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 <net/net_namespace.h>
38 #ifdef CONFIG_DCB
39 #include <net/dcbnl.h>
40 #endif
41 #include <net/netprio_cgroup.h>
42 #include <net/xdp.h>
43
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <linux/hashtable.h>
50 #include <linux/rbtree.h>
51
52 struct netpoll_info;
53 struct device;
54 struct ethtool_ops;
55 struct phy_device;
56 struct dsa_port;
57 struct ip_tunnel_parm;
58 struct macsec_context;
59 struct macsec_ops;
60
61 struct sfp_bus;
62 /* 802.11 specific */
63 struct wireless_dev;
64 /* 802.15.4 specific */
65 struct wpan_dev;
66 struct mpls_dev;
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
69 struct udp_tunnel_nic_info;
70 struct udp_tunnel_nic;
71 struct bpf_prog;
72 struct xdp_buff;
73
74 void synchronize_net(void);
75 void netdev_set_default_ethtool_ops(struct net_device *dev,
76 const struct ethtool_ops *ops);
77
78 /* Backlog congestion levels */
79 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
80 #define NET_RX_DROP 1 /* packet dropped */
81
82 #define MAX_NEST_DEV 8
83
84 /*
85 * Transmit return codes: transmit return codes originate from three different
86 * namespaces:
87 *
88 * - qdisc return codes
89 * - driver transmit return codes
90 * - errno values
91 *
92 * Drivers are allowed to return any one of those in their hard_start_xmit()
93 * function. Real network devices commonly used with qdiscs should only return
94 * the driver transmit return codes though - when qdiscs are used, the actual
95 * transmission happens asynchronously, so the value is not propagated to
96 * higher layers. Virtual network devices transmit synchronously; in this case
97 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
98 * others are propagated to higher layers.
99 */
100
101 /* qdisc ->enqueue() return codes. */
102 #define NET_XMIT_SUCCESS 0x00
103 #define NET_XMIT_DROP 0x01 /* skb dropped */
104 #define NET_XMIT_CN 0x02 /* congestion notification */
105 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
106
107 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
108 * indicates that the device will soon be dropping packets, or already drops
109 * some packets of the same priority; prompting us to send less aggressively. */
110 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
111 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
112
113 /* Driver transmit return codes */
114 #define NETDEV_TX_MASK 0xf0
115
116 enum netdev_tx {
117 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
118 NETDEV_TX_OK = 0x00, /* driver took care of packet */
119 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
120 };
121 typedef enum netdev_tx netdev_tx_t;
122
123 /*
124 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
125 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
126 */
dev_xmit_complete(int rc)127 static inline bool dev_xmit_complete(int rc)
128 {
129 /*
130 * Positive cases with an skb consumed by a driver:
131 * - successful transmission (rc == NETDEV_TX_OK)
132 * - error while transmitting (rc < 0)
133 * - error while queueing to a different device (rc & NET_XMIT_MASK)
134 */
135 if (likely(rc < NET_XMIT_MASK))
136 return true;
137
138 return false;
139 }
140
141 /*
142 * Compute the worst-case header length according to the protocols
143 * used.
144 */
145
146 #if defined(CONFIG_HYPERV_NET)
147 # define LL_MAX_HEADER 128
148 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
149 # if defined(CONFIG_MAC80211_MESH)
150 # define LL_MAX_HEADER 128
151 # else
152 # define LL_MAX_HEADER 96
153 # endif
154 #else
155 # define LL_MAX_HEADER 32
156 #endif
157
158 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
159 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
160 #define MAX_HEADER LL_MAX_HEADER
161 #else
162 #define MAX_HEADER (LL_MAX_HEADER + 48)
163 #endif
164
165 /*
166 * Old network device statistics. Fields are native words
167 * (unsigned long) so they can be read and written atomically.
168 */
169
170 struct net_device_stats {
171 unsigned long rx_packets;
172 unsigned long tx_packets;
173 unsigned long rx_bytes;
174 unsigned long tx_bytes;
175 unsigned long rx_errors;
176 unsigned long tx_errors;
177 unsigned long rx_dropped;
178 unsigned long tx_dropped;
179 unsigned long multicast;
180 unsigned long collisions;
181 unsigned long rx_length_errors;
182 unsigned long rx_over_errors;
183 unsigned long rx_crc_errors;
184 unsigned long rx_frame_errors;
185 unsigned long rx_fifo_errors;
186 unsigned long rx_missed_errors;
187 unsigned long tx_aborted_errors;
188 unsigned long tx_carrier_errors;
189 unsigned long tx_fifo_errors;
190 unsigned long tx_heartbeat_errors;
191 unsigned long tx_window_errors;
192 unsigned long rx_compressed;
193 unsigned long tx_compressed;
194 };
195
196
197 #include <linux/cache.h>
198 #include <linux/skbuff.h>
199
200 #ifdef CONFIG_RPS
201 #include <linux/static_key.h>
202 extern struct static_key_false rps_needed;
203 extern struct static_key_false rfs_needed;
204 #endif
205
206 struct neighbour;
207 struct neigh_parms;
208 struct sk_buff;
209
210 struct netdev_hw_addr {
211 struct list_head list;
212 struct rb_node node;
213 unsigned char addr[MAX_ADDR_LEN];
214 unsigned char type;
215 #define NETDEV_HW_ADDR_T_LAN 1
216 #define NETDEV_HW_ADDR_T_SAN 2
217 #define NETDEV_HW_ADDR_T_UNICAST 3
218 #define NETDEV_HW_ADDR_T_MULTICAST 4
219 bool global_use;
220 int sync_cnt;
221 int refcount;
222 int synced;
223 struct rcu_head rcu_head;
224 };
225
226 struct netdev_hw_addr_list {
227 struct list_head list;
228 int count;
229
230 /* Auxiliary tree for faster lookup on addition and deletion */
231 struct rb_root tree;
232 };
233
234 #define netdev_hw_addr_list_count(l) ((l)->count)
235 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
236 #define netdev_hw_addr_list_for_each(ha, l) \
237 list_for_each_entry(ha, &(l)->list, list)
238
239 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
240 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
241 #define netdev_for_each_uc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
243
244 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
245 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
246 #define netdev_for_each_mc_addr(ha, dev) \
247 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
248
249 struct hh_cache {
250 unsigned int hh_len;
251 seqlock_t hh_lock;
252
253 /* cached hardware header; allow for machine alignment needs. */
254 #define HH_DATA_MOD 16
255 #define HH_DATA_OFF(__len) \
256 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
257 #define HH_DATA_ALIGN(__len) \
258 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
259 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
260 };
261
262 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
263 * Alternative is:
264 * dev->hard_header_len ? (dev->hard_header_len +
265 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
266 *
267 * We could use other alignment values, but we must maintain the
268 * relationship HH alignment <= LL alignment.
269 */
270 #define LL_RESERVED_SPACE(dev) \
271 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
272 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
273 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
274
275 struct header_ops {
276 int (*create) (struct sk_buff *skb, struct net_device *dev,
277 unsigned short type, const void *daddr,
278 const void *saddr, unsigned int len);
279 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
280 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
281 void (*cache_update)(struct hh_cache *hh,
282 const struct net_device *dev,
283 const unsigned char *haddr);
284 bool (*validate)(const char *ll_header, unsigned int len);
285 __be16 (*parse_protocol)(const struct sk_buff *skb);
286 };
287
288 /* These flag bits are private to the generic network queueing
289 * layer; they may not be explicitly referenced by any other
290 * code.
291 */
292
293 enum netdev_state_t {
294 __LINK_STATE_START,
295 __LINK_STATE_PRESENT,
296 __LINK_STATE_NOCARRIER,
297 __LINK_STATE_LINKWATCH_PENDING,
298 __LINK_STATE_DORMANT,
299 __LINK_STATE_TESTING,
300 };
301
302
303 struct gro_list {
304 struct list_head list;
305 int count;
306 };
307
308 /*
309 * size of gro hash buckets, must less than bit number of
310 * napi_struct::gro_bitmask
311 */
312 #define GRO_HASH_BUCKETS 8
313
314 /*
315 * Structure for NAPI scheduling similar to tasklet but with weighting
316 */
317 struct napi_struct {
318 /* The poll_list must only be managed by the entity which
319 * changes the state of the NAPI_STATE_SCHED bit. This means
320 * whoever atomically sets that bit can add this napi_struct
321 * to the per-CPU poll_list, and whoever clears that bit
322 * can remove from the list right before clearing the bit.
323 */
324 struct list_head poll_list;
325
326 unsigned long state;
327 int weight;
328 int defer_hard_irqs_count;
329 unsigned long gro_bitmask;
330 int (*poll)(struct napi_struct *, int);
331 #ifdef CONFIG_NETPOLL
332 int poll_owner;
333 #endif
334 struct net_device *dev;
335 struct gro_list gro_hash[GRO_HASH_BUCKETS];
336 struct sk_buff *skb;
337 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
338 int rx_count; /* length of rx_list */
339 struct hrtimer timer;
340 struct list_head dev_list;
341 struct hlist_node napi_hash_node;
342 unsigned int napi_id;
343 struct task_struct *thread;
344 };
345
346 enum {
347 NAPI_STATE_SCHED, /* Poll is scheduled */
348 NAPI_STATE_MISSED, /* reschedule a napi */
349 NAPI_STATE_DISABLE, /* Disable pending */
350 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
351 NAPI_STATE_LISTED, /* NAPI added to system lists */
352 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
353 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
354 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
355 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
356 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
357 };
358
359 enum {
360 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
361 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
362 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
363 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
364 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
365 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
366 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
367 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
368 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
369 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
370 };
371
372 enum gro_result {
373 GRO_MERGED,
374 GRO_MERGED_FREE,
375 GRO_HELD,
376 GRO_NORMAL,
377 GRO_CONSUMED,
378 };
379 typedef enum gro_result gro_result_t;
380
381 /*
382 * enum rx_handler_result - Possible return values for rx_handlers.
383 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
384 * further.
385 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
386 * case skb->dev was changed by rx_handler.
387 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
388 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
389 *
390 * rx_handlers are functions called from inside __netif_receive_skb(), to do
391 * special processing of the skb, prior to delivery to protocol handlers.
392 *
393 * Currently, a net_device can only have a single rx_handler registered. Trying
394 * to register a second rx_handler will return -EBUSY.
395 *
396 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
397 * To unregister a rx_handler on a net_device, use
398 * netdev_rx_handler_unregister().
399 *
400 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
401 * do with the skb.
402 *
403 * If the rx_handler consumed the skb in some way, it should return
404 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
405 * the skb to be delivered in some other way.
406 *
407 * If the rx_handler changed skb->dev, to divert the skb to another
408 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
409 * new device will be called if it exists.
410 *
411 * If the rx_handler decides the skb should be ignored, it should return
412 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
413 * are registered on exact device (ptype->dev == skb->dev).
414 *
415 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
416 * delivered, it should return RX_HANDLER_PASS.
417 *
418 * A device without a registered rx_handler will behave as if rx_handler
419 * returned RX_HANDLER_PASS.
420 */
421
422 enum rx_handler_result {
423 RX_HANDLER_CONSUMED,
424 RX_HANDLER_ANOTHER,
425 RX_HANDLER_EXACT,
426 RX_HANDLER_PASS,
427 };
428 typedef enum rx_handler_result rx_handler_result_t;
429 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
430
431 void __napi_schedule(struct napi_struct *n);
432 void __napi_schedule_irqoff(struct napi_struct *n);
433
napi_disable_pending(struct napi_struct * n)434 static inline bool napi_disable_pending(struct napi_struct *n)
435 {
436 return test_bit(NAPI_STATE_DISABLE, &n->state);
437 }
438
napi_prefer_busy_poll(struct napi_struct * n)439 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
440 {
441 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &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 int dev_set_threaded(struct net_device *dev, bool threaded);
496
497 /**
498 * napi_disable - prevent NAPI from scheduling
499 * @n: NAPI context
500 *
501 * Stop NAPI from being scheduled on this context.
502 * Waits till any outstanding processing completes.
503 */
504 void napi_disable(struct napi_struct *n);
505
506 void napi_enable(struct napi_struct *n);
507
508 /**
509 * napi_synchronize - wait until NAPI is not running
510 * @n: NAPI context
511 *
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
515 */
napi_synchronize(const struct napi_struct * n)516 static inline void napi_synchronize(const struct napi_struct *n)
517 {
518 if (IS_ENABLED(CONFIG_SMP))
519 while (test_bit(NAPI_STATE_SCHED, &n->state))
520 msleep(1);
521 else
522 barrier();
523 }
524
525 /**
526 * napi_if_scheduled_mark_missed - if napi is running, set the
527 * NAPIF_STATE_MISSED
528 * @n: NAPI context
529 *
530 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
531 * NAPI is scheduled.
532 **/
napi_if_scheduled_mark_missed(struct napi_struct * n)533 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
534 {
535 unsigned long val, new;
536
537 do {
538 val = READ_ONCE(n->state);
539 if (val & NAPIF_STATE_DISABLE)
540 return true;
541
542 if (!(val & NAPIF_STATE_SCHED))
543 return false;
544
545 new = val | NAPIF_STATE_MISSED;
546 } while (cmpxchg(&n->state, val, new) != val);
547
548 return true;
549 }
550
551 enum netdev_queue_state_t {
552 __QUEUE_STATE_DRV_XOFF,
553 __QUEUE_STATE_STACK_XOFF,
554 __QUEUE_STATE_FROZEN,
555 };
556
557 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
558 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
559 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
560
561 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
562 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
563 QUEUE_STATE_FROZEN)
564 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
565 QUEUE_STATE_FROZEN)
566
567 /*
568 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
569 * netif_tx_* functions below are used to manipulate this flag. The
570 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
571 * queue independently. The netif_xmit_*stopped functions below are called
572 * to check if the queue has been stopped by the driver or stack (either
573 * of the XOFF bits are set in the state). Drivers should not need to call
574 * netif_xmit*stopped functions, they should only be using netif_tx_*.
575 */
576
577 struct netdev_queue {
578 /*
579 * read-mostly part
580 */
581 struct net_device *dev;
582 struct Qdisc __rcu *qdisc;
583 struct Qdisc *qdisc_sleeping;
584 #ifdef CONFIG_SYSFS
585 struct kobject kobj;
586 #endif
587 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
588 int numa_node;
589 #endif
590 unsigned long tx_maxrate;
591 /*
592 * Number of TX timeouts for this queue
593 * (/sys/class/net/DEV/Q/trans_timeout)
594 */
595 unsigned long trans_timeout;
596
597 /* Subordinate device that the queue has been assigned to */
598 struct net_device *sb_dev;
599 #ifdef CONFIG_XDP_SOCKETS
600 struct xsk_buff_pool *pool;
601 #endif
602 /*
603 * write-mostly part
604 */
605 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
606 int xmit_lock_owner;
607 /*
608 * Time (in jiffies) of last Tx
609 */
610 unsigned long trans_start;
611
612 unsigned long state;
613
614 #ifdef CONFIG_BQL
615 struct dql dql;
616 #endif
617 } ____cacheline_aligned_in_smp;
618
619 extern int sysctl_fb_tunnels_only_for_init_net;
620 extern int sysctl_devconf_inherit_init_net;
621
622 /*
623 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
624 * == 1 : For initns only
625 * == 2 : For none.
626 */
net_has_fallback_tunnels(const struct net * net)627 static inline bool net_has_fallback_tunnels(const struct net *net)
628 {
629 return !IS_ENABLED(CONFIG_SYSCTL) ||
630 !sysctl_fb_tunnels_only_for_init_net ||
631 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
632 }
633
netdev_queue_numa_node_read(const struct netdev_queue * q)634 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
635 {
636 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
637 return q->numa_node;
638 #else
639 return NUMA_NO_NODE;
640 #endif
641 }
642
netdev_queue_numa_node_write(struct netdev_queue * q,int node)643 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
644 {
645 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
646 q->numa_node = node;
647 #endif
648 }
649
650 #ifdef CONFIG_RPS
651 /*
652 * This structure holds an RPS map which can be of variable length. The
653 * map is an array of CPUs.
654 */
655 struct rps_map {
656 unsigned int len;
657 struct rcu_head rcu;
658 u16 cpus[];
659 };
660 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
661
662 /*
663 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
664 * tail pointer for that CPU's input queue at the time of last enqueue, and
665 * a hardware filter index.
666 */
667 struct rps_dev_flow {
668 u16 cpu;
669 u16 filter;
670 unsigned int last_qtail;
671 };
672 #define RPS_NO_FILTER 0xffff
673
674 /*
675 * The rps_dev_flow_table structure contains a table of flow mappings.
676 */
677 struct rps_dev_flow_table {
678 unsigned int mask;
679 struct rcu_head rcu;
680 struct rps_dev_flow flows[];
681 };
682 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
683 ((_num) * sizeof(struct rps_dev_flow)))
684
685 /*
686 * The rps_sock_flow_table contains mappings of flows to the last CPU
687 * on which they were processed by the application (set in recvmsg).
688 * Each entry is a 32bit value. Upper part is the high-order bits
689 * of flow hash, lower part is CPU number.
690 * rps_cpu_mask is used to partition the space, depending on number of
691 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
692 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
693 * meaning we use 32-6=26 bits for the hash.
694 */
695 struct rps_sock_flow_table {
696 u32 mask;
697
698 u32 ents[] ____cacheline_aligned_in_smp;
699 };
700 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
701
702 #define RPS_NO_CPU 0xffff
703
704 extern u32 rps_cpu_mask;
705 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
706
rps_record_sock_flow(struct rps_sock_flow_table * table,u32 hash)707 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
708 u32 hash)
709 {
710 if (table && hash) {
711 unsigned int index = hash & table->mask;
712 u32 val = hash & ~rps_cpu_mask;
713
714 /* We only give a hint, preemption can change CPU under us */
715 val |= raw_smp_processor_id();
716
717 if (table->ents[index] != val)
718 table->ents[index] = val;
719 }
720 }
721
722 #ifdef CONFIG_RFS_ACCEL
723 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
724 u16 filter_id);
725 #endif
726 #endif /* CONFIG_RPS */
727
728 /* This structure contains an instance of an RX queue. */
729 struct netdev_rx_queue {
730 struct xdp_rxq_info xdp_rxq;
731 #ifdef CONFIG_RPS
732 struct rps_map __rcu *rps_map;
733 struct rps_dev_flow_table __rcu *rps_flow_table;
734 #endif
735 struct kobject kobj;
736 struct net_device *dev;
737 #ifdef CONFIG_XDP_SOCKETS
738 struct xsk_buff_pool *pool;
739 #endif
740 } ____cacheline_aligned_in_smp;
741
742 /*
743 * RX queue sysfs structures and functions.
744 */
745 struct rx_queue_attribute {
746 struct attribute attr;
747 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
748 ssize_t (*store)(struct netdev_rx_queue *queue,
749 const char *buf, size_t len);
750 };
751
752 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
753 enum xps_map_type {
754 XPS_CPUS = 0,
755 XPS_RXQS,
756 XPS_MAPS_MAX,
757 };
758
759 #ifdef CONFIG_XPS
760 /*
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
763 */
764 struct xps_map {
765 unsigned int len;
766 unsigned int alloc_len;
767 struct rcu_head rcu;
768 u16 queues[];
769 };
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
773
774 /*
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
776 *
777 * We keep track of the number of cpus/rxqs used when the struct is allocated,
778 * in nr_ids. This will help not accessing out-of-bound memory.
779 *
780 * We keep track of the number of traffic classes used when the struct is
781 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
782 * not crossing its upper bound, as the original dev->num_tc can be updated in
783 * the meantime.
784 */
785 struct xps_dev_maps {
786 struct rcu_head rcu;
787 unsigned int nr_ids;
788 s16 num_tc;
789 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
790 };
791
792 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
793 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
794
795 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
796 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
797
798 #endif /* CONFIG_XPS */
799
800 #define TC_MAX_QUEUE 16
801 #define TC_BITMASK 15
802 /* HW offloaded queuing disciplines txq count and offset maps */
803 struct netdev_tc_txq {
804 u16 count;
805 u16 offset;
806 };
807
808 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
809 /*
810 * This structure is to hold information about the device
811 * configured to run FCoE protocol stack.
812 */
813 struct netdev_fcoe_hbainfo {
814 char manufacturer[64];
815 char serial_number[64];
816 char hardware_version[64];
817 char driver_version[64];
818 char optionrom_version[64];
819 char firmware_version[64];
820 char model[256];
821 char model_description[256];
822 };
823 #endif
824
825 #define MAX_PHYS_ITEM_ID_LEN 32
826
827 /* This structure holds a unique identifier to identify some
828 * physical item (port for example) used by a netdevice.
829 */
830 struct netdev_phys_item_id {
831 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
832 unsigned char id_len;
833 };
834
netdev_phys_item_id_same(struct netdev_phys_item_id * a,struct netdev_phys_item_id * b)835 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
836 struct netdev_phys_item_id *b)
837 {
838 return a->id_len == b->id_len &&
839 memcmp(a->id, b->id, a->id_len) == 0;
840 }
841
842 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
843 struct sk_buff *skb,
844 struct net_device *sb_dev);
845
846 enum net_device_path_type {
847 DEV_PATH_ETHERNET = 0,
848 DEV_PATH_VLAN,
849 DEV_PATH_BRIDGE,
850 DEV_PATH_PPPOE,
851 DEV_PATH_DSA,
852 };
853
854 struct net_device_path {
855 enum net_device_path_type type;
856 const struct net_device *dev;
857 union {
858 struct {
859 u16 id;
860 __be16 proto;
861 u8 h_dest[ETH_ALEN];
862 } encap;
863 struct {
864 enum {
865 DEV_PATH_BR_VLAN_KEEP,
866 DEV_PATH_BR_VLAN_TAG,
867 DEV_PATH_BR_VLAN_UNTAG,
868 DEV_PATH_BR_VLAN_UNTAG_HW,
869 } vlan_mode;
870 u16 vlan_id;
871 __be16 vlan_proto;
872 } bridge;
873 struct {
874 int port;
875 u16 proto;
876 } dsa;
877 };
878 };
879
880 #define NET_DEVICE_PATH_STACK_MAX 5
881 #define NET_DEVICE_PATH_VLAN_MAX 2
882
883 struct net_device_path_stack {
884 int num_paths;
885 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
886 };
887
888 struct net_device_path_ctx {
889 const struct net_device *dev;
890 const u8 *daddr;
891
892 int num_vlans;
893 struct {
894 u16 id;
895 __be16 proto;
896 } vlan[NET_DEVICE_PATH_VLAN_MAX];
897 };
898
899 enum tc_setup_type {
900 TC_SETUP_QDISC_MQPRIO,
901 TC_SETUP_CLSU32,
902 TC_SETUP_CLSFLOWER,
903 TC_SETUP_CLSMATCHALL,
904 TC_SETUP_CLSBPF,
905 TC_SETUP_BLOCK,
906 TC_SETUP_QDISC_CBS,
907 TC_SETUP_QDISC_RED,
908 TC_SETUP_QDISC_PRIO,
909 TC_SETUP_QDISC_MQ,
910 TC_SETUP_QDISC_ETF,
911 TC_SETUP_ROOT_QDISC,
912 TC_SETUP_QDISC_GRED,
913 TC_SETUP_QDISC_TAPRIO,
914 TC_SETUP_FT,
915 TC_SETUP_QDISC_ETS,
916 TC_SETUP_QDISC_TBF,
917 TC_SETUP_QDISC_FIFO,
918 TC_SETUP_QDISC_HTB,
919 };
920
921 /* These structures hold the attributes of bpf state that are being passed
922 * to the netdevice through the bpf op.
923 */
924 enum bpf_netdev_command {
925 /* Set or clear a bpf program used in the earliest stages of packet
926 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
927 * is responsible for calling bpf_prog_put on any old progs that are
928 * stored. In case of error, the callee need not release the new prog
929 * reference, but on success it takes ownership and must bpf_prog_put
930 * when it is no longer used.
931 */
932 XDP_SETUP_PROG,
933 XDP_SETUP_PROG_HW,
934 /* BPF program for offload callbacks, invoked at program load time. */
935 BPF_OFFLOAD_MAP_ALLOC,
936 BPF_OFFLOAD_MAP_FREE,
937 XDP_SETUP_XSK_POOL,
938 };
939
940 struct bpf_prog_offload_ops;
941 struct netlink_ext_ack;
942 struct xdp_umem;
943 struct xdp_dev_bulk_queue;
944 struct bpf_xdp_link;
945
946 enum bpf_xdp_mode {
947 XDP_MODE_SKB = 0,
948 XDP_MODE_DRV = 1,
949 XDP_MODE_HW = 2,
950 __MAX_XDP_MODE
951 };
952
953 struct bpf_xdp_entity {
954 struct bpf_prog *prog;
955 struct bpf_xdp_link *link;
956 };
957
958 struct netdev_bpf {
959 enum bpf_netdev_command command;
960 union {
961 /* XDP_SETUP_PROG */
962 struct {
963 u32 flags;
964 struct bpf_prog *prog;
965 struct netlink_ext_ack *extack;
966 };
967 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
968 struct {
969 struct bpf_offloaded_map *offmap;
970 };
971 /* XDP_SETUP_XSK_POOL */
972 struct {
973 struct xsk_buff_pool *pool;
974 u16 queue_id;
975 } xsk;
976 };
977 };
978
979 /* Flags for ndo_xsk_wakeup. */
980 #define XDP_WAKEUP_RX (1 << 0)
981 #define XDP_WAKEUP_TX (1 << 1)
982
983 #ifdef CONFIG_XFRM_OFFLOAD
984 struct xfrmdev_ops {
985 int (*xdo_dev_state_add) (struct xfrm_state *x);
986 void (*xdo_dev_state_delete) (struct xfrm_state *x);
987 void (*xdo_dev_state_free) (struct xfrm_state *x);
988 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
989 struct xfrm_state *x);
990 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
991 };
992 #endif
993
994 struct dev_ifalias {
995 struct rcu_head rcuhead;
996 char ifalias[];
997 };
998
999 struct devlink;
1000 struct tlsdev_ops;
1001
1002 struct netdev_name_node {
1003 struct hlist_node hlist;
1004 struct list_head list;
1005 struct net_device *dev;
1006 const char *name;
1007 };
1008
1009 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1010 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1011
1012 struct netdev_net_notifier {
1013 struct list_head list;
1014 struct notifier_block *nb;
1015 };
1016
1017 /*
1018 * This structure defines the management hooks for network devices.
1019 * The following hooks can be defined; unless noted otherwise, they are
1020 * optional and can be filled with a null pointer.
1021 *
1022 * int (*ndo_init)(struct net_device *dev);
1023 * This function is called once when a network device is registered.
1024 * The network device can use this for any late stage initialization
1025 * or semantic validation. It can fail with an error code which will
1026 * be propagated back to register_netdev.
1027 *
1028 * void (*ndo_uninit)(struct net_device *dev);
1029 * This function is called when device is unregistered or when registration
1030 * fails. It is not called if init fails.
1031 *
1032 * int (*ndo_open)(struct net_device *dev);
1033 * This function is called when a network device transitions to the up
1034 * state.
1035 *
1036 * int (*ndo_stop)(struct net_device *dev);
1037 * This function is called when a network device transitions to the down
1038 * state.
1039 *
1040 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1041 * struct net_device *dev);
1042 * Called when a packet needs to be transmitted.
1043 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1044 * the queue before that can happen; it's for obsolete devices and weird
1045 * corner cases, but the stack really does a non-trivial amount
1046 * of useless work if you return NETDEV_TX_BUSY.
1047 * Required; cannot be NULL.
1048 *
1049 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1050 * struct net_device *dev
1051 * netdev_features_t features);
1052 * Called by core transmit path to determine if device is capable of
1053 * performing offload operations on a given packet. This is to give
1054 * the device an opportunity to implement any restrictions that cannot
1055 * be otherwise expressed by feature flags. The check is called with
1056 * the set of features that the stack has calculated and it returns
1057 * those the driver believes to be appropriate.
1058 *
1059 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1060 * struct net_device *sb_dev);
1061 * Called to decide which queue to use when device supports multiple
1062 * transmit queues.
1063 *
1064 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1065 * This function is called to allow device receiver to make
1066 * changes to configuration when multicast or promiscuous is enabled.
1067 *
1068 * void (*ndo_set_rx_mode)(struct net_device *dev);
1069 * This function is called device changes address list filtering.
1070 * If driver handles unicast address filtering, it should set
1071 * IFF_UNICAST_FLT in its priv_flags.
1072 *
1073 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1074 * This function is called when the Media Access Control address
1075 * needs to be changed. If this interface is not defined, the
1076 * MAC address can not be changed.
1077 *
1078 * int (*ndo_validate_addr)(struct net_device *dev);
1079 * Test if Media Access Control address is valid for the device.
1080 *
1081 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1082 * Old-style ioctl entry point. This is used internally by the
1083 * appletalk and ieee802154 subsystems but is no longer called by
1084 * the device ioctl handler.
1085 *
1086 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1087 * Used by the bonding driver for its device specific ioctls:
1088 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1089 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1090 *
1091 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1092 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1093 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1094 *
1095 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1096 * Used to set network devices bus interface parameters. This interface
1097 * is retained for legacy reasons; new devices should use the bus
1098 * interface (PCI) for low level management.
1099 *
1100 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1101 * Called when a user wants to change the Maximum Transfer Unit
1102 * of a device.
1103 *
1104 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1105 * Callback used when the transmitter has not made any progress
1106 * for dev->watchdog ticks.
1107 *
1108 * void (*ndo_get_stats64)(struct net_device *dev,
1109 * struct rtnl_link_stats64 *storage);
1110 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1111 * Called when a user wants to get the network device usage
1112 * statistics. Drivers must do one of the following:
1113 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1114 * rtnl_link_stats64 structure passed by the caller.
1115 * 2. Define @ndo_get_stats to update a net_device_stats structure
1116 * (which should normally be dev->stats) and return a pointer to
1117 * it. The structure may be changed asynchronously only if each
1118 * field is written atomically.
1119 * 3. Update dev->stats asynchronously and atomically, and define
1120 * neither operation.
1121 *
1122 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1123 * Return true if this device supports offload stats of this attr_id.
1124 *
1125 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1126 * void *attr_data)
1127 * Get statistics for offload operations by attr_id. Write it into the
1128 * attr_data pointer.
1129 *
1130 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1131 * If device supports VLAN filtering this function is called when a
1132 * VLAN id is registered.
1133 *
1134 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1135 * If device supports VLAN filtering this function is called when a
1136 * VLAN id is unregistered.
1137 *
1138 * void (*ndo_poll_controller)(struct net_device *dev);
1139 *
1140 * SR-IOV management functions.
1141 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1142 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1143 * u8 qos, __be16 proto);
1144 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1145 * int max_tx_rate);
1146 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1147 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1148 * int (*ndo_get_vf_config)(struct net_device *dev,
1149 * int vf, struct ifla_vf_info *ivf);
1150 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1151 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1152 * struct nlattr *port[]);
1153 *
1154 * Enable or disable the VF ability to query its RSS Redirection Table and
1155 * Hash Key. This is needed since on some devices VF share this information
1156 * with PF and querying it may introduce a theoretical security risk.
1157 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1158 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1159 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1160 * void *type_data);
1161 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1162 * This is always called from the stack with the rtnl lock held and netif
1163 * tx queues stopped. This allows the netdevice to perform queue
1164 * management safely.
1165 *
1166 * Fiber Channel over Ethernet (FCoE) offload functions.
1167 * int (*ndo_fcoe_enable)(struct net_device *dev);
1168 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1169 * so the underlying device can perform whatever needed configuration or
1170 * initialization to support acceleration of FCoE traffic.
1171 *
1172 * int (*ndo_fcoe_disable)(struct net_device *dev);
1173 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1174 * so the underlying device can perform whatever needed clean-ups to
1175 * stop supporting acceleration of FCoE traffic.
1176 *
1177 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1178 * struct scatterlist *sgl, unsigned int sgc);
1179 * Called when the FCoE Initiator wants to initialize an I/O that
1180 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1181 * perform necessary setup and returns 1 to indicate the device is set up
1182 * successfully to perform DDP on this I/O, otherwise this returns 0.
1183 *
1184 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1185 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1186 * indicated by the FC exchange id 'xid', so the underlying device can
1187 * clean up and reuse resources for later DDP requests.
1188 *
1189 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1190 * struct scatterlist *sgl, unsigned int sgc);
1191 * Called when the FCoE Target wants to initialize an I/O that
1192 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1193 * perform necessary setup and returns 1 to indicate the device is set up
1194 * successfully to perform DDP on this I/O, otherwise this returns 0.
1195 *
1196 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1197 * struct netdev_fcoe_hbainfo *hbainfo);
1198 * Called when the FCoE Protocol stack wants information on the underlying
1199 * device. This information is utilized by the FCoE protocol stack to
1200 * register attributes with Fiber Channel management service as per the
1201 * FC-GS Fabric Device Management Information(FDMI) specification.
1202 *
1203 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1204 * Called when the underlying device wants to override default World Wide
1205 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1206 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1207 * protocol stack to use.
1208 *
1209 * RFS acceleration.
1210 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1211 * u16 rxq_index, u32 flow_id);
1212 * Set hardware filter for RFS. rxq_index is the target queue index;
1213 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1214 * Return the filter ID on success, or a negative error code.
1215 *
1216 * Slave management functions (for bridge, bonding, etc).
1217 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1218 * Called to make another netdev an underling.
1219 *
1220 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1221 * Called to release previously enslaved netdev.
1222 *
1223 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1224 * struct sk_buff *skb,
1225 * bool all_slaves);
1226 * Get the xmit slave of master device. If all_slaves is true, function
1227 * assume all the slaves can transmit.
1228 *
1229 * Feature/offload setting functions.
1230 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1231 * netdev_features_t features);
1232 * Adjusts the requested feature flags according to device-specific
1233 * constraints, and returns the resulting flags. Must not modify
1234 * the device state.
1235 *
1236 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1237 * Called to update device configuration to new features. Passed
1238 * feature set might be less than what was returned by ndo_fix_features()).
1239 * Must return >0 or -errno if it changed dev->features itself.
1240 *
1241 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1242 * struct net_device *dev,
1243 * const unsigned char *addr, u16 vid, u16 flags,
1244 * struct netlink_ext_ack *extack);
1245 * Adds an FDB entry to dev for addr.
1246 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1247 * struct net_device *dev,
1248 * const unsigned char *addr, u16 vid)
1249 * Deletes the FDB entry from dev coresponding to addr.
1250 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1251 * struct net_device *dev, struct net_device *filter_dev,
1252 * int *idx)
1253 * Used to add FDB entries to dump requests. Implementers should add
1254 * entries to skb and update idx with the number of entries.
1255 *
1256 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1257 * u16 flags, struct netlink_ext_ack *extack)
1258 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1259 * struct net_device *dev, u32 filter_mask,
1260 * int nlflags)
1261 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1262 * u16 flags);
1263 *
1264 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1265 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1266 * which do not represent real hardware may define this to allow their
1267 * userspace components to manage their virtual carrier state. Devices
1268 * that determine carrier state from physical hardware properties (eg
1269 * network cables) or protocol-dependent mechanisms (eg
1270 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1271 *
1272 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1273 * struct netdev_phys_item_id *ppid);
1274 * Called to get ID of physical port of this device. If driver does
1275 * not implement this, it is assumed that the hw is not able to have
1276 * multiple net devices on single physical port.
1277 *
1278 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1279 * struct netdev_phys_item_id *ppid)
1280 * Called to get the parent ID of the physical port of this device.
1281 *
1282 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1283 * struct net_device *dev)
1284 * Called by upper layer devices to accelerate switching or other
1285 * station functionality into hardware. 'pdev is the lowerdev
1286 * to use for the offload and 'dev' is the net device that will
1287 * back the offload. Returns a pointer to the private structure
1288 * the upper layer will maintain.
1289 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1290 * Called by upper layer device to delete the station created
1291 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1292 * the station and priv is the structure returned by the add
1293 * operation.
1294 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1295 * int queue_index, u32 maxrate);
1296 * Called when a user wants to set a max-rate limitation of specific
1297 * TX queue.
1298 * int (*ndo_get_iflink)(const struct net_device *dev);
1299 * Called to get the iflink value of this device.
1300 * void (*ndo_change_proto_down)(struct net_device *dev,
1301 * bool proto_down);
1302 * This function is used to pass protocol port error state information
1303 * to the switch driver. The switch driver can react to the proto_down
1304 * by doing a phys down on the associated switch port.
1305 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1306 * This function is used to get egress tunnel information for given skb.
1307 * This is useful for retrieving outer tunnel header parameters while
1308 * sampling packet.
1309 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1310 * This function is used to specify the headroom that the skb must
1311 * consider when allocation skb during packet reception. Setting
1312 * appropriate rx headroom value allows avoiding skb head copy on
1313 * forward. Setting a negative value resets the rx headroom to the
1314 * default value.
1315 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1316 * This function is used to set or query state related to XDP on the
1317 * netdevice and manage BPF offload. See definition of
1318 * enum bpf_netdev_command for details.
1319 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1320 * u32 flags);
1321 * This function is used to submit @n XDP packets for transmit on a
1322 * netdevice. Returns number of frames successfully transmitted, frames
1323 * that got dropped are freed/returned via xdp_return_frame().
1324 * Returns negative number, means general error invoking ndo, meaning
1325 * no frames were xmit'ed and core-caller will free all frames.
1326 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1327 * struct xdp_buff *xdp);
1328 * Get the xmit slave of master device based on the xdp_buff.
1329 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1330 * This function is used to wake up the softirq, ksoftirqd or kthread
1331 * responsible for sending and/or receiving packets on a specific
1332 * queue id bound to an AF_XDP socket. The flags field specifies if
1333 * only RX, only Tx, or both should be woken up using the flags
1334 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1335 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1336 * Get devlink port instance associated with a given netdev.
1337 * Called with a reference on the netdevice and devlink locks only,
1338 * rtnl_lock is not held.
1339 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1340 * int cmd);
1341 * Add, change, delete or get information on an IPv4 tunnel.
1342 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1343 * If a device is paired with a peer device, return the peer instance.
1344 * The caller must be under RCU read context.
1345 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1346 * Get the forwarding path to reach the real device from the HW destination address
1347 */
1348 struct net_device_ops {
1349 int (*ndo_init)(struct net_device *dev);
1350 void (*ndo_uninit)(struct net_device *dev);
1351 int (*ndo_open)(struct net_device *dev);
1352 int (*ndo_stop)(struct net_device *dev);
1353 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1354 struct net_device *dev);
1355 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1356 struct net_device *dev,
1357 netdev_features_t features);
1358 u16 (*ndo_select_queue)(struct net_device *dev,
1359 struct sk_buff *skb,
1360 struct net_device *sb_dev);
1361 void (*ndo_change_rx_flags)(struct net_device *dev,
1362 int flags);
1363 void (*ndo_set_rx_mode)(struct net_device *dev);
1364 int (*ndo_set_mac_address)(struct net_device *dev,
1365 void *addr);
1366 int (*ndo_validate_addr)(struct net_device *dev);
1367 int (*ndo_do_ioctl)(struct net_device *dev,
1368 struct ifreq *ifr, int cmd);
1369 int (*ndo_eth_ioctl)(struct net_device *dev,
1370 struct ifreq *ifr, int cmd);
1371 int (*ndo_siocbond)(struct net_device *dev,
1372 struct ifreq *ifr, int cmd);
1373 int (*ndo_siocwandev)(struct net_device *dev,
1374 struct if_settings *ifs);
1375 int (*ndo_siocdevprivate)(struct net_device *dev,
1376 struct ifreq *ifr,
1377 void __user *data, int cmd);
1378 int (*ndo_set_config)(struct net_device *dev,
1379 struct ifmap *map);
1380 int (*ndo_change_mtu)(struct net_device *dev,
1381 int new_mtu);
1382 int (*ndo_neigh_setup)(struct net_device *dev,
1383 struct neigh_parms *);
1384 void (*ndo_tx_timeout) (struct net_device *dev,
1385 unsigned int txqueue);
1386
1387 void (*ndo_get_stats64)(struct net_device *dev,
1388 struct rtnl_link_stats64 *storage);
1389 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1390 int (*ndo_get_offload_stats)(int attr_id,
1391 const struct net_device *dev,
1392 void *attr_data);
1393 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1394
1395 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1396 __be16 proto, u16 vid);
1397 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1398 __be16 proto, u16 vid);
1399 #ifdef CONFIG_NET_POLL_CONTROLLER
1400 void (*ndo_poll_controller)(struct net_device *dev);
1401 int (*ndo_netpoll_setup)(struct net_device *dev,
1402 struct netpoll_info *info);
1403 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1404 #endif
1405 int (*ndo_set_vf_mac)(struct net_device *dev,
1406 int queue, u8 *mac);
1407 int (*ndo_set_vf_vlan)(struct net_device *dev,
1408 int queue, u16 vlan,
1409 u8 qos, __be16 proto);
1410 int (*ndo_set_vf_rate)(struct net_device *dev,
1411 int vf, int min_tx_rate,
1412 int max_tx_rate);
1413 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1414 int vf, bool setting);
1415 int (*ndo_set_vf_trust)(struct net_device *dev,
1416 int vf, bool setting);
1417 int (*ndo_get_vf_config)(struct net_device *dev,
1418 int vf,
1419 struct ifla_vf_info *ivf);
1420 int (*ndo_set_vf_link_state)(struct net_device *dev,
1421 int vf, int link_state);
1422 int (*ndo_get_vf_stats)(struct net_device *dev,
1423 int vf,
1424 struct ifla_vf_stats
1425 *vf_stats);
1426 int (*ndo_set_vf_port)(struct net_device *dev,
1427 int vf,
1428 struct nlattr *port[]);
1429 int (*ndo_get_vf_port)(struct net_device *dev,
1430 int vf, struct sk_buff *skb);
1431 int (*ndo_get_vf_guid)(struct net_device *dev,
1432 int vf,
1433 struct ifla_vf_guid *node_guid,
1434 struct ifla_vf_guid *port_guid);
1435 int (*ndo_set_vf_guid)(struct net_device *dev,
1436 int vf, u64 guid,
1437 int guid_type);
1438 int (*ndo_set_vf_rss_query_en)(
1439 struct net_device *dev,
1440 int vf, bool setting);
1441 int (*ndo_setup_tc)(struct net_device *dev,
1442 enum tc_setup_type type,
1443 void *type_data);
1444 #if IS_ENABLED(CONFIG_FCOE)
1445 int (*ndo_fcoe_enable)(struct net_device *dev);
1446 int (*ndo_fcoe_disable)(struct net_device *dev);
1447 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1448 u16 xid,
1449 struct scatterlist *sgl,
1450 unsigned int sgc);
1451 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1452 u16 xid);
1453 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1454 u16 xid,
1455 struct scatterlist *sgl,
1456 unsigned int sgc);
1457 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1458 struct netdev_fcoe_hbainfo *hbainfo);
1459 #endif
1460
1461 #if IS_ENABLED(CONFIG_LIBFCOE)
1462 #define NETDEV_FCOE_WWNN 0
1463 #define NETDEV_FCOE_WWPN 1
1464 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1465 u64 *wwn, int type);
1466 #endif
1467
1468 #ifdef CONFIG_RFS_ACCEL
1469 int (*ndo_rx_flow_steer)(struct net_device *dev,
1470 const struct sk_buff *skb,
1471 u16 rxq_index,
1472 u32 flow_id);
1473 #endif
1474 int (*ndo_add_slave)(struct net_device *dev,
1475 struct net_device *slave_dev,
1476 struct netlink_ext_ack *extack);
1477 int (*ndo_del_slave)(struct net_device *dev,
1478 struct net_device *slave_dev);
1479 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1480 struct sk_buff *skb,
1481 bool all_slaves);
1482 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1483 struct sock *sk);
1484 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1485 netdev_features_t features);
1486 int (*ndo_set_features)(struct net_device *dev,
1487 netdev_features_t features);
1488 int (*ndo_neigh_construct)(struct net_device *dev,
1489 struct neighbour *n);
1490 void (*ndo_neigh_destroy)(struct net_device *dev,
1491 struct neighbour *n);
1492
1493 int (*ndo_fdb_add)(struct ndmsg *ndm,
1494 struct nlattr *tb[],
1495 struct net_device *dev,
1496 const unsigned char *addr,
1497 u16 vid,
1498 u16 flags,
1499 struct netlink_ext_ack *extack);
1500 int (*ndo_fdb_del)(struct ndmsg *ndm,
1501 struct nlattr *tb[],
1502 struct net_device *dev,
1503 const unsigned char *addr,
1504 u16 vid);
1505 int (*ndo_fdb_dump)(struct sk_buff *skb,
1506 struct netlink_callback *cb,
1507 struct net_device *dev,
1508 struct net_device *filter_dev,
1509 int *idx);
1510 int (*ndo_fdb_get)(struct sk_buff *skb,
1511 struct nlattr *tb[],
1512 struct net_device *dev,
1513 const unsigned char *addr,
1514 u16 vid, u32 portid, u32 seq,
1515 struct netlink_ext_ack *extack);
1516 int (*ndo_bridge_setlink)(struct net_device *dev,
1517 struct nlmsghdr *nlh,
1518 u16 flags,
1519 struct netlink_ext_ack *extack);
1520 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1521 u32 pid, u32 seq,
1522 struct net_device *dev,
1523 u32 filter_mask,
1524 int nlflags);
1525 int (*ndo_bridge_dellink)(struct net_device *dev,
1526 struct nlmsghdr *nlh,
1527 u16 flags);
1528 int (*ndo_change_carrier)(struct net_device *dev,
1529 bool new_carrier);
1530 int (*ndo_get_phys_port_id)(struct net_device *dev,
1531 struct netdev_phys_item_id *ppid);
1532 int (*ndo_get_port_parent_id)(struct net_device *dev,
1533 struct netdev_phys_item_id *ppid);
1534 int (*ndo_get_phys_port_name)(struct net_device *dev,
1535 char *name, size_t len);
1536 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1537 struct net_device *dev);
1538 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1539 void *priv);
1540
1541 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1542 int queue_index,
1543 u32 maxrate);
1544 int (*ndo_get_iflink)(const struct net_device *dev);
1545 int (*ndo_change_proto_down)(struct net_device *dev,
1546 bool proto_down);
1547 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1548 struct sk_buff *skb);
1549 void (*ndo_set_rx_headroom)(struct net_device *dev,
1550 int needed_headroom);
1551 int (*ndo_bpf)(struct net_device *dev,
1552 struct netdev_bpf *bpf);
1553 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1554 struct xdp_frame **xdp,
1555 u32 flags);
1556 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1557 struct xdp_buff *xdp);
1558 int (*ndo_xsk_wakeup)(struct net_device *dev,
1559 u32 queue_id, u32 flags);
1560 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1561 int (*ndo_tunnel_ctl)(struct net_device *dev,
1562 struct ip_tunnel_parm *p, int cmd);
1563 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1564 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1565 struct net_device_path *path);
1566 };
1567
1568 /**
1569 * enum netdev_priv_flags - &struct net_device priv_flags
1570 *
1571 * These are the &struct net_device, they are only set internally
1572 * by drivers and used in the kernel. These flags are invisible to
1573 * userspace; this means that the order of these flags can change
1574 * during any kernel release.
1575 *
1576 * You should have a pretty good reason to be extending these flags.
1577 *
1578 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1579 * @IFF_EBRIDGE: Ethernet bridging device
1580 * @IFF_BONDING: bonding master or slave
1581 * @IFF_ISATAP: ISATAP interface (RFC4214)
1582 * @IFF_WAN_HDLC: WAN HDLC device
1583 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1584 * release skb->dst
1585 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1586 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1587 * @IFF_MACVLAN_PORT: device used as macvlan port
1588 * @IFF_BRIDGE_PORT: device used as bridge port
1589 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1590 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1591 * @IFF_UNICAST_FLT: Supports unicast filtering
1592 * @IFF_TEAM_PORT: device used as team port
1593 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1594 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1595 * change when it's running
1596 * @IFF_MACVLAN: Macvlan device
1597 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1598 * underlying stacked devices
1599 * @IFF_L3MDEV_MASTER: device is an L3 master device
1600 * @IFF_NO_QUEUE: device can run without qdisc attached
1601 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1602 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1603 * @IFF_TEAM: device is a team device
1604 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1605 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1606 * entity (i.e. the master device for bridged veth)
1607 * @IFF_MACSEC: device is a MACsec device
1608 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1609 * @IFF_FAILOVER: device is a failover master device
1610 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1611 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1612 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1613 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1614 * skb_headlen(skb) == 0 (data starts from frag0)
1615 */
1616 enum netdev_priv_flags {
1617 IFF_802_1Q_VLAN = 1<<0,
1618 IFF_EBRIDGE = 1<<1,
1619 IFF_BONDING = 1<<2,
1620 IFF_ISATAP = 1<<3,
1621 IFF_WAN_HDLC = 1<<4,
1622 IFF_XMIT_DST_RELEASE = 1<<5,
1623 IFF_DONT_BRIDGE = 1<<6,
1624 IFF_DISABLE_NETPOLL = 1<<7,
1625 IFF_MACVLAN_PORT = 1<<8,
1626 IFF_BRIDGE_PORT = 1<<9,
1627 IFF_OVS_DATAPATH = 1<<10,
1628 IFF_TX_SKB_SHARING = 1<<11,
1629 IFF_UNICAST_FLT = 1<<12,
1630 IFF_TEAM_PORT = 1<<13,
1631 IFF_SUPP_NOFCS = 1<<14,
1632 IFF_LIVE_ADDR_CHANGE = 1<<15,
1633 IFF_MACVLAN = 1<<16,
1634 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1635 IFF_L3MDEV_MASTER = 1<<18,
1636 IFF_NO_QUEUE = 1<<19,
1637 IFF_OPENVSWITCH = 1<<20,
1638 IFF_L3MDEV_SLAVE = 1<<21,
1639 IFF_TEAM = 1<<22,
1640 IFF_RXFH_CONFIGURED = 1<<23,
1641 IFF_PHONY_HEADROOM = 1<<24,
1642 IFF_MACSEC = 1<<25,
1643 IFF_NO_RX_HANDLER = 1<<26,
1644 IFF_FAILOVER = 1<<27,
1645 IFF_FAILOVER_SLAVE = 1<<28,
1646 IFF_L3MDEV_RX_HANDLER = 1<<29,
1647 IFF_LIVE_RENAME_OK = 1<<30,
1648 IFF_TX_SKB_NO_LINEAR = 1<<31,
1649 };
1650
1651 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1652 #define IFF_EBRIDGE IFF_EBRIDGE
1653 #define IFF_BONDING IFF_BONDING
1654 #define IFF_ISATAP IFF_ISATAP
1655 #define IFF_WAN_HDLC IFF_WAN_HDLC
1656 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1657 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1658 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1659 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1660 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1661 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1662 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1663 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1664 #define IFF_TEAM_PORT IFF_TEAM_PORT
1665 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1666 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1667 #define IFF_MACVLAN IFF_MACVLAN
1668 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1669 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1670 #define IFF_NO_QUEUE IFF_NO_QUEUE
1671 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1672 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1673 #define IFF_TEAM IFF_TEAM
1674 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1675 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1676 #define IFF_MACSEC IFF_MACSEC
1677 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1678 #define IFF_FAILOVER IFF_FAILOVER
1679 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1680 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1681 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1682 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1683
1684 /* Specifies the type of the struct net_device::ml_priv pointer */
1685 enum netdev_ml_priv_type {
1686 ML_PRIV_NONE,
1687 ML_PRIV_CAN,
1688 };
1689
1690 /**
1691 * struct net_device - The DEVICE structure.
1692 *
1693 * Actually, this whole structure is a big mistake. It mixes I/O
1694 * data with strictly "high-level" data, and it has to know about
1695 * almost every data structure used in the INET module.
1696 *
1697 * @name: This is the first field of the "visible" part of this structure
1698 * (i.e. as seen by users in the "Space.c" file). It is the name
1699 * of the interface.
1700 *
1701 * @name_node: Name hashlist node
1702 * @ifalias: SNMP alias
1703 * @mem_end: Shared memory end
1704 * @mem_start: Shared memory start
1705 * @base_addr: Device I/O address
1706 * @irq: Device IRQ number
1707 *
1708 * @state: Generic network queuing layer state, see netdev_state_t
1709 * @dev_list: The global list of network devices
1710 * @napi_list: List entry used for polling NAPI devices
1711 * @unreg_list: List entry when we are unregistering the
1712 * device; see the function unregister_netdev
1713 * @close_list: List entry used when we are closing the device
1714 * @ptype_all: Device-specific packet handlers for all protocols
1715 * @ptype_specific: Device-specific, protocol-specific packet handlers
1716 *
1717 * @adj_list: Directly linked devices, like slaves for bonding
1718 * @features: Currently active device features
1719 * @hw_features: User-changeable features
1720 *
1721 * @wanted_features: User-requested features
1722 * @vlan_features: Mask of features inheritable by VLAN devices
1723 *
1724 * @hw_enc_features: Mask of features inherited by encapsulating devices
1725 * This field indicates what encapsulation
1726 * offloads the hardware is capable of doing,
1727 * and drivers will need to set them appropriately.
1728 *
1729 * @mpls_features: Mask of features inheritable by MPLS
1730 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1731 *
1732 * @ifindex: interface index
1733 * @group: The group the device belongs to
1734 *
1735 * @stats: Statistics struct, which was left as a legacy, use
1736 * rtnl_link_stats64 instead
1737 *
1738 * @rx_dropped: Dropped packets by core network,
1739 * do not use this in drivers
1740 * @tx_dropped: Dropped packets by core network,
1741 * do not use this in drivers
1742 * @rx_nohandler: nohandler dropped packets by core network on
1743 * inactive devices, do not use this in drivers
1744 * @carrier_up_count: Number of times the carrier has been up
1745 * @carrier_down_count: Number of times the carrier has been down
1746 *
1747 * @wireless_handlers: List of functions to handle Wireless Extensions,
1748 * instead of ioctl,
1749 * see <net/iw_handler.h> for details.
1750 * @wireless_data: Instance data managed by the core of wireless extensions
1751 *
1752 * @netdev_ops: Includes several pointers to callbacks,
1753 * if one wants to override the ndo_*() functions
1754 * @ethtool_ops: Management operations
1755 * @l3mdev_ops: Layer 3 master device operations
1756 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1757 * discovery handling. Necessary for e.g. 6LoWPAN.
1758 * @xfrmdev_ops: Transformation offload operations
1759 * @tlsdev_ops: Transport Layer Security offload operations
1760 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1761 * of Layer 2 headers.
1762 *
1763 * @flags: Interface flags (a la BSD)
1764 * @priv_flags: Like 'flags' but invisible to userspace,
1765 * see if.h for the definitions
1766 * @gflags: Global flags ( kept as legacy )
1767 * @padded: How much padding added by alloc_netdev()
1768 * @operstate: RFC2863 operstate
1769 * @link_mode: Mapping policy to operstate
1770 * @if_port: Selectable AUI, TP, ...
1771 * @dma: DMA channel
1772 * @mtu: Interface MTU value
1773 * @min_mtu: Interface Minimum MTU value
1774 * @max_mtu: Interface Maximum MTU value
1775 * @type: Interface hardware type
1776 * @hard_header_len: Maximum hardware header length.
1777 * @min_header_len: Minimum hardware header length
1778 *
1779 * @needed_headroom: Extra headroom the hardware may need, but not in all
1780 * cases can this be guaranteed
1781 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1782 * cases can this be guaranteed. Some cases also use
1783 * LL_MAX_HEADER instead to allocate the skb
1784 *
1785 * interface address info:
1786 *
1787 * @perm_addr: Permanent hw address
1788 * @addr_assign_type: Hw address assignment type
1789 * @addr_len: Hardware address length
1790 * @upper_level: Maximum depth level of upper devices.
1791 * @lower_level: Maximum depth level of lower devices.
1792 * @neigh_priv_len: Used in neigh_alloc()
1793 * @dev_id: Used to differentiate devices that share
1794 * the same link layer address
1795 * @dev_port: Used to differentiate devices that share
1796 * the same function
1797 * @addr_list_lock: XXX: need comments on this one
1798 * @name_assign_type: network interface name assignment type
1799 * @uc_promisc: Counter that indicates promiscuous mode
1800 * has been enabled due to the need to listen to
1801 * additional unicast addresses in a device that
1802 * does not implement ndo_set_rx_mode()
1803 * @uc: unicast mac addresses
1804 * @mc: multicast mac addresses
1805 * @dev_addrs: list of device hw addresses
1806 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1807 * @promiscuity: Number of times the NIC is told to work in
1808 * promiscuous mode; if it becomes 0 the NIC will
1809 * exit promiscuous mode
1810 * @allmulti: Counter, enables or disables allmulticast mode
1811 *
1812 * @vlan_info: VLAN info
1813 * @dsa_ptr: dsa specific data
1814 * @tipc_ptr: TIPC specific data
1815 * @atalk_ptr: AppleTalk link
1816 * @ip_ptr: IPv4 specific data
1817 * @dn_ptr: DECnet specific data
1818 * @ip6_ptr: IPv6 specific data
1819 * @ax25_ptr: AX.25 specific data
1820 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1821 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1822 * device struct
1823 * @mpls_ptr: mpls_dev struct pointer
1824 * @mctp_ptr: MCTP specific data
1825 *
1826 * @dev_addr: Hw address (before bcast,
1827 * because most packets are unicast)
1828 *
1829 * @_rx: Array of RX queues
1830 * @num_rx_queues: Number of RX queues
1831 * allocated at register_netdev() time
1832 * @real_num_rx_queues: Number of RX queues currently active in device
1833 * @xdp_prog: XDP sockets filter program pointer
1834 * @gro_flush_timeout: timeout for GRO layer in NAPI
1835 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1836 * allow to avoid NIC hard IRQ, on busy queues.
1837 *
1838 * @rx_handler: handler for received packets
1839 * @rx_handler_data: XXX: need comments on this one
1840 * @miniq_ingress: ingress/clsact qdisc specific data for
1841 * ingress processing
1842 * @ingress_queue: XXX: need comments on this one
1843 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1844 * @broadcast: hw bcast address
1845 *
1846 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1847 * indexed by RX queue number. Assigned by driver.
1848 * This must only be set if the ndo_rx_flow_steer
1849 * operation is defined
1850 * @index_hlist: Device index hash chain
1851 *
1852 * @_tx: Array of TX queues
1853 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1854 * @real_num_tx_queues: Number of TX queues currently active in device
1855 * @qdisc: Root qdisc from userspace point of view
1856 * @tx_queue_len: Max frames per queue allowed
1857 * @tx_global_lock: XXX: need comments on this one
1858 * @xdp_bulkq: XDP device bulk queue
1859 * @xps_maps: all CPUs/RXQs maps for XPS device
1860 *
1861 * @xps_maps: XXX: need comments on this one
1862 * @miniq_egress: clsact qdisc specific data for
1863 * egress processing
1864 * @qdisc_hash: qdisc hash table
1865 * @watchdog_timeo: Represents the timeout that is used by
1866 * the watchdog (see dev_watchdog())
1867 * @watchdog_timer: List of timers
1868 *
1869 * @proto_down_reason: reason a netdev interface is held down
1870 * @pcpu_refcnt: Number of references to this device
1871 * @dev_refcnt: Number of references to this device
1872 * @todo_list: Delayed register/unregister
1873 * @link_watch_list: XXX: need comments on this one
1874 *
1875 * @reg_state: Register/unregister state machine
1876 * @dismantle: Device is going to be freed
1877 * @rtnl_link_state: This enum represents the phases of creating
1878 * a new link
1879 *
1880 * @needs_free_netdev: Should unregister perform free_netdev?
1881 * @priv_destructor: Called from unregister
1882 * @npinfo: XXX: need comments on this one
1883 * @nd_net: Network namespace this network device is inside
1884 *
1885 * @ml_priv: Mid-layer private
1886 * @ml_priv_type: Mid-layer private type
1887 * @lstats: Loopback statistics
1888 * @tstats: Tunnel statistics
1889 * @dstats: Dummy statistics
1890 * @vstats: Virtual ethernet statistics
1891 *
1892 * @garp_port: GARP
1893 * @mrp_port: MRP
1894 *
1895 * @dev: Class/net/name entry
1896 * @sysfs_groups: Space for optional device, statistics and wireless
1897 * sysfs groups
1898 *
1899 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1900 * @rtnl_link_ops: Rtnl_link_ops
1901 *
1902 * @gso_max_size: Maximum size of generic segmentation offload
1903 * @gso_max_segs: Maximum number of segments that can be passed to the
1904 * NIC for GSO
1905 *
1906 * @dcbnl_ops: Data Center Bridging netlink ops
1907 * @num_tc: Number of traffic classes in the net device
1908 * @tc_to_txq: XXX: need comments on this one
1909 * @prio_tc_map: XXX: need comments on this one
1910 *
1911 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1912 *
1913 * @priomap: XXX: need comments on this one
1914 * @phydev: Physical device may attach itself
1915 * for hardware timestamping
1916 * @sfp_bus: attached &struct sfp_bus structure.
1917 *
1918 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1919 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1920 *
1921 * @proto_down: protocol port state information can be sent to the
1922 * switch driver and used to set the phys state of the
1923 * switch port.
1924 *
1925 * @wol_enabled: Wake-on-LAN is enabled
1926 *
1927 * @threaded: napi threaded mode is enabled
1928 *
1929 * @net_notifier_list: List of per-net netdev notifier block
1930 * that follow this device when it is moved
1931 * to another network namespace.
1932 *
1933 * @macsec_ops: MACsec offloading ops
1934 *
1935 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1936 * offload capabilities of the device
1937 * @udp_tunnel_nic: UDP tunnel offload state
1938 * @xdp_state: stores info on attached XDP BPF programs
1939 *
1940 * @nested_level: Used as as a parameter of spin_lock_nested() of
1941 * dev->addr_list_lock.
1942 * @unlink_list: As netif_addr_lock() can be called recursively,
1943 * keep a list of interfaces to be deleted.
1944 *
1945 * FIXME: cleanup struct net_device such that network protocol info
1946 * moves out.
1947 */
1948
1949 struct net_device {
1950 char name[IFNAMSIZ];
1951 struct netdev_name_node *name_node;
1952 struct dev_ifalias __rcu *ifalias;
1953 /*
1954 * I/O specific fields
1955 * FIXME: Merge these and struct ifmap into one
1956 */
1957 unsigned long mem_end;
1958 unsigned long mem_start;
1959 unsigned long base_addr;
1960
1961 /*
1962 * Some hardware also needs these fields (state,dev_list,
1963 * napi_list,unreg_list,close_list) but they are not
1964 * part of the usual set specified in Space.c.
1965 */
1966
1967 unsigned long state;
1968
1969 struct list_head dev_list;
1970 struct list_head napi_list;
1971 struct list_head unreg_list;
1972 struct list_head close_list;
1973 struct list_head ptype_all;
1974 struct list_head ptype_specific;
1975
1976 struct {
1977 struct list_head upper;
1978 struct list_head lower;
1979 } adj_list;
1980
1981 /* Read-mostly cache-line for fast-path access */
1982 unsigned int flags;
1983 unsigned int priv_flags;
1984 const struct net_device_ops *netdev_ops;
1985 int ifindex;
1986 unsigned short gflags;
1987 unsigned short hard_header_len;
1988
1989 /* Note : dev->mtu is often read without holding a lock.
1990 * Writers usually hold RTNL.
1991 * It is recommended to use READ_ONCE() to annotate the reads,
1992 * and to use WRITE_ONCE() to annotate the writes.
1993 */
1994 unsigned int mtu;
1995 unsigned short needed_headroom;
1996 unsigned short needed_tailroom;
1997
1998 netdev_features_t features;
1999 netdev_features_t hw_features;
2000 netdev_features_t wanted_features;
2001 netdev_features_t vlan_features;
2002 netdev_features_t hw_enc_features;
2003 netdev_features_t mpls_features;
2004 netdev_features_t gso_partial_features;
2005
2006 unsigned int min_mtu;
2007 unsigned int max_mtu;
2008 unsigned short type;
2009 unsigned char min_header_len;
2010 unsigned char name_assign_type;
2011
2012 int group;
2013
2014 struct net_device_stats stats; /* not used by modern drivers */
2015
2016 atomic_long_t rx_dropped;
2017 atomic_long_t tx_dropped;
2018 atomic_long_t rx_nohandler;
2019
2020 /* Stats to monitor link on/off, flapping */
2021 atomic_t carrier_up_count;
2022 atomic_t carrier_down_count;
2023
2024 #ifdef CONFIG_WIRELESS_EXT
2025 const struct iw_handler_def *wireless_handlers;
2026 struct iw_public_data *wireless_data;
2027 #endif
2028 const struct ethtool_ops *ethtool_ops;
2029 #ifdef CONFIG_NET_L3_MASTER_DEV
2030 const struct l3mdev_ops *l3mdev_ops;
2031 #endif
2032 #if IS_ENABLED(CONFIG_IPV6)
2033 const struct ndisc_ops *ndisc_ops;
2034 #endif
2035
2036 #ifdef CONFIG_XFRM_OFFLOAD
2037 const struct xfrmdev_ops *xfrmdev_ops;
2038 #endif
2039
2040 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2041 const struct tlsdev_ops *tlsdev_ops;
2042 #endif
2043
2044 const struct header_ops *header_ops;
2045
2046 unsigned char operstate;
2047 unsigned char link_mode;
2048
2049 unsigned char if_port;
2050 unsigned char dma;
2051
2052 /* Interface address info. */
2053 unsigned char perm_addr[MAX_ADDR_LEN];
2054 unsigned char addr_assign_type;
2055 unsigned char addr_len;
2056 unsigned char upper_level;
2057 unsigned char lower_level;
2058
2059 unsigned short neigh_priv_len;
2060 unsigned short dev_id;
2061 unsigned short dev_port;
2062 unsigned short padded;
2063
2064 spinlock_t addr_list_lock;
2065 int irq;
2066
2067 struct netdev_hw_addr_list uc;
2068 struct netdev_hw_addr_list mc;
2069 struct netdev_hw_addr_list dev_addrs;
2070
2071 #ifdef CONFIG_SYSFS
2072 struct kset *queues_kset;
2073 #endif
2074 #ifdef CONFIG_LOCKDEP
2075 struct list_head unlink_list;
2076 #endif
2077 unsigned int promiscuity;
2078 unsigned int allmulti;
2079 bool uc_promisc;
2080 #ifdef CONFIG_LOCKDEP
2081 unsigned char nested_level;
2082 #endif
2083
2084
2085 /* Protocol-specific pointers */
2086
2087 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2088 struct vlan_info __rcu *vlan_info;
2089 #endif
2090 #if IS_ENABLED(CONFIG_NET_DSA)
2091 struct dsa_port *dsa_ptr;
2092 #endif
2093 #if IS_ENABLED(CONFIG_TIPC)
2094 struct tipc_bearer __rcu *tipc_ptr;
2095 #endif
2096 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2097 void *atalk_ptr;
2098 #endif
2099 struct in_device __rcu *ip_ptr;
2100 #if IS_ENABLED(CONFIG_DECNET)
2101 struct dn_dev __rcu *dn_ptr;
2102 #endif
2103 struct inet6_dev __rcu *ip6_ptr;
2104 #if IS_ENABLED(CONFIG_AX25)
2105 void *ax25_ptr;
2106 #endif
2107 struct wireless_dev *ieee80211_ptr;
2108 struct wpan_dev *ieee802154_ptr;
2109 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2110 struct mpls_dev __rcu *mpls_ptr;
2111 #endif
2112 #if IS_ENABLED(CONFIG_MCTP)
2113 struct mctp_dev __rcu *mctp_ptr;
2114 #endif
2115
2116 /*
2117 * Cache lines mostly used on receive path (including eth_type_trans())
2118 */
2119 /* Interface address info used in eth_type_trans() */
2120 unsigned char *dev_addr;
2121
2122 struct netdev_rx_queue *_rx;
2123 unsigned int num_rx_queues;
2124 unsigned int real_num_rx_queues;
2125
2126 struct bpf_prog __rcu *xdp_prog;
2127 unsigned long gro_flush_timeout;
2128 int napi_defer_hard_irqs;
2129 rx_handler_func_t __rcu *rx_handler;
2130 void __rcu *rx_handler_data;
2131
2132 #ifdef CONFIG_NET_CLS_ACT
2133 struct mini_Qdisc __rcu *miniq_ingress;
2134 #endif
2135 struct netdev_queue __rcu *ingress_queue;
2136 #ifdef CONFIG_NETFILTER_INGRESS
2137 struct nf_hook_entries __rcu *nf_hooks_ingress;
2138 #endif
2139
2140 unsigned char broadcast[MAX_ADDR_LEN];
2141 #ifdef CONFIG_RFS_ACCEL
2142 struct cpu_rmap *rx_cpu_rmap;
2143 #endif
2144 struct hlist_node index_hlist;
2145
2146 /*
2147 * Cache lines mostly used on transmit path
2148 */
2149 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2150 unsigned int num_tx_queues;
2151 unsigned int real_num_tx_queues;
2152 struct Qdisc *qdisc;
2153 unsigned int tx_queue_len;
2154 spinlock_t tx_global_lock;
2155
2156 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2157
2158 #ifdef CONFIG_XPS
2159 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2160 #endif
2161 #ifdef CONFIG_NET_CLS_ACT
2162 struct mini_Qdisc __rcu *miniq_egress;
2163 #endif
2164
2165 #ifdef CONFIG_NET_SCHED
2166 DECLARE_HASHTABLE (qdisc_hash, 4);
2167 #endif
2168 /* These may be needed for future network-power-down code. */
2169 struct timer_list watchdog_timer;
2170 int watchdog_timeo;
2171
2172 u32 proto_down_reason;
2173
2174 struct list_head todo_list;
2175
2176 #ifdef CONFIG_PCPU_DEV_REFCNT
2177 int __percpu *pcpu_refcnt;
2178 #else
2179 refcount_t dev_refcnt;
2180 #endif
2181
2182 struct list_head link_watch_list;
2183
2184 enum { NETREG_UNINITIALIZED=0,
2185 NETREG_REGISTERED, /* completed register_netdevice */
2186 NETREG_UNREGISTERING, /* called unregister_netdevice */
2187 NETREG_UNREGISTERED, /* completed unregister todo */
2188 NETREG_RELEASED, /* called free_netdev */
2189 NETREG_DUMMY, /* dummy device for NAPI poll */
2190 } reg_state:8;
2191
2192 bool dismantle;
2193
2194 enum {
2195 RTNL_LINK_INITIALIZED,
2196 RTNL_LINK_INITIALIZING,
2197 } rtnl_link_state:16;
2198
2199 bool needs_free_netdev;
2200 void (*priv_destructor)(struct net_device *dev);
2201
2202 #ifdef CONFIG_NETPOLL
2203 struct netpoll_info __rcu *npinfo;
2204 #endif
2205
2206 possible_net_t nd_net;
2207
2208 /* mid-layer private */
2209 void *ml_priv;
2210 enum netdev_ml_priv_type ml_priv_type;
2211
2212 union {
2213 struct pcpu_lstats __percpu *lstats;
2214 struct pcpu_sw_netstats __percpu *tstats;
2215 struct pcpu_dstats __percpu *dstats;
2216 };
2217
2218 #if IS_ENABLED(CONFIG_GARP)
2219 struct garp_port __rcu *garp_port;
2220 #endif
2221 #if IS_ENABLED(CONFIG_MRP)
2222 struct mrp_port __rcu *mrp_port;
2223 #endif
2224
2225 struct device dev;
2226 const struct attribute_group *sysfs_groups[4];
2227 const struct attribute_group *sysfs_rx_queue_group;
2228
2229 const struct rtnl_link_ops *rtnl_link_ops;
2230
2231 /* for setting kernel sock attribute on TCP connection setup */
2232 #define GSO_MAX_SIZE 65536
2233 unsigned int gso_max_size;
2234 #define GSO_MAX_SEGS 65535
2235 u16 gso_max_segs;
2236
2237 #ifdef CONFIG_DCB
2238 const struct dcbnl_rtnl_ops *dcbnl_ops;
2239 #endif
2240 s16 num_tc;
2241 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2242 u8 prio_tc_map[TC_BITMASK + 1];
2243
2244 #if IS_ENABLED(CONFIG_FCOE)
2245 unsigned int fcoe_ddp_xid;
2246 #endif
2247 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2248 struct netprio_map __rcu *priomap;
2249 #endif
2250 struct phy_device *phydev;
2251 struct sfp_bus *sfp_bus;
2252 struct lock_class_key *qdisc_tx_busylock;
2253 struct lock_class_key *qdisc_running_key;
2254 bool proto_down;
2255 unsigned wol_enabled:1;
2256 unsigned threaded:1;
2257
2258 struct list_head net_notifier_list;
2259
2260 #if IS_ENABLED(CONFIG_MACSEC)
2261 /* MACsec management functions */
2262 const struct macsec_ops *macsec_ops;
2263 #endif
2264 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2265 struct udp_tunnel_nic *udp_tunnel_nic;
2266
2267 /* protected by rtnl_lock */
2268 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2269 };
2270 #define to_net_dev(d) container_of(d, struct net_device, dev)
2271
netif_elide_gro(const struct net_device * dev)2272 static inline bool netif_elide_gro(const struct net_device *dev)
2273 {
2274 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2275 return true;
2276 return false;
2277 }
2278
2279 #define NETDEV_ALIGN 32
2280
2281 static inline
netdev_get_prio_tc_map(const struct net_device * dev,u32 prio)2282 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2283 {
2284 return dev->prio_tc_map[prio & TC_BITMASK];
2285 }
2286
2287 static inline
netdev_set_prio_tc_map(struct net_device * dev,u8 prio,u8 tc)2288 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2289 {
2290 if (tc >= dev->num_tc)
2291 return -EINVAL;
2292
2293 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2294 return 0;
2295 }
2296
2297 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2298 void netdev_reset_tc(struct net_device *dev);
2299 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2300 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2301
2302 static inline
netdev_get_num_tc(struct net_device * dev)2303 int netdev_get_num_tc(struct net_device *dev)
2304 {
2305 return dev->num_tc;
2306 }
2307
net_prefetch(void * p)2308 static inline void net_prefetch(void *p)
2309 {
2310 prefetch(p);
2311 #if L1_CACHE_BYTES < 128
2312 prefetch((u8 *)p + L1_CACHE_BYTES);
2313 #endif
2314 }
2315
net_prefetchw(void * p)2316 static inline void net_prefetchw(void *p)
2317 {
2318 prefetchw(p);
2319 #if L1_CACHE_BYTES < 128
2320 prefetchw((u8 *)p + L1_CACHE_BYTES);
2321 #endif
2322 }
2323
2324 void netdev_unbind_sb_channel(struct net_device *dev,
2325 struct net_device *sb_dev);
2326 int netdev_bind_sb_channel_queue(struct net_device *dev,
2327 struct net_device *sb_dev,
2328 u8 tc, u16 count, u16 offset);
2329 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
netdev_get_sb_channel(struct net_device * dev)2330 static inline int netdev_get_sb_channel(struct net_device *dev)
2331 {
2332 return max_t(int, -dev->num_tc, 0);
2333 }
2334
2335 static inline
netdev_get_tx_queue(const struct net_device * dev,unsigned int index)2336 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2337 unsigned int index)
2338 {
2339 return &dev->_tx[index];
2340 }
2341
skb_get_tx_queue(const struct net_device * dev,const struct sk_buff * skb)2342 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2343 const struct sk_buff *skb)
2344 {
2345 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2346 }
2347
netdev_for_each_tx_queue(struct net_device * dev,void (* f)(struct net_device *,struct netdev_queue *,void *),void * arg)2348 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2349 void (*f)(struct net_device *,
2350 struct netdev_queue *,
2351 void *),
2352 void *arg)
2353 {
2354 unsigned int i;
2355
2356 for (i = 0; i < dev->num_tx_queues; i++)
2357 f(dev, &dev->_tx[i], arg);
2358 }
2359
2360 #define netdev_lockdep_set_classes(dev) \
2361 { \
2362 static struct lock_class_key qdisc_tx_busylock_key; \
2363 static struct lock_class_key qdisc_running_key; \
2364 static struct lock_class_key qdisc_xmit_lock_key; \
2365 static struct lock_class_key dev_addr_list_lock_key; \
2366 unsigned int i; \
2367 \
2368 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2369 (dev)->qdisc_running_key = &qdisc_running_key; \
2370 lockdep_set_class(&(dev)->addr_list_lock, \
2371 &dev_addr_list_lock_key); \
2372 for (i = 0; i < (dev)->num_tx_queues; i++) \
2373 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2374 &qdisc_xmit_lock_key); \
2375 }
2376
2377 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2378 struct net_device *sb_dev);
2379 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2380 struct sk_buff *skb,
2381 struct net_device *sb_dev);
2382
2383 /* returns the headroom that the master device needs to take in account
2384 * when forwarding to this dev
2385 */
netdev_get_fwd_headroom(struct net_device * dev)2386 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2387 {
2388 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2389 }
2390
netdev_set_rx_headroom(struct net_device * dev,int new_hr)2391 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2392 {
2393 if (dev->netdev_ops->ndo_set_rx_headroom)
2394 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2395 }
2396
2397 /* set the device rx headroom to the dev's default */
netdev_reset_rx_headroom(struct net_device * dev)2398 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2399 {
2400 netdev_set_rx_headroom(dev, -1);
2401 }
2402
netdev_get_ml_priv(struct net_device * dev,enum netdev_ml_priv_type type)2403 static inline void *netdev_get_ml_priv(struct net_device *dev,
2404 enum netdev_ml_priv_type type)
2405 {
2406 if (dev->ml_priv_type != type)
2407 return NULL;
2408
2409 return dev->ml_priv;
2410 }
2411
netdev_set_ml_priv(struct net_device * dev,void * ml_priv,enum netdev_ml_priv_type type)2412 static inline void netdev_set_ml_priv(struct net_device *dev,
2413 void *ml_priv,
2414 enum netdev_ml_priv_type type)
2415 {
2416 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2417 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2418 dev->ml_priv_type, type);
2419 WARN(!dev->ml_priv_type && dev->ml_priv,
2420 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2421
2422 dev->ml_priv = ml_priv;
2423 dev->ml_priv_type = type;
2424 }
2425
2426 /*
2427 * Net namespace inlines
2428 */
2429 static inline
dev_net(const struct net_device * dev)2430 struct net *dev_net(const struct net_device *dev)
2431 {
2432 return read_pnet(&dev->nd_net);
2433 }
2434
2435 static inline
dev_net_set(struct net_device * dev,struct net * net)2436 void dev_net_set(struct net_device *dev, struct net *net)
2437 {
2438 write_pnet(&dev->nd_net, net);
2439 }
2440
2441 /**
2442 * netdev_priv - access network device private data
2443 * @dev: network device
2444 *
2445 * Get network device private data
2446 */
netdev_priv(const struct net_device * dev)2447 static inline void *netdev_priv(const struct net_device *dev)
2448 {
2449 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2450 }
2451
2452 /* Set the sysfs physical device reference for the network logical device
2453 * if set prior to registration will cause a symlink during initialization.
2454 */
2455 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2456
2457 /* Set the sysfs device type for the network logical device to allow
2458 * fine-grained identification of different network device types. For
2459 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2460 */
2461 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2462
2463 /* Default NAPI poll() weight
2464 * Device drivers are strongly advised to not use bigger value
2465 */
2466 #define NAPI_POLL_WEIGHT 64
2467
2468 /**
2469 * netif_napi_add - initialize a NAPI context
2470 * @dev: network device
2471 * @napi: NAPI context
2472 * @poll: polling function
2473 * @weight: default weight
2474 *
2475 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2476 * *any* of the other NAPI-related functions.
2477 */
2478 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2479 int (*poll)(struct napi_struct *, int), int weight);
2480
2481 /**
2482 * netif_tx_napi_add - initialize a NAPI context
2483 * @dev: network device
2484 * @napi: NAPI context
2485 * @poll: polling function
2486 * @weight: default weight
2487 *
2488 * This variant of netif_napi_add() should be used from drivers using NAPI
2489 * to exclusively poll a TX queue.
2490 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2491 */
netif_tx_napi_add(struct net_device * dev,struct napi_struct * napi,int (* poll)(struct napi_struct *,int),int weight)2492 static inline void netif_tx_napi_add(struct net_device *dev,
2493 struct napi_struct *napi,
2494 int (*poll)(struct napi_struct *, int),
2495 int weight)
2496 {
2497 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2498 netif_napi_add(dev, napi, poll, weight);
2499 }
2500
2501 /**
2502 * __netif_napi_del - remove a NAPI context
2503 * @napi: NAPI context
2504 *
2505 * Warning: caller must observe RCU grace period before freeing memory
2506 * containing @napi. Drivers might want to call this helper to combine
2507 * all the needed RCU grace periods into a single one.
2508 */
2509 void __netif_napi_del(struct napi_struct *napi);
2510
2511 /**
2512 * netif_napi_del - remove a NAPI context
2513 * @napi: NAPI context
2514 *
2515 * netif_napi_del() removes a NAPI context from the network device NAPI list
2516 */
netif_napi_del(struct napi_struct * napi)2517 static inline void netif_napi_del(struct napi_struct *napi)
2518 {
2519 __netif_napi_del(napi);
2520 synchronize_net();
2521 }
2522
2523 struct napi_gro_cb {
2524 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2525 void *frag0;
2526
2527 /* Length of frag0. */
2528 unsigned int frag0_len;
2529
2530 /* This indicates where we are processing relative to skb->data. */
2531 int data_offset;
2532
2533 /* This is non-zero if the packet cannot be merged with the new skb. */
2534 u16 flush;
2535
2536 /* Save the IP ID here and check when we get to the transport layer */
2537 u16 flush_id;
2538
2539 /* Number of segments aggregated. */
2540 u16 count;
2541
2542 /* Start offset for remote checksum offload */
2543 u16 gro_remcsum_start;
2544
2545 /* jiffies when first packet was created/queued */
2546 unsigned long age;
2547
2548 /* Used in ipv6_gro_receive() and foo-over-udp */
2549 u16 proto;
2550
2551 /* This is non-zero if the packet may be of the same flow. */
2552 u8 same_flow:1;
2553
2554 /* Used in tunnel GRO receive */
2555 u8 encap_mark:1;
2556
2557 /* GRO checksum is valid */
2558 u8 csum_valid:1;
2559
2560 /* Number of checksums via CHECKSUM_UNNECESSARY */
2561 u8 csum_cnt:3;
2562
2563 /* Free the skb? */
2564 u8 free:2;
2565 #define NAPI_GRO_FREE 1
2566 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2567
2568 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2569 u8 is_ipv6:1;
2570
2571 /* Used in GRE, set in fou/gue_gro_receive */
2572 u8 is_fou:1;
2573
2574 /* Used to determine if flush_id can be ignored */
2575 u8 is_atomic:1;
2576
2577 /* Number of gro_receive callbacks this packet already went through */
2578 u8 recursion_counter:4;
2579
2580 /* GRO is done by frag_list pointer chaining. */
2581 u8 is_flist:1;
2582
2583 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2584 __wsum csum;
2585
2586 /* used in skb_gro_receive() slow path */
2587 struct sk_buff *last;
2588 };
2589
2590 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2591
2592 #define GRO_RECURSION_LIMIT 15
gro_recursion_inc_test(struct sk_buff * skb)2593 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2594 {
2595 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2596 }
2597
2598 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2599 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2600 struct list_head *head,
2601 struct sk_buff *skb)
2602 {
2603 if (unlikely(gro_recursion_inc_test(skb))) {
2604 NAPI_GRO_CB(skb)->flush |= 1;
2605 return NULL;
2606 }
2607
2608 return cb(head, skb);
2609 }
2610
2611 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2612 struct sk_buff *);
2613 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2614 struct sock *sk,
2615 struct list_head *head,
2616 struct sk_buff *skb)
2617 {
2618 if (unlikely(gro_recursion_inc_test(skb))) {
2619 NAPI_GRO_CB(skb)->flush |= 1;
2620 return NULL;
2621 }
2622
2623 return cb(sk, head, skb);
2624 }
2625
2626 struct packet_type {
2627 __be16 type; /* This is really htons(ether_type). */
2628 bool ignore_outgoing;
2629 struct net_device *dev; /* NULL is wildcarded here */
2630 int (*func) (struct sk_buff *,
2631 struct net_device *,
2632 struct packet_type *,
2633 struct net_device *);
2634 void (*list_func) (struct list_head *,
2635 struct packet_type *,
2636 struct net_device *);
2637 bool (*id_match)(struct packet_type *ptype,
2638 struct sock *sk);
2639 void *af_packet_priv;
2640 struct list_head list;
2641 };
2642
2643 struct offload_callbacks {
2644 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2645 netdev_features_t features);
2646 struct sk_buff *(*gro_receive)(struct list_head *head,
2647 struct sk_buff *skb);
2648 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2649 };
2650
2651 struct packet_offload {
2652 __be16 type; /* This is really htons(ether_type). */
2653 u16 priority;
2654 struct offload_callbacks callbacks;
2655 struct list_head list;
2656 };
2657
2658 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2659 struct pcpu_sw_netstats {
2660 u64 rx_packets;
2661 u64 rx_bytes;
2662 u64 tx_packets;
2663 u64 tx_bytes;
2664 struct u64_stats_sync syncp;
2665 } __aligned(4 * sizeof(u64));
2666
2667 struct pcpu_lstats {
2668 u64_stats_t packets;
2669 u64_stats_t bytes;
2670 struct u64_stats_sync syncp;
2671 } __aligned(2 * sizeof(u64));
2672
2673 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2674
dev_sw_netstats_rx_add(struct net_device * dev,unsigned int len)2675 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2676 {
2677 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2678
2679 u64_stats_update_begin(&tstats->syncp);
2680 tstats->rx_bytes += len;
2681 tstats->rx_packets++;
2682 u64_stats_update_end(&tstats->syncp);
2683 }
2684
dev_sw_netstats_tx_add(struct net_device * dev,unsigned int packets,unsigned int len)2685 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2686 unsigned int packets,
2687 unsigned int len)
2688 {
2689 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2690
2691 u64_stats_update_begin(&tstats->syncp);
2692 tstats->tx_bytes += len;
2693 tstats->tx_packets += packets;
2694 u64_stats_update_end(&tstats->syncp);
2695 }
2696
dev_lstats_add(struct net_device * dev,unsigned int len)2697 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2698 {
2699 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2700
2701 u64_stats_update_begin(&lstats->syncp);
2702 u64_stats_add(&lstats->bytes, len);
2703 u64_stats_inc(&lstats->packets);
2704 u64_stats_update_end(&lstats->syncp);
2705 }
2706
2707 #define __netdev_alloc_pcpu_stats(type, gfp) \
2708 ({ \
2709 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2710 if (pcpu_stats) { \
2711 int __cpu; \
2712 for_each_possible_cpu(__cpu) { \
2713 typeof(type) *stat; \
2714 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2715 u64_stats_init(&stat->syncp); \
2716 } \
2717 } \
2718 pcpu_stats; \
2719 })
2720
2721 #define netdev_alloc_pcpu_stats(type) \
2722 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2723
2724 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2725 ({ \
2726 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2727 if (pcpu_stats) { \
2728 int __cpu; \
2729 for_each_possible_cpu(__cpu) { \
2730 typeof(type) *stat; \
2731 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2732 u64_stats_init(&stat->syncp); \
2733 } \
2734 } \
2735 pcpu_stats; \
2736 })
2737
2738 enum netdev_lag_tx_type {
2739 NETDEV_LAG_TX_TYPE_UNKNOWN,
2740 NETDEV_LAG_TX_TYPE_RANDOM,
2741 NETDEV_LAG_TX_TYPE_BROADCAST,
2742 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2743 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2744 NETDEV_LAG_TX_TYPE_HASH,
2745 };
2746
2747 enum netdev_lag_hash {
2748 NETDEV_LAG_HASH_NONE,
2749 NETDEV_LAG_HASH_L2,
2750 NETDEV_LAG_HASH_L34,
2751 NETDEV_LAG_HASH_L23,
2752 NETDEV_LAG_HASH_E23,
2753 NETDEV_LAG_HASH_E34,
2754 NETDEV_LAG_HASH_VLAN_SRCMAC,
2755 NETDEV_LAG_HASH_UNKNOWN,
2756 };
2757
2758 struct netdev_lag_upper_info {
2759 enum netdev_lag_tx_type tx_type;
2760 enum netdev_lag_hash hash_type;
2761 };
2762
2763 struct netdev_lag_lower_state_info {
2764 u8 link_up : 1,
2765 tx_enabled : 1;
2766 };
2767
2768 #include <linux/notifier.h>
2769
2770 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2771 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2772 * adding new types.
2773 */
2774 enum netdev_cmd {
2775 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2776 NETDEV_DOWN,
2777 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2778 detected a hardware crash and restarted
2779 - we can use this eg to kick tcp sessions
2780 once done */
2781 NETDEV_CHANGE, /* Notify device state change */
2782 NETDEV_REGISTER,
2783 NETDEV_UNREGISTER,
2784 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2785 NETDEV_CHANGEADDR, /* notify after the address change */
2786 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2787 NETDEV_GOING_DOWN,
2788 NETDEV_CHANGENAME,
2789 NETDEV_FEAT_CHANGE,
2790 NETDEV_BONDING_FAILOVER,
2791 NETDEV_PRE_UP,
2792 NETDEV_PRE_TYPE_CHANGE,
2793 NETDEV_POST_TYPE_CHANGE,
2794 NETDEV_POST_INIT,
2795 NETDEV_RELEASE,
2796 NETDEV_NOTIFY_PEERS,
2797 NETDEV_JOIN,
2798 NETDEV_CHANGEUPPER,
2799 NETDEV_RESEND_IGMP,
2800 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2801 NETDEV_CHANGEINFODATA,
2802 NETDEV_BONDING_INFO,
2803 NETDEV_PRECHANGEUPPER,
2804 NETDEV_CHANGELOWERSTATE,
2805 NETDEV_UDP_TUNNEL_PUSH_INFO,
2806 NETDEV_UDP_TUNNEL_DROP_INFO,
2807 NETDEV_CHANGE_TX_QUEUE_LEN,
2808 NETDEV_CVLAN_FILTER_PUSH_INFO,
2809 NETDEV_CVLAN_FILTER_DROP_INFO,
2810 NETDEV_SVLAN_FILTER_PUSH_INFO,
2811 NETDEV_SVLAN_FILTER_DROP_INFO,
2812 };
2813 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2814
2815 int register_netdevice_notifier(struct notifier_block *nb);
2816 int unregister_netdevice_notifier(struct notifier_block *nb);
2817 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2818 int unregister_netdevice_notifier_net(struct net *net,
2819 struct notifier_block *nb);
2820 int register_netdevice_notifier_dev_net(struct net_device *dev,
2821 struct notifier_block *nb,
2822 struct netdev_net_notifier *nn);
2823 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2824 struct notifier_block *nb,
2825 struct netdev_net_notifier *nn);
2826
2827 struct netdev_notifier_info {
2828 struct net_device *dev;
2829 struct netlink_ext_ack *extack;
2830 };
2831
2832 struct netdev_notifier_info_ext {
2833 struct netdev_notifier_info info; /* must be first */
2834 union {
2835 u32 mtu;
2836 } ext;
2837 };
2838
2839 struct netdev_notifier_change_info {
2840 struct netdev_notifier_info info; /* must be first */
2841 unsigned int flags_changed;
2842 };
2843
2844 struct netdev_notifier_changeupper_info {
2845 struct netdev_notifier_info info; /* must be first */
2846 struct net_device *upper_dev; /* new upper dev */
2847 bool master; /* is upper dev master */
2848 bool linking; /* is the notification for link or unlink */
2849 void *upper_info; /* upper dev info */
2850 };
2851
2852 struct netdev_notifier_changelowerstate_info {
2853 struct netdev_notifier_info info; /* must be first */
2854 void *lower_state_info; /* is lower dev state */
2855 };
2856
2857 struct netdev_notifier_pre_changeaddr_info {
2858 struct netdev_notifier_info info; /* must be first */
2859 const unsigned char *dev_addr;
2860 };
2861
netdev_notifier_info_init(struct netdev_notifier_info * info,struct net_device * dev)2862 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2863 struct net_device *dev)
2864 {
2865 info->dev = dev;
2866 info->extack = NULL;
2867 }
2868
2869 static inline struct net_device *
netdev_notifier_info_to_dev(const struct netdev_notifier_info * info)2870 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2871 {
2872 return info->dev;
2873 }
2874
2875 static inline struct netlink_ext_ack *
netdev_notifier_info_to_extack(const struct netdev_notifier_info * info)2876 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2877 {
2878 return info->extack;
2879 }
2880
2881 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2882
2883
2884 extern rwlock_t dev_base_lock; /* Device list lock */
2885
2886 #define for_each_netdev(net, d) \
2887 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2888 #define for_each_netdev_reverse(net, d) \
2889 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2890 #define for_each_netdev_rcu(net, d) \
2891 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2892 #define for_each_netdev_safe(net, d, n) \
2893 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2894 #define for_each_netdev_continue(net, d) \
2895 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2896 #define for_each_netdev_continue_reverse(net, d) \
2897 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2898 dev_list)
2899 #define for_each_netdev_continue_rcu(net, d) \
2900 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2901 #define for_each_netdev_in_bond_rcu(bond, slave) \
2902 for_each_netdev_rcu(&init_net, slave) \
2903 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2904 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2905
next_net_device(struct net_device * dev)2906 static inline struct net_device *next_net_device(struct net_device *dev)
2907 {
2908 struct list_head *lh;
2909 struct net *net;
2910
2911 net = dev_net(dev);
2912 lh = dev->dev_list.next;
2913 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2914 }
2915
next_net_device_rcu(struct net_device * dev)2916 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2917 {
2918 struct list_head *lh;
2919 struct net *net;
2920
2921 net = dev_net(dev);
2922 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2923 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2924 }
2925
first_net_device(struct net * net)2926 static inline struct net_device *first_net_device(struct net *net)
2927 {
2928 return list_empty(&net->dev_base_head) ? NULL :
2929 net_device_entry(net->dev_base_head.next);
2930 }
2931
first_net_device_rcu(struct net * net)2932 static inline struct net_device *first_net_device_rcu(struct net *net)
2933 {
2934 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2935
2936 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2937 }
2938
2939 int netdev_boot_setup_check(struct net_device *dev);
2940 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2941 const char *hwaddr);
2942 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2943 void dev_add_pack(struct packet_type *pt);
2944 void dev_remove_pack(struct packet_type *pt);
2945 void __dev_remove_pack(struct packet_type *pt);
2946 void dev_add_offload(struct packet_offload *po);
2947 void dev_remove_offload(struct packet_offload *po);
2948
2949 int dev_get_iflink(const struct net_device *dev);
2950 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2951 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2952 struct net_device_path_stack *stack);
2953 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2954 unsigned short mask);
2955 struct net_device *dev_get_by_name(struct net *net, const char *name);
2956 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2957 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2958 int dev_alloc_name(struct net_device *dev, const char *name);
2959 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2960 void dev_close(struct net_device *dev);
2961 void dev_close_many(struct list_head *head, bool unlink);
2962 void dev_disable_lro(struct net_device *dev);
2963 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2964 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2965 struct net_device *sb_dev);
2966 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2967 struct net_device *sb_dev);
2968
2969 int dev_queue_xmit(struct sk_buff *skb);
2970 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2971 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2972
dev_direct_xmit(struct sk_buff * skb,u16 queue_id)2973 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2974 {
2975 int ret;
2976
2977 ret = __dev_direct_xmit(skb, queue_id);
2978 if (!dev_xmit_complete(ret))
2979 kfree_skb(skb);
2980 return ret;
2981 }
2982
2983 int register_netdevice(struct net_device *dev);
2984 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2985 void unregister_netdevice_many(struct list_head *head);
unregister_netdevice(struct net_device * dev)2986 static inline void unregister_netdevice(struct net_device *dev)
2987 {
2988 unregister_netdevice_queue(dev, NULL);
2989 }
2990
2991 int netdev_refcnt_read(const struct net_device *dev);
2992 void free_netdev(struct net_device *dev);
2993 void netdev_freemem(struct net_device *dev);
2994 int init_dummy_netdev(struct net_device *dev);
2995
2996 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2997 struct sk_buff *skb,
2998 bool all_slaves);
2999 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3000 struct sock *sk);
3001 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3002 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3003 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3004 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3005 int netdev_get_name(struct net *net, char *name, int ifindex);
3006 int dev_restart(struct net_device *dev);
3007 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3008 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3009
skb_gro_offset(const struct sk_buff * skb)3010 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3011 {
3012 return NAPI_GRO_CB(skb)->data_offset;
3013 }
3014
skb_gro_len(const struct sk_buff * skb)3015 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3016 {
3017 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3018 }
3019
skb_gro_pull(struct sk_buff * skb,unsigned int len)3020 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3021 {
3022 NAPI_GRO_CB(skb)->data_offset += len;
3023 }
3024
skb_gro_header_fast(struct sk_buff * skb,unsigned int offset)3025 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3026 unsigned int offset)
3027 {
3028 return NAPI_GRO_CB(skb)->frag0 + offset;
3029 }
3030
skb_gro_header_hard(struct sk_buff * skb,unsigned int hlen)3031 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3032 {
3033 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3034 }
3035
skb_gro_frag0_invalidate(struct sk_buff * skb)3036 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3037 {
3038 NAPI_GRO_CB(skb)->frag0 = NULL;
3039 NAPI_GRO_CB(skb)->frag0_len = 0;
3040 }
3041
skb_gro_header_slow(struct sk_buff * skb,unsigned int hlen,unsigned int offset)3042 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3043 unsigned int offset)
3044 {
3045 if (!pskb_may_pull(skb, hlen))
3046 return NULL;
3047
3048 skb_gro_frag0_invalidate(skb);
3049 return skb->data + offset;
3050 }
3051
skb_gro_network_header(struct sk_buff * skb)3052 static inline void *skb_gro_network_header(struct sk_buff *skb)
3053 {
3054 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3055 skb_network_offset(skb);
3056 }
3057
skb_gro_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3058 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3059 const void *start, unsigned int len)
3060 {
3061 if (NAPI_GRO_CB(skb)->csum_valid)
3062 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3063 csum_partial(start, len, 0));
3064 }
3065
3066 /* GRO checksum functions. These are logical equivalents of the normal
3067 * checksum functions (in skbuff.h) except that they operate on the GRO
3068 * offsets and fields in sk_buff.
3069 */
3070
3071 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3072
skb_at_gro_remcsum_start(struct sk_buff * skb)3073 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3074 {
3075 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3076 }
3077
__skb_gro_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3078 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3079 bool zero_okay,
3080 __sum16 check)
3081 {
3082 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3083 skb_checksum_start_offset(skb) <
3084 skb_gro_offset(skb)) &&
3085 !skb_at_gro_remcsum_start(skb) &&
3086 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3087 (!zero_okay || check));
3088 }
3089
__skb_gro_checksum_validate_complete(struct sk_buff * skb,__wsum psum)3090 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3091 __wsum psum)
3092 {
3093 if (NAPI_GRO_CB(skb)->csum_valid &&
3094 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3095 return 0;
3096
3097 NAPI_GRO_CB(skb)->csum = psum;
3098
3099 return __skb_gro_checksum_complete(skb);
3100 }
3101
skb_gro_incr_csum_unnecessary(struct sk_buff * skb)3102 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3103 {
3104 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3105 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3106 NAPI_GRO_CB(skb)->csum_cnt--;
3107 } else {
3108 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3109 * verified a new top level checksum or an encapsulated one
3110 * during GRO. This saves work if we fallback to normal path.
3111 */
3112 __skb_incr_checksum_unnecessary(skb);
3113 }
3114 }
3115
3116 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3117 compute_pseudo) \
3118 ({ \
3119 __sum16 __ret = 0; \
3120 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3121 __ret = __skb_gro_checksum_validate_complete(skb, \
3122 compute_pseudo(skb, proto)); \
3123 if (!__ret) \
3124 skb_gro_incr_csum_unnecessary(skb); \
3125 __ret; \
3126 })
3127
3128 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3129 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3130
3131 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3132 compute_pseudo) \
3133 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3134
3135 #define skb_gro_checksum_simple_validate(skb) \
3136 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3137
__skb_gro_checksum_convert_check(struct sk_buff * skb)3138 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3139 {
3140 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3141 !NAPI_GRO_CB(skb)->csum_valid);
3142 }
3143
__skb_gro_checksum_convert(struct sk_buff * skb,__wsum pseudo)3144 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3145 __wsum pseudo)
3146 {
3147 NAPI_GRO_CB(skb)->csum = ~pseudo;
3148 NAPI_GRO_CB(skb)->csum_valid = 1;
3149 }
3150
3151 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3152 do { \
3153 if (__skb_gro_checksum_convert_check(skb)) \
3154 __skb_gro_checksum_convert(skb, \
3155 compute_pseudo(skb, proto)); \
3156 } while (0)
3157
3158 struct gro_remcsum {
3159 int offset;
3160 __wsum delta;
3161 };
3162
skb_gro_remcsum_init(struct gro_remcsum * grc)3163 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3164 {
3165 grc->offset = 0;
3166 grc->delta = 0;
3167 }
3168
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)3169 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3170 unsigned int off, size_t hdrlen,
3171 int start, int offset,
3172 struct gro_remcsum *grc,
3173 bool nopartial)
3174 {
3175 __wsum delta;
3176 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3177
3178 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3179
3180 if (!nopartial) {
3181 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3182 return ptr;
3183 }
3184
3185 ptr = skb_gro_header_fast(skb, off);
3186 if (skb_gro_header_hard(skb, off + plen)) {
3187 ptr = skb_gro_header_slow(skb, off + plen, off);
3188 if (!ptr)
3189 return NULL;
3190 }
3191
3192 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3193 start, offset);
3194
3195 /* Adjust skb->csum since we changed the packet */
3196 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3197
3198 grc->offset = off + hdrlen + offset;
3199 grc->delta = delta;
3200
3201 return ptr;
3202 }
3203
skb_gro_remcsum_cleanup(struct sk_buff * skb,struct gro_remcsum * grc)3204 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3205 struct gro_remcsum *grc)
3206 {
3207 void *ptr;
3208 size_t plen = grc->offset + sizeof(u16);
3209
3210 if (!grc->delta)
3211 return;
3212
3213 ptr = skb_gro_header_fast(skb, grc->offset);
3214 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3215 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3216 if (!ptr)
3217 return;
3218 }
3219
3220 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3221 }
3222
3223 #ifdef CONFIG_XFRM_OFFLOAD
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3224 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3225 {
3226 if (PTR_ERR(pp) != -EINPROGRESS)
3227 NAPI_GRO_CB(skb)->flush |= flush;
3228 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3229 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3230 struct sk_buff *pp,
3231 int flush,
3232 struct gro_remcsum *grc)
3233 {
3234 if (PTR_ERR(pp) != -EINPROGRESS) {
3235 NAPI_GRO_CB(skb)->flush |= flush;
3236 skb_gro_remcsum_cleanup(skb, grc);
3237 skb->remcsum_offload = 0;
3238 }
3239 }
3240 #else
skb_gro_flush_final(struct sk_buff * skb,struct sk_buff * pp,int flush)3241 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3242 {
3243 NAPI_GRO_CB(skb)->flush |= flush;
3244 }
skb_gro_flush_final_remcsum(struct sk_buff * skb,struct sk_buff * pp,int flush,struct gro_remcsum * grc)3245 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3246 struct sk_buff *pp,
3247 int flush,
3248 struct gro_remcsum *grc)
3249 {
3250 NAPI_GRO_CB(skb)->flush |= flush;
3251 skb_gro_remcsum_cleanup(skb, grc);
3252 skb->remcsum_offload = 0;
3253 }
3254 #endif
3255
dev_hard_header(struct sk_buff * skb,struct net_device * dev,unsigned short type,const void * daddr,const void * saddr,unsigned int len)3256 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3257 unsigned short type,
3258 const void *daddr, const void *saddr,
3259 unsigned int len)
3260 {
3261 if (!dev->header_ops || !dev->header_ops->create)
3262 return 0;
3263
3264 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3265 }
3266
dev_parse_header(const struct sk_buff * skb,unsigned char * haddr)3267 static inline int dev_parse_header(const struct sk_buff *skb,
3268 unsigned char *haddr)
3269 {
3270 const struct net_device *dev = skb->dev;
3271
3272 if (!dev->header_ops || !dev->header_ops->parse)
3273 return 0;
3274 return dev->header_ops->parse(skb, haddr);
3275 }
3276
dev_parse_header_protocol(const struct sk_buff * skb)3277 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3278 {
3279 const struct net_device *dev = skb->dev;
3280
3281 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3282 return 0;
3283 return dev->header_ops->parse_protocol(skb);
3284 }
3285
3286 /* ll_header must have at least hard_header_len allocated */
dev_validate_header(const struct net_device * dev,char * ll_header,int len)3287 static inline bool dev_validate_header(const struct net_device *dev,
3288 char *ll_header, int len)
3289 {
3290 if (likely(len >= dev->hard_header_len))
3291 return true;
3292 if (len < dev->min_header_len)
3293 return false;
3294
3295 if (capable(CAP_SYS_RAWIO)) {
3296 memset(ll_header + len, 0, dev->hard_header_len - len);
3297 return true;
3298 }
3299
3300 if (dev->header_ops && dev->header_ops->validate)
3301 return dev->header_ops->validate(ll_header, len);
3302
3303 return false;
3304 }
3305
dev_has_header(const struct net_device * dev)3306 static inline bool dev_has_header(const struct net_device *dev)
3307 {
3308 return dev->header_ops && dev->header_ops->create;
3309 }
3310
3311 #ifdef CONFIG_NET_FLOW_LIMIT
3312 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3313 struct sd_flow_limit {
3314 u64 count;
3315 unsigned int num_buckets;
3316 unsigned int history_head;
3317 u16 history[FLOW_LIMIT_HISTORY];
3318 u8 buckets[];
3319 };
3320
3321 extern int netdev_flow_limit_table_len;
3322 #endif /* CONFIG_NET_FLOW_LIMIT */
3323
3324 /*
3325 * Incoming packets are placed on per-CPU queues
3326 */
3327 struct softnet_data {
3328 struct list_head poll_list;
3329 struct sk_buff_head process_queue;
3330
3331 /* stats */
3332 unsigned int processed;
3333 unsigned int time_squeeze;
3334 unsigned int received_rps;
3335 #ifdef CONFIG_RPS
3336 struct softnet_data *rps_ipi_list;
3337 #endif
3338 #ifdef CONFIG_NET_FLOW_LIMIT
3339 struct sd_flow_limit __rcu *flow_limit;
3340 #endif
3341 struct Qdisc *output_queue;
3342 struct Qdisc **output_queue_tailp;
3343 struct sk_buff *completion_queue;
3344 #ifdef CONFIG_XFRM_OFFLOAD
3345 struct sk_buff_head xfrm_backlog;
3346 #endif
3347 /* written and read only by owning cpu: */
3348 struct {
3349 u16 recursion;
3350 u8 more;
3351 } xmit;
3352 #ifdef CONFIG_RPS
3353 /* input_queue_head should be written by cpu owning this struct,
3354 * and only read by other cpus. Worth using a cache line.
3355 */
3356 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3357
3358 /* Elements below can be accessed between CPUs for RPS/RFS */
3359 call_single_data_t csd ____cacheline_aligned_in_smp;
3360 struct softnet_data *rps_ipi_next;
3361 unsigned int cpu;
3362 unsigned int input_queue_tail;
3363 #endif
3364 unsigned int dropped;
3365 struct sk_buff_head input_pkt_queue;
3366 struct napi_struct backlog;
3367
3368 };
3369
input_queue_head_incr(struct softnet_data * sd)3370 static inline void input_queue_head_incr(struct softnet_data *sd)
3371 {
3372 #ifdef CONFIG_RPS
3373 sd->input_queue_head++;
3374 #endif
3375 }
3376
input_queue_tail_incr_save(struct softnet_data * sd,unsigned int * qtail)3377 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3378 unsigned int *qtail)
3379 {
3380 #ifdef CONFIG_RPS
3381 *qtail = ++sd->input_queue_tail;
3382 #endif
3383 }
3384
3385 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3386
dev_recursion_level(void)3387 static inline int dev_recursion_level(void)
3388 {
3389 return this_cpu_read(softnet_data.xmit.recursion);
3390 }
3391
3392 #define XMIT_RECURSION_LIMIT 8
dev_xmit_recursion(void)3393 static inline bool dev_xmit_recursion(void)
3394 {
3395 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3396 XMIT_RECURSION_LIMIT);
3397 }
3398
dev_xmit_recursion_inc(void)3399 static inline void dev_xmit_recursion_inc(void)
3400 {
3401 __this_cpu_inc(softnet_data.xmit.recursion);
3402 }
3403
dev_xmit_recursion_dec(void)3404 static inline void dev_xmit_recursion_dec(void)
3405 {
3406 __this_cpu_dec(softnet_data.xmit.recursion);
3407 }
3408
3409 void __netif_schedule(struct Qdisc *q);
3410 void netif_schedule_queue(struct netdev_queue *txq);
3411
netif_tx_schedule_all(struct net_device * dev)3412 static inline void netif_tx_schedule_all(struct net_device *dev)
3413 {
3414 unsigned int i;
3415
3416 for (i = 0; i < dev->num_tx_queues; i++)
3417 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3418 }
3419
netif_tx_start_queue(struct netdev_queue * dev_queue)3420 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3421 {
3422 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3423 }
3424
3425 /**
3426 * netif_start_queue - allow transmit
3427 * @dev: network device
3428 *
3429 * Allow upper layers to call the device hard_start_xmit routine.
3430 */
netif_start_queue(struct net_device * dev)3431 static inline void netif_start_queue(struct net_device *dev)
3432 {
3433 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3434 }
3435
netif_tx_start_all_queues(struct net_device * dev)3436 static inline void netif_tx_start_all_queues(struct net_device *dev)
3437 {
3438 unsigned int i;
3439
3440 for (i = 0; i < dev->num_tx_queues; i++) {
3441 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3442 netif_tx_start_queue(txq);
3443 }
3444 }
3445
3446 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3447
3448 /**
3449 * netif_wake_queue - restart transmit
3450 * @dev: network device
3451 *
3452 * Allow upper layers to call the device hard_start_xmit routine.
3453 * Used for flow control when transmit resources are available.
3454 */
netif_wake_queue(struct net_device * dev)3455 static inline void netif_wake_queue(struct net_device *dev)
3456 {
3457 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3458 }
3459
netif_tx_wake_all_queues(struct net_device * dev)3460 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3461 {
3462 unsigned int i;
3463
3464 for (i = 0; i < dev->num_tx_queues; i++) {
3465 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3466 netif_tx_wake_queue(txq);
3467 }
3468 }
3469
netif_tx_stop_queue(struct netdev_queue * dev_queue)3470 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3471 {
3472 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3473 }
3474
3475 /**
3476 * netif_stop_queue - stop transmitted packets
3477 * @dev: network device
3478 *
3479 * Stop upper layers calling the device hard_start_xmit routine.
3480 * Used for flow control when transmit resources are unavailable.
3481 */
netif_stop_queue(struct net_device * dev)3482 static inline void netif_stop_queue(struct net_device *dev)
3483 {
3484 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3485 }
3486
3487 void netif_tx_stop_all_queues(struct net_device *dev);
3488
netif_tx_queue_stopped(const struct netdev_queue * dev_queue)3489 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3490 {
3491 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3492 }
3493
3494 /**
3495 * netif_queue_stopped - test if transmit queue is flowblocked
3496 * @dev: network device
3497 *
3498 * Test if transmit queue on device is currently unable to send.
3499 */
netif_queue_stopped(const struct net_device * dev)3500 static inline bool netif_queue_stopped(const struct net_device *dev)
3501 {
3502 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3503 }
3504
netif_xmit_stopped(const struct netdev_queue * dev_queue)3505 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3506 {
3507 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3508 }
3509
3510 static inline bool
netif_xmit_frozen_or_stopped(const struct netdev_queue * dev_queue)3511 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3512 {
3513 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3514 }
3515
3516 static inline bool
netif_xmit_frozen_or_drv_stopped(const struct netdev_queue * dev_queue)3517 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3518 {
3519 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3520 }
3521
3522 /**
3523 * netdev_queue_set_dql_min_limit - set dql minimum limit
3524 * @dev_queue: pointer to transmit queue
3525 * @min_limit: dql minimum limit
3526 *
3527 * Forces xmit_more() to return true until the minimum threshold
3528 * defined by @min_limit is reached (or until the tx queue is
3529 * empty). Warning: to be use with care, misuse will impact the
3530 * latency.
3531 */
netdev_queue_set_dql_min_limit(struct netdev_queue * dev_queue,unsigned int min_limit)3532 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3533 unsigned int min_limit)
3534 {
3535 #ifdef CONFIG_BQL
3536 dev_queue->dql.min_limit = min_limit;
3537 #endif
3538 }
3539
3540 /**
3541 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3542 * @dev_queue: pointer to transmit queue
3543 *
3544 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3545 * to give appropriate hint to the CPU.
3546 */
netdev_txq_bql_enqueue_prefetchw(struct netdev_queue * dev_queue)3547 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3548 {
3549 #ifdef CONFIG_BQL
3550 prefetchw(&dev_queue->dql.num_queued);
3551 #endif
3552 }
3553
3554 /**
3555 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3556 * @dev_queue: pointer to transmit queue
3557 *
3558 * BQL enabled drivers might use this helper in their TX completion path,
3559 * to give appropriate hint to the CPU.
3560 */
netdev_txq_bql_complete_prefetchw(struct netdev_queue * dev_queue)3561 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3562 {
3563 #ifdef CONFIG_BQL
3564 prefetchw(&dev_queue->dql.limit);
3565 #endif
3566 }
3567
netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes)3568 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3569 unsigned int bytes)
3570 {
3571 #ifdef CONFIG_BQL
3572 dql_queued(&dev_queue->dql, bytes);
3573
3574 if (likely(dql_avail(&dev_queue->dql) >= 0))
3575 return;
3576
3577 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3578
3579 /*
3580 * The XOFF flag must be set before checking the dql_avail below,
3581 * because in netdev_tx_completed_queue we update the dql_completed
3582 * before checking the XOFF flag.
3583 */
3584 smp_mb();
3585
3586 /* check again in case another CPU has just made room avail */
3587 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3588 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3589 #endif
3590 }
3591
3592 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3593 * that they should not test BQL status themselves.
3594 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3595 * skb of a batch.
3596 * Returns true if the doorbell must be used to kick the NIC.
3597 */
__netdev_tx_sent_queue(struct netdev_queue * dev_queue,unsigned int bytes,bool xmit_more)3598 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3599 unsigned int bytes,
3600 bool xmit_more)
3601 {
3602 if (xmit_more) {
3603 #ifdef CONFIG_BQL
3604 dql_queued(&dev_queue->dql, bytes);
3605 #endif
3606 return netif_tx_queue_stopped(dev_queue);
3607 }
3608 netdev_tx_sent_queue(dev_queue, bytes);
3609 return true;
3610 }
3611
3612 /**
3613 * netdev_sent_queue - report the number of bytes queued to hardware
3614 * @dev: network device
3615 * @bytes: number of bytes queued to the hardware device queue
3616 *
3617 * Report the number of bytes queued for sending/completion to the network
3618 * device hardware queue. @bytes should be a good approximation and should
3619 * exactly match netdev_completed_queue() @bytes
3620 */
netdev_sent_queue(struct net_device * dev,unsigned int bytes)3621 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3622 {
3623 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3624 }
3625
__netdev_sent_queue(struct net_device * dev,unsigned int bytes,bool xmit_more)3626 static inline bool __netdev_sent_queue(struct net_device *dev,
3627 unsigned int bytes,
3628 bool xmit_more)
3629 {
3630 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3631 xmit_more);
3632 }
3633
netdev_tx_completed_queue(struct netdev_queue * dev_queue,unsigned int pkts,unsigned int bytes)3634 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3635 unsigned int pkts, unsigned int bytes)
3636 {
3637 #ifdef CONFIG_BQL
3638 if (unlikely(!bytes))
3639 return;
3640
3641 dql_completed(&dev_queue->dql, bytes);
3642
3643 /*
3644 * Without the memory barrier there is a small possiblity that
3645 * netdev_tx_sent_queue will miss the update and cause the queue to
3646 * be stopped forever
3647 */
3648 smp_mb();
3649
3650 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3651 return;
3652
3653 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3654 netif_schedule_queue(dev_queue);
3655 #endif
3656 }
3657
3658 /**
3659 * netdev_completed_queue - report bytes and packets completed by device
3660 * @dev: network device
3661 * @pkts: actual number of packets sent over the medium
3662 * @bytes: actual number of bytes sent over the medium
3663 *
3664 * Report the number of bytes and packets transmitted by the network device
3665 * hardware queue over the physical medium, @bytes must exactly match the
3666 * @bytes amount passed to netdev_sent_queue()
3667 */
netdev_completed_queue(struct net_device * dev,unsigned int pkts,unsigned int bytes)3668 static inline void netdev_completed_queue(struct net_device *dev,
3669 unsigned int pkts, unsigned int bytes)
3670 {
3671 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3672 }
3673
netdev_tx_reset_queue(struct netdev_queue * q)3674 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3675 {
3676 #ifdef CONFIG_BQL
3677 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3678 dql_reset(&q->dql);
3679 #endif
3680 }
3681
3682 /**
3683 * netdev_reset_queue - reset the packets and bytes count of a network device
3684 * @dev_queue: network device
3685 *
3686 * Reset the bytes and packet count of a network device and clear the
3687 * software flow control OFF bit for this network device
3688 */
netdev_reset_queue(struct net_device * dev_queue)3689 static inline void netdev_reset_queue(struct net_device *dev_queue)
3690 {
3691 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3692 }
3693
3694 /**
3695 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3696 * @dev: network device
3697 * @queue_index: given tx queue index
3698 *
3699 * Returns 0 if given tx queue index >= number of device tx queues,
3700 * otherwise returns the originally passed tx queue index.
3701 */
netdev_cap_txqueue(struct net_device * dev,u16 queue_index)3702 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3703 {
3704 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3705 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3706 dev->name, queue_index,
3707 dev->real_num_tx_queues);
3708 return 0;
3709 }
3710
3711 return queue_index;
3712 }
3713
3714 /**
3715 * netif_running - test if up
3716 * @dev: network device
3717 *
3718 * Test if the device has been brought up.
3719 */
netif_running(const struct net_device * dev)3720 static inline bool netif_running(const struct net_device *dev)
3721 {
3722 return test_bit(__LINK_STATE_START, &dev->state);
3723 }
3724
3725 /*
3726 * Routines to manage the subqueues on a device. We only need start,
3727 * stop, and a check if it's stopped. All other device management is
3728 * done at the overall netdevice level.
3729 * Also test the device if we're multiqueue.
3730 */
3731
3732 /**
3733 * netif_start_subqueue - allow sending packets on subqueue
3734 * @dev: network device
3735 * @queue_index: sub queue index
3736 *
3737 * Start individual transmit queue of a device with multiple transmit queues.
3738 */
netif_start_subqueue(struct net_device * dev,u16 queue_index)3739 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3740 {
3741 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3742
3743 netif_tx_start_queue(txq);
3744 }
3745
3746 /**
3747 * netif_stop_subqueue - stop sending packets on subqueue
3748 * @dev: network device
3749 * @queue_index: sub queue index
3750 *
3751 * Stop individual transmit queue of a device with multiple transmit queues.
3752 */
netif_stop_subqueue(struct net_device * dev,u16 queue_index)3753 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3754 {
3755 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3756 netif_tx_stop_queue(txq);
3757 }
3758
3759 /**
3760 * __netif_subqueue_stopped - test status of subqueue
3761 * @dev: network device
3762 * @queue_index: sub queue index
3763 *
3764 * Check individual transmit queue of a device with multiple transmit queues.
3765 */
__netif_subqueue_stopped(const struct net_device * dev,u16 queue_index)3766 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3767 u16 queue_index)
3768 {
3769 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3770
3771 return netif_tx_queue_stopped(txq);
3772 }
3773
3774 /**
3775 * netif_subqueue_stopped - test status of subqueue
3776 * @dev: network device
3777 * @skb: sub queue buffer pointer
3778 *
3779 * Check individual transmit queue of a device with multiple transmit queues.
3780 */
netif_subqueue_stopped(const struct net_device * dev,struct sk_buff * skb)3781 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3782 struct sk_buff *skb)
3783 {
3784 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3785 }
3786
3787 /**
3788 * netif_wake_subqueue - allow sending packets on subqueue
3789 * @dev: network device
3790 * @queue_index: sub queue index
3791 *
3792 * Resume individual transmit queue of a device with multiple transmit queues.
3793 */
netif_wake_subqueue(struct net_device * dev,u16 queue_index)3794 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3795 {
3796 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3797
3798 netif_tx_wake_queue(txq);
3799 }
3800
3801 #ifdef CONFIG_XPS
3802 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3803 u16 index);
3804 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3805 u16 index, enum xps_map_type type);
3806
3807 /**
3808 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3809 * @j: CPU/Rx queue index
3810 * @mask: bitmask of all cpus/rx queues
3811 * @nr_bits: number of bits in the bitmask
3812 *
3813 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3814 */
netif_attr_test_mask(unsigned long j,const unsigned long * mask,unsigned int nr_bits)3815 static inline bool netif_attr_test_mask(unsigned long j,
3816 const unsigned long *mask,
3817 unsigned int nr_bits)
3818 {
3819 cpu_max_bits_warn(j, nr_bits);
3820 return test_bit(j, mask);
3821 }
3822
3823 /**
3824 * netif_attr_test_online - Test for online CPU/Rx queue
3825 * @j: CPU/Rx queue index
3826 * @online_mask: bitmask for CPUs/Rx queues that are online
3827 * @nr_bits: number of bits in the bitmask
3828 *
3829 * Returns true if a CPU/Rx queue is online.
3830 */
netif_attr_test_online(unsigned long j,const unsigned long * online_mask,unsigned int nr_bits)3831 static inline bool netif_attr_test_online(unsigned long j,
3832 const unsigned long *online_mask,
3833 unsigned int nr_bits)
3834 {
3835 cpu_max_bits_warn(j, nr_bits);
3836
3837 if (online_mask)
3838 return test_bit(j, online_mask);
3839
3840 return (j < nr_bits);
3841 }
3842
3843 /**
3844 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3845 * @n: CPU/Rx queue index
3846 * @srcp: the cpumask/Rx queue mask pointer
3847 * @nr_bits: number of bits in the bitmask
3848 *
3849 * Returns >= nr_bits if no further CPUs/Rx queues set.
3850 */
netif_attrmask_next(int n,const unsigned long * srcp,unsigned int nr_bits)3851 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3852 unsigned int nr_bits)
3853 {
3854 /* -1 is a legal arg here. */
3855 if (n != -1)
3856 cpu_max_bits_warn(n, nr_bits);
3857
3858 if (srcp)
3859 return find_next_bit(srcp, nr_bits, n + 1);
3860
3861 return n + 1;
3862 }
3863
3864 /**
3865 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3866 * @n: CPU/Rx queue index
3867 * @src1p: the first CPUs/Rx queues mask pointer
3868 * @src2p: the second CPUs/Rx queues mask pointer
3869 * @nr_bits: number of bits in the bitmask
3870 *
3871 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3872 */
netif_attrmask_next_and(int n,const unsigned long * src1p,const unsigned long * src2p,unsigned int nr_bits)3873 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3874 const unsigned long *src2p,
3875 unsigned int nr_bits)
3876 {
3877 /* -1 is a legal arg here. */
3878 if (n != -1)
3879 cpu_max_bits_warn(n, nr_bits);
3880
3881 if (src1p && src2p)
3882 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3883 else if (src1p)
3884 return find_next_bit(src1p, nr_bits, n + 1);
3885 else if (src2p)
3886 return find_next_bit(src2p, nr_bits, n + 1);
3887
3888 return n + 1;
3889 }
3890 #else
netif_set_xps_queue(struct net_device * dev,const struct cpumask * mask,u16 index)3891 static inline int netif_set_xps_queue(struct net_device *dev,
3892 const struct cpumask *mask,
3893 u16 index)
3894 {
3895 return 0;
3896 }
3897
__netif_set_xps_queue(struct net_device * dev,const unsigned long * mask,u16 index,enum xps_map_type type)3898 static inline int __netif_set_xps_queue(struct net_device *dev,
3899 const unsigned long *mask,
3900 u16 index, enum xps_map_type type)
3901 {
3902 return 0;
3903 }
3904 #endif
3905
3906 /**
3907 * netif_is_multiqueue - test if device has multiple transmit queues
3908 * @dev: network device
3909 *
3910 * Check if device has multiple transmit queues
3911 */
netif_is_multiqueue(const struct net_device * dev)3912 static inline bool netif_is_multiqueue(const struct net_device *dev)
3913 {
3914 return dev->num_tx_queues > 1;
3915 }
3916
3917 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3918
3919 #ifdef CONFIG_SYSFS
3920 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3921 #else
netif_set_real_num_rx_queues(struct net_device * dev,unsigned int rxqs)3922 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3923 unsigned int rxqs)
3924 {
3925 dev->real_num_rx_queues = rxqs;
3926 return 0;
3927 }
3928 #endif
3929 int netif_set_real_num_queues(struct net_device *dev,
3930 unsigned int txq, unsigned int rxq);
3931
3932 static inline struct netdev_rx_queue *
__netif_get_rx_queue(struct net_device * dev,unsigned int rxq)3933 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3934 {
3935 return dev->_rx + rxq;
3936 }
3937
3938 #ifdef CONFIG_SYSFS
get_netdev_rx_queue_index(struct netdev_rx_queue * queue)3939 static inline unsigned int get_netdev_rx_queue_index(
3940 struct netdev_rx_queue *queue)
3941 {
3942 struct net_device *dev = queue->dev;
3943 int index = queue - dev->_rx;
3944
3945 BUG_ON(index >= dev->num_rx_queues);
3946 return index;
3947 }
3948 #endif
3949
3950 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3951 int netif_get_num_default_rss_queues(void);
3952
3953 enum skb_free_reason {
3954 SKB_REASON_CONSUMED,
3955 SKB_REASON_DROPPED,
3956 };
3957
3958 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3959 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3960
3961 /*
3962 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3963 * interrupt context or with hardware interrupts being disabled.
3964 * (in_hardirq() || irqs_disabled())
3965 *
3966 * We provide four helpers that can be used in following contexts :
3967 *
3968 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3969 * replacing kfree_skb(skb)
3970 *
3971 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3972 * Typically used in place of consume_skb(skb) in TX completion path
3973 *
3974 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3975 * replacing kfree_skb(skb)
3976 *
3977 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3978 * and consumed a packet. Used in place of consume_skb(skb)
3979 */
dev_kfree_skb_irq(struct sk_buff * skb)3980 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3981 {
3982 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3983 }
3984
dev_consume_skb_irq(struct sk_buff * skb)3985 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3986 {
3987 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3988 }
3989
dev_kfree_skb_any(struct sk_buff * skb)3990 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3991 {
3992 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3993 }
3994
dev_consume_skb_any(struct sk_buff * skb)3995 static inline void dev_consume_skb_any(struct sk_buff *skb)
3996 {
3997 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3998 }
3999
4000 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
4001 struct bpf_prog *xdp_prog);
4002 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
4003 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
4004 int netif_rx(struct sk_buff *skb);
4005 int netif_rx_ni(struct sk_buff *skb);
4006 int netif_rx_any_context(struct sk_buff *skb);
4007 int netif_receive_skb(struct sk_buff *skb);
4008 int netif_receive_skb_core(struct sk_buff *skb);
4009 void netif_receive_skb_list(struct list_head *head);
4010 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4011 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4012 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4013 gro_result_t napi_gro_frags(struct napi_struct *napi);
4014 struct packet_offload *gro_find_receive_by_type(__be16 type);
4015 struct packet_offload *gro_find_complete_by_type(__be16 type);
4016
napi_free_frags(struct napi_struct * napi)4017 static inline void napi_free_frags(struct napi_struct *napi)
4018 {
4019 kfree_skb(napi->skb);
4020 napi->skb = NULL;
4021 }
4022
4023 bool netdev_is_rx_handler_busy(struct net_device *dev);
4024 int netdev_rx_handler_register(struct net_device *dev,
4025 rx_handler_func_t *rx_handler,
4026 void *rx_handler_data);
4027 void netdev_rx_handler_unregister(struct net_device *dev);
4028
4029 bool dev_valid_name(const char *name);
is_socket_ioctl_cmd(unsigned int cmd)4030 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
4031 {
4032 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4033 }
4034 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4035 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4036 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4037 void __user *data, bool *need_copyout);
4038 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4039 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4040 unsigned int dev_get_flags(const struct net_device *);
4041 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4042 struct netlink_ext_ack *extack);
4043 int dev_change_flags(struct net_device *dev, unsigned int flags,
4044 struct netlink_ext_ack *extack);
4045 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4046 unsigned int gchanges);
4047 int dev_change_name(struct net_device *, const char *);
4048 int dev_set_alias(struct net_device *, const char *, size_t);
4049 int dev_get_alias(const struct net_device *, char *, size_t);
4050 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4051 const char *pat, int new_ifindex);
4052 static inline
dev_change_net_namespace(struct net_device * dev,struct net * net,const char * pat)4053 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4054 const char *pat)
4055 {
4056 return __dev_change_net_namespace(dev, net, pat, 0);
4057 }
4058 int __dev_set_mtu(struct net_device *, int);
4059 int dev_validate_mtu(struct net_device *dev, int mtu,
4060 struct netlink_ext_ack *extack);
4061 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4062 struct netlink_ext_ack *extack);
4063 int dev_set_mtu(struct net_device *, int);
4064 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4065 void dev_set_group(struct net_device *, int);
4066 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4067 struct netlink_ext_ack *extack);
4068 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4069 struct netlink_ext_ack *extack);
4070 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4071 struct netlink_ext_ack *extack);
4072 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4073 int dev_change_carrier(struct net_device *, bool new_carrier);
4074 int dev_get_phys_port_id(struct net_device *dev,
4075 struct netdev_phys_item_id *ppid);
4076 int dev_get_phys_port_name(struct net_device *dev,
4077 char *name, size_t len);
4078 int dev_get_port_parent_id(struct net_device *dev,
4079 struct netdev_phys_item_id *ppid, bool recurse);
4080 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4081 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4082 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4083 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4084 u32 value);
4085 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4086 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4087 struct netdev_queue *txq, int *ret);
4088
4089 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4090 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4091 int fd, int expected_fd, u32 flags);
4092 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4093 u8 dev_xdp_prog_count(struct net_device *dev);
4094 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4095
4096 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4097 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4098 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4099 bool is_skb_forwardable(const struct net_device *dev,
4100 const struct sk_buff *skb);
4101
__is_skb_forwardable(const struct net_device * dev,const struct sk_buff * skb,const bool check_mtu)4102 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4103 const struct sk_buff *skb,
4104 const bool check_mtu)
4105 {
4106 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4107 unsigned int len;
4108
4109 if (!(dev->flags & IFF_UP))
4110 return false;
4111
4112 if (!check_mtu)
4113 return true;
4114
4115 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4116 if (skb->len <= len)
4117 return true;
4118
4119 /* if TSO is enabled, we don't care about the length as the packet
4120 * could be forwarded without being segmented before
4121 */
4122 if (skb_is_gso(skb))
4123 return true;
4124
4125 return false;
4126 }
4127
____dev_forward_skb(struct net_device * dev,struct sk_buff * skb,const bool check_mtu)4128 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4129 struct sk_buff *skb,
4130 const bool check_mtu)
4131 {
4132 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4133 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4134 atomic_long_inc(&dev->rx_dropped);
4135 kfree_skb(skb);
4136 return NET_RX_DROP;
4137 }
4138
4139 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4140 skb->priority = 0;
4141 return 0;
4142 }
4143
4144 bool dev_nit_active(struct net_device *dev);
4145 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4146
4147 extern int netdev_budget;
4148 extern unsigned int netdev_budget_usecs;
4149
4150 /* Called by rtnetlink.c:rtnl_unlock() */
4151 void netdev_run_todo(void);
4152
4153 /**
4154 * dev_put - release reference to device
4155 * @dev: network device
4156 *
4157 * Release reference to device to allow it to be freed.
4158 */
dev_put(struct net_device * dev)4159 static inline void dev_put(struct net_device *dev)
4160 {
4161 if (dev) {
4162 #ifdef CONFIG_PCPU_DEV_REFCNT
4163 this_cpu_dec(*dev->pcpu_refcnt);
4164 #else
4165 refcount_dec(&dev->dev_refcnt);
4166 #endif
4167 }
4168 }
4169
4170 /**
4171 * dev_hold - get reference to device
4172 * @dev: network device
4173 *
4174 * Hold reference to device to keep it from being freed.
4175 */
dev_hold(struct net_device * dev)4176 static inline void dev_hold(struct net_device *dev)
4177 {
4178 if (dev) {
4179 #ifdef CONFIG_PCPU_DEV_REFCNT
4180 this_cpu_inc(*dev->pcpu_refcnt);
4181 #else
4182 refcount_inc(&dev->dev_refcnt);
4183 #endif
4184 }
4185 }
4186
4187 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4188 * and _off may be called from IRQ context, but it is caller
4189 * who is responsible for serialization of these calls.
4190 *
4191 * The name carrier is inappropriate, these functions should really be
4192 * called netif_lowerlayer_*() because they represent the state of any
4193 * kind of lower layer not just hardware media.
4194 */
4195
4196 void linkwatch_init_dev(struct net_device *dev);
4197 void linkwatch_fire_event(struct net_device *dev);
4198 void linkwatch_forget_dev(struct net_device *dev);
4199
4200 /**
4201 * netif_carrier_ok - test if carrier present
4202 * @dev: network device
4203 *
4204 * Check if carrier is present on device
4205 */
netif_carrier_ok(const struct net_device * dev)4206 static inline bool netif_carrier_ok(const struct net_device *dev)
4207 {
4208 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4209 }
4210
4211 unsigned long dev_trans_start(struct net_device *dev);
4212
4213 void __netdev_watchdog_up(struct net_device *dev);
4214
4215 void netif_carrier_on(struct net_device *dev);
4216 void netif_carrier_off(struct net_device *dev);
4217 void netif_carrier_event(struct net_device *dev);
4218
4219 /**
4220 * netif_dormant_on - mark device as dormant.
4221 * @dev: network device
4222 *
4223 * Mark device as dormant (as per RFC2863).
4224 *
4225 * The dormant state indicates that the relevant interface is not
4226 * actually in a condition to pass packets (i.e., it is not 'up') but is
4227 * in a "pending" state, waiting for some external event. For "on-
4228 * demand" interfaces, this new state identifies the situation where the
4229 * interface is waiting for events to place it in the up state.
4230 */
netif_dormant_on(struct net_device * dev)4231 static inline void netif_dormant_on(struct net_device *dev)
4232 {
4233 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4234 linkwatch_fire_event(dev);
4235 }
4236
4237 /**
4238 * netif_dormant_off - set device as not dormant.
4239 * @dev: network device
4240 *
4241 * Device is not in dormant state.
4242 */
netif_dormant_off(struct net_device * dev)4243 static inline void netif_dormant_off(struct net_device *dev)
4244 {
4245 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4246 linkwatch_fire_event(dev);
4247 }
4248
4249 /**
4250 * netif_dormant - test if device is dormant
4251 * @dev: network device
4252 *
4253 * Check if device is dormant.
4254 */
netif_dormant(const struct net_device * dev)4255 static inline bool netif_dormant(const struct net_device *dev)
4256 {
4257 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4258 }
4259
4260
4261 /**
4262 * netif_testing_on - mark device as under test.
4263 * @dev: network device
4264 *
4265 * Mark device as under test (as per RFC2863).
4266 *
4267 * The testing state indicates that some test(s) must be performed on
4268 * the interface. After completion, of the test, the interface state
4269 * will change to up, dormant, or down, as appropriate.
4270 */
netif_testing_on(struct net_device * dev)4271 static inline void netif_testing_on(struct net_device *dev)
4272 {
4273 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4274 linkwatch_fire_event(dev);
4275 }
4276
4277 /**
4278 * netif_testing_off - set device as not under test.
4279 * @dev: network device
4280 *
4281 * Device is not in testing state.
4282 */
netif_testing_off(struct net_device * dev)4283 static inline void netif_testing_off(struct net_device *dev)
4284 {
4285 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4286 linkwatch_fire_event(dev);
4287 }
4288
4289 /**
4290 * netif_testing - test if device is under test
4291 * @dev: network device
4292 *
4293 * Check if device is under test
4294 */
netif_testing(const struct net_device * dev)4295 static inline bool netif_testing(const struct net_device *dev)
4296 {
4297 return test_bit(__LINK_STATE_TESTING, &dev->state);
4298 }
4299
4300
4301 /**
4302 * netif_oper_up - test if device is operational
4303 * @dev: network device
4304 *
4305 * Check if carrier is operational
4306 */
netif_oper_up(const struct net_device * dev)4307 static inline bool netif_oper_up(const struct net_device *dev)
4308 {
4309 return (dev->operstate == IF_OPER_UP ||
4310 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4311 }
4312
4313 /**
4314 * netif_device_present - is device available or removed
4315 * @dev: network device
4316 *
4317 * Check if device has not been removed from system.
4318 */
netif_device_present(const struct net_device * dev)4319 static inline bool netif_device_present(const struct net_device *dev)
4320 {
4321 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4322 }
4323
4324 void netif_device_detach(struct net_device *dev);
4325
4326 void netif_device_attach(struct net_device *dev);
4327
4328 /*
4329 * Network interface message level settings
4330 */
4331
4332 enum {
4333 NETIF_MSG_DRV_BIT,
4334 NETIF_MSG_PROBE_BIT,
4335 NETIF_MSG_LINK_BIT,
4336 NETIF_MSG_TIMER_BIT,
4337 NETIF_MSG_IFDOWN_BIT,
4338 NETIF_MSG_IFUP_BIT,
4339 NETIF_MSG_RX_ERR_BIT,
4340 NETIF_MSG_TX_ERR_BIT,
4341 NETIF_MSG_TX_QUEUED_BIT,
4342 NETIF_MSG_INTR_BIT,
4343 NETIF_MSG_TX_DONE_BIT,
4344 NETIF_MSG_RX_STATUS_BIT,
4345 NETIF_MSG_PKTDATA_BIT,
4346 NETIF_MSG_HW_BIT,
4347 NETIF_MSG_WOL_BIT,
4348
4349 /* When you add a new bit above, update netif_msg_class_names array
4350 * in net/ethtool/common.c
4351 */
4352 NETIF_MSG_CLASS_COUNT,
4353 };
4354 /* Both ethtool_ops interface and internal driver implementation use u32 */
4355 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4356
4357 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4358 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4359
4360 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4361 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4362 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4363 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4364 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4365 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4366 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4367 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4368 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4369 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4370 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4371 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4372 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4373 #define NETIF_MSG_HW __NETIF_MSG(HW)
4374 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4375
4376 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4377 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4378 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4379 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4380 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4381 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4382 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4383 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4384 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4385 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4386 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4387 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4388 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4389 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4390 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4391
netif_msg_init(int debug_value,int default_msg_enable_bits)4392 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4393 {
4394 /* use default */
4395 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4396 return default_msg_enable_bits;
4397 if (debug_value == 0) /* no output */
4398 return 0;
4399 /* set low N bits */
4400 return (1U << debug_value) - 1;
4401 }
4402
__netif_tx_lock(struct netdev_queue * txq,int cpu)4403 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4404 {
4405 spin_lock(&txq->_xmit_lock);
4406 txq->xmit_lock_owner = cpu;
4407 }
4408
__netif_tx_acquire(struct netdev_queue * txq)4409 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4410 {
4411 __acquire(&txq->_xmit_lock);
4412 return true;
4413 }
4414
__netif_tx_release(struct netdev_queue * txq)4415 static inline void __netif_tx_release(struct netdev_queue *txq)
4416 {
4417 __release(&txq->_xmit_lock);
4418 }
4419
__netif_tx_lock_bh(struct netdev_queue * txq)4420 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4421 {
4422 spin_lock_bh(&txq->_xmit_lock);
4423 txq->xmit_lock_owner = smp_processor_id();
4424 }
4425
__netif_tx_trylock(struct netdev_queue * txq)4426 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4427 {
4428 bool ok = spin_trylock(&txq->_xmit_lock);
4429 if (likely(ok))
4430 txq->xmit_lock_owner = smp_processor_id();
4431 return ok;
4432 }
4433
__netif_tx_unlock(struct netdev_queue * txq)4434 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4435 {
4436 txq->xmit_lock_owner = -1;
4437 spin_unlock(&txq->_xmit_lock);
4438 }
4439
__netif_tx_unlock_bh(struct netdev_queue * txq)4440 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4441 {
4442 txq->xmit_lock_owner = -1;
4443 spin_unlock_bh(&txq->_xmit_lock);
4444 }
4445
txq_trans_update(struct netdev_queue * txq)4446 static inline void txq_trans_update(struct netdev_queue *txq)
4447 {
4448 if (txq->xmit_lock_owner != -1)
4449 txq->trans_start = jiffies;
4450 }
4451
4452 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
netif_trans_update(struct net_device * dev)4453 static inline void netif_trans_update(struct net_device *dev)
4454 {
4455 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4456
4457 if (txq->trans_start != jiffies)
4458 txq->trans_start = jiffies;
4459 }
4460
4461 /**
4462 * netif_tx_lock - grab network device transmit lock
4463 * @dev: network device
4464 *
4465 * Get network device transmit lock
4466 */
netif_tx_lock(struct net_device * dev)4467 static inline void netif_tx_lock(struct net_device *dev)
4468 {
4469 unsigned int i;
4470 int cpu;
4471
4472 spin_lock(&dev->tx_global_lock);
4473 cpu = smp_processor_id();
4474 for (i = 0; i < dev->num_tx_queues; i++) {
4475 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4476
4477 /* We are the only thread of execution doing a
4478 * freeze, but we have to grab the _xmit_lock in
4479 * order to synchronize with threads which are in
4480 * the ->hard_start_xmit() handler and already
4481 * checked the frozen bit.
4482 */
4483 __netif_tx_lock(txq, cpu);
4484 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4485 __netif_tx_unlock(txq);
4486 }
4487 }
4488
netif_tx_lock_bh(struct net_device * dev)4489 static inline void netif_tx_lock_bh(struct net_device *dev)
4490 {
4491 local_bh_disable();
4492 netif_tx_lock(dev);
4493 }
4494
netif_tx_unlock(struct net_device * dev)4495 static inline void netif_tx_unlock(struct net_device *dev)
4496 {
4497 unsigned int i;
4498
4499 for (i = 0; i < dev->num_tx_queues; i++) {
4500 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4501
4502 /* No need to grab the _xmit_lock here. If the
4503 * queue is not stopped for another reason, we
4504 * force a schedule.
4505 */
4506 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4507 netif_schedule_queue(txq);
4508 }
4509 spin_unlock(&dev->tx_global_lock);
4510 }
4511
netif_tx_unlock_bh(struct net_device * dev)4512 static inline void netif_tx_unlock_bh(struct net_device *dev)
4513 {
4514 netif_tx_unlock(dev);
4515 local_bh_enable();
4516 }
4517
4518 #define HARD_TX_LOCK(dev, txq, cpu) { \
4519 if ((dev->features & NETIF_F_LLTX) == 0) { \
4520 __netif_tx_lock(txq, cpu); \
4521 } else { \
4522 __netif_tx_acquire(txq); \
4523 } \
4524 }
4525
4526 #define HARD_TX_TRYLOCK(dev, txq) \
4527 (((dev->features & NETIF_F_LLTX) == 0) ? \
4528 __netif_tx_trylock(txq) : \
4529 __netif_tx_acquire(txq))
4530
4531 #define HARD_TX_UNLOCK(dev, txq) { \
4532 if ((dev->features & NETIF_F_LLTX) == 0) { \
4533 __netif_tx_unlock(txq); \
4534 } else { \
4535 __netif_tx_release(txq); \
4536 } \
4537 }
4538
netif_tx_disable(struct net_device * dev)4539 static inline void netif_tx_disable(struct net_device *dev)
4540 {
4541 unsigned int i;
4542 int cpu;
4543
4544 local_bh_disable();
4545 cpu = smp_processor_id();
4546 spin_lock(&dev->tx_global_lock);
4547 for (i = 0; i < dev->num_tx_queues; i++) {
4548 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4549
4550 __netif_tx_lock(txq, cpu);
4551 netif_tx_stop_queue(txq);
4552 __netif_tx_unlock(txq);
4553 }
4554 spin_unlock(&dev->tx_global_lock);
4555 local_bh_enable();
4556 }
4557
netif_addr_lock(struct net_device * dev)4558 static inline void netif_addr_lock(struct net_device *dev)
4559 {
4560 unsigned char nest_level = 0;
4561
4562 #ifdef CONFIG_LOCKDEP
4563 nest_level = dev->nested_level;
4564 #endif
4565 spin_lock_nested(&dev->addr_list_lock, nest_level);
4566 }
4567
netif_addr_lock_bh(struct net_device * dev)4568 static inline void netif_addr_lock_bh(struct net_device *dev)
4569 {
4570 unsigned char nest_level = 0;
4571
4572 #ifdef CONFIG_LOCKDEP
4573 nest_level = dev->nested_level;
4574 #endif
4575 local_bh_disable();
4576 spin_lock_nested(&dev->addr_list_lock, nest_level);
4577 }
4578
netif_addr_unlock(struct net_device * dev)4579 static inline void netif_addr_unlock(struct net_device *dev)
4580 {
4581 spin_unlock(&dev->addr_list_lock);
4582 }
4583
netif_addr_unlock_bh(struct net_device * dev)4584 static inline void netif_addr_unlock_bh(struct net_device *dev)
4585 {
4586 spin_unlock_bh(&dev->addr_list_lock);
4587 }
4588
4589 /*
4590 * dev_addrs walker. Should be used only for read access. Call with
4591 * rcu_read_lock held.
4592 */
4593 #define for_each_dev_addr(dev, ha) \
4594 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4595
4596 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4597
4598 void ether_setup(struct net_device *dev);
4599
4600 /* Support for loadable net-drivers */
4601 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4602 unsigned char name_assign_type,
4603 void (*setup)(struct net_device *),
4604 unsigned int txqs, unsigned int rxqs);
4605 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4606 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4607
4608 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4609 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4610 count)
4611
4612 int register_netdev(struct net_device *dev);
4613 void unregister_netdev(struct net_device *dev);
4614
4615 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4616
4617 /* General hardware address lists handling functions */
4618 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4619 struct netdev_hw_addr_list *from_list, int addr_len);
4620 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4621 struct netdev_hw_addr_list *from_list, int addr_len);
4622 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4623 struct net_device *dev,
4624 int (*sync)(struct net_device *, const unsigned char *),
4625 int (*unsync)(struct net_device *,
4626 const unsigned char *));
4627 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4628 struct net_device *dev,
4629 int (*sync)(struct net_device *,
4630 const unsigned char *, int),
4631 int (*unsync)(struct net_device *,
4632 const unsigned char *, int));
4633 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4634 struct net_device *dev,
4635 int (*unsync)(struct net_device *,
4636 const unsigned char *, int));
4637 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4638 struct net_device *dev,
4639 int (*unsync)(struct net_device *,
4640 const unsigned char *));
4641 void __hw_addr_init(struct netdev_hw_addr_list *list);
4642
4643 /* Functions used for device addresses handling */
4644 static inline void
__dev_addr_set(struct net_device * dev,const u8 * addr,size_t len)4645 __dev_addr_set(struct net_device *dev, const u8 *addr, size_t len)
4646 {
4647 memcpy(dev->dev_addr, addr, len);
4648 }
4649
dev_addr_set(struct net_device * dev,const u8 * addr)4650 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4651 {
4652 __dev_addr_set(dev, addr, dev->addr_len);
4653 }
4654
4655 static inline void
dev_addr_mod(struct net_device * dev,unsigned int offset,const u8 * addr,size_t len)4656 dev_addr_mod(struct net_device *dev, unsigned int offset,
4657 const u8 *addr, size_t len)
4658 {
4659 memcpy(&dev->dev_addr[offset], addr, len);
4660 }
4661
4662 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4663 unsigned char addr_type);
4664 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4665 unsigned char addr_type);
4666 void dev_addr_flush(struct net_device *dev);
4667 int dev_addr_init(struct net_device *dev);
4668
4669 /* Functions used for unicast addresses handling */
4670 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4671 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4672 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4673 int dev_uc_sync(struct net_device *to, struct net_device *from);
4674 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4675 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4676 void dev_uc_flush(struct net_device *dev);
4677 void dev_uc_init(struct net_device *dev);
4678
4679 /**
4680 * __dev_uc_sync - Synchonize device's unicast list
4681 * @dev: device to sync
4682 * @sync: function to call if address should be added
4683 * @unsync: function to call if address should be removed
4684 *
4685 * Add newly added addresses to the interface, and release
4686 * addresses that have been deleted.
4687 */
__dev_uc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4688 static inline int __dev_uc_sync(struct net_device *dev,
4689 int (*sync)(struct net_device *,
4690 const unsigned char *),
4691 int (*unsync)(struct net_device *,
4692 const unsigned char *))
4693 {
4694 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4695 }
4696
4697 /**
4698 * __dev_uc_unsync - Remove synchronized addresses from device
4699 * @dev: device to sync
4700 * @unsync: function to call if address should be removed
4701 *
4702 * Remove all addresses that were added to the device by dev_uc_sync().
4703 */
__dev_uc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4704 static inline void __dev_uc_unsync(struct net_device *dev,
4705 int (*unsync)(struct net_device *,
4706 const unsigned char *))
4707 {
4708 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4709 }
4710
4711 /* Functions used for multicast addresses handling */
4712 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4713 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4714 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4715 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4716 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4717 int dev_mc_sync(struct net_device *to, struct net_device *from);
4718 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4719 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4720 void dev_mc_flush(struct net_device *dev);
4721 void dev_mc_init(struct net_device *dev);
4722
4723 /**
4724 * __dev_mc_sync - Synchonize device's multicast list
4725 * @dev: device to sync
4726 * @sync: function to call if address should be added
4727 * @unsync: function to call if address should be removed
4728 *
4729 * Add newly added addresses to the interface, and release
4730 * addresses that have been deleted.
4731 */
__dev_mc_sync(struct net_device * dev,int (* sync)(struct net_device *,const unsigned char *),int (* unsync)(struct net_device *,const unsigned char *))4732 static inline int __dev_mc_sync(struct net_device *dev,
4733 int (*sync)(struct net_device *,
4734 const unsigned char *),
4735 int (*unsync)(struct net_device *,
4736 const unsigned char *))
4737 {
4738 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4739 }
4740
4741 /**
4742 * __dev_mc_unsync - Remove synchronized addresses from device
4743 * @dev: device to sync
4744 * @unsync: function to call if address should be removed
4745 *
4746 * Remove all addresses that were added to the device by dev_mc_sync().
4747 */
__dev_mc_unsync(struct net_device * dev,int (* unsync)(struct net_device *,const unsigned char *))4748 static inline void __dev_mc_unsync(struct net_device *dev,
4749 int (*unsync)(struct net_device *,
4750 const unsigned char *))
4751 {
4752 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4753 }
4754
4755 /* Functions used for secondary unicast and multicast support */
4756 void dev_set_rx_mode(struct net_device *dev);
4757 void __dev_set_rx_mode(struct net_device *dev);
4758 int dev_set_promiscuity(struct net_device *dev, int inc);
4759 int dev_set_allmulti(struct net_device *dev, int inc);
4760 void netdev_state_change(struct net_device *dev);
4761 void __netdev_notify_peers(struct net_device *dev);
4762 void netdev_notify_peers(struct net_device *dev);
4763 void netdev_features_change(struct net_device *dev);
4764 /* Load a device via the kmod */
4765 void dev_load(struct net *net, const char *name);
4766 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4767 struct rtnl_link_stats64 *storage);
4768 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4769 const struct net_device_stats *netdev_stats);
4770 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4771 const struct pcpu_sw_netstats __percpu *netstats);
4772 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4773
4774 extern int netdev_max_backlog;
4775 extern int netdev_tstamp_prequeue;
4776 extern int netdev_unregister_timeout_secs;
4777 extern int weight_p;
4778 extern int dev_weight_rx_bias;
4779 extern int dev_weight_tx_bias;
4780 extern int dev_rx_weight;
4781 extern int dev_tx_weight;
4782 extern int gro_normal_batch;
4783
4784 enum {
4785 NESTED_SYNC_IMM_BIT,
4786 NESTED_SYNC_TODO_BIT,
4787 };
4788
4789 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4790 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4791
4792 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4793 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4794
4795 struct netdev_nested_priv {
4796 unsigned char flags;
4797 void *data;
4798 };
4799
4800 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4801 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4802 struct list_head **iter);
4803 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4804 struct list_head **iter);
4805
4806 #ifdef CONFIG_LOCKDEP
4807 static LIST_HEAD(net_unlink_list);
4808
net_unlink_todo(struct net_device * dev)4809 static inline void net_unlink_todo(struct net_device *dev)
4810 {
4811 if (list_empty(&dev->unlink_list))
4812 list_add_tail(&dev->unlink_list, &net_unlink_list);
4813 }
4814 #endif
4815
4816 /* iterate through upper list, must be called under RCU read lock */
4817 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4818 for (iter = &(dev)->adj_list.upper, \
4819 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4820 updev; \
4821 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4822
4823 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4824 int (*fn)(struct net_device *upper_dev,
4825 struct netdev_nested_priv *priv),
4826 struct netdev_nested_priv *priv);
4827
4828 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4829 struct net_device *upper_dev);
4830
4831 bool netdev_has_any_upper_dev(struct net_device *dev);
4832
4833 void *netdev_lower_get_next_private(struct net_device *dev,
4834 struct list_head **iter);
4835 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4836 struct list_head **iter);
4837
4838 #define netdev_for_each_lower_private(dev, priv, iter) \
4839 for (iter = (dev)->adj_list.lower.next, \
4840 priv = netdev_lower_get_next_private(dev, &(iter)); \
4841 priv; \
4842 priv = netdev_lower_get_next_private(dev, &(iter)))
4843
4844 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4845 for (iter = &(dev)->adj_list.lower, \
4846 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4847 priv; \
4848 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4849
4850 void *netdev_lower_get_next(struct net_device *dev,
4851 struct list_head **iter);
4852
4853 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4854 for (iter = (dev)->adj_list.lower.next, \
4855 ldev = netdev_lower_get_next(dev, &(iter)); \
4856 ldev; \
4857 ldev = netdev_lower_get_next(dev, &(iter)))
4858
4859 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4860 struct list_head **iter);
4861 int netdev_walk_all_lower_dev(struct net_device *dev,
4862 int (*fn)(struct net_device *lower_dev,
4863 struct netdev_nested_priv *priv),
4864 struct netdev_nested_priv *priv);
4865 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4866 int (*fn)(struct net_device *lower_dev,
4867 struct netdev_nested_priv *priv),
4868 struct netdev_nested_priv *priv);
4869
4870 void *netdev_adjacent_get_private(struct list_head *adj_list);
4871 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4872 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4873 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4874 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4875 struct netlink_ext_ack *extack);
4876 int netdev_master_upper_dev_link(struct net_device *dev,
4877 struct net_device *upper_dev,
4878 void *upper_priv, void *upper_info,
4879 struct netlink_ext_ack *extack);
4880 void netdev_upper_dev_unlink(struct net_device *dev,
4881 struct net_device *upper_dev);
4882 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4883 struct net_device *new_dev,
4884 struct net_device *dev,
4885 struct netlink_ext_ack *extack);
4886 void netdev_adjacent_change_commit(struct net_device *old_dev,
4887 struct net_device *new_dev,
4888 struct net_device *dev);
4889 void netdev_adjacent_change_abort(struct net_device *old_dev,
4890 struct net_device *new_dev,
4891 struct net_device *dev);
4892 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4893 void *netdev_lower_dev_get_private(struct net_device *dev,
4894 struct net_device *lower_dev);
4895 void netdev_lower_state_changed(struct net_device *lower_dev,
4896 void *lower_state_info);
4897
4898 /* RSS keys are 40 or 52 bytes long */
4899 #define NETDEV_RSS_KEY_LEN 52
4900 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4901 void netdev_rss_key_fill(void *buffer, size_t len);
4902
4903 int skb_checksum_help(struct sk_buff *skb);
4904 int skb_crc32c_csum_help(struct sk_buff *skb);
4905 int skb_csum_hwoffload_help(struct sk_buff *skb,
4906 const netdev_features_t features);
4907
4908 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4909 netdev_features_t features, bool tx_path);
4910 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4911 netdev_features_t features);
4912
4913 struct netdev_bonding_info {
4914 ifslave slave;
4915 ifbond master;
4916 };
4917
4918 struct netdev_notifier_bonding_info {
4919 struct netdev_notifier_info info; /* must be first */
4920 struct netdev_bonding_info bonding_info;
4921 };
4922
4923 void netdev_bonding_info_change(struct net_device *dev,
4924 struct netdev_bonding_info *bonding_info);
4925
4926 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4927 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4928 #else
ethtool_notify(struct net_device * dev,unsigned int cmd,const void * data)4929 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4930 const void *data)
4931 {
4932 }
4933 #endif
4934
4935 static inline
skb_gso_segment(struct sk_buff * skb,netdev_features_t features)4936 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4937 {
4938 return __skb_gso_segment(skb, features, true);
4939 }
4940 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4941
can_checksum_protocol(netdev_features_t features,__be16 protocol)4942 static inline bool can_checksum_protocol(netdev_features_t features,
4943 __be16 protocol)
4944 {
4945 if (protocol == htons(ETH_P_FCOE))
4946 return !!(features & NETIF_F_FCOE_CRC);
4947
4948 /* Assume this is an IP checksum (not SCTP CRC) */
4949
4950 if (features & NETIF_F_HW_CSUM) {
4951 /* Can checksum everything */
4952 return true;
4953 }
4954
4955 switch (protocol) {
4956 case htons(ETH_P_IP):
4957 return !!(features & NETIF_F_IP_CSUM);
4958 case htons(ETH_P_IPV6):
4959 return !!(features & NETIF_F_IPV6_CSUM);
4960 default:
4961 return false;
4962 }
4963 }
4964
4965 #ifdef CONFIG_BUG
4966 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4967 #else
netdev_rx_csum_fault(struct net_device * dev,struct sk_buff * skb)4968 static inline void netdev_rx_csum_fault(struct net_device *dev,
4969 struct sk_buff *skb)
4970 {
4971 }
4972 #endif
4973 /* rx skb timestamps */
4974 void net_enable_timestamp(void);
4975 void net_disable_timestamp(void);
4976
4977 #ifdef CONFIG_PROC_FS
4978 int __init dev_proc_init(void);
4979 #else
4980 #define dev_proc_init() 0
4981 #endif
4982
__netdev_start_xmit(const struct net_device_ops * ops,struct sk_buff * skb,struct net_device * dev,bool more)4983 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4984 struct sk_buff *skb, struct net_device *dev,
4985 bool more)
4986 {
4987 __this_cpu_write(softnet_data.xmit.more, more);
4988 return ops->ndo_start_xmit(skb, dev);
4989 }
4990
netdev_xmit_more(void)4991 static inline bool netdev_xmit_more(void)
4992 {
4993 return __this_cpu_read(softnet_data.xmit.more);
4994 }
4995
netdev_start_xmit(struct sk_buff * skb,struct net_device * dev,struct netdev_queue * txq,bool more)4996 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4997 struct netdev_queue *txq, bool more)
4998 {
4999 const struct net_device_ops *ops = dev->netdev_ops;
5000 netdev_tx_t rc;
5001
5002 rc = __netdev_start_xmit(ops, skb, dev, more);
5003 if (rc == NETDEV_TX_OK)
5004 txq_trans_update(txq);
5005
5006 return rc;
5007 }
5008
5009 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5010 const void *ns);
5011 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5012 const void *ns);
5013
5014 extern const struct kobj_ns_type_operations net_ns_type_operations;
5015
5016 const char *netdev_drivername(const struct net_device *dev);
5017
5018 void linkwatch_run_queue(void);
5019
netdev_intersect_features(netdev_features_t f1,netdev_features_t f2)5020 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5021 netdev_features_t f2)
5022 {
5023 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5024 if (f1 & NETIF_F_HW_CSUM)
5025 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5026 else
5027 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5028 }
5029
5030 return f1 & f2;
5031 }
5032
netdev_get_wanted_features(struct net_device * dev)5033 static inline netdev_features_t netdev_get_wanted_features(
5034 struct net_device *dev)
5035 {
5036 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5037 }
5038 netdev_features_t netdev_increment_features(netdev_features_t all,
5039 netdev_features_t one, netdev_features_t mask);
5040
5041 /* Allow TSO being used on stacked device :
5042 * Performing the GSO segmentation before last device
5043 * is a performance improvement.
5044 */
netdev_add_tso_features(netdev_features_t features,netdev_features_t mask)5045 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5046 netdev_features_t mask)
5047 {
5048 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5049 }
5050
5051 int __netdev_update_features(struct net_device *dev);
5052 void netdev_update_features(struct net_device *dev);
5053 void netdev_change_features(struct net_device *dev);
5054
5055 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5056 struct net_device *dev);
5057
5058 netdev_features_t passthru_features_check(struct sk_buff *skb,
5059 struct net_device *dev,
5060 netdev_features_t features);
5061 netdev_features_t netif_skb_features(struct sk_buff *skb);
5062
net_gso_ok(netdev_features_t features,int gso_type)5063 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5064 {
5065 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5066
5067 /* check flags correspondence */
5068 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5069 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5070 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5071 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5072 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5073 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5074 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5075 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5076 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5077 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5078 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5079 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5080 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5081 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5082 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5083 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5084 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5085 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5086 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5087
5088 return (features & feature) == feature;
5089 }
5090
skb_gso_ok(struct sk_buff * skb,netdev_features_t features)5091 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5092 {
5093 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5094 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5095 }
5096
netif_needs_gso(struct sk_buff * skb,netdev_features_t features)5097 static inline bool netif_needs_gso(struct sk_buff *skb,
5098 netdev_features_t features)
5099 {
5100 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5101 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5102 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5103 }
5104
netif_set_gso_max_size(struct net_device * dev,unsigned int size)5105 static inline void netif_set_gso_max_size(struct net_device *dev,
5106 unsigned int size)
5107 {
5108 dev->gso_max_size = size;
5109 }
5110
skb_gso_error_unwind(struct sk_buff * skb,__be16 protocol,int pulled_hlen,u16 mac_offset,int mac_len)5111 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5112 int pulled_hlen, u16 mac_offset,
5113 int mac_len)
5114 {
5115 skb->protocol = protocol;
5116 skb->encapsulation = 1;
5117 skb_push(skb, pulled_hlen);
5118 skb_reset_transport_header(skb);
5119 skb->mac_header = mac_offset;
5120 skb->network_header = skb->mac_header + mac_len;
5121 skb->mac_len = mac_len;
5122 }
5123
netif_is_macsec(const struct net_device * dev)5124 static inline bool netif_is_macsec(const struct net_device *dev)
5125 {
5126 return dev->priv_flags & IFF_MACSEC;
5127 }
5128
netif_is_macvlan(const struct net_device * dev)5129 static inline bool netif_is_macvlan(const struct net_device *dev)
5130 {
5131 return dev->priv_flags & IFF_MACVLAN;
5132 }
5133
netif_is_macvlan_port(const struct net_device * dev)5134 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5135 {
5136 return dev->priv_flags & IFF_MACVLAN_PORT;
5137 }
5138
netif_is_bond_master(const struct net_device * dev)5139 static inline bool netif_is_bond_master(const struct net_device *dev)
5140 {
5141 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5142 }
5143
netif_is_bond_slave(const struct net_device * dev)5144 static inline bool netif_is_bond_slave(const struct net_device *dev)
5145 {
5146 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5147 }
5148
netif_supports_nofcs(struct net_device * dev)5149 static inline bool netif_supports_nofcs(struct net_device *dev)
5150 {
5151 return dev->priv_flags & IFF_SUPP_NOFCS;
5152 }
5153
netif_has_l3_rx_handler(const struct net_device * dev)5154 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5155 {
5156 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5157 }
5158
netif_is_l3_master(const struct net_device * dev)5159 static inline bool netif_is_l3_master(const struct net_device *dev)
5160 {
5161 return dev->priv_flags & IFF_L3MDEV_MASTER;
5162 }
5163
netif_is_l3_slave(const struct net_device * dev)5164 static inline bool netif_is_l3_slave(const struct net_device *dev)
5165 {
5166 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5167 }
5168
netif_is_bridge_master(const struct net_device * dev)5169 static inline bool netif_is_bridge_master(const struct net_device *dev)
5170 {
5171 return dev->priv_flags & IFF_EBRIDGE;
5172 }
5173
netif_is_bridge_port(const struct net_device * dev)5174 static inline bool netif_is_bridge_port(const struct net_device *dev)
5175 {
5176 return dev->priv_flags & IFF_BRIDGE_PORT;
5177 }
5178
netif_is_ovs_master(const struct net_device * dev)5179 static inline bool netif_is_ovs_master(const struct net_device *dev)
5180 {
5181 return dev->priv_flags & IFF_OPENVSWITCH;
5182 }
5183
netif_is_ovs_port(const struct net_device * dev)5184 static inline bool netif_is_ovs_port(const struct net_device *dev)
5185 {
5186 return dev->priv_flags & IFF_OVS_DATAPATH;
5187 }
5188
netif_is_any_bridge_port(const struct net_device * dev)5189 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5190 {
5191 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5192 }
5193
netif_is_team_master(const struct net_device * dev)5194 static inline bool netif_is_team_master(const struct net_device *dev)
5195 {
5196 return dev->priv_flags & IFF_TEAM;
5197 }
5198
netif_is_team_port(const struct net_device * dev)5199 static inline bool netif_is_team_port(const struct net_device *dev)
5200 {
5201 return dev->priv_flags & IFF_TEAM_PORT;
5202 }
5203
netif_is_lag_master(const struct net_device * dev)5204 static inline bool netif_is_lag_master(const struct net_device *dev)
5205 {
5206 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5207 }
5208
netif_is_lag_port(const struct net_device * dev)5209 static inline bool netif_is_lag_port(const struct net_device *dev)
5210 {
5211 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5212 }
5213
netif_is_rxfh_configured(const struct net_device * dev)5214 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5215 {
5216 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5217 }
5218
netif_is_failover(const struct net_device * dev)5219 static inline bool netif_is_failover(const struct net_device *dev)
5220 {
5221 return dev->priv_flags & IFF_FAILOVER;
5222 }
5223
netif_is_failover_slave(const struct net_device * dev)5224 static inline bool netif_is_failover_slave(const struct net_device *dev)
5225 {
5226 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5227 }
5228
5229 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
netif_keep_dst(struct net_device * dev)5230 static inline void netif_keep_dst(struct net_device *dev)
5231 {
5232 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5233 }
5234
5235 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
netif_reduces_vlan_mtu(struct net_device * dev)5236 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5237 {
5238 /* TODO: reserve and use an additional IFF bit, if we get more users */
5239 return dev->priv_flags & IFF_MACSEC;
5240 }
5241
5242 extern struct pernet_operations __net_initdata loopback_net_ops;
5243
5244 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5245
5246 /* netdev_printk helpers, similar to dev_printk */
5247
netdev_name(const struct net_device * dev)5248 static inline const char *netdev_name(const struct net_device *dev)
5249 {
5250 if (!dev->name[0] || strchr(dev->name, '%'))
5251 return "(unnamed net_device)";
5252 return dev->name;
5253 }
5254
netdev_unregistering(const struct net_device * dev)5255 static inline bool netdev_unregistering(const struct net_device *dev)
5256 {
5257 return dev->reg_state == NETREG_UNREGISTERING;
5258 }
5259
netdev_reg_state(const struct net_device * dev)5260 static inline const char *netdev_reg_state(const struct net_device *dev)
5261 {
5262 switch (dev->reg_state) {
5263 case NETREG_UNINITIALIZED: return " (uninitialized)";
5264 case NETREG_REGISTERED: return "";
5265 case NETREG_UNREGISTERING: return " (unregistering)";
5266 case NETREG_UNREGISTERED: return " (unregistered)";
5267 case NETREG_RELEASED: return " (released)";
5268 case NETREG_DUMMY: return " (dummy)";
5269 }
5270
5271 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5272 return " (unknown)";
5273 }
5274
5275 __printf(3, 4) __cold
5276 void netdev_printk(const char *level, const struct net_device *dev,
5277 const char *format, ...);
5278 __printf(2, 3) __cold
5279 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5280 __printf(2, 3) __cold
5281 void netdev_alert(const struct net_device *dev, const char *format, ...);
5282 __printf(2, 3) __cold
5283 void netdev_crit(const struct net_device *dev, const char *format, ...);
5284 __printf(2, 3) __cold
5285 void netdev_err(const struct net_device *dev, const char *format, ...);
5286 __printf(2, 3) __cold
5287 void netdev_warn(const struct net_device *dev, const char *format, ...);
5288 __printf(2, 3) __cold
5289 void netdev_notice(const struct net_device *dev, const char *format, ...);
5290 __printf(2, 3) __cold
5291 void netdev_info(const struct net_device *dev, const char *format, ...);
5292
5293 #define netdev_level_once(level, dev, fmt, ...) \
5294 do { \
5295 static bool __print_once __read_mostly; \
5296 \
5297 if (!__print_once) { \
5298 __print_once = true; \
5299 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5300 } \
5301 } while (0)
5302
5303 #define netdev_emerg_once(dev, fmt, ...) \
5304 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5305 #define netdev_alert_once(dev, fmt, ...) \
5306 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5307 #define netdev_crit_once(dev, fmt, ...) \
5308 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5309 #define netdev_err_once(dev, fmt, ...) \
5310 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5311 #define netdev_warn_once(dev, fmt, ...) \
5312 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5313 #define netdev_notice_once(dev, fmt, ...) \
5314 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5315 #define netdev_info_once(dev, fmt, ...) \
5316 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5317
5318 #define MODULE_ALIAS_NETDEV(device) \
5319 MODULE_ALIAS("netdev-" device)
5320
5321 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5322 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5323 #define netdev_dbg(__dev, format, args...) \
5324 do { \
5325 dynamic_netdev_dbg(__dev, format, ##args); \
5326 } while (0)
5327 #elif defined(DEBUG)
5328 #define netdev_dbg(__dev, format, args...) \
5329 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5330 #else
5331 #define netdev_dbg(__dev, format, args...) \
5332 ({ \
5333 if (0) \
5334 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5335 })
5336 #endif
5337
5338 #if defined(VERBOSE_DEBUG)
5339 #define netdev_vdbg netdev_dbg
5340 #else
5341
5342 #define netdev_vdbg(dev, format, args...) \
5343 ({ \
5344 if (0) \
5345 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5346 0; \
5347 })
5348 #endif
5349
5350 /*
5351 * netdev_WARN() acts like dev_printk(), but with the key difference
5352 * of using a WARN/WARN_ON to get the message out, including the
5353 * file/line information and a backtrace.
5354 */
5355 #define netdev_WARN(dev, format, args...) \
5356 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5357 netdev_reg_state(dev), ##args)
5358
5359 #define netdev_WARN_ONCE(dev, format, args...) \
5360 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5361 netdev_reg_state(dev), ##args)
5362
5363 /* netif printk helpers, similar to netdev_printk */
5364
5365 #define netif_printk(priv, type, level, dev, fmt, args...) \
5366 do { \
5367 if (netif_msg_##type(priv)) \
5368 netdev_printk(level, (dev), fmt, ##args); \
5369 } while (0)
5370
5371 #define netif_level(level, priv, type, dev, fmt, args...) \
5372 do { \
5373 if (netif_msg_##type(priv)) \
5374 netdev_##level(dev, fmt, ##args); \
5375 } while (0)
5376
5377 #define netif_emerg(priv, type, dev, fmt, args...) \
5378 netif_level(emerg, priv, type, dev, fmt, ##args)
5379 #define netif_alert(priv, type, dev, fmt, args...) \
5380 netif_level(alert, priv, type, dev, fmt, ##args)
5381 #define netif_crit(priv, type, dev, fmt, args...) \
5382 netif_level(crit, priv, type, dev, fmt, ##args)
5383 #define netif_err(priv, type, dev, fmt, args...) \
5384 netif_level(err, priv, type, dev, fmt, ##args)
5385 #define netif_warn(priv, type, dev, fmt, args...) \
5386 netif_level(warn, priv, type, dev, fmt, ##args)
5387 #define netif_notice(priv, type, dev, fmt, args...) \
5388 netif_level(notice, priv, type, dev, fmt, ##args)
5389 #define netif_info(priv, type, dev, fmt, args...) \
5390 netif_level(info, priv, type, dev, fmt, ##args)
5391
5392 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5393 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5394 #define netif_dbg(priv, type, netdev, format, args...) \
5395 do { \
5396 if (netif_msg_##type(priv)) \
5397 dynamic_netdev_dbg(netdev, format, ##args); \
5398 } while (0)
5399 #elif defined(DEBUG)
5400 #define netif_dbg(priv, type, dev, format, args...) \
5401 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5402 #else
5403 #define netif_dbg(priv, type, dev, format, args...) \
5404 ({ \
5405 if (0) \
5406 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5407 0; \
5408 })
5409 #endif
5410
5411 /* if @cond then downgrade to debug, else print at @level */
5412 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5413 do { \
5414 if (cond) \
5415 netif_dbg(priv, type, netdev, fmt, ##args); \
5416 else \
5417 netif_ ## level(priv, type, netdev, fmt, ##args); \
5418 } while (0)
5419
5420 #if defined(VERBOSE_DEBUG)
5421 #define netif_vdbg netif_dbg
5422 #else
5423 #define netif_vdbg(priv, type, dev, format, args...) \
5424 ({ \
5425 if (0) \
5426 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5427 0; \
5428 })
5429 #endif
5430
5431 /*
5432 * The list of packet types we will receive (as opposed to discard)
5433 * and the routines to invoke.
5434 *
5435 * Why 16. Because with 16 the only overlap we get on a hash of the
5436 * low nibble of the protocol value is RARP/SNAP/X.25.
5437 *
5438 * 0800 IP
5439 * 0001 802.3
5440 * 0002 AX.25
5441 * 0004 802.2
5442 * 8035 RARP
5443 * 0005 SNAP
5444 * 0805 X.25
5445 * 0806 ARP
5446 * 8137 IPX
5447 * 0009 Localtalk
5448 * 86DD IPv6
5449 */
5450 #define PTYPE_HASH_SIZE (16)
5451 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5452
5453 extern struct list_head ptype_all __read_mostly;
5454 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5455
5456 extern struct net_device *blackhole_netdev;
5457
5458 #endif /* _LINUX_NETDEVICE_H */
5459