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