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