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