer address
1697 * @dev_port: Used to differentiate devices that share
1698 * the same function
1699 * @addr_list_lock: XXX: need comments on this one
1700 * @uc_promisc: Counter that indicates promiscuous mode
1701 * has been enabled due to the need to listen to
1702 * additional unicast addresses in a device that
1703 * does not implement ndo_set_rx_mode()
1704 * @uc: unicast mac addresses
1705 * @mc: multicast mac addresses
1706 * @dev_addrs: list of device hw addresses
1707 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1708 * @promiscuity: Number of times the NIC is told to work in
1709 * promiscuous mode; if it becomes 0 the NIC will
1710 * exit promiscuous mode
1711 * @allmulti: Counter, enables or disables allmulticast mode
1712 *
1713 * @vlan_info: VLAN info
1714 * @dsa_ptr: dsa specific data
1715 * @tipc_ptr: TIPC specific data
1716 * @atalk_ptr: AppleTalk link
1717 * @ip_ptr: IPv4 specific data
1718 * @ip6_ptr: IPv6 specific data
1719 * @ax25_ptr: AX.25 specific data
1720 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1721 *
1722 * @dev_addr: Hw address (before bcast,
1723 * because most packets are unicast)
1724 *
1725 * @_rx: Array of RX queues
1726 * @num_rx_queues: Number of RX queues
1727 * allocated at register_netdev() time
1728 * @real_num_rx_queues: Number of RX queues currently active in device
1729 *
1730 * @rx_handler: handler for received packets
1731 * @rx_handler_data: XXX: need comments on this one
1732 * @miniq_ingress: ingress/clsact qdisc specific data for
1733 * ingress processing
1734 * @ingress_queue: XXX: need comments on this one
1735 * @broadcast: hw bcast address
1736 *
1737 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1738 * indexed by RX queue number. Assigned by driver.
1739 * This must only be set if the ndo_rx_flow_steer
1740 * operation is defined
1741 * @index_hlist: Device index hash chain
1742 *
1743 * @_tx: Array of TX queues
1744 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1745 * @real_num_tx_queues: Number of TX queues currently active in device
1746 * @qdisc: Root qdisc from userspace point of view
1747 * @tx_queue_len: Max frames per queue allowed
1748 * @tx_global_lock: XXX: need comments on this one
1749 *
1750 * @xps_maps: XXX: need comments on this one
1751 * @miniq_egress: clsact qdisc specific data for
1752 * egress processing
1753 * @watchdog_timeo: Represents the timeout that is used by
1754 * the watchdog (see dev_watchdog())
1755 * @watchdog_timer: List of timers
1756 *
1757 * @pcpu_refcnt: Number of references to this device
1758 * @todo_list: Delayed register/unregister
1759 * @link_watch_list: XXX: need comments on this one
1760 *
1761 * @reg_state: Register/unregister state machine
1762 * @dismantle: Device is going to be freed
1763 * @rtnl_link_state: This enum represents the phases of creating
1764 * a new link
1765 *
1766 * @needs_free_netdev: Should unregister perform free_netdev?
1767 * @priv_destructor: Called from unregister
1768 * @npinfo: XXX: need comments on this one
1769 * @nd_net: Network namespace this network device is inside
1770 *
1771 * @ml_priv: Mid-layer private
1772 * @lstats: Loopback statistics
1773 * @tstats: Tunnel statistics
1774 * @dstats: Dummy statistics
1775 * @vstats: Virtual ethernet statistics
1776 *
1777 * @garp_port: GARP
1778 * @mrp_port: MRP
1779 *
1780 * @dev: Class/net/name entry
1781 * @sysfs_groups: Space for optional device, statistics and wireless
1782 * sysfs groups
1783 *
1784 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1785 * @rtnl_link_ops: Rtnl_link_ops
1786 *
1787 * @gso_max_size: Maximum size of generic segmentation offload
1788 * @gso_max_segs: Maximum number of segments that can be passed to the
1789 * NIC for GSO
1790 *
1791 * @dcbnl_ops: Data Center Bridging netlink ops
1792 * @num_tc: Number of traffic classes in the net device
1793 * @tc_to_txq: XXX: need comments on this one
1794 * @prio_tc_map: XXX: need comments on this one
1795 *
1796 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1797 *
1798 * @priomap: XXX: need comments on this one
1799 * @phydev: Physical device may attach itself
1800 * for hardware timestamping
1801 * @sfp_bus: attached &struct sfp_bus structure.
1802 * @qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1803 * spinlock
1804 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1805 * @qdisc_xmit_lock_key: lockdep class annotating
1806 * netdev_queue->_xmit_lock spinlock
1807 * @addr_list_lock_key: lockdep class annotating
1808 * net_device->addr_list_lock spinlock
1809 *
1810 * @proto_down: protocol port state information can be sent to the
1811 * switch driver and used to set the phys state of the
1812 * switch port.
1813 *
1814 * @wol_enabled: Wake-on-LAN is enabled
1815 *
1816 * FIXME: cleanup struct net_device such that network protocol info
1817 * moves out.
1818 */
1819
1820 struct net_device {
1821 char name[IFNAMSIZ];
1822 struct hlist_node name_hlist;
1823 struct dev_ifalias __rcu *ifalias;
1824 /*
1825 * I/O specific fields
1826 * FIXME: Merge these and struct ifmap into one
1827 */
1828 unsigned long mem_end;
1829 unsigned long mem_start;
1830 unsigned long base_addr;
1831 int irq;
1832
1833 /*
1834 * Some hardware also needs these fields (state,dev_list,
1835 * napi_list,unreg_list,close_list) but they are not
1836 * part of the usual set specified in Space.c.
1837 */
1838
1839 unsigned long state;
1840
1841 struct list_head dev_list;
1842 struct list_head napi_list;
1843 struct list_head unreg_list;
1844 struct list_head close_list;
1845 struct list_head ptype_all;
1846 struct list_head ptype_specific;
1847
1848 struct {
1849 struct list_head upper;
1850 struct list_head lower;
1851 } adj_list;
1852
1853 netdev_features_t features;
1854 netdev_features_t hw_features;
1855 netdev_features_t wanted_features;
1856 netdev_features_t vlan_features;
1857 netdev_features_t hw_enc_features;
1858 netdev_features_t mpls_features;
1859 netdev_features_t gso_partial_features;
1860
1861 int ifindex;
1862 int group;
1863
1864 struct net_device_stats stats;
1865
1866 atomic_long_t rx_dropped;
1867 atomic_long_t tx_dropped;
1868 atomic_long_t rx_nohandler;
1869
1870 /* Stats to monitor link on/off, flapping */
1871 atomic_t carrier_up_count;
1872 atomic_t carrier_down_count;
1873
1874 #ifdef CONFIG_WIRELESS_EXT
1875 const struct iw_handler_def *wireless_handlers;
1876 struct iw_public_data *wireless_data;
1877 #endif
1878 const struct net_device_ops *netdev_ops;
1879 const struct ethtool_ops *ethtool_ops;
1880 #ifdef CONFIG_NET_L3_MASTER_DEV
1881 const struct l3mdev_ops *l3mdev_ops;
1882 #endif
1883 #if IS_ENABLED(CONFIG_IPV6)
1884 const struct ndisc_ops *ndisc_ops;
1885 #endif
1886
1887 #ifdef CONFIG_XFRM_OFFLOAD
1888 const struct xfrmdev_ops *xfrmdev_ops;
1889 #endif
1890
1891 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1892 const struct tlsdev_ops *tlsdev_ops;
1893 #endif
1894
1895 const struct header_ops *header_ops;
1896
1897 unsigned int flags;
1898 unsigned int priv_flags;
1899
1900 unsigned short gflags;
1901 unsigned short padded;
1902
1903 unsigned char operstate;
1904 unsigned char link_mode;
1905
1906 unsigned char if_port;
1907 unsigned char dma;
1908
1909 /* Note : dev->mtu is often read without holding a lock.
1910 * Writers usually hold RTNL.
1911 * It is recommended to use READ_ONCE() to annotate the reads,
1912 * and to use WRITE_ONCE() to annotate the writes.
1913 */
1914 unsigned int mtu;
1915 unsigned int min_mtu;
1916 unsigned int max_mtu;
1917 unsigned short type;
1918 unsigned short hard_header_len;
1919 unsigned char min_header_len;
1920
1921 unsigned short needed_headroom;
1922 unsigned short needed_tailroom;
1923
1924 /* Interface address info. */
1925 unsigned char perm_addr[MAX_ADDR_LEN];
1926 unsigned char addr_assign_type;
1927 unsigned char addr_len;
1928 unsigned char upper_level;
1929 unsigned char lower_level;
1930 unsigned short neigh_priv_len;
1931 unsigned short dev_id;
1932 unsigned short dev_port;
1933 spinlock_t addr_list_lock;
1934 unsigned char name_assign_type;
1935 bool uc_promisc;
1936 struct netdev_hw_addr_list uc;
1937 struct netdev_hw_addr_list mc;
1938 struct netdev_hw_addr_list dev_addrs;
1939
1940 #ifdef CONFIG_SYSFS
1941 struct kset *queues_kset;
1942 #endif
1943 unsigned int promiscuity;
1944 unsigned int allmulti;
1945
1946
1947 /* Protocol-specific pointers */
1948
1949 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1950 struct vlan_info __rcu *vlan_info;
1951 #endif
1952 #if IS_ENABLED(CONFIG_NET_DSA)
1953 struct dsa_port *dsa_ptr;
1954 #endif
1955 #if IS_ENABLED(CONFIG_TIPC)
1956 struct tipc_bearer __rcu *tipc_ptr;
1957 #endif
1958 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1959 void *atalk_ptr;
1960 #endif
1961 struct in_device __rcu *ip_ptr;
1962 struct inet6_dev __rcu *ip6_ptr;
1963 #if IS_ENABLED(CONFIG_AX25)
1964 void *ax25_ptr;
1965 #endif
1966 struct wireless_dev *ieee80211_ptr;
1967 struct wpan_dev *ieee802154_ptr;
1968 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1969 struct mpls_dev __rcu *mpls_ptr;
1970 #endif
1971
1972 /*
1973 * Cache lines mostly used on receive path (including eth_type_trans())
1974 */
1975 /* Interface address info used in eth_type_trans() */
1976 unsigned char *dev_addr;
1977
1978 struct netdev_rx_queue *_rx;
1979 unsigned int num_rx_queues;
1980 unsigned int real_num_rx_queues;
1981
1982 struct bpf_prog __rcu *xdp_prog;
1983 unsigned long gro_flush_timeout;
1984 rx_handler_func_t __rcu *rx_handler;
1985 void __rcu *rx_handler_data;
1986
1987 #ifdef CONFIG_NET_CLS_ACT
1988 struct mini_Qdisc __rcu *miniq_ingress;
1989 #endif
1990 struct netdev_queue __rcu *ingress_queue;
1991 #ifdef CONFIG_NETFILTER_INGRESS
1992 struct nf_hook_entries __rcu *nf_hooks_ingress;
1993 #endif
1994
1995 unsigned char broadcast[MAX_ADDR_LEN];
1996 #ifdef CONFIG_RFS_ACCEL
1997 struct cpu_rmap *rx_cpu_rmap;
1998 #endif
1999 struct hlist_node index_hlist;
2000
2001 /*
2002 * Cache lines mostly used on transmit path
2003 */
2004 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2005 unsigned int num_tx_queues;
2006 unsigned int real_num_tx_queues;
2007 struct Qdisc __rcu *qdisc;
2008 #ifdef CONFIG_NET_SCHED
2009 DECLARE_HASHTABLE (qdisc_hash, 4);
2010 #endif
2011 unsigned int tx_queue_len;
2012 spinlock_t tx_global_lock;
2013 int watchdog_timeo;
2014
2015 #ifdef CONFIG_XPS
2016 struct xps_dev_maps __rcu *xps_cpus_map;
2017 struct xps_dev_maps __rcu *xps_rxqs_map;
2018 #endif
2019 #ifdef CONFIG_NET_CLS_ACT
2020 struct mini_Qdisc __rcu *miniq_egress;
2021 #endif
2022
2023 /* These may be needed for future network-power-down code. */
2024 struct timer_list watchdog_timer;
2025
2026 int __percpu *pcpu_refcnt;
2027 struct list_head todo_list;
2028
2029 struct list_head link_watch_list;
2030
2031 enum { NETREG_UNINITIALIZED=0,
2032 NETREG_REGISTERED, /* completed register_netdevice */
2033 NETREG_UNREGISTERING, /* called unregister_netdevice */
2034 NETREG_UNREGISTERED, /* completed unregister todo */
2035 NETREG_RELEASED, /* called free_netdev */
2036 NETREG_DUMMY, /* dummy device for NAPI poll */
2037 } reg_state:8;
2038
2039 bool dismantle;
2040
2041 enum {
2042 RTNL_LINK_INITIALIZED,
2043 RTNL_LINK_INITIALIZING,
2044 } rtnl_link_state:16;
2045
2046 bool needs_free_netdev;
2047 void (*priv_destructor)(struct net_device *dev);
2048
2049 #ifdef CONFIG_NETPOLL
2050 struct netpoll_info __rcu *npinfo;
2051 #endif
2052
2053 possible_net_t nd_net;
2054
2055 /* mid-layer private */
2056 union {
2057 void *ml_priv;
2058 struct pcpu_lstats __percpu *lstats;
2059 struct pcpu_sw_netstats __percpu *tstats;
2060 struct pcpu_dstats __percpu *dstats;
2061 };
2062
2063 #if IS_ENABLED(CONFIG_GARP)
2064 struct garp_port __rcu *garp_port;
2065 #endif
2066 #if IS_ENABLED(CONFIG_MRP)
2067 struct mrp_port __rcu *mrp_port;
2068 #endif
2069
2070 struct device dev;
2071 const struct attribute_group *sysfs_groups[4];
2072 const struct attribute_group *sysfs_rx_queue_group;
2073
2074 const struct rtnl_link_ops *rtnl_link_ops;
2075
2076 /* for setting kernel sock attribute on TCP connection setup */
2077 #define GSO_MAX_SIZE 65536
2078 unsigned int gso_max_size;
2079 #define GSO_MAX_SEGS 65535
2080 u16 gso_max_segs;
2081
2082 #ifdef CONFIG_DCB
2083 const struct dcbnl_rtnl_ops *dcbnl_ops;
2084 #endif
2085 s16 num_tc;
2086 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2087 u8 prio_tc_map[TC_BITMASK + 1];
2088
2089 #if IS_ENABLED(CONFIG_FCOE)
2090 unsigned int fcoe_ddp_xid;
2091 #endif
2092 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2093 struct netprio_map __rcu *priomap;
2094 #endif
2095 struct phy_device *phydev;
2096 struct sfp_bus *sfp_bus;
2097 struct lock_class_key qdisc_tx_busylock_key;
2098 struct lock_class_key qdisc_running_key;
2099 struct lock_class_key qdisc_xmit_lock_key;
2100 struct lock_class_key addr_list_lock_key;
2101 bool proto_down;
2102 unsigned wol_enabled:1;
2103
2104 ANDROID_KABI_RESERVE(1);
2105 ANDROID_KABI_RESERVE(2);
2106 ANDROID_KABI_RESERVE(3);
2107 ANDROID_KABI_RESERVE(4);
2108 ANDROID_KABI_RESERVE(5);
2109 ANDROID_KABI_RESERVE(6);
2110 ANDROID_KABI_RESERVE(7);
2111 ANDROID_KABI_RESERVE(8);
2112
2113 };
2114 #define to_net_dev(d) container_of(d, struct net_device, dev)
2115
netif_elide_gro(const struct net_device * dev)2116 static inline bool netif_elide_gro(const struct net_device *dev)
2117 {
2118 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2119 return true;
2120 return false;
2121 }
2122
2123 #define NETDEV_ALIGN 32
2124
2125 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2126 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2127 {
2128 return dev->prio_tc_map[prio & TC_BITMASK];
2129 }
2130
2131 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2132 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2133 {
2134 if (tc >= dev->num_tc)
2135 return -EINVAL;
2136
2137 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2138 return 0;
2139 }
2140
2141 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2142 void netdev_reset_tc(struct net_device *dev);
2143 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2144 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2145
2146 static inline
netdev_get_num_tc(struct net_device * dev)2147 int netdev_get_num_tc(struct net_device *dev)
2148 {
2149 return dev->num_tc;
2150 }
2151
2152 void netdev_unbind_sb_channel(struct net_device *dev,
2153 struct net_device *sb_dev);
2154 int netdev_bind_sb_channel_queue(struct net_device *dev,
2155 struct net_device *sb_dev,
2156 u8 tc, u16 count, u16 offset);
2157 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2158 static inline int netdev_get_sb_channel(struct net_device *dev)
2159 {
2160 return max_t(int, -dev->num_tc, 0);
2161 }
2162
2163 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2164 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2165 unsigned int index)
2166 {
2167 return &dev->_tx[index];
2168 }
2169
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2170 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2171 const struct sk_buff *skb)
2172 {
2173 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2174 }
2175
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2176 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2177 void (*f)(struct net_device *,
2178 struct netdev_queue *,
2179 void *),
2180 void *arg)
2181 {
2182 unsigned int i;
2183
2184 for (i = 0; i < dev->num_tx_queues; i++)
2185 f(dev, &dev->_tx[i], arg);
2186 }
2187
2188 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2189 struct net_device *sb_dev);
2190 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2191 struct sk_buff *skb,
2192 struct net_device *sb_dev);
2193
2194 /* returns the headroom that the master device needs to take in account
2195 * when forwarding to this dev
2196 */
netdev_get_fwd_headroom(struct net_device * dev)2197 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2198 {
2199 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2200 }
2201
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2202 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2203 {
2204 if (dev->netdev_ops->ndo_set_rx_headroom)
2205 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2206 }
2207
2208 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2209 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2210 {
2211 netdev_set_rx_headroom(dev, -1);
2212 }
2213
2214 /*
2215 * Net namespace inlines
2216 */
2217 static inline
dev_net(const struct net_device * dev)2218 struct net *dev_net(const struct net_device *dev)
2219 {
2220 return read_pnet(&dev->nd_net);
2221 }
2222
2223 static inline
dev_net_set(struct net_device * dev,struct net * net)2224 void dev_net_set(struct net_device *dev, struct net *net)
2225 {
2226 write_pnet(&dev->nd_net, net);
2227 }
2228
2229 /**
2230 * netdev_priv - access network device private data
2231 * @dev: network device
2232 *
2233 * Get network device private data
2234 */
netdev_priv(const struct net_device * dev)2235 static inline void *netdev_priv(const struct net_device *dev)
2236 {
2237 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2238 }
2239
2240 /* Set the sysfs physical device reference for the network logical device
2241 * if set prior to registration will cause a symlink during initialization.
2242 */
2243 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2244
2245 /* Set the sysfs device type for the network logical device to allow
2246 * fine-grained identification of different network device types. For
2247 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2248 */
2249 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2250
2251 /* Default NAPI poll() weight
2252 * Device drivers are strongly advised to not use bigger value
2253 */
2254 #define NAPI_POLL_WEIGHT 64
2255
2256 /**
2257 * netif_napi_add - initialize a NAPI context
2258 * @dev: network device
2259 * @napi: NAPI context
2260 * @poll: polling function
2261 * @weight: default weight
2262 *
2263 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2264 * *any* of the other NAPI-related functions.
2265 */
2266 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2267 int (*poll)(struct napi_struct *, int), int weight);
2268
2269 /**
2270 * netif_tx_napi_add - initialize a NAPI context
2271 * @dev: network device
2272 * @napi: NAPI context
2273 * @poll: polling function
2274 * @weight: default weight
2275 *
2276 * This variant of netif_napi_add() should be used from drivers using NAPI
2277 * to exclusively poll a TX queue.
2278 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2279 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2280 static inline void netif_tx_napi_add(struct net_device *dev,
2281 struct napi_struct *napi,
2282 int (*poll)(struct napi_struct *, int),
2283 int weight)
2284 {
2285 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2286 netif_napi_add(dev, napi, poll, weight);
2287 }
2288
2289 /**
2290 * netif_napi_del - remove a NAPI context
2291 * @napi: NAPI context
2292 *
2293 * netif_napi_del() removes a NAPI context from the network device NAPI list
2294 */
2295 void netif_napi_del(struct napi_struct *napi);
2296
2297 struct napi_gro_cb {
2298 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2299 void *frag0;
2300
2301 /* Length of frag0. */
2302 unsigned int frag0_len;
2303
2304 /* This indicates where we are processing relative to skb->data. */
2305 int data_offset;
2306
2307 /* This is non-zero if the packet cannot be merged with the new skb. */
2308 u16 flush;
2309
2310 /* Save the IP ID here and check when we get to the transport layer */
2311 u16 flush_id;
2312
2313 /* Number of segments aggregated. */
2314 u16 count;
2315
2316 /* Start offset for remote checksum offload */
2317 u16 gro_remcsum_start;
2318
2319 /* jiffies when first packet was created/queued */
2320 unsigned long age;
2321
2322 /* Used in ipv6_gro_receive() and foo-over-udp */
2323 u16 proto;
2324
2325 /* This is non-zero if the packet may be of the same flow. */
2326 u8 same_flow:1;
2327
2328 /* Used in tunnel GRO receive */
2329 u8 encap_mark:1;
2330
2331 /* GRO checksum is valid */
2332 u8 csum_valid:1;
2333
2334 /* Number of checksums via CHECKSUM_UNNECESSARY */
2335 u8 csum_cnt:3;
2336
2337 /* Free the skb? */
2338 u8 free:2;
2339 #define NAPI_GRO_FREE 1
2340 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2341
2342 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2343 u8 is_ipv6:1;
2344
2345 /* Used in GRE, set in fou/gue_gro_receive */
2346 u8 is_fou:1;
2347
2348 /* Used to determine if flush_id can be ignored */
2349 u8 is_atomic:1;
2350
2351 /* Number of gro_receive callbacks this packet already went through */
2352 u8 recursion_counter:4;
2353
2354 /* GRO is done by frag_list pointer chaining. */
2355 u8 is_flist:1;
2356
2357 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2358 __wsum csum;
2359
2360 /* used in skb_gro_receive() slow path */
2361 struct sk_buff *last;
2362 };
2363
2364 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2365
2366 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2367 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2368 {
2369 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2370 }
2371
2372 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
call_gro_receive(gro_receive_t cb,struct list_head * head,struct sk_buff * skb)2373 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2374 struct list_head *head,
2375 struct sk_buff *skb)
2376 {
2377 if (unlikely(gro_recursion_inc_test(skb))) {
2378 NAPI_GRO_CB(skb)->flush |= 1;
2379 return NULL;
2380 }
2381
2382 return cb(head, skb);
2383 }
2384
2385 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2386 struct sk_buff *);
call_gro_receive_sk(gro_receive_sk_t cb,struct sock * sk,struct list_head * head,struct sk_buff * skb)2387 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2388 struct sock *sk,
2389 struct list_head *head,
2390 struct sk_buff *skb)
2391 {
2392 if (unlikely(gro_recursion_inc_test(skb))) {
2393 NAPI_GRO_CB(skb)->flush |= 1;
2394 return NULL;
2395 }
2396
2397 return cb(sk, head, skb);
2398 }
2399
2400 struct packet_type {
2401 __be16 type; /* This is really htons(ether_type). */
2402 bool ignore_outgoing;
2403 struct net_device *dev; /* NULL is wildcarded here */
2404 int (*func) (struct sk_buff *,
2405 struct net_device *,
2406 struct packet_type *,
2407 struct net_device *);
2408 void (*list_func) (struct list_head *,
2409 struct packet_type *,
2410 struct net_device *);
2411 bool (*id_match)(struct packet_type *ptype,
2412 struct sock *sk);
2413 void *af_packet_priv;
2414 struct list_head list;
2415
2416 ANDROID_KABI_RESERVE(1);
2417 ANDROID_KABI_RESERVE(2);
2418 ANDROID_KABI_RESERVE(3);
2419 ANDROID_KABI_RESERVE(4);
2420 };
2421
2422 struct offload_callbacks {
2423 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2424 netdev_features_t features);
2425 struct sk_buff *(*gro_receive)(struct list_head *head,
2426 struct sk_buff *skb);
2427 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2428 };
2429
2430 struct packet_offload {
2431 __be16 type; /* This is really htons(ether_type). */
2432 u16 priority;
2433 struct offload_callbacks callbacks;
2434 struct list_head list;
2435 };
2436
2437 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2438 struct pcpu_sw_netstats {
2439 u64 rx_packets;
2440 u64 rx_bytes;
2441 u64 tx_packets;
2442 u64 tx_bytes;
2443 struct u64_stats_sync syncp;
2444 } __aligned(4 * sizeof(u64));
2445
2446 struct pcpu_lstats {
2447 u64 packets;
2448 u64 bytes;
2449 struct u64_stats_sync syncp;
2450 } __aligned(2 * sizeof(u64));
2451
2452 #define __netdev_alloc_pcpu_stats(type, gfp) \
2453 ({ \
2454 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2455 if (pcpu_stats) { \
2456 int __cpu; \
2457 for_each_possible_cpu(__cpu) { \
2458 typeof(type) *stat; \
2459 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2460 u64_stats_init(&stat->syncp); \
2461 } \
2462 } \
2463 pcpu_stats; \
2464 })
2465
2466 #define netdev_alloc_pcpu_stats(type) \
2467 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2468
2469 enum netdev_lag_tx_type {
2470 NETDEV_LAG_TX_TYPE_UNKNOWN,
2471 NETDEV_LAG_TX_TYPE_RANDOM,
2472 NETDEV_LAG_TX_TYPE_BROADCAST,
2473 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2474 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2475 NETDEV_LAG_TX_TYPE_HASH,
2476 };
2477
2478 enum netdev_lag_hash {
2479 NETDEV_LAG_HASH_NONE,
2480 NETDEV_LAG_HASH_L2,
2481 NETDEV_LAG_HASH_L34,
2482 NETDEV_LAG_HASH_L23,
2483 NETDEV_LAG_HASH_E23,
2484 NETDEV_LAG_HASH_E34,
2485 NETDEV_LAG_HASH_UNKNOWN,
2486 };
2487
2488 struct netdev_lag_upper_info {
2489 enum netdev_lag_tx_type tx_type;
2490 enum netdev_lag_hash hash_type;
2491 };
2492
2493 struct netdev_lag_lower_state_info {
2494 u8 link_up : 1,
2495 tx_enabled : 1;
2496 };
2497
2498 #include <linux/notifier.h>
2499
2500 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2501 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2502 * adding new types.
2503 */
2504 enum netdev_cmd {
2505 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2506 NETDEV_DOWN,
2507 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2508 detected a hardware crash and restarted
2509 - we can use this eg to kick tcp sessions
2510 once done */
2511 NETDEV_CHANGE, /* Notify device state change */
2512 NETDEV_REGISTER,
2513 NETDEV_UNREGISTER,
2514 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2515 NETDEV_CHANGEADDR, /* notify after the address change */
2516 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2517 NETDEV_GOING_DOWN,
2518 NETDEV_CHANGENAME,
2519 NETDEV_FEAT_CHANGE,
2520 NETDEV_BONDING_FAILOVER,
2521 NETDEV_PRE_UP,
2522 NETDEV_PRE_TYPE_CHANGE,
2523 NETDEV_POST_TYPE_CHANGE,
2524 NETDEV_POST_INIT,
2525 NETDEV_RELEASE,
2526 NETDEV_NOTIFY_PEERS,
2527 NETDEV_JOIN,
2528 NETDEV_CHANGEUPPER,
2529 NETDEV_RESEND_IGMP,
2530 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2531 NETDEV_CHANGEINFODATA,
2532 NETDEV_BONDING_INFO,
2533 NETDEV_PRECHANGEUPPER,
2534 NETDEV_CHANGELOWERSTATE,
2535 NETDEV_UDP_TUNNEL_PUSH_INFO,
2536 NETDEV_UDP_TUNNEL_DROP_INFO,
2537 NETDEV_CHANGE_TX_QUEUE_LEN,
2538 NETDEV_CVLAN_FILTER_PUSH_INFO,
2539 NETDEV_CVLAN_FILTER_DROP_INFO,
2540 NETDEV_SVLAN_FILTER_PUSH_INFO,
2541 NETDEV_SVLAN_FILTER_DROP_INFO,
2542 };
2543 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2544
2545 int register_netdevice_notifier(struct notifier_block *nb);
2546 int unregister_netdevice_notifier(struct notifier_block *nb);
2547
2548 struct netdev_notifier_info {
2549 struct net_device *dev;
2550 struct netlink_ext_ack *extack;
2551 };
2552
2553 struct netdev_notifier_info_ext {
2554 struct netdev_notifier_info info; /* must be first */
2555 union {
2556 u32 mtu;
2557 } ext;
2558 };
2559
2560 struct netdev_notifier_change_info {
2561 struct netdev_notifier_info info; /* must be first */
2562 unsigned int flags_changed;
2563 };
2564
2565 struct netdev_notifier_changeupper_info {
2566 struct netdev_notifier_info info; /* must be first */
2567 struct net_device *upper_dev; /* new upper dev */
2568 bool master; /* is upper dev master */
2569 bool linking; /* is the notification for link or unlink */
2570 void *upper_info; /* upper dev info */
2571 };
2572
2573 struct netdev_notifier_changelowerstate_info {
2574 struct netdev_notifier_info info; /* must be first */
2575 void *lower_state_info; /* is lower dev state */
2576 };
2577
2578 struct netdev_notifier_pre_changeaddr_info {
2579 struct netdev_notifier_info info; /* must be first */
2580 const unsigned char *dev_addr;
2581 };
2582
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2583 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2584 struct net_device *dev)
2585 {
2586 info->dev = dev;
2587 info->extack = NULL;
2588 }
2589
2590 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2591 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2592 {
2593 return info->dev;
2594 }
2595
2596 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2597 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2598 {
2599 return info->extack;
2600 }
2601
2602 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2603
2604
2605 extern rwlock_t dev_base_lock; /* Device list lock */
2606
2607 #define for_each_netdev(net, d) \
2608 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2609 #define for_each_netdev_reverse(net, d) \
2610 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2611 #define for_each_netdev_rcu(net, d) \
2612 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2613 #define for_each_netdev_safe(net, d, n) \
2614 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2615 #define for_each_netdev_continue(net, d) \
2616 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2617 #define for_each_netdev_continue_rcu(net, d) \
2618 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2619 #define for_each_netdev_in_bond_rcu(bond, slave) \
2620 for_each_netdev_rcu(&init_net, slave) \
2621 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2622 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2623
next_net_device(struct net_device * dev)2624 static inline struct net_device *next_net_device(struct net_device *dev)
2625 {
2626 struct list_head *lh;
2627 struct net *net;
2628
2629 net = dev_net(dev);
2630 lh = dev->dev_list.next;
2631 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2632 }
2633
next_net_device_rcu(struct net_device * dev)2634 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2635 {
2636 struct list_head *lh;
2637 struct net *net;
2638
2639 net = dev_net(dev);
2640 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2641 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2642 }
2643
first_net_device(struct net * net)2644 static inline struct net_device *first_net_device(struct net *net)
2645 {
2646 return list_empty(&net->dev_base_head) ? NULL :
2647 net_device_entry(net->dev_base_head.next);
2648 }
2649
first_net_device_rcu(struct net * net)2650 static inline struct net_device *first_net_device_rcu(struct net *net)
2651 {
2652 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2653
2654 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2655 }
2656
2657 int netdev_boot_setup_check(struct net_device *dev);
2658 unsigned long netdev_boot_base(const char *prefix, int unit);
2659 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2660 const char *hwaddr);
2661 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2662 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2663 void dev_add_pack(struct packet_type *pt);
2664 void dev_remove_pack(struct packet_type *pt);
2665 void __dev_remove_pack(struct packet_type *pt);
2666 void dev_add_offload(struct packet_offload *po);
2667 void dev_remove_offload(struct packet_offload *po);
2668
2669 int dev_get_iflink(const struct net_device *dev);
2670 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2671 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2672 unsigned short mask);
2673 struct net_device *dev_get_by_name(struct net *net, const char *name);
2674 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2675 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2676 int dev_alloc_name(struct net_device *dev, const char *name);
2677 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2678 void dev_close(struct net_device *dev);
2679 void dev_close_many(struct list_head *head, bool unlink);
2680 void dev_disable_lro(struct net_device *dev);
2681 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2682 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2683 struct net_device *sb_dev);
2684 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2685 struct net_device *sb_dev);
2686 int dev_queue_xmit(struct sk_buff *skb);
2687 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2688 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2689 int register_netdevice(struct net_device *dev);
2690 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2691 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2692 static inline void unregister_netdevice(struct net_device *dev)
2693 {
2694 unregister_netdevice_queue(dev, NULL);
2695 }
2696
2697 int netdev_refcnt_read(const struct net_device *dev);
2698 void free_netdev(struct net_device *dev);
2699 void netdev_freemem(struct net_device *dev);
2700 void synchronize_net(void);
2701 int init_dummy_netdev(struct net_device *dev);
2702
2703 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2704 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2705 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2706 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2707 int netdev_get_name(struct net *net, char *name, int ifindex);
2708 int dev_restart(struct net_device *dev);
2709 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2710 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2711
skb_gro_offset(const struct sk_buff * skb)2712 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2713 {
2714 return NAPI_GRO_CB(skb)->data_offset;
2715 }
2716
skb_gro_len(const struct sk_buff * skb)2717 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2718 {
2719 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2720 }
2721
skb_gro_pull(struct sk_buff * skb,unsigned int len)2722 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2723 {
2724 NAPI_GRO_CB(skb)->data_offset += len;
2725 }
2726
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)2727 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2728 unsigned int offset)
2729 {
2730 return NAPI_GRO_CB(skb)->frag0 + offset;
2731 }
2732
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)2733 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2734 {
2735 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2736 }
2737
skb_gro_frag0_invalidate(struct sk_buff * skb)2738 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2739 {
2740 NAPI_GRO_CB(skb)->frag0 = NULL;
2741 NAPI_GRO_CB(skb)->frag0_len = 0;
2742 }
2743
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)2744 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2745 unsigned int offset)
2746 {
2747 if (!pskb_may_pull(skb, hlen))
2748 return NULL;
2749
2750 skb_gro_frag0_invalidate(skb);
2751 return skb->data + offset;
2752 }
2753
skb_gro_network_header(struct sk_buff * skb)2754 static inline void *skb_gro_network_header(struct sk_buff *skb)
2755 {
2756 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2757 skb_network_offset(skb);
2758 }
2759
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)2760 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2761 const void *start, unsigned int len)
2762 {
2763 if (NAPI_GRO_CB(skb)->csum_valid)
2764 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2765 csum_partial(start, len, 0));
2766 }
2767
2768 /* GRO checksum functions. These are logical equivalents of the normal
2769 * checksum functions (in skbuff.h) except that they operate on the GRO
2770 * offsets and fields in sk_buff.
2771 */
2772
2773 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2774
skb_at_gro_remcsum_start(struct sk_buff * skb)2775 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2776 {
2777 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2778 }
2779
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)2780 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2781 bool zero_okay,
2782 __sum16 check)
2783 {
2784 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2785 skb_checksum_start_offset(skb) <
2786 skb_gro_offset(skb)) &&
2787 !skb_at_gro_remcsum_start(skb) &&
2788 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2789 (!zero_okay || check));
2790 }
2791
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)2792 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2793 __wsum psum)
2794 {
2795 if (NAPI_GRO_CB(skb)->csum_valid &&
2796 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2797 return 0;
2798
2799 NAPI_GRO_CB(skb)->csum = psum;
2800
2801 return __skb_gro_checksum_complete(skb);
2802 }
2803
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)2804 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2805 {
2806 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2807 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2808 NAPI_GRO_CB(skb)->csum_cnt--;
2809 } else {
2810 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2811 * verified a new top level checksum or an encapsulated one
2812 * during GRO. This saves work if we fallback to normal path.
2813 */
2814 __skb_incr_checksum_unnecessary(skb);
2815 }
2816 }
2817
2818 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2819 compute_pseudo) \
2820 ({ \
2821 __sum16 __ret = 0; \
2822 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2823 __ret = __skb_gro_checksum_validate_complete(skb, \
2824 compute_pseudo(skb, proto)); \
2825 if (!__ret) \
2826 skb_gro_incr_csum_unnecessary(skb); \
2827 __ret; \
2828 })
2829
2830 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2831 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2832
2833 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2834 compute_pseudo) \
2835 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2836
2837 #define skb_gro_checksum_simple_validate(skb) \
2838 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2839
__skb_gro_checksum_convert_check(struct sk_buff * skb)2840 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2841 {
2842 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2843 !NAPI_GRO_CB(skb)->csum_valid);
2844 }
2845
__skb_gro_checksum_convert(struct sk_buff * skb,__sum16 check,__wsum pseudo)2846 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2847 __sum16 check, __wsum pseudo)
2848 {
2849 NAPI_GRO_CB(skb)->csum = ~pseudo;
2850 NAPI_GRO_CB(skb)->csum_valid = 1;
2851 }
2852
2853 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2854 do { \
2855 if (__skb_gro_checksum_convert_check(skb)) \
2856 __skb_gro_checksum_convert(skb, check, \
2857 compute_pseudo(skb, proto)); \
2858 } while (0)
2859
2860 struct gro_remcsum {
2861 int offset;
2862 __wsum delta;
2863 };
2864
skb_gro_remcsum_init(struct gro_remcsum * grc)2865 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2866 {
2867 grc->offset = 0;
2868 grc->delta = 0;
2869 }
2870
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)2871 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2872 unsigned int off, size_t hdrlen,
2873 int start, int offset,
2874 struct gro_remcsum *grc,
2875 bool nopartial)
2876 {
2877 __wsum delta;
2878 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2879
2880 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2881
2882 if (!nopartial) {
2883 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2884 return ptr;
2885 }
2886
2887 ptr = skb_gro_header_fast(skb, off);
2888 if (skb_gro_header_hard(skb, off + plen)) {
2889 ptr = skb_gro_header_slow(skb, off + plen, off);
2890 if (!ptr)
2891 return NULL;
2892 }
2893
2894 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2895 start, offset);
2896
2897 /* Adjust skb->csum since we changed the packet */
2898 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2899
2900 grc->offset = off + hdrlen + offset;
2901 grc->delta = delta;
2902
2903 return ptr;
2904 }
2905
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)2906 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2907 struct gro_remcsum *grc)
2908 {
2909 void *ptr;
2910 size_t plen = grc->offset + sizeof(u16);
2911
2912 if (!grc->delta)
2913 return;
2914
2915 ptr = skb_gro_header_fast(skb, grc->offset);
2916 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2917 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2918 if (!ptr)
2919 return;
2920 }
2921
2922 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2923 }
2924
2925 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)2926 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2927 {
2928 if (PTR_ERR(pp) != -EINPROGRESS)
2929 NAPI_GRO_CB(skb)->flush |= flush;
2930 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)2931 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2932 struct sk_buff *pp,
2933 int flush,
2934 struct gro_remcsum *grc)
2935 {
2936 if (PTR_ERR(pp) != -EINPROGRESS) {
2937 NAPI_GRO_CB(skb)->flush |= flush;
2938 skb_gro_remcsum_cleanup(skb, grc);
2939 skb->remcsum_offload = 0;
2940 }
2941 }
2942 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)2943 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2944 {
2945 NAPI_GRO_CB(skb)->flush |= flush;
2946 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)2947 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2948 struct sk_buff *pp,
2949 int flush,
2950 struct gro_remcsum *grc)
2951 {
2952 NAPI_GRO_CB(skb)->flush |= flush;
2953 skb_gro_remcsum_cleanup(skb, grc);
2954 skb->remcsum_offload = 0;
2955 }
2956 #endif
2957
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)2958 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2959 unsigned short type,
2960 const void *daddr, const void *saddr,
2961 unsigned int len)
2962 {
2963 if (!dev->header_ops || !dev->header_ops->create)
2964 return 0;
2965
2966 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2967 }
2968
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)2969 static inline int dev_parse_header(const struct sk_buff *skb,
2970 unsigned char *haddr)
2971 {
2972 const struct net_device *dev = skb->dev;
2973
2974 if (!dev->header_ops || !dev->header_ops->parse)
2975 return 0;
2976 return dev->header_ops->parse(skb, haddr);
2977 }
2978
dev_parse_header_protocol(const struct sk_buff * skb)2979 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2980 {
2981 const struct net_device *dev = skb->dev;
2982
2983 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2984 return 0;
2985 return dev->header_ops->parse_protocol(skb);
2986 }
2987
2988 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)2989 static inline bool dev_validate_header(const struct net_device *dev,
2990 char *ll_header, int len)
2991 {
2992 if (likely(len >= dev->hard_header_len))
2993 return true;
2994 if (len < dev->min_header_len)
2995 return false;
2996
2997 if (capable(CAP_SYS_RAWIO)) {
2998 memset(ll_header + len, 0, dev->hard_header_len - len);
2999 return true;
3000 }
3001
3002 if (dev->header_ops && dev->header_ops->validate)
3003 return dev->header_ops->validate(ll_header, len);
3004
3005 return false;
3006 }
3007
3008 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3009 int len, int size);
3010 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
unregister_gifconf(unsigned int family)3011 static inline int unregister_gifconf(unsigned int family)
3012 {
3013 return register_gifconf(family, NULL);
3014 }
3015
3016 #ifdef CONFIG_NET_FLOW_LIMIT
3017 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3018 struct sd_flow_limit {
3019 u64 count;
3020 unsigned int num_buckets;
3021 unsigned int history_head;
3022 u16 history[FLOW_LIMIT_HISTORY];
3023 u8 buckets[];
3024 };
3025
3026 extern int netdev_flow_limit_table_len;
3027 #endif /* CONFIG_NET_FLOW_LIMIT */
3028
3029 /*
3030 * Incoming packets are placed on per-CPU queues
3031 */
3032 struct softnet_data {
3033 struct list_head poll_list;
3034 struct sk_buff_head process_queue;
3035
3036 /* stats */
3037 unsigned int processed;
3038 unsigned int time_squeeze;
3039 unsigned int received_rps;
3040 #ifdef CONFIG_RPS
3041 struct softnet_data *rps_ipi_list;
3042 #endif
3043 #ifdef CONFIG_NET_FLOW_LIMIT
3044 struct sd_flow_limit __rcu *flow_limit;
3045 #endif
3046 struct Qdisc *output_queue;
3047 struct Qdisc **output_queue_tailp;
3048 struct sk_buff *completion_queue;
3049 #ifdef CONFIG_XFRM_OFFLOAD
3050 struct sk_buff_head xfrm_backlog;
3051 #endif
3052 /* written and read only by owning cpu: */
3053 struct {
3054 u16 recursion;
3055 u8 more;
3056 } xmit;
3057 #ifdef CONFIG_RPS
3058 /* input_queue_head should be written by cpu owning this struct,
3059 * and only read by other cpus. Worth using a cache line.
3060 */
3061 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3062
3063 /* Elements below can be accessed between CPUs for RPS/RFS */
3064 call_single_data_t csd ____cacheline_aligned_in_smp;
3065 struct softnet_data *rps_ipi_next;
3066 unsigned int cpu;
3067 unsigned int input_queue_tail;
3068 #endif
3069 unsigned int dropped;
3070 struct sk_buff_head input_pkt_queue;
3071 struct napi_struct backlog;
3072
3073 };
3074
input_queue_head_incr(struct softnet_data * sd)3075 static inline void input_queue_head_incr(struct softnet_data *sd)
3076 {
3077 #ifdef CONFIG_RPS
3078 sd->input_queue_head++;
3079 #endif
3080 }
3081
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3082 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3083 unsigned int *qtail)
3084 {
3085 #ifdef CONFIG_RPS
3086 *qtail = ++sd->input_queue_tail;
3087 #endif
3088 }
3089
3090 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3091
dev_recursion_level(void)3092 static inline int dev_recursion_level(void)
3093 {
3094 return this_cpu_read(softnet_data.xmit.recursion);
3095 }
3096
3097 #define XMIT_RECURSION_LIMIT 8
dev_xmit_recursion(void)3098 static inline bool dev_xmit_recursion(void)
3099 {
3100 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3101 XMIT_RECURSION_LIMIT);
3102 }
3103
dev_xmit_recursion_inc(void)3104 static inline void dev_xmit_recursion_inc(void)
3105 {
3106 __this_cpu_inc(softnet_data.xmit.recursion);
3107 }
3108
dev_xmit_recursion_dec(void)3109 static inline void dev_xmit_recursion_dec(void)
3110 {
3111 __this_cpu_dec(softnet_data.xmit.recursion);
3112 }
3113
3114 void __netif_schedule(struct Qdisc *q);
3115 void netif_schedule_queue(struct netdev_queue *txq);
3116
netif_tx_schedule_all(struct net_device * dev)3117 static inline void netif_tx_schedule_all(struct net_device *dev)
3118 {
3119 unsigned int i;
3120
3121 for (i = 0; i < dev->num_tx_queues; i++)
3122 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3123 }
3124
netif_tx_start_queue(struct netdev_queue * dev_queue)3125 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3126 {
3127 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3128 }
3129
3130 /**
3131 * netif_start_queue - allow transmit
3132 * @dev: network device
3133 *
3134 * Allow upper layers to call the device hard_start_xmit routine.
3135 */
netif_start_queue(struct net_device * dev)3136 static inline void netif_start_queue(struct net_device *dev)
3137 {
3138 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3139 }
3140
netif_tx_start_all_queues(struct net_device * dev)3141 static inline void netif_tx_start_all_queues(struct net_device *dev)
3142 {
3143 unsigned int i;
3144
3145 for (i = 0; i < dev->num_tx_queues; i++) {
3146 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3147 netif_tx_start_queue(txq);
3148 }
3149 }
3150
3151 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3152
3153 /**
3154 * netif_wake_queue - restart transmit
3155 * @dev: network device
3156 *
3157 * Allow upper layers to call the device hard_start_xmit routine.
3158 * Used for flow control when transmit resources are available.
3159 */
netif_wake_queue(struct net_device * dev)3160 static inline void netif_wake_queue(struct net_device *dev)
3161 {
3162 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3163 }
3164
netif_tx_wake_all_queues(struct net_device * dev)3165 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3166 {
3167 unsigned int i;
3168
3169 for (i = 0; i < dev->num_tx_queues; i++) {
3170 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3171 netif_tx_wake_queue(txq);
3172 }
3173 }
3174
netif_tx_stop_queue(struct netdev_queue * dev_queue)3175 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3176 {
3177 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3178 }
3179
3180 /**
3181 * netif_stop_queue - stop transmitted packets
3182 * @dev: network device
3183 *
3184 * Stop upper layers calling the device hard_start_xmit routine.
3185 * Used for flow control when transmit resources are unavailable.
3186 */
netif_stop_queue(struct net_device * dev)3187 static inline void netif_stop_queue(struct net_device *dev)
3188 {
3189 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3190 }
3191
3192 void netif_tx_stop_all_queues(struct net_device *dev);
3193 void netdev_update_lockdep_key(struct net_device *dev);
3194
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3195 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3196 {
3197 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3198 }
3199
3200 /**
3201 * netif_queue_stopped - test if transmit queue is flowblocked
3202 * @dev: network device
3203 *
3204 * Test if transmit queue on device is currently unable to send.
3205 */
netif_queue_stopped(const struct net_device * dev)3206 static inline bool netif_queue_stopped(const struct net_device *dev)
3207 {
3208 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3209 }
3210
netif_xmit_stopped(const struct netdev_queue * dev_queue)3211 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3212 {
3213 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3214 }
3215
3216 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3217 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3218 {
3219 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3220 }
3221
3222 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3223 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3224 {
3225 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3226 }
3227
3228 /**
3229 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3230 * @dev_queue: pointer to transmit queue
3231 *
3232 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3233 * to give appropriate hint to the CPU.
3234 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3235 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3236 {
3237 #ifdef CONFIG_BQL
3238 prefetchw(&dev_queue->dql.num_queued);
3239 #endif
3240 }
3241
3242 /**
3243 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3244 * @dev_queue: pointer to transmit queue
3245 *
3246 * BQL enabled drivers might use this helper in their TX completion path,
3247 * to give appropriate hint to the CPU.
3248 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3249 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3250 {
3251 #ifdef CONFIG_BQL
3252 prefetchw(&dev_queue->dql.limit);
3253 #endif
3254 }
3255
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3256 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3257 unsigned int bytes)
3258 {
3259 #ifdef CONFIG_BQL
3260 dql_queued(&dev_queue->dql, bytes);
3261
3262 if (likely(dql_avail(&dev_queue->dql) >= 0))
3263 return;
3264
3265 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3266
3267 /*
3268 * The XOFF flag must be set before checking the dql_avail below,
3269 * because in netdev_tx_completed_queue we update the dql_completed
3270 * before checking the XOFF flag.
3271 */
3272 smp_mb();
3273
3274 /* check again in case another CPU has just made room avail */
3275 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3276 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3277 #endif
3278 }
3279
3280 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3281 * that they should not test BQL status themselves.
3282 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3283 * skb of a batch.
3284 * Returns true if the doorbell must be used to kick the NIC.
3285 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3286 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3287 unsigned int bytes,
3288 bool xmit_more)
3289 {
3290 if (xmit_more) {
3291 #ifdef CONFIG_BQL
3292 dql_queued(&dev_queue->dql, bytes);
3293 #endif
3294 return netif_tx_queue_stopped(dev_queue);
3295 }
3296 netdev_tx_sent_queue(dev_queue, bytes);
3297 return true;
3298 }
3299
3300 /**
3301 * netdev_sent_queue - report the number of bytes queued to hardware
3302 * @dev: network device
3303 * @bytes: number of bytes queued to the hardware device queue
3304 *
3305 * Report the number of bytes queued for sending/completion to the network
3306 * device hardware queue. @bytes should be a good approximation and should
3307 * exactly match netdev_completed_queue() @bytes
3308 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3309 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3310 {
3311 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3312 }
3313
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3314 static inline bool __netdev_sent_queue(struct net_device *dev,
3315 unsigned int bytes,
3316 bool xmit_more)
3317 {
3318 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3319 xmit_more);
3320 }
3321
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3322 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3323 unsigned int pkts, unsigned int bytes)
3324 {
3325 #ifdef CONFIG_BQL
3326 if (unlikely(!bytes))
3327 return;
3328
3329 dql_completed(&dev_queue->dql, bytes);
3330
3331 /*
3332 * Without the memory barrier there is a small possiblity that
3333 * netdev_tx_sent_queue will miss the update and cause the queue to
3334 * be stopped forever
3335 */
3336 smp_mb();
3337
3338 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3339 return;
3340
3341 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3342 netif_schedule_queue(dev_queue);
3343 #endif
3344 }
3345
3346 /**
3347 * netdev_completed_queue - report bytes and packets completed by device
3348 * @dev: network device
3349 * @pkts: actual number of packets sent over the medium
3350 * @bytes: actual number of bytes sent over the medium
3351 *
3352 * Report the number of bytes and packets transmitted by the network device
3353 * hardware queue over the physical medium, @bytes must exactly match the
3354 * @bytes amount passed to netdev_sent_queue()
3355 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3356 static inline void netdev_completed_queue(struct net_device *dev,
3357 unsigned int pkts, unsigned int bytes)
3358 {
3359 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3360 }
3361
netdev_tx_reset_queue(struct netdev_queue * q)3362 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3363 {
3364 #ifdef CONFIG_BQL
3365 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3366 dql_reset(&q->dql);
3367 #endif
3368 }
3369
3370 /**
3371 * netdev_reset_queue - reset the packets and bytes count of a network device
3372 * @dev_queue: network device
3373 *
3374 * Reset the bytes and packet count of a network device and clear the
3375 * software flow control OFF bit for this network device
3376 */
netdev_reset_queue(struct net_device * dev_queue)3377 static inline void netdev_reset_queue(struct net_device *dev_queue)
3378 {
3379 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3380 }
3381
3382 /**
3383 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3384 * @dev: network device
3385 * @queue_index: given tx queue index
3386 *
3387 * Returns 0 if given tx queue index >= number of device tx queues,
3388 * otherwise returns the originally passed tx queue index.
3389 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3390 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3391 {
3392 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3393 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3394 dev->name, queue_index,
3395 dev->real_num_tx_queues);
3396 return 0;
3397 }
3398
3399 return queue_index;
3400 }
3401
3402 /**
3403 * netif_running - test if up
3404 * @dev: network device
3405 *
3406 * Test if the device has been brought up.
3407 */
netif_running(const struct net_device * dev)3408 static inline bool netif_running(const struct net_device *dev)
3409 {
3410 return test_bit(__LINK_STATE_START, &dev->state);
3411 }
3412
3413 /*
3414 * Routines to manage the subqueues on a device. We only need start,
3415 * stop, and a check if it's stopped. All other device management is
3416 * done at the overall netdevice level.
3417 * Also test the device if we're multiqueue.
3418 */
3419
3420 /**
3421 * netif_start_subqueue - allow sending packets on subqueue
3422 * @dev: network device
3423 * @queue_index: sub queue index
3424 *
3425 * Start individual transmit queue of a device with multiple transmit queues.
3426 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3427 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3428 {
3429 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3430
3431 netif_tx_start_queue(txq);
3432 }
3433
3434 /**
3435 * netif_stop_subqueue - stop sending packets on subqueue
3436 * @dev: network device
3437 * @queue_index: sub queue index
3438 *
3439 * Stop individual transmit queue of a device with multiple transmit queues.
3440 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3441 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3442 {
3443 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3444 netif_tx_stop_queue(txq);
3445 }
3446
3447 /**
3448 * netif_subqueue_stopped - test status of subqueue
3449 * @dev: network device
3450 * @queue_index: sub queue index
3451 *
3452 * Check individual transmit queue of a device with multiple transmit queues.
3453 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3454 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3455 u16 queue_index)
3456 {
3457 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3458
3459 return netif_tx_queue_stopped(txq);
3460 }
3461
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3462 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3463 struct sk_buff *skb)
3464 {
3465 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3466 }
3467
3468 /**
3469 * netif_wake_subqueue - allow sending packets on subqueue
3470 * @dev: network device
3471 * @queue_index: sub queue index
3472 *
3473 * Resume individual transmit queue of a device with multiple transmit queues.
3474 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3475 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3476 {
3477 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3478
3479 netif_tx_wake_queue(txq);
3480 }
3481
3482 #ifdef CONFIG_XPS
3483 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3484 u16 index);
3485 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3486 u16 index, bool is_rxqs_map);
3487
3488 /**
3489 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3490 * @j: CPU/Rx queue index
3491 * @mask: bitmask of all cpus/rx queues
3492 * @nr_bits: number of bits in the bitmask
3493 *
3494 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3495 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3496 static inline bool netif_attr_test_mask(unsigned long j,
3497 const unsigned long *mask,
3498 unsigned int nr_bits)
3499 {
3500 cpu_max_bits_warn(j, nr_bits);
3501 return test_bit(j, mask);
3502 }
3503
3504 /**
3505 * netif_attr_test_online - Test for online CPU/Rx queue
3506 * @j: CPU/Rx queue index
3507 * @online_mask: bitmask for CPUs/Rx queues that are online
3508 * @nr_bits: number of bits in the bitmask
3509 *
3510 * Returns true if a CPU/Rx queue is online.
3511 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3512 static inline bool netif_attr_test_online(unsigned long j,
3513 const unsigned long *online_mask,
3514 unsigned int nr_bits)
3515 {
3516 cpu_max_bits_warn(j, nr_bits);
3517
3518 if (online_mask)
3519 return test_bit(j, online_mask);
3520
3521 return (j < nr_bits);
3522 }
3523
3524 /**
3525 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3526 * @n: CPU/Rx queue index
3527 * @srcp: the cpumask/Rx queue mask pointer
3528 * @nr_bits: number of bits in the bitmask
3529 *
3530 * Returns >= nr_bits if no further CPUs/Rx queues set.
3531 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3532 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3533 unsigned int nr_bits)
3534 {
3535 /* -1 is a legal arg here. */
3536 if (n != -1)
3537 cpu_max_bits_warn(n, nr_bits);
3538
3539 if (srcp)
3540 return find_next_bit(srcp, nr_bits, n + 1);
3541
3542 return n + 1;
3543 }
3544
3545 /**
3546 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3547 * @n: CPU/Rx queue index
3548 * @src1p: the first CPUs/Rx queues mask pointer
3549 * @src2p: the second CPUs/Rx queues mask pointer
3550 * @nr_bits: number of bits in the bitmask
3551 *
3552 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3553 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3554 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3555 const unsigned long *src2p,
3556 unsigned int nr_bits)
3557 {
3558 /* -1 is a legal arg here. */
3559 if (n != -1)
3560 cpu_max_bits_warn(n, nr_bits);
3561
3562 if (src1p && src2p)
3563 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3564 else if (src1p)
3565 return find_next_bit(src1p, nr_bits, n + 1);
3566 else if (src2p)
3567 return find_next_bit(src2p, nr_bits, n + 1);
3568
3569 return n + 1;
3570 }
3571 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3572 static inline int netif_set_xps_queue(struct net_device *dev,
3573 const struct cpumask *mask,
3574 u16 index)
3575 {
3576 return 0;
3577 }
3578
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,bool is_rxqs_map)3579 static inline int __netif_set_xps_queue(struct net_device *dev,
3580 const unsigned long *mask,
3581 u16 index, bool is_rxqs_map)
3582 {
3583 return 0;
3584 }
3585 #endif
3586
3587 /**
3588 * netif_is_multiqueue - test if device has multiple transmit queues
3589 * @dev: network device
3590 *
3591 * Check if device has multiple transmit queues
3592 */
netif_is_multiqueue(const struct net_device * dev)3593 static inline bool netif_is_multiqueue(const struct net_device *dev)
3594 {
3595 return dev->num_tx_queues > 1;
3596 }
3597
3598 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3599
3600 #ifdef CONFIG_SYSFS
3601 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3602 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3603 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3604 unsigned int rxqs)
3605 {
3606 dev->real_num_rx_queues = rxqs;
3607 return 0;
3608 }
3609 #endif
3610
3611 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3612 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3613 {
3614 return dev->_rx + rxq;
3615 }
3616
3617 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3618 static inline unsigned int get_netdev_rx_queue_index(
3619 struct netdev_rx_queue *queue)
3620 {
3621 struct net_device *dev = queue->dev;
3622 int index = queue - dev->_rx;
3623
3624 BUG_ON(index >= dev->num_rx_queues);
3625 return index;
3626 }
3627 #endif
3628
3629 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3630 int netif_get_num_default_rss_queues(void);
3631
3632 enum skb_free_reason {
3633 SKB_REASON_CONSUMED,
3634 SKB_REASON_DROPPED,
3635 };
3636
3637 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3638 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3639
3640 /*
3641 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3642 * interrupt context or with hardware interrupts being disabled.
3643 * (in_irq() || irqs_disabled())
3644 *
3645 * We provide four helpers that can be used in following contexts :
3646 *
3647 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3648 * replacing kfree_skb(skb)
3649 *
3650 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3651 * Typically used in place of consume_skb(skb) in TX completion path
3652 *
3653 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3654 * replacing kfree_skb(skb)
3655 *
3656 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3657 * and consumed a packet. Used in place of consume_skb(skb)
3658 */
dev_kfree_skb_irq(struct sk_buff * skb)3659 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3660 {
3661 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3662 }
3663
dev_consume_skb_irq(struct sk_buff * skb)3664 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3665 {
3666 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3667 }
3668
dev_kfree_skb_any(struct sk_buff * skb)3669 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3670 {
3671 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3672 }
3673
dev_consume_skb_any(struct sk_buff * skb)3674 static inline void dev_consume_skb_any(struct sk_buff *skb)
3675 {
3676 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3677 }
3678
3679 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3680 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3681 int netif_rx(struct sk_buff *skb);
3682 int netif_rx_ni(struct sk_buff *skb);
3683 int netif_receive_skb(struct sk_buff *skb);
3684 int netif_receive_skb_core(struct sk_buff *skb);
3685 void netif_receive_skb_list(struct list_head *head);
3686 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3687 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3688 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3689 gro_result_t napi_gro_frags(struct napi_struct *napi);
3690 struct packet_offload *gro_find_receive_by_type(__be16 type);
3691 struct packet_offload *gro_find_complete_by_type(__be16 type);
3692
napi_free_frags(struct napi_struct * napi)3693 static inline void napi_free_frags(struct napi_struct *napi)
3694 {
3695 kfree_skb(napi->skb);
3696 napi->skb = NULL;
3697 }
3698
3699 bool netdev_is_rx_handler_busy(struct net_device *dev);
3700 int netdev_rx_handler_register(struct net_device *dev,
3701 rx_handler_func_t *rx_handler,
3702 void *rx_handler_data);
3703 void netdev_rx_handler_unregister(struct net_device *dev);
3704
3705 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)3706 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3707 {
3708 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3709 }
3710 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3711 bool *need_copyout);
3712 int dev_ifconf(struct net *net, struct ifconf *, int);
3713 int dev_ethtool(struct net *net, struct ifreq *);
3714 unsigned int dev_get_flags(const struct net_device *);
3715 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3716 struct netlink_ext_ack *extack);
3717 int dev_change_flags(struct net_device *dev, unsigned int flags,
3718 struct netlink_ext_ack *extack);
3719 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3720 unsigned int gchanges);
3721 int dev_change_name(struct net_device *, const char *);
3722 int dev_set_alias(struct net_device *, const char *, size_t);
3723 int dev_get_alias(const struct net_device *, char *, size_t);
3724 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3725 int __dev_set_mtu(struct net_device *, int);
3726 int dev_validate_mtu(struct net_device *dev, int mtu,
3727 struct netlink_ext_ack *extack);
3728 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3729 struct netlink_ext_ack *extack);
3730 int dev_set_mtu(struct net_device *, int);
3731 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3732 void dev_set_group(struct net_device *, int);
3733 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3734 struct netlink_ext_ack *extack);
3735 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3736 struct netlink_ext_ack *extack);
3737 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3738 struct netlink_ext_ack *extack);
3739 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3740 int dev_change_carrier(struct net_device *, bool new_carrier);
3741 int dev_get_phys_port_id(struct net_device *dev,
3742 struct netdev_phys_item_id *ppid);
3743 int dev_get_phys_port_name(struct net_device *dev,
3744 char *name, size_t len);
3745 int dev_get_port_parent_id(struct net_device *dev,
3746 struct netdev_phys_item_id *ppid, bool recurse);
3747 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3748 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3749 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3750 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3751 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3752 struct netdev_queue *txq, int *ret);
3753
3754 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3755 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3756 int fd, u32 flags);
3757 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3758 enum bpf_netdev_command cmd);
3759 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3760
3761 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3762 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3763 bool is_skb_forwardable(const struct net_device *dev,
3764 const struct sk_buff *skb);
3765
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb)3766 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3767 struct sk_buff *skb)
3768 {
3769 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3770 unlikely(!is_skb_forwardable(dev, skb))) {
3771 atomic_long_inc(&dev->rx_dropped);
3772 kfree_skb(skb);
3773 return NET_RX_DROP;
3774 }
3775
3776 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3777 skb->priority = 0;
3778 return 0;
3779 }
3780
3781 bool dev_nit_active(struct net_device *dev);
3782 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3783
3784 extern int netdev_budget;
3785 extern unsigned int netdev_budget_usecs;
3786
3787 /* Called by rtnetlink.c:rtnl_unlock() */
3788 void netdev_run_todo(void);
3789
3790 /**
3791 * dev_put - release reference to device
3792 * @dev: network device
3793 *
3794 * Release reference to device to allow it to be freed.
3795 */
dev_put(struct net_device * dev)3796 static inline void dev_put(struct net_device *dev)
3797 {
3798 if (dev)
3799 this_cpu_dec(*dev->pcpu_refcnt);
3800 }
3801
3802 /**
3803 * dev_hold - get reference to device
3804 * @dev: network device
3805 *
3806 * Hold reference to device to keep it from being freed.
3807 */
dev_hold(struct net_device * dev)3808 static inline void dev_hold(struct net_device *dev)
3809 {
3810 if (dev)
3811 this_cpu_inc(*dev->pcpu_refcnt);
3812 }
3813
3814 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3815 * and _off may be called from IRQ context, but it is caller
3816 * who is responsible for serialization of these calls.
3817 *
3818 * The name carrier is inappropriate, these functions should really be
3819 * called netif_lowerlayer_*() because they represent the state of any
3820 * kind of lower layer not just hardware media.
3821 */
3822
3823 void linkwatch_init_dev(struct net_device *dev);
3824 void linkwatch_fire_event(struct net_device *dev);
3825 void linkwatch_forget_dev(struct net_device *dev);
3826
3827 /**
3828 * netif_carrier_ok - test if carrier present
3829 * @dev: network device
3830 *
3831 * Check if carrier is present on device
3832 */
netif_carrier_ok(const struct net_device * dev)3833 static inline bool netif_carrier_ok(const struct net_device *dev)
3834 {
3835 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3836 }
3837
3838 unsigned long dev_trans_start(struct net_device *dev);
3839
3840 void __netdev_watchdog_up(struct net_device *dev);
3841
3842 void netif_carrier_on(struct net_device *dev);
3843
3844 void netif_carrier_off(struct net_device *dev);
3845
3846 /**
3847 * netif_dormant_on - mark device as dormant.
3848 * @dev: network device
3849 *
3850 * Mark device as dormant (as per RFC2863).
3851 *
3852 * The dormant state indicates that the relevant interface is not
3853 * actually in a condition to pass packets (i.e., it is not 'up') but is
3854 * in a "pending" state, waiting for some external event. For "on-
3855 * demand" interfaces, this new state identifies the situation where the
3856 * interface is waiting for events to place it in the up state.
3857 */
netif_dormant_on(struct net_device * dev)3858 static inline void netif_dormant_on(struct net_device *dev)
3859 {
3860 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3861 linkwatch_fire_event(dev);
3862 }
3863
3864 /**
3865 * netif_dormant_off - set device as not dormant.
3866 * @dev: network device
3867 *
3868 * Device is not in dormant state.
3869 */
netif_dormant_off(struct net_device * dev)3870 static inline void netif_dormant_off(struct net_device *dev)
3871 {
3872 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3873 linkwatch_fire_event(dev);
3874 }
3875
3876 /**
3877 * netif_dormant - test if device is dormant
3878 * @dev: network device
3879 *
3880 * Check if device is dormant.
3881 */
netif_dormant(const struct net_device * dev)3882 static inline bool netif_dormant(const struct net_device *dev)
3883 {
3884 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3885 }
3886
3887
3888 /**
3889 * netif_oper_up - test if device is operational
3890 * @dev: network device
3891 *
3892 * Check if carrier is operational
3893 */
netif_oper_up(const struct net_device * dev)3894 static inline bool netif_oper_up(const struct net_device *dev)
3895 {
3896 return (dev->operstate == IF_OPER_UP ||
3897 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3898 }
3899
3900 /**
3901 * netif_device_present - is device available or removed
3902 * @dev: network device
3903 *
3904 * Check if device has not been removed from system.
3905 */
netif_device_present(struct net_device * dev)3906 static inline bool netif_device_present(struct net_device *dev)
3907 {
3908 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3909 }
3910
3911 void netif_device_detach(struct net_device *dev);
3912
3913 void netif_device_attach(struct net_device *dev);
3914
3915 /*
3916 * Network interface message level settings
3917 */
3918
3919 enum {
3920 NETIF_MSG_DRV = 0x0001,
3921 NETIF_MSG_PROBE = 0x0002,
3922 NETIF_MSG_LINK = 0x0004,
3923 NETIF_MSG_TIMER = 0x0008,
3924 NETIF_MSG_IFDOWN = 0x0010,
3925 NETIF_MSG_IFUP = 0x0020,
3926 NETIF_MSG_RX_ERR = 0x0040,
3927 NETIF_MSG_TX_ERR = 0x0080,
3928 NETIF_MSG_TX_QUEUED = 0x0100,
3929 NETIF_MSG_INTR = 0x0200,
3930 NETIF_MSG_TX_DONE = 0x0400,
3931 NETIF_MSG_RX_STATUS = 0x0800,
3932 NETIF_MSG_PKTDATA = 0x1000,
3933 NETIF_MSG_HW = 0x2000,
3934 NETIF_MSG_WOL = 0x4000,
3935 };
3936
3937 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3938 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3939 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3940 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3941 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3942 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3943 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3944 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3945 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3946 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3947 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3948 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3949 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3950 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3951 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3952
netif_msg_init(int debug_value,int default_msg_enable_bits)3953 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3954 {
3955 /* use default */
3956 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3957 return default_msg_enable_bits;
3958 if (debug_value == 0) /* no output */
3959 return 0;
3960 /* set low N bits */
3961 return (1U << debug_value) - 1;
3962 }
3963
__netif_tx_lock(struct netdev_queue * txq,int cpu)3964 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3965 {
3966 spin_lock(&txq->_xmit_lock);
3967 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
3968 WRITE_ONCE(txq->xmit_lock_owner, cpu);
3969 }
3970
__netif_tx_acquire(struct netdev_queue * txq)3971 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3972 {
3973 __acquire(&txq->_xmit_lock);
3974 return true;
3975 }
3976
__netif_tx_release(struct netdev_queue * txq)3977 static inline void __netif_tx_release(struct netdev_queue *txq)
3978 {
3979 __release(&txq->_xmit_lock);
3980 }
3981
__netif_tx_lock_bh(struct netdev_queue * txq)3982 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3983 {
3984 spin_lock_bh(&txq->_xmit_lock);
3985 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
3986 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
3987 }
3988
__netif_tx_trylock(struct netdev_queue * txq)3989 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3990 {
3991 bool ok = spin_trylock(&txq->_xmit_lock);
3992
3993 if (likely(ok)) {
3994 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
3995 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
3996 }
3997 return ok;
3998 }
3999
__netif_tx_unlock(struct netdev_queue * txq)4000 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4001 {
4002 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4003 WRITE_ONCE(txq->xmit_lock_owner, -1);
4004 spin_unlock(&txq->_xmit_lock);
4005 }
4006
__netif_tx_unlock_bh(struct netdev_queue * txq)4007 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4008 {
4009 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4010 WRITE_ONCE(txq->xmit_lock_owner, -1);
4011 spin_unlock_bh(&txq->_xmit_lock);
4012 }
4013
txq_trans_update(struct netdev_queue * txq)4014 static inline void txq_trans_update(struct netdev_queue *txq)
4015 {
4016 if (txq->xmit_lock_owner != -1)
4017 txq->trans_start = jiffies;
4018 }
4019
4020 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4021 static inline void netif_trans_update(struct net_device *dev)
4022 {
4023 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4024
4025 if (txq->trans_start != jiffies)
4026 txq->trans_start = jiffies;
4027 }
4028
4029 /**
4030 * netif_tx_lock - grab network device transmit lock
4031 * @dev: network device
4032 *
4033 * Get network device transmit lock
4034 */
netif_tx_lock(struct net_device * dev)4035 static inline void netif_tx_lock(struct net_device *dev)
4036 {
4037 unsigned int i;
4038 int cpu;
4039
4040 spin_lock(&dev->tx_global_lock);
4041 cpu = smp_processor_id();
4042 for (i = 0; i < dev->num_tx_queues; i++) {
4043 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4044
4045 /* We are the only thread of execution doing a
4046 * freeze, but we have to grab the _xmit_lock in
4047 * order to synchronize with threads which are in
4048 * the ->hard_start_xmit() handler and already
4049 * checked the frozen bit.
4050 */
4051 __netif_tx_lock(txq, cpu);
4052 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4053 __netif_tx_unlock(txq);
4054 }
4055 }
4056
netif_tx_lock_bh(struct net_device * dev)4057 static inline void netif_tx_lock_bh(struct net_device *dev)
4058 {
4059 local_bh_disable();
4060 netif_tx_lock(dev);
4061 }
4062
netif_tx_unlock(struct net_device * dev)4063 static inline void netif_tx_unlock(struct net_device *dev)
4064 {
4065 unsigned int i;
4066
4067 for (i = 0; i < dev->num_tx_queues; i++) {
4068 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4069
4070 /* No need to grab the _xmit_lock here. If the
4071 * queue is not stopped for another reason, we
4072 * force a schedule.
4073 */
4074 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4075 netif_schedule_queue(txq);
4076 }
4077 spin_unlock(&dev->tx_global_lock);
4078 }
4079
netif_tx_unlock_bh(struct net_device * dev)4080 static inline void netif_tx_unlock_bh(struct net_device *dev)
4081 {
4082 netif_tx_unlock(dev);
4083 local_bh_enable();
4084 }
4085
4086 #define HARD_TX_LOCK(dev, txq, cpu) { \
4087 if ((dev->features & NETIF_F_LLTX) == 0) { \
4088 __netif_tx_lock(txq, cpu); \
4089 } else { \
4090 __netif_tx_acquire(txq); \
4091 } \
4092 }
4093
4094 #define HARD_TX_TRYLOCK(dev, txq) \
4095 (((dev->features & NETIF_F_LLTX) == 0) ? \
4096 __netif_tx_trylock(txq) : \
4097 __netif_tx_acquire(txq))
4098
4099 #define HARD_TX_UNLOCK(dev, txq) { \
4100 if ((dev->features & NETIF_F_LLTX) == 0) { \
4101 __netif_tx_unlock(txq); \
4102 } else { \
4103 __netif_tx_release(txq); \
4104 } \
4105 }
4106
netif_tx_disable(struct net_device * dev)4107 static inline void netif_tx_disable(struct net_device *dev)
4108 {
4109 unsigned int i;
4110 int cpu;
4111
4112 local_bh_disable();
4113 cpu = smp_processor_id();
4114 spin_lock(&dev->tx_global_lock);
4115 for (i = 0; i < dev->num_tx_queues; i++) {
4116 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4117
4118 __netif_tx_lock(txq, cpu);
4119 netif_tx_stop_queue(txq);
4120 __netif_tx_unlock(txq);
4121 }
4122 spin_unlock(&dev->tx_global_lock);
4123 local_bh_enable();
4124 }
4125
netif_addr_lock(struct net_device * dev)4126 static inline void netif_addr_lock(struct net_device *dev)
4127 {
4128 spin_lock(&dev->addr_list_lock);
4129 }
4130
netif_addr_lock_bh(struct net_device * dev)4131 static inline void netif_addr_lock_bh(struct net_device *dev)
4132 {
4133 spin_lock_bh(&dev->addr_list_lock);
4134 }
4135
netif_addr_unlock(struct net_device * dev)4136 static inline void netif_addr_unlock(struct net_device *dev)
4137 {
4138 spin_unlock(&dev->addr_list_lock);
4139 }
4140
netif_addr_unlock_bh(struct net_device * dev)4141 static inline void netif_addr_unlock_bh(struct net_device *dev)
4142 {
4143 spin_unlock_bh(&dev->addr_list_lock);
4144 }
4145
4146 /*
4147 * dev_addrs walker. Should be used only for read access. Call with
4148 * rcu_read_lock held.
4149 */
4150 #define for_each_dev_addr(dev, ha) \
4151 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4152
4153 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4154
4155 void ether_setup(struct net_device *dev);
4156
4157 /* Support for loadable net-drivers */
4158 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4159 unsigned char name_assign_type,
4160 void (*setup)(struct net_device *),
4161 unsigned int txqs, unsigned int rxqs);
4162 int dev_get_valid_name(struct net *net, struct net_device *dev,
4163 const char *name);
4164
4165 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4166 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4167
4168 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4169 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4170 count)
4171
4172 int register_netdev(struct net_device *dev);
4173 void unregister_netdev(struct net_device *dev);
4174
4175 /* General hardware address lists handling functions */
4176 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4177 struct netdev_hw_addr_list *from_list, int addr_len);
4178 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4179 struct netdev_hw_addr_list *from_list, int addr_len);
4180 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4181 struct net_device *dev,
4182 int (*sync)(struct net_device *, const unsigned char *),
4183 int (*unsync)(struct net_device *,
4184 const unsigned char *));
4185 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4186 struct net_device *dev,
4187 int (*sync)(struct net_device *,
4188 const unsigned char *, int),
4189 int (*unsync)(struct net_device *,
4190 const unsigned char *, int));
4191 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4192 struct net_device *dev,
4193 int (*unsync)(struct net_device *,
4194 const unsigned char *, int));
4195 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4196 struct net_device *dev,
4197 int (*unsync)(struct net_device *,
4198 const unsigned char *));
4199 void __hw_addr_init(struct netdev_hw_addr_list *list);
4200
4201 /* Functions used for device addresses handling */
4202 static inline void
__dev_addr_set(struct net_device * dev,const u8 * addr,size_t len)4203 __dev_addr_set(struct net_device *dev, const u8 *addr, size_t len)
4204 {
4205 memcpy(dev->dev_addr, addr, len);
4206 }
4207
dev_addr_set(struct net_device * dev,const u8 * addr)4208 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4209 {
4210 __dev_addr_set(dev, addr, dev->addr_len);
4211 }
4212
4213 static inline void
dev_addr_mod(struct net_device * dev,unsigned int offset,const u8 * addr,size_t len)4214 dev_addr_mod(struct net_device *dev, unsigned int offset,
4215 const u8 *addr, size_t len)
4216 {
4217 memcpy(&dev->dev_addr[offset], addr, len);
4218 }
4219
4220 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4221 unsigned char addr_type);
4222 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4223 unsigned char addr_type);
4224 void dev_addr_flush(struct net_device *dev);
4225 int dev_addr_init(struct net_device *dev);
4226
4227 /* Functions used for unicast addresses handling */
4228 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4229 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4230 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4231 int dev_uc_sync(struct net_device *to, struct net_device *from);
4232 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4233 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4234 void dev_uc_flush(struct net_device *dev);
4235 void dev_uc_init(struct net_device *dev);
4236
4237 /**
4238 * __dev_uc_sync - Synchonize device's unicast list
4239 * @dev: device to sync
4240 * @sync: function to call if address should be added
4241 * @unsync: function to call if address should be removed
4242 *
4243 * Add newly added addresses to the interface, and release
4244 * addresses that have been deleted.
4245 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4246 static inline int __dev_uc_sync(struct net_device *dev,
4247 int (*sync)(struct net_device *,
4248 const unsigned char *),
4249 int (*unsync)(struct net_device *,
4250 const unsigned char *))
4251 {
4252 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4253 }
4254
4255 /**
4256 * __dev_uc_unsync - Remove synchronized addresses from device
4257 * @dev: device to sync
4258 * @unsync: function to call if address should be removed
4259 *
4260 * Remove all addresses that were added to the device by dev_uc_sync().
4261 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4262 static inline void __dev_uc_unsync(struct net_device *dev,
4263 int (*unsync)(struct net_device *,
4264 const unsigned char *))
4265 {
4266 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4267 }
4268
4269 /* Functions used for multicast addresses handling */
4270 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4271 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4272 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4273 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4274 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4275 int dev_mc_sync(struct net_device *to, struct net_device *from);
4276 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4277 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4278 void dev_mc_flush(struct net_device *dev);
4279 void dev_mc_init(struct net_device *dev);
4280
4281 /**
4282 * __dev_mc_sync - Synchonize device's multicast list
4283 * @dev: device to sync
4284 * @sync: function to call if address should be added
4285 * @unsync: function to call if address should be removed
4286 *
4287 * Add newly added addresses to the interface, and release
4288 * addresses that have been deleted.
4289 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4290 static inline int __dev_mc_sync(struct net_device *dev,
4291 int (*sync)(struct net_device *,
4292 const unsigned char *),
4293 int (*unsync)(struct net_device *,
4294 const unsigned char *))
4295 {
4296 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4297 }
4298
4299 /**
4300 * __dev_mc_unsync - Remove synchronized addresses from device
4301 * @dev: device to sync
4302 * @unsync: function to call if address should be removed
4303 *
4304 * Remove all addresses that were added to the device by dev_mc_sync().
4305 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4306 static inline void __dev_mc_unsync(struct net_device *dev,
4307 int (*unsync)(struct net_device *,
4308 const unsigned char *))
4309 {
4310 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4311 }
4312
4313 /* Functions used for secondary unicast and multicast support */
4314 void dev_set_rx_mode(struct net_device *dev);
4315 void __dev_set_rx_mode(struct net_device *dev);
4316 int dev_set_promiscuity(struct net_device *dev, int inc);
4317 int dev_set_allmulti(struct net_device *dev, int inc);
4318 void netdev_state_change(struct net_device *dev);
4319 void netdev_notify_peers(struct net_device *dev);
4320 void netdev_features_change(struct net_device *dev);
4321 /* Load a device via the kmod */
4322 void dev_load(struct net *net, const char *name);
4323 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4324 struct rtnl_link_stats64 *storage);
4325 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4326 const struct net_device_stats *netdev_stats);
4327
4328 extern int netdev_max_backlog;
4329 extern int netdev_tstamp_prequeue;
4330 extern int weight_p;
4331 extern int dev_weight_rx_bias;
4332 extern int dev_weight_tx_bias;
4333 extern int dev_rx_weight;
4334 extern int dev_tx_weight;
4335 extern int gro_normal_batch;
4336
4337 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4338 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4339 struct list_head **iter);
4340 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4341 struct list_head **iter);
4342
4343 /* iterate through upper list, must be called under RCU read lock */
4344 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4345 for (iter = &(dev)->adj_list.upper, \
4346 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4347 updev; \
4348 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4349
4350 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4351 int (*fn)(struct net_device *upper_dev,
4352 void *data),
4353 void *data);
4354
4355 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4356 struct net_device *upper_dev);
4357
4358 bool netdev_has_any_upper_dev(struct net_device *dev);
4359
4360 void *netdev_lower_get_next_private(struct net_device *dev,
4361 struct list_head **iter);
4362 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4363 struct list_head **iter);
4364
4365 #define netdev_for_each_lower_private(dev, priv, iter) \
4366 for (iter = (dev)->adj_list.lower.next, \
4367 priv = netdev_lower_get_next_private(dev, &(iter)); \
4368 priv; \
4369 priv = netdev_lower_get_next_private(dev, &(iter)))
4370
4371 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4372 for (iter = &(dev)->adj_list.lower, \
4373 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4374 priv; \
4375 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4376
4377 void *netdev_lower_get_next(struct net_device *dev,
4378 struct list_head **iter);
4379
4380 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4381 for (iter = (dev)->adj_list.lower.next, \
4382 ldev = netdev_lower_get_next(dev, &(iter)); \
4383 ldev; \
4384 ldev = netdev_lower_get_next(dev, &(iter)))
4385
4386 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4387 struct list_head **iter);
4388 int netdev_walk_all_lower_dev(struct net_device *dev,
4389 int (*fn)(struct net_device *lower_dev,
4390 void *data),
4391 void *data);
4392 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4393 int (*fn)(struct net_device *lower_dev,
4394 void *data),
4395 void *data);
4396
4397 void *netdev_adjacent_get_private(struct list_head *adj_list);
4398 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4399 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4400 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4401 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4402 struct netlink_ext_ack *extack);
4403 int netdev_master_upper_dev_link(struct net_device *dev,
4404 struct net_device *upper_dev,
4405 void *upper_priv, void *upper_info,
4406 struct netlink_ext_ack *extack);
4407 void netdev_upper_dev_unlink(struct net_device *dev,
4408 struct net_device *upper_dev);
4409 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4410 struct net_device *new_dev,
4411 struct net_device *dev,
4412 struct netlink_ext_ack *extack);
4413 void netdev_adjacent_change_commit(struct net_device *old_dev,
4414 struct net_device *new_dev,
4415 struct net_device *dev);
4416 void netdev_adjacent_change_abort(struct net_device *old_dev,
4417 struct net_device *new_dev,
4418 struct net_device *dev);
4419 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4420 void *netdev_lower_dev_get_private(struct net_device *dev,
4421 struct net_device *lower_dev);
4422 void netdev_lower_state_changed(struct net_device *lower_dev,
4423 void *lower_state_info);
4424
4425 /* RSS keys are 40 or 52 bytes long */
4426 #define NETDEV_RSS_KEY_LEN 52
4427 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4428 void netdev_rss_key_fill(void *buffer, size_t len);
4429
4430 int skb_checksum_help(struct sk_buff *skb);
4431 int skb_crc32c_csum_help(struct sk_buff *skb);
4432 int skb_csum_hwoffload_help(struct sk_buff *skb,
4433 const netdev_features_t features);
4434
4435 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4436 netdev_features_t features, bool tx_path);
4437 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4438 netdev_features_t features);
4439
4440 struct netdev_bonding_info {
4441 ifslave slave;
4442 ifbond master;
4443 };
4444
4445 struct netdev_notifier_bonding_info {
4446 struct netdev_notifier_info info; /* must be first */
4447 struct netdev_bonding_info bonding_info;
4448 };
4449
4450 void netdev_bonding_info_change(struct net_device *dev,
4451 struct netdev_bonding_info *bonding_info);
4452
4453 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4454 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4455 {
4456 return __skb_gso_segment(skb, features, true);
4457 }
4458 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4459
can_checksum_protocol(netdev_features_t features,__be16 protocol)4460 static inline bool can_checksum_protocol(netdev_features_t features,
4461 __be16 protocol)
4462 {
4463 if (protocol == htons(ETH_P_FCOE))
4464 return !!(features & NETIF_F_FCOE_CRC);
4465
4466 /* Assume this is an IP checksum (not SCTP CRC) */
4467
4468 if (features & NETIF_F_HW_CSUM) {
4469 /* Can checksum everything */
4470 return true;
4471 }
4472
4473 switch (protocol) {
4474 case htons(ETH_P_IP):
4475 return !!(features & NETIF_F_IP_CSUM);
4476 case htons(ETH_P_IPV6):
4477 return !!(features & NETIF_F_IPV6_CSUM);
4478 default:
4479 return false;
4480 }
4481 }
4482
4483 #ifdef CONFIG_BUG
4484 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4485 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4486 static inline void netdev_rx_csum_fault(struct net_device *dev,
4487 struct sk_buff *skb)
4488 {
4489 }
4490 #endif
4491 /* rx skb timestamps */
4492 void net_enable_timestamp(void);
4493 void net_disable_timestamp(void);
4494
4495 #ifdef CONFIG_PROC_FS
4496 int __init dev_proc_init(void);
4497 #else
4498 #define dev_proc_init() 0
4499 #endif
4500
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4501 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4502 struct sk_buff *skb, struct net_device *dev,
4503 bool more)
4504 {
4505 __this_cpu_write(softnet_data.xmit.more, more);
4506 return ops->ndo_start_xmit(skb, dev);
4507 }
4508
netdev_xmit_more(void)4509 static inline bool netdev_xmit_more(void)
4510 {
4511 return __this_cpu_read(softnet_data.xmit.more);
4512 }
4513
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4514 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4515 struct netdev_queue *txq, bool more)
4516 {
4517 const struct net_device_ops *ops = dev->netdev_ops;
4518 netdev_tx_t rc;
4519
4520 rc = __netdev_start_xmit(ops, skb, dev, more);
4521 if (rc == NETDEV_TX_OK)
4522 txq_trans_update(txq);
4523
4524 return rc;
4525 }
4526
4527 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4528 const void *ns);
4529 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4530 const void *ns);
4531
netdev_class_create_file(const struct class_attribute * class_attr)4532 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4533 {
4534 return netdev_class_create_file_ns(class_attr, NULL);
4535 }
4536
netdev_class_remove_file(const struct class_attribute * class_attr)4537 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4538 {
4539 netdev_class_remove_file_ns(class_attr, NULL);
4540 }
4541
4542 extern const struct kobj_ns_type_operations net_ns_type_operations;
4543
4544 const char *netdev_drivername(const struct net_device *dev);
4545
4546 void linkwatch_run_queue(void);
4547
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)4548 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4549 netdev_features_t f2)
4550 {
4551 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4552 if (f1 & NETIF_F_HW_CSUM)
4553 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4554 else
4555 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4556 }
4557
4558 return f1 & f2;
4559 }
4560
netdev_get_wanted_features(struct net_device * dev)4561 static inline netdev_features_t netdev_get_wanted_features(
4562 struct net_device *dev)
4563 {
4564 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4565 }
4566 netdev_features_t netdev_increment_features(netdev_features_t all,
4567 netdev_features_t one, netdev_features_t mask);
4568
4569 /* Allow TSO being used on stacked device :
4570 * Performing the GSO segmentation before last device
4571 * is a performance improvement.
4572 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)4573 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4574 netdev_features_t mask)
4575 {
4576 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4577 }
4578
4579 int __netdev_update_features(struct net_device *dev);
4580 void netdev_update_features(struct net_device *dev);
4581 void netdev_change_features(struct net_device *dev);
4582
4583 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4584 struct net_device *dev);
4585
4586 netdev_features_t passthru_features_check(struct sk_buff *skb,
4587 struct net_device *dev,
4588 netdev_features_t features);
4589 netdev_features_t netif_skb_features(struct sk_buff *skb);
4590
net_gso_ok(netdev_features_t features,int gso_type)4591 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4592 {
4593 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4594
4595 /* check flags correspondence */
4596 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4597 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4598 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4599 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4600 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4601 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4602 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4603 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4604 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4605 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4606 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4607 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4608 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4609 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4610 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4611 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4612 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4613 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4614 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4615
4616 return (features & feature) == feature;
4617 }
4618
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)4619 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4620 {
4621 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4622 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4623 }
4624
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)4625 static inline bool netif_needs_gso(struct sk_buff *skb,
4626 netdev_features_t features)
4627 {
4628 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4629 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4630 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4631 }
4632
netif_set_gso_max_size(struct net_device * dev,unsigned int size)4633 static inline void netif_set_gso_max_size(struct net_device *dev,
4634 unsigned int size)
4635 {
4636 dev->gso_max_size = size;
4637 }
4638
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)4639 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4640 int pulled_hlen, u16 mac_offset,
4641 int mac_len)
4642 {
4643 skb->protocol = protocol;
4644 skb->encapsulation = 1;
4645 skb_push(skb, pulled_hlen);
4646 skb_reset_transport_header(skb);
4647 skb->mac_header = mac_offset;
4648 skb->network_header = skb->mac_header + mac_len;
4649 skb->mac_len = mac_len;
4650 }
4651
netif_is_macsec(const struct net_device * dev)4652 static inline bool netif_is_macsec(const struct net_device *dev)
4653 {
4654 return dev->priv_flags & IFF_MACSEC;
4655 }
4656
netif_is_macvlan(const struct net_device * dev)4657 static inline bool netif_is_macvlan(const struct net_device *dev)
4658 {
4659 return dev->priv_flags & IFF_MACVLAN;
4660 }
4661
netif_is_macvlan_port(const struct net_device * dev)4662 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4663 {
4664 return dev->priv_flags & IFF_MACVLAN_PORT;
4665 }
4666
netif_is_bond_master(const struct net_device * dev)4667 static inline bool netif_is_bond_master(const struct net_device *dev)
4668 {
4669 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4670 }
4671
netif_is_bond_slave(const struct net_device * dev)4672 static inline bool netif_is_bond_slave(const struct net_device *dev)
4673 {
4674 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4675 }
4676
netif_supports_nofcs(struct net_device * dev)4677 static inline bool netif_supports_nofcs(struct net_device *dev)
4678 {
4679 return dev->priv_flags & IFF_SUPP_NOFCS;
4680 }
4681
netif_has_l3_rx_handler(const struct net_device * dev)4682 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4683 {
4684 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4685 }
4686
netif_is_l3_master(const struct net_device * dev)4687 static inline bool netif_is_l3_master(const struct net_device *dev)
4688 {
4689 return dev->priv_flags & IFF_L3MDEV_MASTER;
4690 }
4691
netif_is_l3_slave(const struct net_device * dev)4692 static inline bool netif_is_l3_slave(const struct net_device *dev)
4693 {
4694 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4695 }
4696
netif_is_bridge_master(const struct net_device * dev)4697 static inline bool netif_is_bridge_master(const struct net_device *dev)
4698 {
4699 return dev->priv_flags & IFF_EBRIDGE;
4700 }
4701
netif_is_bridge_port(const struct net_device * dev)4702 static inline bool netif_is_bridge_port(const struct net_device *dev)
4703 {
4704 return dev->priv_flags & IFF_BRIDGE_PORT;
4705 }
4706
netif_is_ovs_master(const struct net_device * dev)4707 static inline bool netif_is_ovs_master(const struct net_device *dev)
4708 {
4709 return dev->priv_flags & IFF_OPENVSWITCH;
4710 }
4711
netif_is_ovs_port(const struct net_device * dev)4712 static inline bool netif_is_ovs_port(const struct net_device *dev)
4713 {
4714 return dev->priv_flags & IFF_OVS_DATAPATH;
4715 }
4716
netif_is_team_master(const struct net_device * dev)4717 static inline bool netif_is_team_master(const struct net_device *dev)
4718 {
4719 return dev->priv_flags & IFF_TEAM;
4720 }
4721
netif_is_team_port(const struct net_device * dev)4722 static inline bool netif_is_team_port(const struct net_device *dev)
4723 {
4724 return dev->priv_flags & IFF_TEAM_PORT;
4725 }
4726
netif_is_lag_master(const struct net_device * dev)4727 static inline bool netif_is_lag_master(const struct net_device *dev)
4728 {
4729 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4730 }
4731
netif_is_lag_port(const struct net_device * dev)4732 static inline bool netif_is_lag_port(const struct net_device *dev)
4733 {
4734 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4735 }
4736
netif_is_rxfh_configured(const struct net_device * dev)4737 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4738 {
4739 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4740 }
4741
netif_is_failover(const struct net_device * dev)4742 static inline bool netif_is_failover(const struct net_device *dev)
4743 {
4744 return dev->priv_flags & IFF_FAILOVER;
4745 }
4746
netif_is_failover_slave(const struct net_device * dev)4747 static inline bool netif_is_failover_slave(const struct net_device *dev)
4748 {
4749 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4750 }
4751
4752 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)4753 static inline void netif_keep_dst(struct net_device *dev)
4754 {
4755 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4756 }
4757
4758 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)4759 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4760 {
4761 /* TODO: reserve and use an additional IFF bit, if we get more users */
4762 return dev->priv_flags & IFF_MACSEC;
4763 }
4764
4765 extern struct pernet_operations __net_initdata loopback_net_ops;
4766
4767 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4768
4769 /* netdev_printk helpers, similar to dev_printk */
4770
netdev_name(const struct net_device * dev)4771 static inline const char *netdev_name(const struct net_device *dev)
4772 {
4773 if (!dev->name[0] || strchr(dev->name, '%'))
4774 return "(unnamed net_device)";
4775 return dev->name;
4776 }
4777
netdev_unregistering(const struct net_device * dev)4778 static inline bool netdev_unregistering(const struct net_device *dev)
4779 {
4780 return dev->reg_state == NETREG_UNREGISTERING;
4781 }
4782
netdev_reg_state(const struct net_device * dev)4783 static inline const char *netdev_reg_state(const struct net_device *dev)
4784 {
4785 switch (dev->reg_state) {
4786 case NETREG_UNINITIALIZED: return " (uninitialized)";
4787 case NETREG_REGISTERED: return "";
4788 case NETREG_UNREGISTERING: return " (unregistering)";
4789 case NETREG_UNREGISTERED: return " (unregistered)";
4790 case NETREG_RELEASED: return " (released)";
4791 case NETREG_DUMMY: return " (dummy)";
4792 }
4793
4794 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4795 return " (unknown)";
4796 }
4797
4798 __printf(3, 4) __cold
4799 void netdev_printk(const char *level, const struct net_device *dev,
4800 const char *format, ...);
4801 __printf(2, 3) __cold
4802 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4803 __printf(2, 3) __cold
4804 void netdev_alert(const struct net_device *dev, const char *format, ...);
4805 __printf(2, 3) __cold
4806 void netdev_crit(const struct net_device *dev, const char *format, ...);
4807 __printf(2, 3) __cold
4808 void netdev_err(const struct net_device *dev, const char *format, ...);
4809 __printf(2, 3) __cold
4810 void netdev_warn(const struct net_device *dev, const char *format, ...);
4811 __printf(2, 3) __cold
4812 void netdev_notice(const struct net_device *dev, const char *format, ...);
4813 __printf(2, 3) __cold
4814 void netdev_info(const struct net_device *dev, const char *format, ...);
4815
4816 #define netdev_level_once(level, dev, fmt, ...) \
4817 do { \
4818 static bool __print_once __read_mostly; \
4819 \
4820 if (!__print_once) { \
4821 __print_once = true; \
4822 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4823 } \
4824 } while (0)
4825
4826 #define netdev_emerg_once(dev, fmt, ...) \
4827 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4828 #define netdev_alert_once(dev, fmt, ...) \
4829 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4830 #define netdev_crit_once(dev, fmt, ...) \
4831 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4832 #define netdev_err_once(dev, fmt, ...) \
4833 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4834 #define netdev_warn_once(dev, fmt, ...) \
4835 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4836 #define netdev_notice_once(dev, fmt, ...) \
4837 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4838 #define netdev_info_once(dev, fmt, ...) \
4839 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4840
4841 #define MODULE_ALIAS_NETDEV(device) \
4842 MODULE_ALIAS("netdev-" device)
4843
4844 #if defined(CONFIG_DYNAMIC_DEBUG) || \
4845 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
4846 #define netdev_dbg(__dev, format, args...) \
4847 do { \
4848 dynamic_netdev_dbg(__dev, format, ##args); \
4849 } while (0)
4850 #elif defined(DEBUG)
4851 #define netdev_dbg(__dev, format, args...) \
4852 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4853 #else
4854 #define netdev_dbg(__dev, format, args...) \
4855 ({ \
4856 if (0) \
4857 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4858 })
4859 #endif
4860
4861 #if defined(VERBOSE_DEBUG)
4862 #define netdev_vdbg netdev_dbg
4863 #else
4864
4865 #define netdev_vdbg(dev, format, args...) \
4866 ({ \
4867 if (0) \
4868 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4869 0; \
4870 })
4871 #endif
4872
4873 /*
4874 * netdev_WARN() acts like dev_printk(), but with the key difference
4875 * of using a WARN/WARN_ON to get the message out, including the
4876 * file/line information and a backtrace.
4877 */
4878 #define netdev_WARN(dev, format, args...) \
4879 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4880 netdev_reg_state(dev), ##args)
4881
4882 #define netdev_WARN_ONCE(dev, format, args...) \
4883 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4884 netdev_reg_state(dev), ##args)
4885
4886 /* netif printk helpers, similar to netdev_printk */
4887
4888 #define netif_printk(priv, type, level, dev, fmt, args...) \
4889 do { \
4890 if (netif_msg_##type(priv)) \
4891 netdev_printk(level, (dev), fmt, ##args); \
4892 } while (0)
4893
4894 #define netif_level(level, priv, type, dev, fmt, args...) \
4895 do { \
4896 if (netif_msg_##type(priv)) \
4897 netdev_##level(dev, fmt, ##args); \
4898 } while (0)
4899
4900 #define netif_emerg(priv, type, dev, fmt, args...) \
4901 netif_level(emerg, priv, type, dev, fmt, ##args)
4902 #define netif_alert(priv, type, dev, fmt, args...) \
4903 netif_level(alert, priv, type, dev, fmt, ##args)
4904 #define netif_crit(priv, type, dev, fmt, args...) \
4905 netif_level(crit, priv, type, dev, fmt, ##args)
4906 #define netif_err(priv, type, dev, fmt, args...) \
4907 netif_level(err, priv, type, dev, fmt, ##args)
4908 #define netif_warn(priv, type, dev, fmt, args...) \
4909 netif_level(warn, priv, type, dev, fmt, ##args)
4910 #define netif_notice(priv, type, dev, fmt, args...) \
4911 netif_level(notice, priv, type, dev, fmt, ##args)
4912 #define netif_info(priv, type, dev, fmt, args...) \
4913 netif_level(info, priv, type, dev, fmt, ##args)
4914
4915 #if defined(CONFIG_DYNAMIC_DEBUG) || \
4916 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
4917 #define netif_dbg(priv, type, netdev, format, args...) \
4918 do { \
4919 if (netif_msg_##type(priv)) \
4920 dynamic_netdev_dbg(netdev, format, ##args); \
4921 } while (0)
4922 #elif defined(DEBUG)
4923 #define netif_dbg(priv, type, dev, format, args...) \
4924 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4925 #else
4926 #define netif_dbg(priv, type, dev, format, args...) \
4927 ({ \
4928 if (0) \
4929 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4930 0; \
4931 })
4932 #endif
4933
4934 /* if @cond then downgrade to debug, else print at @level */
4935 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4936 do { \
4937 if (cond) \
4938 netif_dbg(priv, type, netdev, fmt, ##args); \
4939 else \
4940 netif_ ## level(priv, type, netdev, fmt, ##args); \
4941 } while (0)
4942
4943 #if defined(VERBOSE_DEBUG)
4944 #define netif_vdbg netif_dbg
4945 #else
4946 #define netif_vdbg(priv, type, dev, format, args...) \
4947 ({ \
4948 if (0) \
4949 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4950 0; \
4951 })
4952 #endif
4953
4954 /*
4955 * The list of packet types we will receive (as opposed to discard)
4956 * and the routines to invoke.
4957 *
4958 * Why 16. Because with 16 the only overlap we get on a hash of the
4959 * low nibble of the protocol value is RARP/SNAP/X.25.
4960 *
4961 * 0800 IP
4962 * 0001 802.3
4963 * 0002 AX.25
4964 * 0004 802.2
4965 * 8035 RARP
4966 * 0005 SNAP
4967 * 0805 X.25
4968 * 0806 ARP
4969 * 8137 IPX
4970 * 0009 Localtalk
4971 * 86DD IPv6
4972 */
4973 #define PTYPE_HASH_SIZE (16)
4974 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4975
4976 extern struct net_device *blackhole_netdev;
4977
4978 #endif /* _LINUX_NETDEVICE_H */
4979