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