Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for the Interfaces handler.
8 *
9 * Version: @(#)dev.h 1.0.10 08/12/93
10 *
11 * Authors: Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 *
19 * Moved to /usr/include/linux for NET3
20 */
21#ifndef _LINUX_NETDEVICE_H
22#define _LINUX_NETDEVICE_H
23
24#include <linux/timer.h>
25#include <linux/bug.h>
26#include <linux/delay.h>
27#include <linux/atomic.h>
28#include <linux/prefetch.h>
29#include <asm/cache.h>
30#include <asm/byteorder.h>
31
32#include <linux/percpu.h>
33#include <linux/rculist.h>
34#include <linux/workqueue.h>
35#include <linux/dynamic_queue_limits.h>
36
37#include <net/net_namespace.h>
38#ifdef CONFIG_DCB
39#include <net/dcbnl.h>
40#endif
41#include <net/netprio_cgroup.h>
42#include <net/xdp.h>
43
44#include <linux/netdev_features.h>
45#include <linux/neighbour.h>
46#include <uapi/linux/netdevice.h>
47#include <uapi/linux/if_bonding.h>
48#include <uapi/linux/pkt_cls.h>
49#include <linux/hashtable.h>
50
51struct netpoll_info;
52struct device;
53struct ethtool_ops;
54struct phy_device;
55struct dsa_port;
56struct ip_tunnel_parm;
57struct macsec_context;
58struct macsec_ops;
59
60struct sfp_bus;
61/* 802.11 specific */
62struct wireless_dev;
63/* 802.15.4 specific */
64struct wpan_dev;
65struct mpls_dev;
66/* UDP Tunnel offloads */
67struct udp_tunnel_info;
68struct udp_tunnel_nic_info;
69struct udp_tunnel_nic;
70struct bpf_prog;
71struct xdp_buff;
72
73void synchronize_net(void);
74void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
76
77/* Backlog congestion levels */
78#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79#define NET_RX_DROP 1 /* packet dropped */
80
81#define MAX_NEST_DEV 8
82
83/*
84 * Transmit return codes: transmit return codes originate from three different
85 * namespaces:
86 *
87 * - qdisc return codes
88 * - driver transmit return codes
89 * - errno values
90 *
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
98 */
99
100/* qdisc ->enqueue() return codes. */
101#define NET_XMIT_SUCCESS 0x00
102#define NET_XMIT_DROP 0x01 /* skb dropped */
103#define NET_XMIT_CN 0x02 /* congestion notification */
104#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
105
106/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
111
112/* Driver transmit return codes */
113#define NETDEV_TX_MASK 0xf0
114
115enum netdev_tx {
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
119};
120typedef enum netdev_tx netdev_tx_t;
121
122/*
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
125 */
126static inline bool dev_xmit_complete(int rc)
127{
128 /*
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
133 */
134 if (likely(rc < NET_XMIT_MASK))
135 return true;
136
137 return false;
138}
139
140/*
141 * Compute the worst-case header length according to the protocols
142 * used.
143 */
144
145#if defined(CONFIG_HYPERV_NET)
146# define LL_MAX_HEADER 128
147#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148# if defined(CONFIG_MAC80211_MESH)
149# define LL_MAX_HEADER 128
150# else
151# define LL_MAX_HEADER 96
152# endif
153#else
154# define LL_MAX_HEADER 32
155#endif
156
157#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159#define MAX_HEADER LL_MAX_HEADER
160#else
161#define MAX_HEADER (LL_MAX_HEADER + 48)
162#endif
163
164/*
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
167 */
168
169struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
193};
194
195
196#include <linux/cache.h>
197#include <linux/skbuff.h>
198
199#ifdef CONFIG_RPS
200#include <linux/static_key.h>
201extern struct static_key_false rps_needed;
202extern struct static_key_false rfs_needed;
203#endif
204
205struct neighbour;
206struct neigh_parms;
207struct sk_buff;
208
209struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
212 unsigned char type;
213#define NETDEV_HW_ADDR_T_LAN 1
214#define NETDEV_HW_ADDR_T_SAN 2
215#define NETDEV_HW_ADDR_T_UNICAST 3
216#define NETDEV_HW_ADDR_T_MULTICAST 4
217 bool global_use;
218 int sync_cnt;
219 int refcount;
220 int synced;
221 struct rcu_head rcu_head;
222};
223
224struct netdev_hw_addr_list {
225 struct list_head list;
226 int count;
227};
228
229#define netdev_hw_addr_list_count(l) ((l)->count)
230#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231#define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
233
234#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236#define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
238
239#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241#define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243
244struct hh_cache {
245 unsigned int hh_len;
246 seqlock_t hh_lock;
247
248 /* cached hardware header; allow for machine alignment needs. */
249#define HH_DATA_MOD 16
250#define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252#define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
255};
256
257/* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
258 * Alternative is:
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
261 *
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
264 */
265#define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
269
270struct header_ops {
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
281};
282
283/* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
285 * code.
286 */
287
288enum netdev_state_t {
289 __LINK_STATE_START,
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
295};
296
297
298/*
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
301 */
302struct netdev_boot_setup {
303 char name[IFNAMSIZ];
304 struct ifmap map;
305};
306#define NETDEV_BOOT_SETUP_MAX 8
307
308int __init netdev_boot_setup(char *str);
309
310struct gro_list {
311 struct list_head list;
312 int count;
313};
314
315/*
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
318 */
319#define GRO_HASH_BUCKETS 8
320
321/*
322 * Structure for NAPI scheduling similar to tasklet but with weighting
323 */
324struct napi_struct {
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
330 */
331 struct list_head poll_list;
332
333 unsigned long state;
334 int weight;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338#ifdef CONFIG_NETPOLL
339 int poll_owner;
340#endif
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
343 struct sk_buff *skb;
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
350 struct task_struct *thread;
351};
352
353enum {
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_MISSED, /* reschedule a napi */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
358 NAPI_STATE_LISTED, /* NAPI added to system lists */
359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
363};
364
365enum {
366 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
367 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
368 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
369 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
370 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
371 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
372 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
373 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
374 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
375};
376
377enum gro_result {
378 GRO_MERGED,
379 GRO_MERGED_FREE,
380 GRO_HELD,
381 GRO_NORMAL,
382 GRO_CONSUMED,
383};
384typedef enum gro_result gro_result_t;
385
386/*
387 * enum rx_handler_result - Possible return values for rx_handlers.
388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
389 * further.
390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
391 * case skb->dev was changed by rx_handler.
392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
394 *
395 * rx_handlers are functions called from inside __netif_receive_skb(), to do
396 * special processing of the skb, prior to delivery to protocol handlers.
397 *
398 * Currently, a net_device can only have a single rx_handler registered. Trying
399 * to register a second rx_handler will return -EBUSY.
400 *
401 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
402 * To unregister a rx_handler on a net_device, use
403 * netdev_rx_handler_unregister().
404 *
405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
406 * do with the skb.
407 *
408 * If the rx_handler consumed the skb in some way, it should return
409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
410 * the skb to be delivered in some other way.
411 *
412 * If the rx_handler changed skb->dev, to divert the skb to another
413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
414 * new device will be called if it exists.
415 *
416 * If the rx_handler decides the skb should be ignored, it should return
417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
418 * are registered on exact device (ptype->dev == skb->dev).
419 *
420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
421 * delivered, it should return RX_HANDLER_PASS.
422 *
423 * A device without a registered rx_handler will behave as if rx_handler
424 * returned RX_HANDLER_PASS.
425 */
426
427enum rx_handler_result {
428 RX_HANDLER_CONSUMED,
429 RX_HANDLER_ANOTHER,
430 RX_HANDLER_EXACT,
431 RX_HANDLER_PASS,
432};
433typedef enum rx_handler_result rx_handler_result_t;
434typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
435
436void __napi_schedule(struct napi_struct *n);
437void __napi_schedule_irqoff(struct napi_struct *n);
438
439static inline bool napi_disable_pending(struct napi_struct *n)
440{
441 return test_bit(NAPI_STATE_DISABLE, &n->state);
442}
443
444static inline bool napi_prefer_busy_poll(struct napi_struct *n)
445{
446 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
447}
448
449bool napi_schedule_prep(struct napi_struct *n);
450
451/**
452 * napi_schedule - schedule NAPI poll
453 * @n: NAPI context
454 *
455 * Schedule NAPI poll routine to be called if it is not already
456 * running.
457 */
458static inline void napi_schedule(struct napi_struct *n)
459{
460 if (napi_schedule_prep(n))
461 __napi_schedule(n);
462}
463
464/**
465 * napi_schedule_irqoff - schedule NAPI poll
466 * @n: NAPI context
467 *
468 * Variant of napi_schedule(), assuming hard irqs are masked.
469 */
470static inline void napi_schedule_irqoff(struct napi_struct *n)
471{
472 if (napi_schedule_prep(n))
473 __napi_schedule_irqoff(n);
474}
475
476/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
477static inline bool napi_reschedule(struct napi_struct *napi)
478{
479 if (napi_schedule_prep(napi)) {
480 __napi_schedule(napi);
481 return true;
482 }
483 return false;
484}
485
486bool napi_complete_done(struct napi_struct *n, int work_done);
487/**
488 * napi_complete - NAPI processing complete
489 * @n: NAPI context
490 *
491 * Mark NAPI processing as complete.
492 * Consider using napi_complete_done() instead.
493 * Return false if device should avoid rearming interrupts.
494 */
495static inline bool napi_complete(struct napi_struct *n)
496{
497 return napi_complete_done(n, 0);
498}
499
500int dev_set_threaded(struct net_device *dev, bool threaded);
501
502/**
503 * napi_disable - prevent NAPI from scheduling
504 * @n: NAPI context
505 *
506 * Stop NAPI from being scheduled on this context.
507 * Waits till any outstanding processing completes.
508 */
509void napi_disable(struct napi_struct *n);
510
511void napi_enable(struct napi_struct *n);
512
513/**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: NAPI context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521static inline void napi_synchronize(const struct napi_struct *n)
522{
523 if (IS_ENABLED(CONFIG_SMP))
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
525 msleep(1);
526 else
527 barrier();
528}
529
530/**
531 * napi_if_scheduled_mark_missed - if napi is running, set the
532 * NAPIF_STATE_MISSED
533 * @n: NAPI context
534 *
535 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
536 * NAPI is scheduled.
537 **/
538static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
539{
540 unsigned long val, new;
541
542 do {
543 val = READ_ONCE(n->state);
544 if (val & NAPIF_STATE_DISABLE)
545 return true;
546
547 if (!(val & NAPIF_STATE_SCHED))
548 return false;
549
550 new = val | NAPIF_STATE_MISSED;
551 } while (cmpxchg(&n->state, val, new) != val);
552
553 return true;
554}
555
556enum netdev_queue_state_t {
557 __QUEUE_STATE_DRV_XOFF,
558 __QUEUE_STATE_STACK_XOFF,
559 __QUEUE_STATE_FROZEN,
560};
561
562#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
563#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
564#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
565
566#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
567#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
568 QUEUE_STATE_FROZEN)
569#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
570 QUEUE_STATE_FROZEN)
571
572/*
573 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
574 * netif_tx_* functions below are used to manipulate this flag. The
575 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
576 * queue independently. The netif_xmit_*stopped functions below are called
577 * to check if the queue has been stopped by the driver or stack (either
578 * of the XOFF bits are set in the state). Drivers should not need to call
579 * netif_xmit*stopped functions, they should only be using netif_tx_*.
580 */
581
582struct netdev_queue {
583/*
584 * read-mostly part
585 */
586 struct net_device *dev;
587 struct Qdisc __rcu *qdisc;
588 struct Qdisc *qdisc_sleeping;
589#ifdef CONFIG_SYSFS
590 struct kobject kobj;
591#endif
592#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
593 int numa_node;
594#endif
595 unsigned long tx_maxrate;
596 /*
597 * Number of TX timeouts for this queue
598 * (/sys/class/net/DEV/Q/trans_timeout)
599 */
600 unsigned long trans_timeout;
601
602 /* Subordinate device that the queue has been assigned to */
603 struct net_device *sb_dev;
604#ifdef CONFIG_XDP_SOCKETS
605 struct xsk_buff_pool *pool;
606#endif
607/*
608 * write-mostly part
609 */
610 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
611 int xmit_lock_owner;
612 /*
613 * Time (in jiffies) of last Tx
614 */
615 unsigned long trans_start;
616
617 unsigned long state;
618
619#ifdef CONFIG_BQL
620 struct dql dql;
621#endif
622} ____cacheline_aligned_in_smp;
623
624extern int sysctl_fb_tunnels_only_for_init_net;
625extern int sysctl_devconf_inherit_init_net;
626
627/*
628 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
629 * == 1 : For initns only
630 * == 2 : For none.
631 */
632static inline bool net_has_fallback_tunnels(const struct net *net)
633{
634 return !IS_ENABLED(CONFIG_SYSCTL) ||
635 !sysctl_fb_tunnels_only_for_init_net ||
636 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
637}
638
639static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
640{
641#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
642 return q->numa_node;
643#else
644 return NUMA_NO_NODE;
645#endif
646}
647
648static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
649{
650#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
651 q->numa_node = node;
652#endif
653}
654
655#ifdef CONFIG_RPS
656/*
657 * This structure holds an RPS map which can be of variable length. The
658 * map is an array of CPUs.
659 */
660struct rps_map {
661 unsigned int len;
662 struct rcu_head rcu;
663 u16 cpus[];
664};
665#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
666
667/*
668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
669 * tail pointer for that CPU's input queue at the time of last enqueue, and
670 * a hardware filter index.
671 */
672struct rps_dev_flow {
673 u16 cpu;
674 u16 filter;
675 unsigned int last_qtail;
676};
677#define RPS_NO_FILTER 0xffff
678
679/*
680 * The rps_dev_flow_table structure contains a table of flow mappings.
681 */
682struct rps_dev_flow_table {
683 unsigned int mask;
684 struct rcu_head rcu;
685 struct rps_dev_flow flows[];
686};
687#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
688 ((_num) * sizeof(struct rps_dev_flow)))
689
690/*
691 * The rps_sock_flow_table contains mappings of flows to the last CPU
692 * on which they were processed by the application (set in recvmsg).
693 * Each entry is a 32bit value. Upper part is the high-order bits
694 * of flow hash, lower part is CPU number.
695 * rps_cpu_mask is used to partition the space, depending on number of
696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
698 * meaning we use 32-6=26 bits for the hash.
699 */
700struct rps_sock_flow_table {
701 u32 mask;
702
703 u32 ents[] ____cacheline_aligned_in_smp;
704};
705#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
706
707#define RPS_NO_CPU 0xffff
708
709extern u32 rps_cpu_mask;
710extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
711
712static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
713 u32 hash)
714{
715 if (table && hash) {
716 unsigned int index = hash & table->mask;
717 u32 val = hash & ~rps_cpu_mask;
718
719 /* We only give a hint, preemption can change CPU under us */
720 val |= raw_smp_processor_id();
721
722 if (table->ents[index] != val)
723 table->ents[index] = val;
724 }
725}
726
727#ifdef CONFIG_RFS_ACCEL
728bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
729 u16 filter_id);
730#endif
731#endif /* CONFIG_RPS */
732
733/* This structure contains an instance of an RX queue. */
734struct netdev_rx_queue {
735#ifdef CONFIG_RPS
736 struct rps_map __rcu *rps_map;
737 struct rps_dev_flow_table __rcu *rps_flow_table;
738#endif
739 struct kobject kobj;
740 struct net_device *dev;
741 struct xdp_rxq_info xdp_rxq;
742#ifdef CONFIG_XDP_SOCKETS
743 struct xsk_buff_pool *pool;
744#endif
745} ____cacheline_aligned_in_smp;
746
747/*
748 * RX queue sysfs structures and functions.
749 */
750struct rx_queue_attribute {
751 struct attribute attr;
752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
753 ssize_t (*store)(struct netdev_rx_queue *queue,
754 const char *buf, size_t len);
755};
756
757#ifdef CONFIG_XPS
758/*
759 * This structure holds an XPS map which can be of variable length. The
760 * map is an array of queues.
761 */
762struct xps_map {
763 unsigned int len;
764 unsigned int alloc_len;
765 struct rcu_head rcu;
766 u16 queues[];
767};
768#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
769#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
770 - sizeof(struct xps_map)) / sizeof(u16))
771
772/*
773 * This structure holds all XPS maps for device. Maps are indexed by CPU.
774 */
775struct xps_dev_maps {
776 struct rcu_head rcu;
777 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
778};
779
780#define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
781 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
782
783#define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
784 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
785
786#endif /* CONFIG_XPS */
787
788#define TC_MAX_QUEUE 16
789#define TC_BITMASK 15
790/* HW offloaded queuing disciplines txq count and offset maps */
791struct netdev_tc_txq {
792 u16 count;
793 u16 offset;
794};
795
796#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
797/*
798 * This structure is to hold information about the device
799 * configured to run FCoE protocol stack.
800 */
801struct netdev_fcoe_hbainfo {
802 char manufacturer[64];
803 char serial_number[64];
804 char hardware_version[64];
805 char driver_version[64];
806 char optionrom_version[64];
807 char firmware_version[64];
808 char model[256];
809 char model_description[256];
810};
811#endif
812
813#define MAX_PHYS_ITEM_ID_LEN 32
814
815/* This structure holds a unique identifier to identify some
816 * physical item (port for example) used by a netdevice.
817 */
818struct netdev_phys_item_id {
819 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
820 unsigned char id_len;
821};
822
823static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
824 struct netdev_phys_item_id *b)
825{
826 return a->id_len == b->id_len &&
827 memcmp(a->id, b->id, a->id_len) == 0;
828}
829
830typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
831 struct sk_buff *skb,
832 struct net_device *sb_dev);
833
834enum tc_setup_type {
835 TC_SETUP_QDISC_MQPRIO,
836 TC_SETUP_CLSU32,
837 TC_SETUP_CLSFLOWER,
838 TC_SETUP_CLSMATCHALL,
839 TC_SETUP_CLSBPF,
840 TC_SETUP_BLOCK,
841 TC_SETUP_QDISC_CBS,
842 TC_SETUP_QDISC_RED,
843 TC_SETUP_QDISC_PRIO,
844 TC_SETUP_QDISC_MQ,
845 TC_SETUP_QDISC_ETF,
846 TC_SETUP_ROOT_QDISC,
847 TC_SETUP_QDISC_GRED,
848 TC_SETUP_QDISC_TAPRIO,
849 TC_SETUP_FT,
850 TC_SETUP_QDISC_ETS,
851 TC_SETUP_QDISC_TBF,
852 TC_SETUP_QDISC_FIFO,
853 TC_SETUP_QDISC_HTB,
854};
855
856/* These structures hold the attributes of bpf state that are being passed
857 * to the netdevice through the bpf op.
858 */
859enum bpf_netdev_command {
860 /* Set or clear a bpf program used in the earliest stages of packet
861 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
862 * is responsible for calling bpf_prog_put on any old progs that are
863 * stored. In case of error, the callee need not release the new prog
864 * reference, but on success it takes ownership and must bpf_prog_put
865 * when it is no longer used.
866 */
867 XDP_SETUP_PROG,
868 XDP_SETUP_PROG_HW,
869 /* BPF program for offload callbacks, invoked at program load time. */
870 BPF_OFFLOAD_MAP_ALLOC,
871 BPF_OFFLOAD_MAP_FREE,
872 XDP_SETUP_XSK_POOL,
873};
874
875struct bpf_prog_offload_ops;
876struct netlink_ext_ack;
877struct xdp_umem;
878struct xdp_dev_bulk_queue;
879struct bpf_xdp_link;
880
881enum bpf_xdp_mode {
882 XDP_MODE_SKB = 0,
883 XDP_MODE_DRV = 1,
884 XDP_MODE_HW = 2,
885 __MAX_XDP_MODE
886};
887
888struct bpf_xdp_entity {
889 struct bpf_prog *prog;
890 struct bpf_xdp_link *link;
891};
892
893struct netdev_bpf {
894 enum bpf_netdev_command command;
895 union {
896 /* XDP_SETUP_PROG */
897 struct {
898 u32 flags;
899 struct bpf_prog *prog;
900 struct netlink_ext_ack *extack;
901 };
902 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
903 struct {
904 struct bpf_offloaded_map *offmap;
905 };
906 /* XDP_SETUP_XSK_POOL */
907 struct {
908 struct xsk_buff_pool *pool;
909 u16 queue_id;
910 } xsk;
911 };
912};
913
914/* Flags for ndo_xsk_wakeup. */
915#define XDP_WAKEUP_RX (1 << 0)
916#define XDP_WAKEUP_TX (1 << 1)
917
918#ifdef CONFIG_XFRM_OFFLOAD
919struct xfrmdev_ops {
920 int (*xdo_dev_state_add) (struct xfrm_state *x);
921 void (*xdo_dev_state_delete) (struct xfrm_state *x);
922 void (*xdo_dev_state_free) (struct xfrm_state *x);
923 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
924 struct xfrm_state *x);
925 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
926};
927#endif
928
929struct dev_ifalias {
930 struct rcu_head rcuhead;
931 char ifalias[];
932};
933
934struct devlink;
935struct tlsdev_ops;
936
937struct netdev_name_node {
938 struct hlist_node hlist;
939 struct list_head list;
940 struct net_device *dev;
941 const char *name;
942};
943
944int netdev_name_node_alt_create(struct net_device *dev, const char *name);
945int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
946
947struct netdev_net_notifier {
948 struct list_head list;
949 struct notifier_block *nb;
950};
951
952/*
953 * This structure defines the management hooks for network devices.
954 * The following hooks can be defined; unless noted otherwise, they are
955 * optional and can be filled with a null pointer.
956 *
957 * int (*ndo_init)(struct net_device *dev);
958 * This function is called once when a network device is registered.
959 * The network device can use this for any late stage initialization
960 * or semantic validation. It can fail with an error code which will
961 * be propagated back to register_netdev.
962 *
963 * void (*ndo_uninit)(struct net_device *dev);
964 * This function is called when device is unregistered or when registration
965 * fails. It is not called if init fails.
966 *
967 * int (*ndo_open)(struct net_device *dev);
968 * This function is called when a network device transitions to the up
969 * state.
970 *
971 * int (*ndo_stop)(struct net_device *dev);
972 * This function is called when a network device transitions to the down
973 * state.
974 *
975 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
976 * struct net_device *dev);
977 * Called when a packet needs to be transmitted.
978 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
979 * the queue before that can happen; it's for obsolete devices and weird
980 * corner cases, but the stack really does a non-trivial amount
981 * of useless work if you return NETDEV_TX_BUSY.
982 * Required; cannot be NULL.
983 *
984 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
985 * struct net_device *dev
986 * netdev_features_t features);
987 * Called by core transmit path to determine if device is capable of
988 * performing offload operations on a given packet. This is to give
989 * the device an opportunity to implement any restrictions that cannot
990 * be otherwise expressed by feature flags. The check is called with
991 * the set of features that the stack has calculated and it returns
992 * those the driver believes to be appropriate.
993 *
994 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
995 * struct net_device *sb_dev);
996 * Called to decide which queue to use when device supports multiple
997 * transmit queues.
998 *
999 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1000 * This function is called to allow device receiver to make
1001 * changes to configuration when multicast or promiscuous is enabled.
1002 *
1003 * void (*ndo_set_rx_mode)(struct net_device *dev);
1004 * This function is called device changes address list filtering.
1005 * If driver handles unicast address filtering, it should set
1006 * IFF_UNICAST_FLT in its priv_flags.
1007 *
1008 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1009 * This function is called when the Media Access Control address
1010 * needs to be changed. If this interface is not defined, the
1011 * MAC address can not be changed.
1012 *
1013 * int (*ndo_validate_addr)(struct net_device *dev);
1014 * Test if Media Access Control address is valid for the device.
1015 *
1016 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1017 * Called when a user requests an ioctl which can't be handled by
1018 * the generic interface code. If not defined ioctls return
1019 * not supported error code.
1020 *
1021 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1022 * Used to set network devices bus interface parameters. This interface
1023 * is retained for legacy reasons; new devices should use the bus
1024 * interface (PCI) for low level management.
1025 *
1026 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1027 * Called when a user wants to change the Maximum Transfer Unit
1028 * of a device.
1029 *
1030 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1031 * Callback used when the transmitter has not made any progress
1032 * for dev->watchdog ticks.
1033 *
1034 * void (*ndo_get_stats64)(struct net_device *dev,
1035 * struct rtnl_link_stats64 *storage);
1036 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1037 * Called when a user wants to get the network device usage
1038 * statistics. Drivers must do one of the following:
1039 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1040 * rtnl_link_stats64 structure passed by the caller.
1041 * 2. Define @ndo_get_stats to update a net_device_stats structure
1042 * (which should normally be dev->stats) and return a pointer to
1043 * it. The structure may be changed asynchronously only if each
1044 * field is written atomically.
1045 * 3. Update dev->stats asynchronously and atomically, and define
1046 * neither operation.
1047 *
1048 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1049 * Return true if this device supports offload stats of this attr_id.
1050 *
1051 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1052 * void *attr_data)
1053 * Get statistics for offload operations by attr_id. Write it into the
1054 * attr_data pointer.
1055 *
1056 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1057 * If device supports VLAN filtering this function is called when a
1058 * VLAN id is registered.
1059 *
1060 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1061 * If device supports VLAN filtering this function is called when a
1062 * VLAN id is unregistered.
1063 *
1064 * void (*ndo_poll_controller)(struct net_device *dev);
1065 *
1066 * SR-IOV management functions.
1067 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1068 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1069 * u8 qos, __be16 proto);
1070 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1071 * int max_tx_rate);
1072 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1073 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1074 * int (*ndo_get_vf_config)(struct net_device *dev,
1075 * int vf, struct ifla_vf_info *ivf);
1076 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1077 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1078 * struct nlattr *port[]);
1079 *
1080 * Enable or disable the VF ability to query its RSS Redirection Table and
1081 * Hash Key. This is needed since on some devices VF share this information
1082 * with PF and querying it may introduce a theoretical security risk.
1083 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1084 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1085 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1086 * void *type_data);
1087 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1088 * This is always called from the stack with the rtnl lock held and netif
1089 * tx queues stopped. This allows the netdevice to perform queue
1090 * management safely.
1091 *
1092 * Fiber Channel over Ethernet (FCoE) offload functions.
1093 * int (*ndo_fcoe_enable)(struct net_device *dev);
1094 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1095 * so the underlying device can perform whatever needed configuration or
1096 * initialization to support acceleration of FCoE traffic.
1097 *
1098 * int (*ndo_fcoe_disable)(struct net_device *dev);
1099 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1100 * so the underlying device can perform whatever needed clean-ups to
1101 * stop supporting acceleration of FCoE traffic.
1102 *
1103 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1104 * struct scatterlist *sgl, unsigned int sgc);
1105 * Called when the FCoE Initiator wants to initialize an I/O that
1106 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1107 * perform necessary setup and returns 1 to indicate the device is set up
1108 * successfully to perform DDP on this I/O, otherwise this returns 0.
1109 *
1110 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1111 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1112 * indicated by the FC exchange id 'xid', so the underlying device can
1113 * clean up and reuse resources for later DDP requests.
1114 *
1115 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1116 * struct scatterlist *sgl, unsigned int sgc);
1117 * Called when the FCoE Target wants to initialize an I/O that
1118 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1119 * perform necessary setup and returns 1 to indicate the device is set up
1120 * successfully to perform DDP on this I/O, otherwise this returns 0.
1121 *
1122 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1123 * struct netdev_fcoe_hbainfo *hbainfo);
1124 * Called when the FCoE Protocol stack wants information on the underlying
1125 * device. This information is utilized by the FCoE protocol stack to
1126 * register attributes with Fiber Channel management service as per the
1127 * FC-GS Fabric Device Management Information(FDMI) specification.
1128 *
1129 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1130 * Called when the underlying device wants to override default World Wide
1131 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1132 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1133 * protocol stack to use.
1134 *
1135 * RFS acceleration.
1136 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1137 * u16 rxq_index, u32 flow_id);
1138 * Set hardware filter for RFS. rxq_index is the target queue index;
1139 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1140 * Return the filter ID on success, or a negative error code.
1141 *
1142 * Slave management functions (for bridge, bonding, etc).
1143 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1144 * Called to make another netdev an underling.
1145 *
1146 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1147 * Called to release previously enslaved netdev.
1148 *
1149 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1150 * struct sk_buff *skb,
1151 * bool all_slaves);
1152 * Get the xmit slave of master device. If all_slaves is true, function
1153 * assume all the slaves can transmit.
1154 *
1155 * Feature/offload setting functions.
1156 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1157 * netdev_features_t features);
1158 * Adjusts the requested feature flags according to device-specific
1159 * constraints, and returns the resulting flags. Must not modify
1160 * the device state.
1161 *
1162 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1163 * Called to update device configuration to new features. Passed
1164 * feature set might be less than what was returned by ndo_fix_features()).
1165 * Must return >0 or -errno if it changed dev->features itself.
1166 *
1167 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1168 * struct net_device *dev,
1169 * const unsigned char *addr, u16 vid, u16 flags,
1170 * struct netlink_ext_ack *extack);
1171 * Adds an FDB entry to dev for addr.
1172 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1173 * struct net_device *dev,
1174 * const unsigned char *addr, u16 vid)
1175 * Deletes the FDB entry from dev coresponding to addr.
1176 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1177 * struct net_device *dev, struct net_device *filter_dev,
1178 * int *idx)
1179 * Used to add FDB entries to dump requests. Implementers should add
1180 * entries to skb and update idx with the number of entries.
1181 *
1182 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1183 * u16 flags, struct netlink_ext_ack *extack)
1184 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1185 * struct net_device *dev, u32 filter_mask,
1186 * int nlflags)
1187 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1188 * u16 flags);
1189 *
1190 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1191 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1192 * which do not represent real hardware may define this to allow their
1193 * userspace components to manage their virtual carrier state. Devices
1194 * that determine carrier state from physical hardware properties (eg
1195 * network cables) or protocol-dependent mechanisms (eg
1196 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1197 *
1198 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1199 * struct netdev_phys_item_id *ppid);
1200 * Called to get ID of physical port of this device. If driver does
1201 * not implement this, it is assumed that the hw is not able to have
1202 * multiple net devices on single physical port.
1203 *
1204 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1205 * struct netdev_phys_item_id *ppid)
1206 * Called to get the parent ID of the physical port of this device.
1207 *
1208 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1209 * struct net_device *dev)
1210 * Called by upper layer devices to accelerate switching or other
1211 * station functionality into hardware. 'pdev is the lowerdev
1212 * to use for the offload and 'dev' is the net device that will
1213 * back the offload. Returns a pointer to the private structure
1214 * the upper layer will maintain.
1215 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1216 * Called by upper layer device to delete the station created
1217 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1218 * the station and priv is the structure returned by the add
1219 * operation.
1220 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1221 * int queue_index, u32 maxrate);
1222 * Called when a user wants to set a max-rate limitation of specific
1223 * TX queue.
1224 * int (*ndo_get_iflink)(const struct net_device *dev);
1225 * Called to get the iflink value of this device.
1226 * void (*ndo_change_proto_down)(struct net_device *dev,
1227 * bool proto_down);
1228 * This function is used to pass protocol port error state information
1229 * to the switch driver. The switch driver can react to the proto_down
1230 * by doing a phys down on the associated switch port.
1231 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1232 * This function is used to get egress tunnel information for given skb.
1233 * This is useful for retrieving outer tunnel header parameters while
1234 * sampling packet.
1235 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1236 * This function is used to specify the headroom that the skb must
1237 * consider when allocation skb during packet reception. Setting
1238 * appropriate rx headroom value allows avoiding skb head copy on
1239 * forward. Setting a negative value resets the rx headroom to the
1240 * default value.
1241 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1242 * This function is used to set or query state related to XDP on the
1243 * netdevice and manage BPF offload. See definition of
1244 * enum bpf_netdev_command for details.
1245 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1246 * u32 flags);
1247 * This function is used to submit @n XDP packets for transmit on a
1248 * netdevice. Returns number of frames successfully transmitted, frames
1249 * that got dropped are freed/returned via xdp_return_frame().
1250 * Returns negative number, means general error invoking ndo, meaning
1251 * no frames were xmit'ed and core-caller will free all frames.
1252 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1253 * This function is used to wake up the softirq, ksoftirqd or kthread
1254 * responsible for sending and/or receiving packets on a specific
1255 * queue id bound to an AF_XDP socket. The flags field specifies if
1256 * only RX, only Tx, or both should be woken up using the flags
1257 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1258 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1259 * Get devlink port instance associated with a given netdev.
1260 * Called with a reference on the netdevice and devlink locks only,
1261 * rtnl_lock is not held.
1262 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1263 * int cmd);
1264 * Add, change, delete or get information on an IPv4 tunnel.
1265 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1266 * If a device is paired with a peer device, return the peer instance.
1267 * The caller must be under RCU read context.
1268 */
1269struct net_device_ops {
1270 int (*ndo_init)(struct net_device *dev);
1271 void (*ndo_uninit)(struct net_device *dev);
1272 int (*ndo_open)(struct net_device *dev);
1273 int (*ndo_stop)(struct net_device *dev);
1274 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1275 struct net_device *dev);
1276 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1277 struct net_device *dev,
1278 netdev_features_t features);
1279 u16 (*ndo_select_queue)(struct net_device *dev,
1280 struct sk_buff *skb,
1281 struct net_device *sb_dev);
1282 void (*ndo_change_rx_flags)(struct net_device *dev,
1283 int flags);
1284 void (*ndo_set_rx_mode)(struct net_device *dev);
1285 int (*ndo_set_mac_address)(struct net_device *dev,
1286 void *addr);
1287 int (*ndo_validate_addr)(struct net_device *dev);
1288 int (*ndo_do_ioctl)(struct net_device *dev,
1289 struct ifreq *ifr, int cmd);
1290 int (*ndo_set_config)(struct net_device *dev,
1291 struct ifmap *map);
1292 int (*ndo_change_mtu)(struct net_device *dev,
1293 int new_mtu);
1294 int (*ndo_neigh_setup)(struct net_device *dev,
1295 struct neigh_parms *);
1296 void (*ndo_tx_timeout) (struct net_device *dev,
1297 unsigned int txqueue);
1298
1299 void (*ndo_get_stats64)(struct net_device *dev,
1300 struct rtnl_link_stats64 *storage);
1301 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1302 int (*ndo_get_offload_stats)(int attr_id,
1303 const struct net_device *dev,
1304 void *attr_data);
1305 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1306
1307 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1308 __be16 proto, u16 vid);
1309 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1310 __be16 proto, u16 vid);
1311#ifdef CONFIG_NET_POLL_CONTROLLER
1312 void (*ndo_poll_controller)(struct net_device *dev);
1313 int (*ndo_netpoll_setup)(struct net_device *dev,
1314 struct netpoll_info *info);
1315 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1316#endif
1317 int (*ndo_set_vf_mac)(struct net_device *dev,
1318 int queue, u8 *mac);
1319 int (*ndo_set_vf_vlan)(struct net_device *dev,
1320 int queue, u16 vlan,
1321 u8 qos, __be16 proto);
1322 int (*ndo_set_vf_rate)(struct net_device *dev,
1323 int vf, int min_tx_rate,
1324 int max_tx_rate);
1325 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1326 int vf, bool setting);
1327 int (*ndo_set_vf_trust)(struct net_device *dev,
1328 int vf, bool setting);
1329 int (*ndo_get_vf_config)(struct net_device *dev,
1330 int vf,
1331 struct ifla_vf_info *ivf);
1332 int (*ndo_set_vf_link_state)(struct net_device *dev,
1333 int vf, int link_state);
1334 int (*ndo_get_vf_stats)(struct net_device *dev,
1335 int vf,
1336 struct ifla_vf_stats
1337 *vf_stats);
1338 int (*ndo_set_vf_port)(struct net_device *dev,
1339 int vf,
1340 struct nlattr *port[]);
1341 int (*ndo_get_vf_port)(struct net_device *dev,
1342 int vf, struct sk_buff *skb);
1343 int (*ndo_get_vf_guid)(struct net_device *dev,
1344 int vf,
1345 struct ifla_vf_guid *node_guid,
1346 struct ifla_vf_guid *port_guid);
1347 int (*ndo_set_vf_guid)(struct net_device *dev,
1348 int vf, u64 guid,
1349 int guid_type);
1350 int (*ndo_set_vf_rss_query_en)(
1351 struct net_device *dev,
1352 int vf, bool setting);
1353 int (*ndo_setup_tc)(struct net_device *dev,
1354 enum tc_setup_type type,
1355 void *type_data);
1356#if IS_ENABLED(CONFIG_FCOE)
1357 int (*ndo_fcoe_enable)(struct net_device *dev);
1358 int (*ndo_fcoe_disable)(struct net_device *dev);
1359 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1360 u16 xid,
1361 struct scatterlist *sgl,
1362 unsigned int sgc);
1363 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1364 u16 xid);
1365 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1366 u16 xid,
1367 struct scatterlist *sgl,
1368 unsigned int sgc);
1369 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1370 struct netdev_fcoe_hbainfo *hbainfo);
1371#endif
1372
1373#if IS_ENABLED(CONFIG_LIBFCOE)
1374#define NETDEV_FCOE_WWNN 0
1375#define NETDEV_FCOE_WWPN 1
1376 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1377 u64 *wwn, int type);
1378#endif
1379
1380#ifdef CONFIG_RFS_ACCEL
1381 int (*ndo_rx_flow_steer)(struct net_device *dev,
1382 const struct sk_buff *skb,
1383 u16 rxq_index,
1384 u32 flow_id);
1385#endif
1386 int (*ndo_add_slave)(struct net_device *dev,
1387 struct net_device *slave_dev,
1388 struct netlink_ext_ack *extack);
1389 int (*ndo_del_slave)(struct net_device *dev,
1390 struct net_device *slave_dev);
1391 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1392 struct sk_buff *skb,
1393 bool all_slaves);
1394 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1395 struct sock *sk);
1396 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1397 netdev_features_t features);
1398 int (*ndo_set_features)(struct net_device *dev,
1399 netdev_features_t features);
1400 int (*ndo_neigh_construct)(struct net_device *dev,
1401 struct neighbour *n);
1402 void (*ndo_neigh_destroy)(struct net_device *dev,
1403 struct neighbour *n);
1404
1405 int (*ndo_fdb_add)(struct ndmsg *ndm,
1406 struct nlattr *tb[],
1407 struct net_device *dev,
1408 const unsigned char *addr,
1409 u16 vid,
1410 u16 flags,
1411 struct netlink_ext_ack *extack);
1412 int (*ndo_fdb_del)(struct ndmsg *ndm,
1413 struct nlattr *tb[],
1414 struct net_device *dev,
1415 const unsigned char *addr,
1416 u16 vid);
1417 int (*ndo_fdb_dump)(struct sk_buff *skb,
1418 struct netlink_callback *cb,
1419 struct net_device *dev,
1420 struct net_device *filter_dev,
1421 int *idx);
1422 int (*ndo_fdb_get)(struct sk_buff *skb,
1423 struct nlattr *tb[],
1424 struct net_device *dev,
1425 const unsigned char *addr,
1426 u16 vid, u32 portid, u32 seq,
1427 struct netlink_ext_ack *extack);
1428 int (*ndo_bridge_setlink)(struct net_device *dev,
1429 struct nlmsghdr *nlh,
1430 u16 flags,
1431 struct netlink_ext_ack *extack);
1432 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1433 u32 pid, u32 seq,
1434 struct net_device *dev,
1435 u32 filter_mask,
1436 int nlflags);
1437 int (*ndo_bridge_dellink)(struct net_device *dev,
1438 struct nlmsghdr *nlh,
1439 u16 flags);
1440 int (*ndo_change_carrier)(struct net_device *dev,
1441 bool new_carrier);
1442 int (*ndo_get_phys_port_id)(struct net_device *dev,
1443 struct netdev_phys_item_id *ppid);
1444 int (*ndo_get_port_parent_id)(struct net_device *dev,
1445 struct netdev_phys_item_id *ppid);
1446 int (*ndo_get_phys_port_name)(struct net_device *dev,
1447 char *name, size_t len);
1448 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1449 struct net_device *dev);
1450 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1451 void *priv);
1452
1453 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1454 int queue_index,
1455 u32 maxrate);
1456 int (*ndo_get_iflink)(const struct net_device *dev);
1457 int (*ndo_change_proto_down)(struct net_device *dev,
1458 bool proto_down);
1459 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1460 struct sk_buff *skb);
1461 void (*ndo_set_rx_headroom)(struct net_device *dev,
1462 int needed_headroom);
1463 int (*ndo_bpf)(struct net_device *dev,
1464 struct netdev_bpf *bpf);
1465 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1466 struct xdp_frame **xdp,
1467 u32 flags);
1468 int (*ndo_xsk_wakeup)(struct net_device *dev,
1469 u32 queue_id, u32 flags);
1470 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1471 int (*ndo_tunnel_ctl)(struct net_device *dev,
1472 struct ip_tunnel_parm *p, int cmd);
1473 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1474};
1475
1476/**
1477 * enum netdev_priv_flags - &struct net_device priv_flags
1478 *
1479 * These are the &struct net_device, they are only set internally
1480 * by drivers and used in the kernel. These flags are invisible to
1481 * userspace; this means that the order of these flags can change
1482 * during any kernel release.
1483 *
1484 * You should have a pretty good reason to be extending these flags.
1485 *
1486 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1487 * @IFF_EBRIDGE: Ethernet bridging device
1488 * @IFF_BONDING: bonding master or slave
1489 * @IFF_ISATAP: ISATAP interface (RFC4214)
1490 * @IFF_WAN_HDLC: WAN HDLC device
1491 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1492 * release skb->dst
1493 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1494 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1495 * @IFF_MACVLAN_PORT: device used as macvlan port
1496 * @IFF_BRIDGE_PORT: device used as bridge port
1497 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1498 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1499 * @IFF_UNICAST_FLT: Supports unicast filtering
1500 * @IFF_TEAM_PORT: device used as team port
1501 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1502 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1503 * change when it's running
1504 * @IFF_MACVLAN: Macvlan device
1505 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1506 * underlying stacked devices
1507 * @IFF_L3MDEV_MASTER: device is an L3 master device
1508 * @IFF_NO_QUEUE: device can run without qdisc attached
1509 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1510 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1511 * @IFF_TEAM: device is a team device
1512 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1513 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1514 * entity (i.e. the master device for bridged veth)
1515 * @IFF_MACSEC: device is a MACsec device
1516 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1517 * @IFF_FAILOVER: device is a failover master device
1518 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1519 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1520 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1521 */
1522enum netdev_priv_flags {
1523 IFF_802_1Q_VLAN = 1<<0,
1524 IFF_EBRIDGE = 1<<1,
1525 IFF_BONDING = 1<<2,
1526 IFF_ISATAP = 1<<3,
1527 IFF_WAN_HDLC = 1<<4,
1528 IFF_XMIT_DST_RELEASE = 1<<5,
1529 IFF_DONT_BRIDGE = 1<<6,
1530 IFF_DISABLE_NETPOLL = 1<<7,
1531 IFF_MACVLAN_PORT = 1<<8,
1532 IFF_BRIDGE_PORT = 1<<9,
1533 IFF_OVS_DATAPATH = 1<<10,
1534 IFF_TX_SKB_SHARING = 1<<11,
1535 IFF_UNICAST_FLT = 1<<12,
1536 IFF_TEAM_PORT = 1<<13,
1537 IFF_SUPP_NOFCS = 1<<14,
1538 IFF_LIVE_ADDR_CHANGE = 1<<15,
1539 IFF_MACVLAN = 1<<16,
1540 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1541 IFF_L3MDEV_MASTER = 1<<18,
1542 IFF_NO_QUEUE = 1<<19,
1543 IFF_OPENVSWITCH = 1<<20,
1544 IFF_L3MDEV_SLAVE = 1<<21,
1545 IFF_TEAM = 1<<22,
1546 IFF_RXFH_CONFIGURED = 1<<23,
1547 IFF_PHONY_HEADROOM = 1<<24,
1548 IFF_MACSEC = 1<<25,
1549 IFF_NO_RX_HANDLER = 1<<26,
1550 IFF_FAILOVER = 1<<27,
1551 IFF_FAILOVER_SLAVE = 1<<28,
1552 IFF_L3MDEV_RX_HANDLER = 1<<29,
1553 IFF_LIVE_RENAME_OK = 1<<30,
1554};
1555
1556#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1557#define IFF_EBRIDGE IFF_EBRIDGE
1558#define IFF_BONDING IFF_BONDING
1559#define IFF_ISATAP IFF_ISATAP
1560#define IFF_WAN_HDLC IFF_WAN_HDLC
1561#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1562#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1563#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1564#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1565#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1566#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1567#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1568#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1569#define IFF_TEAM_PORT IFF_TEAM_PORT
1570#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1571#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1572#define IFF_MACVLAN IFF_MACVLAN
1573#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1574#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1575#define IFF_NO_QUEUE IFF_NO_QUEUE
1576#define IFF_OPENVSWITCH IFF_OPENVSWITCH
1577#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1578#define IFF_TEAM IFF_TEAM
1579#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1580#define IFF_MACSEC IFF_MACSEC
1581#define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1582#define IFF_FAILOVER IFF_FAILOVER
1583#define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1584#define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1585#define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1586
1587/* Specifies the type of the struct net_device::ml_priv pointer */
1588enum netdev_ml_priv_type {
1589 ML_PRIV_NONE,
1590 ML_PRIV_CAN,
1591};
1592
1593/**
1594 * struct net_device - The DEVICE structure.
1595 *
1596 * Actually, this whole structure is a big mistake. It mixes I/O
1597 * data with strictly "high-level" data, and it has to know about
1598 * almost every data structure used in the INET module.
1599 *
1600 * @name: This is the first field of the "visible" part of this structure
1601 * (i.e. as seen by users in the "Space.c" file). It is the name
1602 * of the interface.
1603 *
1604 * @name_node: Name hashlist node
1605 * @ifalias: SNMP alias
1606 * @mem_end: Shared memory end
1607 * @mem_start: Shared memory start
1608 * @base_addr: Device I/O address
1609 * @irq: Device IRQ number
1610 *
1611 * @state: Generic network queuing layer state, see netdev_state_t
1612 * @dev_list: The global list of network devices
1613 * @napi_list: List entry used for polling NAPI devices
1614 * @unreg_list: List entry when we are unregistering the
1615 * device; see the function unregister_netdev
1616 * @close_list: List entry used when we are closing the device
1617 * @ptype_all: Device-specific packet handlers for all protocols
1618 * @ptype_specific: Device-specific, protocol-specific packet handlers
1619 *
1620 * @adj_list: Directly linked devices, like slaves for bonding
1621 * @features: Currently active device features
1622 * @hw_features: User-changeable features
1623 *
1624 * @wanted_features: User-requested features
1625 * @vlan_features: Mask of features inheritable by VLAN devices
1626 *
1627 * @hw_enc_features: Mask of features inherited by encapsulating devices
1628 * This field indicates what encapsulation
1629 * offloads the hardware is capable of doing,
1630 * and drivers will need to set them appropriately.
1631 *
1632 * @mpls_features: Mask of features inheritable by MPLS
1633 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1634 *
1635 * @ifindex: interface index
1636 * @group: The group the device belongs to
1637 *
1638 * @stats: Statistics struct, which was left as a legacy, use
1639 * rtnl_link_stats64 instead
1640 *
1641 * @rx_dropped: Dropped packets by core network,
1642 * do not use this in drivers
1643 * @tx_dropped: Dropped packets by core network,
1644 * do not use this in drivers
1645 * @rx_nohandler: nohandler dropped packets by core network on
1646 * inactive devices, do not use this in drivers
1647 * @carrier_up_count: Number of times the carrier has been up
1648 * @carrier_down_count: Number of times the carrier has been down
1649 *
1650 * @wireless_handlers: List of functions to handle Wireless Extensions,
1651 * instead of ioctl,
1652 * see <net/iw_handler.h> for details.
1653 * @wireless_data: Instance data managed by the core of wireless extensions
1654 *
1655 * @netdev_ops: Includes several pointers to callbacks,
1656 * if one wants to override the ndo_*() functions
1657 * @ethtool_ops: Management operations
1658 * @l3mdev_ops: Layer 3 master device operations
1659 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1660 * discovery handling. Necessary for e.g. 6LoWPAN.
1661 * @xfrmdev_ops: Transformation offload operations
1662 * @tlsdev_ops: Transport Layer Security offload operations
1663 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1664 * of Layer 2 headers.
1665 *
1666 * @flags: Interface flags (a la BSD)
1667 * @priv_flags: Like 'flags' but invisible to userspace,
1668 * see if.h for the definitions
1669 * @gflags: Global flags ( kept as legacy )
1670 * @padded: How much padding added by alloc_netdev()
1671 * @operstate: RFC2863 operstate
1672 * @link_mode: Mapping policy to operstate
1673 * @if_port: Selectable AUI, TP, ...
1674 * @dma: DMA channel
1675 * @mtu: Interface MTU value
1676 * @min_mtu: Interface Minimum MTU value
1677 * @max_mtu: Interface Maximum MTU value
1678 * @type: Interface hardware type
1679 * @hard_header_len: Maximum hardware header length.
1680 * @min_header_len: Minimum hardware header length
1681 *
1682 * @needed_headroom: Extra headroom the hardware may need, but not in all
1683 * cases can this be guaranteed
1684 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1685 * cases can this be guaranteed. Some cases also use
1686 * LL_MAX_HEADER instead to allocate the skb
1687 *
1688 * interface address info:
1689 *
1690 * @perm_addr: Permanent hw address
1691 * @addr_assign_type: Hw address assignment type
1692 * @addr_len: Hardware address length
1693 * @upper_level: Maximum depth level of upper devices.
1694 * @lower_level: Maximum depth level of lower devices.
1695 * @neigh_priv_len: Used in neigh_alloc()
1696 * @dev_id: Used to differentiate devices that share
1697 * the same link layer address
1698 * @dev_port: Used to differentiate devices that share
1699 * the same function
1700 * @addr_list_lock: XXX: need comments on this one
1701 * @name_assign_type: network interface name assignment type
1702 * @uc_promisc: Counter that indicates promiscuous mode
1703 * has been enabled due to the need to listen to
1704 * additional unicast addresses in a device that
1705 * does not implement ndo_set_rx_mode()
1706 * @uc: unicast mac addresses
1707 * @mc: multicast mac addresses
1708 * @dev_addrs: list of device hw addresses
1709 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1710 * @promiscuity: Number of times the NIC is told to work in
1711 * promiscuous mode; if it becomes 0 the NIC will
1712 * exit promiscuous mode
1713 * @allmulti: Counter, enables or disables allmulticast mode
1714 *
1715 * @vlan_info: VLAN info
1716 * @dsa_ptr: dsa specific data
1717 * @tipc_ptr: TIPC specific data
1718 * @atalk_ptr: AppleTalk link
1719 * @ip_ptr: IPv4 specific data
1720 * @dn_ptr: DECnet specific data
1721 * @ip6_ptr: IPv6 specific data
1722 * @ax25_ptr: AX.25 specific data
1723 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1724 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1725 * device struct
1726 * @mpls_ptr: mpls_dev struct pointer
1727 *
1728 * @dev_addr: Hw address (before bcast,
1729 * because most packets are unicast)
1730 *
1731 * @_rx: Array of RX queues
1732 * @num_rx_queues: Number of RX queues
1733 * allocated at register_netdev() time
1734 * @real_num_rx_queues: Number of RX queues currently active in device
1735 * @xdp_prog: XDP sockets filter program pointer
1736 * @gro_flush_timeout: timeout for GRO layer in NAPI
1737 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1738 * allow to avoid NIC hard IRQ, on busy queues.
1739 *
1740 * @rx_handler: handler for received packets
1741 * @rx_handler_data: XXX: need comments on this one
1742 * @miniq_ingress: ingress/clsact qdisc specific data for
1743 * ingress processing
1744 * @ingress_queue: XXX: need comments on this one
1745 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1746 * @broadcast: hw bcast address
1747 *
1748 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1749 * indexed by RX queue number. Assigned by driver.
1750 * This must only be set if the ndo_rx_flow_steer
1751 * operation is defined
1752 * @index_hlist: Device index hash chain
1753 *
1754 * @_tx: Array of TX queues
1755 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1756 * @real_num_tx_queues: Number of TX queues currently active in device
1757 * @qdisc: Root qdisc from userspace point of view
1758 * @tx_queue_len: Max frames per queue allowed
1759 * @tx_global_lock: XXX: need comments on this one
1760 * @xdp_bulkq: XDP device bulk queue
1761 * @xps_cpus_map: all CPUs map for XPS device
1762 * @xps_rxqs_map: all RXQs map for XPS device
1763 *
1764 * @xps_maps: XXX: need comments on this one
1765 * @miniq_egress: clsact qdisc specific data for
1766 * egress processing
1767 * @qdisc_hash: qdisc hash table
1768 * @watchdog_timeo: Represents the timeout that is used by
1769 * the watchdog (see dev_watchdog())
1770 * @watchdog_timer: List of timers
1771 *
1772 * @proto_down_reason: reason a netdev interface is held down
1773 * @pcpu_refcnt: Number of references to this device
1774 * @todo_list: Delayed register/unregister
1775 * @link_watch_list: XXX: need comments on this one
1776 *
1777 * @reg_state: Register/unregister state machine
1778 * @dismantle: Device is going to be freed
1779 * @rtnl_link_state: This enum represents the phases of creating
1780 * a new link
1781 *
1782 * @needs_free_netdev: Should unregister perform free_netdev?
1783 * @priv_destructor: Called from unregister
1784 * @npinfo: XXX: need comments on this one
1785 * @nd_net: Network namespace this network device is inside
1786 *
1787 * @ml_priv: Mid-layer private
1788 * @ml_priv_type: Mid-layer private type
1789 * @lstats: Loopback statistics
1790 * @tstats: Tunnel statistics
1791 * @dstats: Dummy statistics
1792 * @vstats: Virtual ethernet statistics
1793 *
1794 * @garp_port: GARP
1795 * @mrp_port: MRP
1796 *
1797 * @dev: Class/net/name entry
1798 * @sysfs_groups: Space for optional device, statistics and wireless
1799 * sysfs groups
1800 *
1801 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1802 * @rtnl_link_ops: Rtnl_link_ops
1803 *
1804 * @gso_max_size: Maximum size of generic segmentation offload
1805 * @gso_max_segs: Maximum number of segments that can be passed to the
1806 * NIC for GSO
1807 *
1808 * @dcbnl_ops: Data Center Bridging netlink ops
1809 * @num_tc: Number of traffic classes in the net device
1810 * @tc_to_txq: XXX: need comments on this one
1811 * @prio_tc_map: XXX: need comments on this one
1812 *
1813 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1814 *
1815 * @priomap: XXX: need comments on this one
1816 * @phydev: Physical device may attach itself
1817 * for hardware timestamping
1818 * @sfp_bus: attached &struct sfp_bus structure.
1819 *
1820 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1821 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1822 *
1823 * @proto_down: protocol port state information can be sent to the
1824 * switch driver and used to set the phys state of the
1825 * switch port.
1826 *
1827 * @wol_enabled: Wake-on-LAN is enabled
1828 *
1829 * @threaded: napi threaded mode is enabled
1830 *
1831 * @net_notifier_list: List of per-net netdev notifier block
1832 * that follow this device when it is moved
1833 * to another network namespace.
1834 *
1835 * @macsec_ops: MACsec offloading ops
1836 *
1837 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1838 * offload capabilities of the device
1839 * @udp_tunnel_nic: UDP tunnel offload state
1840 * @xdp_state: stores info on attached XDP BPF programs
1841 *
1842 * @nested_level: Used as as a parameter of spin_lock_nested() of
1843 * dev->addr_list_lock.
1844 * @unlink_list: As netif_addr_lock() can be called recursively,
1845 * keep a list of interfaces to be deleted.
1846 *
1847 * FIXME: cleanup struct net_device such that network protocol info
1848 * moves out.
1849 */
1850
1851struct net_device {
1852 char name[IFNAMSIZ];
1853 struct netdev_name_node *name_node;
1854 struct dev_ifalias __rcu *ifalias;
1855 /*
1856 * I/O specific fields
1857 * FIXME: Merge these and struct ifmap into one
1858 */
1859 unsigned long mem_end;
1860 unsigned long mem_start;
1861 unsigned long base_addr;
1862
1863 /*
1864 * Some hardware also needs these fields (state,dev_list,
1865 * napi_list,unreg_list,close_list) but they are not
1866 * part of the usual set specified in Space.c.
1867 */
1868
1869 unsigned long state;
1870
1871 struct list_head dev_list;
1872 struct list_head napi_list;
1873 struct list_head unreg_list;
1874 struct list_head close_list;
1875 struct list_head ptype_all;
1876 struct list_head ptype_specific;
1877
1878 struct {
1879 struct list_head upper;
1880 struct list_head lower;
1881 } adj_list;
1882
1883 /* Read-mostly cache-line for fast-path access */
1884 unsigned int flags;
1885 unsigned int priv_flags;
1886 const struct net_device_ops *netdev_ops;
1887 int ifindex;
1888 unsigned short gflags;
1889 unsigned short hard_header_len;
1890
1891 /* Note : dev->mtu is often read without holding a lock.
1892 * Writers usually hold RTNL.
1893 * It is recommended to use READ_ONCE() to annotate the reads,
1894 * and to use WRITE_ONCE() to annotate the writes.
1895 */
1896 unsigned int mtu;
1897 unsigned short needed_headroom;
1898 unsigned short needed_tailroom;
1899
1900 netdev_features_t features;
1901 netdev_features_t hw_features;
1902 netdev_features_t wanted_features;
1903 netdev_features_t vlan_features;
1904 netdev_features_t hw_enc_features;
1905 netdev_features_t mpls_features;
1906 netdev_features_t gso_partial_features;
1907
1908 unsigned int min_mtu;
1909 unsigned int max_mtu;
1910 unsigned short type;
1911 unsigned char min_header_len;
1912 unsigned char name_assign_type;
1913
1914 int group;
1915
1916 struct net_device_stats stats; /* not used by modern drivers */
1917
1918 atomic_long_t rx_dropped;
1919 atomic_long_t tx_dropped;
1920 atomic_long_t rx_nohandler;
1921
1922 /* Stats to monitor link on/off, flapping */
1923 atomic_t carrier_up_count;
1924 atomic_t carrier_down_count;
1925
1926#ifdef CONFIG_WIRELESS_EXT
1927 const struct iw_handler_def *wireless_handlers;
1928 struct iw_public_data *wireless_data;
1929#endif
1930 const struct ethtool_ops *ethtool_ops;
1931#ifdef CONFIG_NET_L3_MASTER_DEV
1932 const struct l3mdev_ops *l3mdev_ops;
1933#endif
1934#if IS_ENABLED(CONFIG_IPV6)
1935 const struct ndisc_ops *ndisc_ops;
1936#endif
1937
1938#ifdef CONFIG_XFRM_OFFLOAD
1939 const struct xfrmdev_ops *xfrmdev_ops;
1940#endif
1941
1942#if IS_ENABLED(CONFIG_TLS_DEVICE)
1943 const struct tlsdev_ops *tlsdev_ops;
1944#endif
1945
1946 const struct header_ops *header_ops;
1947
1948 unsigned char operstate;
1949 unsigned char link_mode;
1950
1951 unsigned char if_port;
1952 unsigned char dma;
1953
1954 /* Interface address info. */
1955 unsigned char perm_addr[MAX_ADDR_LEN];
1956 unsigned char addr_assign_type;
1957 unsigned char addr_len;
1958 unsigned char upper_level;
1959 unsigned char lower_level;
1960
1961 unsigned short neigh_priv_len;
1962 unsigned short dev_id;
1963 unsigned short dev_port;
1964 unsigned short padded;
1965
1966 spinlock_t addr_list_lock;
1967 int irq;
1968
1969 struct netdev_hw_addr_list uc;
1970 struct netdev_hw_addr_list mc;
1971 struct netdev_hw_addr_list dev_addrs;
1972
1973#ifdef CONFIG_SYSFS
1974 struct kset *queues_kset;
1975#endif
1976#ifdef CONFIG_LOCKDEP
1977 struct list_head unlink_list;
1978#endif
1979 unsigned int promiscuity;
1980 unsigned int allmulti;
1981 bool uc_promisc;
1982#ifdef CONFIG_LOCKDEP
1983 unsigned char nested_level;
1984#endif
1985
1986
1987 /* Protocol-specific pointers */
1988
1989#if IS_ENABLED(CONFIG_VLAN_8021Q)
1990 struct vlan_info __rcu *vlan_info;
1991#endif
1992#if IS_ENABLED(CONFIG_NET_DSA)
1993 struct dsa_port *dsa_ptr;
1994#endif
1995#if IS_ENABLED(CONFIG_TIPC)
1996 struct tipc_bearer __rcu *tipc_ptr;
1997#endif
1998#if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1999 void *atalk_ptr;
2000#endif
2001 struct in_device __rcu *ip_ptr;
2002#if IS_ENABLED(CONFIG_DECNET)
2003 struct dn_dev __rcu *dn_ptr;
2004#endif
2005 struct inet6_dev __rcu *ip6_ptr;
2006#if IS_ENABLED(CONFIG_AX25)
2007 void *ax25_ptr;
2008#endif
2009 struct wireless_dev *ieee80211_ptr;
2010 struct wpan_dev *ieee802154_ptr;
2011#if IS_ENABLED(CONFIG_MPLS_ROUTING)
2012 struct mpls_dev __rcu *mpls_ptr;
2013#endif
2014
2015/*
2016 * Cache lines mostly used on receive path (including eth_type_trans())
2017 */
2018 /* Interface address info used in eth_type_trans() */
2019 unsigned char *dev_addr;
2020
2021 struct netdev_rx_queue *_rx;
2022 unsigned int num_rx_queues;
2023 unsigned int real_num_rx_queues;
2024
2025 struct bpf_prog __rcu *xdp_prog;
2026 unsigned long gro_flush_timeout;
2027 int napi_defer_hard_irqs;
2028 rx_handler_func_t __rcu *rx_handler;
2029 void __rcu *rx_handler_data;
2030
2031#ifdef CONFIG_NET_CLS_ACT
2032 struct mini_Qdisc __rcu *miniq_ingress;
2033#endif
2034 struct netdev_queue __rcu *ingress_queue;
2035#ifdef CONFIG_NETFILTER_INGRESS
2036 struct nf_hook_entries __rcu *nf_hooks_ingress;
2037#endif
2038
2039 unsigned char broadcast[MAX_ADDR_LEN];
2040#ifdef CONFIG_RFS_ACCEL
2041 struct cpu_rmap *rx_cpu_rmap;
2042#endif
2043 struct hlist_node index_hlist;
2044
2045/*
2046 * Cache lines mostly used on transmit path
2047 */
2048 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2049 unsigned int num_tx_queues;
2050 unsigned int real_num_tx_queues;
2051 struct Qdisc *qdisc;
2052 unsigned int tx_queue_len;
2053 spinlock_t tx_global_lock;
2054
2055 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2056
2057#ifdef CONFIG_XPS
2058 struct xps_dev_maps __rcu *xps_cpus_map;
2059 struct xps_dev_maps __rcu *xps_rxqs_map;
2060#endif
2061#ifdef CONFIG_NET_CLS_ACT
2062 struct mini_Qdisc __rcu *miniq_egress;
2063#endif
2064
2065#ifdef CONFIG_NET_SCHED
2066 DECLARE_HASHTABLE (qdisc_hash, 4);
2067#endif
2068 /* These may be needed for future network-power-down code. */
2069 struct timer_list watchdog_timer;
2070 int watchdog_timeo;
2071
2072 u32 proto_down_reason;
2073
2074 struct list_head todo_list;
2075 int __percpu *pcpu_refcnt;
2076
2077 struct list_head link_watch_list;
2078
2079 enum { NETREG_UNINITIALIZED=0,
2080 NETREG_REGISTERED, /* completed register_netdevice */
2081 NETREG_UNREGISTERING, /* called unregister_netdevice */
2082 NETREG_UNREGISTERED, /* completed unregister todo */
2083 NETREG_RELEASED, /* called free_netdev */
2084 NETREG_DUMMY, /* dummy device for NAPI poll */
2085 } reg_state:8;
2086
2087 bool dismantle;
2088
2089 enum {
2090 RTNL_LINK_INITIALIZED,
2091 RTNL_LINK_INITIALIZING,
2092 } rtnl_link_state:16;
2093
2094 bool needs_free_netdev;
2095 void (*priv_destructor)(struct net_device *dev);
2096
2097#ifdef CONFIG_NETPOLL
2098 struct netpoll_info __rcu *npinfo;
2099#endif
2100
2101 possible_net_t nd_net;
2102
2103 /* mid-layer private */
2104 void *ml_priv;
2105 enum netdev_ml_priv_type ml_priv_type;
2106
2107 union {
2108 struct pcpu_lstats __percpu *lstats;
2109 struct pcpu_sw_netstats __percpu *tstats;
2110 struct pcpu_dstats __percpu *dstats;
2111 };
2112
2113#if IS_ENABLED(CONFIG_GARP)
2114 struct garp_port __rcu *garp_port;
2115#endif
2116#if IS_ENABLED(CONFIG_MRP)
2117 struct mrp_port __rcu *mrp_port;
2118#endif
2119
2120 struct device dev;
2121 const struct attribute_group *sysfs_groups[4];
2122 const struct attribute_group *sysfs_rx_queue_group;
2123
2124 const struct rtnl_link_ops *rtnl_link_ops;
2125
2126 /* for setting kernel sock attribute on TCP connection setup */
2127#define GSO_MAX_SIZE 65536
2128 unsigned int gso_max_size;
2129#define GSO_MAX_SEGS 65535
2130 u16 gso_max_segs;
2131
2132#ifdef CONFIG_DCB
2133 const struct dcbnl_rtnl_ops *dcbnl_ops;
2134#endif
2135 s16 num_tc;
2136 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2137 u8 prio_tc_map[TC_BITMASK + 1];
2138
2139#if IS_ENABLED(CONFIG_FCOE)
2140 unsigned int fcoe_ddp_xid;
2141#endif
2142#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2143 struct netprio_map __rcu *priomap;
2144#endif
2145 struct phy_device *phydev;
2146 struct sfp_bus *sfp_bus;
2147 struct lock_class_key *qdisc_tx_busylock;
2148 struct lock_class_key *qdisc_running_key;
2149 bool proto_down;
2150 unsigned wol_enabled:1;
2151 unsigned threaded:1;
2152
2153 struct list_head net_notifier_list;
2154
2155#if IS_ENABLED(CONFIG_MACSEC)
2156 /* MACsec management functions */
2157 const struct macsec_ops *macsec_ops;
2158#endif
2159 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2160 struct udp_tunnel_nic *udp_tunnel_nic;
2161
2162 /* protected by rtnl_lock */
2163 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2164};
2165#define to_net_dev(d) container_of(d, struct net_device, dev)
2166
2167static inline bool netif_elide_gro(const struct net_device *dev)
2168{
2169 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2170 return true;
2171 return false;
2172}
2173
2174#define NETDEV_ALIGN 32
2175
2176static inline
2177int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2178{
2179 return dev->prio_tc_map[prio & TC_BITMASK];
2180}
2181
2182static inline
2183int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2184{
2185 if (tc >= dev->num_tc)
2186 return -EINVAL;
2187
2188 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2189 return 0;
2190}
2191
2192int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2193void netdev_reset_tc(struct net_device *dev);
2194int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2195int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2196
2197static inline
2198int netdev_get_num_tc(struct net_device *dev)
2199{
2200 return dev->num_tc;
2201}
2202
2203static inline void net_prefetch(void *p)
2204{
2205 prefetch(p);
2206#if L1_CACHE_BYTES < 128
2207 prefetch((u8 *)p + L1_CACHE_BYTES);
2208#endif
2209}
2210
2211static inline void net_prefetchw(void *p)
2212{
2213 prefetchw(p);
2214#if L1_CACHE_BYTES < 128
2215 prefetchw((u8 *)p + L1_CACHE_BYTES);
2216#endif
2217}
2218
2219void netdev_unbind_sb_channel(struct net_device *dev,
2220 struct net_device *sb_dev);
2221int netdev_bind_sb_channel_queue(struct net_device *dev,
2222 struct net_device *sb_dev,
2223 u8 tc, u16 count, u16 offset);
2224int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2225static inline int netdev_get_sb_channel(struct net_device *dev)
2226{
2227 return max_t(int, -dev->num_tc, 0);
2228}
2229
2230static inline
2231struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2232 unsigned int index)
2233{
2234 return &dev->_tx[index];
2235}
2236
2237static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2238 const struct sk_buff *skb)
2239{
2240 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2241}
2242
2243static inline void netdev_for_each_tx_queue(struct net_device *dev,
2244 void (*f)(struct net_device *,
2245 struct netdev_queue *,
2246 void *),
2247 void *arg)
2248{
2249 unsigned int i;
2250
2251 for (i = 0; i < dev->num_tx_queues; i++)
2252 f(dev, &dev->_tx[i], arg);
2253}
2254
2255#define netdev_lockdep_set_classes(dev) \
2256{ \
2257 static struct lock_class_key qdisc_tx_busylock_key; \
2258 static struct lock_class_key qdisc_running_key; \
2259 static struct lock_class_key qdisc_xmit_lock_key; \
2260 static struct lock_class_key dev_addr_list_lock_key; \
2261 unsigned int i; \
2262 \
2263 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2264 (dev)->qdisc_running_key = &qdisc_running_key; \
2265 lockdep_set_class(&(dev)->addr_list_lock, \
2266 &dev_addr_list_lock_key); \
2267 for (i = 0; i < (dev)->num_tx_queues; i++) \
2268 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2269 &qdisc_xmit_lock_key); \
2270}
2271
2272u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2273 struct net_device *sb_dev);
2274struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2275 struct sk_buff *skb,
2276 struct net_device *sb_dev);
2277
2278/* returns the headroom that the master device needs to take in account
2279 * when forwarding to this dev
2280 */
2281static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2282{
2283 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2284}
2285
2286static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2287{
2288 if (dev->netdev_ops->ndo_set_rx_headroom)
2289 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2290}
2291
2292/* set the device rx headroom to the dev's default */
2293static inline void netdev_reset_rx_headroom(struct net_device *dev)
2294{
2295 netdev_set_rx_headroom(dev, -1);
2296}
2297
2298static inline void *netdev_get_ml_priv(struct net_device *dev,
2299 enum netdev_ml_priv_type type)
2300{
2301 if (dev->ml_priv_type != type)
2302 return NULL;
2303
2304 return dev->ml_priv;
2305}
2306
2307static inline void netdev_set_ml_priv(struct net_device *dev,
2308 void *ml_priv,
2309 enum netdev_ml_priv_type type)
2310{
2311 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2312 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2313 dev->ml_priv_type, type);
2314 WARN(!dev->ml_priv_type && dev->ml_priv,
2315 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2316
2317 dev->ml_priv = ml_priv;
2318 dev->ml_priv_type = type;
2319}
2320
2321/*
2322 * Net namespace inlines
2323 */
2324static inline
2325struct net *dev_net(const struct net_device *dev)
2326{
2327 return read_pnet(&dev->nd_net);
2328}
2329
2330static inline
2331void dev_net_set(struct net_device *dev, struct net *net)
2332{
2333 write_pnet(&dev->nd_net, net);
2334}
2335
2336/**
2337 * netdev_priv - access network device private data
2338 * @dev: network device
2339 *
2340 * Get network device private data
2341 */
2342static inline void *netdev_priv(const struct net_device *dev)
2343{
2344 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2345}
2346
2347/* Set the sysfs physical device reference for the network logical device
2348 * if set prior to registration will cause a symlink during initialization.
2349 */
2350#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2351
2352/* Set the sysfs device type for the network logical device to allow
2353 * fine-grained identification of different network device types. For
2354 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2355 */
2356#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2357
2358/* Default NAPI poll() weight
2359 * Device drivers are strongly advised to not use bigger value
2360 */
2361#define NAPI_POLL_WEIGHT 64
2362
2363/**
2364 * netif_napi_add - initialize a NAPI context
2365 * @dev: network device
2366 * @napi: NAPI context
2367 * @poll: polling function
2368 * @weight: default weight
2369 *
2370 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2371 * *any* of the other NAPI-related functions.
2372 */
2373void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2374 int (*poll)(struct napi_struct *, int), int weight);
2375
2376/**
2377 * netif_tx_napi_add - initialize a NAPI context
2378 * @dev: network device
2379 * @napi: NAPI context
2380 * @poll: polling function
2381 * @weight: default weight
2382 *
2383 * This variant of netif_napi_add() should be used from drivers using NAPI
2384 * to exclusively poll a TX queue.
2385 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2386 */
2387static inline void netif_tx_napi_add(struct net_device *dev,
2388 struct napi_struct *napi,
2389 int (*poll)(struct napi_struct *, int),
2390 int weight)
2391{
2392 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2393 netif_napi_add(dev, napi, poll, weight);
2394}
2395
2396/**
2397 * __netif_napi_del - remove a NAPI context
2398 * @napi: NAPI context
2399 *
2400 * Warning: caller must observe RCU grace period before freeing memory
2401 * containing @napi. Drivers might want to call this helper to combine
2402 * all the needed RCU grace periods into a single one.
2403 */
2404void __netif_napi_del(struct napi_struct *napi);
2405
2406/**
2407 * netif_napi_del - remove a NAPI context
2408 * @napi: NAPI context
2409 *
2410 * netif_napi_del() removes a NAPI context from the network device NAPI list
2411 */
2412static inline void netif_napi_del(struct napi_struct *napi)
2413{
2414 __netif_napi_del(napi);
2415 synchronize_net();
2416}
2417
2418struct napi_gro_cb {
2419 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2420 void *frag0;
2421
2422 /* Length of frag0. */
2423 unsigned int frag0_len;
2424
2425 /* This indicates where we are processing relative to skb->data. */
2426 int data_offset;
2427
2428 /* This is non-zero if the packet cannot be merged with the new skb. */
2429 u16 flush;
2430
2431 /* Save the IP ID here and check when we get to the transport layer */
2432 u16 flush_id;
2433
2434 /* Number of segments aggregated. */
2435 u16 count;
2436
2437 /* Start offset for remote checksum offload */
2438 u16 gro_remcsum_start;
2439
2440 /* jiffies when first packet was created/queued */
2441 unsigned long age;
2442
2443 /* Used in ipv6_gro_receive() and foo-over-udp */
2444 u16 proto;
2445
2446 /* This is non-zero if the packet may be of the same flow. */
2447 u8 same_flow:1;
2448
2449 /* Used in tunnel GRO receive */
2450 u8 encap_mark:1;
2451
2452 /* GRO checksum is valid */
2453 u8 csum_valid:1;
2454
2455 /* Number of checksums via CHECKSUM_UNNECESSARY */
2456 u8 csum_cnt:3;
2457
2458 /* Free the skb? */
2459 u8 free:2;
2460#define NAPI_GRO_FREE 1
2461#define NAPI_GRO_FREE_STOLEN_HEAD 2
2462
2463 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2464 u8 is_ipv6:1;
2465
2466 /* Used in GRE, set in fou/gue_gro_receive */
2467 u8 is_fou:1;
2468
2469 /* Used to determine if flush_id can be ignored */
2470 u8 is_atomic:1;
2471
2472 /* Number of gro_receive callbacks this packet already went through */
2473 u8 recursion_counter:4;
2474
2475 /* GRO is done by frag_list pointer chaining. */
2476 u8 is_flist:1;
2477
2478 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2479 __wsum csum;
2480
2481 /* used in skb_gro_receive() slow path */
2482 struct sk_buff *last;
2483};
2484
2485#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2486
2487#define GRO_RECURSION_LIMIT 15
2488static inline int gro_recursion_inc_test(struct sk_buff *skb)
2489{
2490 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2491}
2492
2493typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2494static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2495 struct list_head *head,
2496 struct sk_buff *skb)
2497{
2498 if (unlikely(gro_recursion_inc_test(skb))) {
2499 NAPI_GRO_CB(skb)->flush |= 1;
2500 return NULL;
2501 }
2502
2503 return cb(head, skb);
2504}
2505
2506typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2507 struct sk_buff *);
2508static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2509 struct sock *sk,
2510 struct list_head *head,
2511 struct sk_buff *skb)
2512{
2513 if (unlikely(gro_recursion_inc_test(skb))) {
2514 NAPI_GRO_CB(skb)->flush |= 1;
2515 return NULL;
2516 }
2517
2518 return cb(sk, head, skb);
2519}
2520
2521struct packet_type {
2522 __be16 type; /* This is really htons(ether_type). */
2523 bool ignore_outgoing;
2524 struct net_device *dev; /* NULL is wildcarded here */
2525 int (*func) (struct sk_buff *,
2526 struct net_device *,
2527 struct packet_type *,
2528 struct net_device *);
2529 void (*list_func) (struct list_head *,
2530 struct packet_type *,
2531 struct net_device *);
2532 bool (*id_match)(struct packet_type *ptype,
2533 struct sock *sk);
2534 void *af_packet_priv;
2535 struct list_head list;
2536};
2537
2538struct offload_callbacks {
2539 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2540 netdev_features_t features);
2541 struct sk_buff *(*gro_receive)(struct list_head *head,
2542 struct sk_buff *skb);
2543 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2544};
2545
2546struct packet_offload {
2547 __be16 type; /* This is really htons(ether_type). */
2548 u16 priority;
2549 struct offload_callbacks callbacks;
2550 struct list_head list;
2551};
2552
2553/* often modified stats are per-CPU, other are shared (netdev->stats) */
2554struct pcpu_sw_netstats {
2555 u64 rx_packets;
2556 u64 rx_bytes;
2557 u64 tx_packets;
2558 u64 tx_bytes;
2559 struct u64_stats_sync syncp;
2560} __aligned(4 * sizeof(u64));
2561
2562struct pcpu_lstats {
2563 u64_stats_t packets;
2564 u64_stats_t bytes;
2565 struct u64_stats_sync syncp;
2566} __aligned(2 * sizeof(u64));
2567
2568void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2569
2570static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2571{
2572 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2573
2574 u64_stats_update_begin(&tstats->syncp);
2575 tstats->rx_bytes += len;
2576 tstats->rx_packets++;
2577 u64_stats_update_end(&tstats->syncp);
2578}
2579
2580static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2581 unsigned int packets,
2582 unsigned int len)
2583{
2584 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2585
2586 u64_stats_update_begin(&tstats->syncp);
2587 tstats->tx_bytes += len;
2588 tstats->tx_packets += packets;
2589 u64_stats_update_end(&tstats->syncp);
2590}
2591
2592static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2593{
2594 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2595
2596 u64_stats_update_begin(&lstats->syncp);
2597 u64_stats_add(&lstats->bytes, len);
2598 u64_stats_inc(&lstats->packets);
2599 u64_stats_update_end(&lstats->syncp);
2600}
2601
2602#define __netdev_alloc_pcpu_stats(type, gfp) \
2603({ \
2604 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2605 if (pcpu_stats) { \
2606 int __cpu; \
2607 for_each_possible_cpu(__cpu) { \
2608 typeof(type) *stat; \
2609 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2610 u64_stats_init(&stat->syncp); \
2611 } \
2612 } \
2613 pcpu_stats; \
2614})
2615
2616#define netdev_alloc_pcpu_stats(type) \
2617 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2618
2619#define devm_netdev_alloc_pcpu_stats(dev, type) \
2620({ \
2621 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2622 if (pcpu_stats) { \
2623 int __cpu; \
2624 for_each_possible_cpu(__cpu) { \
2625 typeof(type) *stat; \
2626 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2627 u64_stats_init(&stat->syncp); \
2628 } \
2629 } \
2630 pcpu_stats; \
2631})
2632
2633enum netdev_lag_tx_type {
2634 NETDEV_LAG_TX_TYPE_UNKNOWN,
2635 NETDEV_LAG_TX_TYPE_RANDOM,
2636 NETDEV_LAG_TX_TYPE_BROADCAST,
2637 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2638 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2639 NETDEV_LAG_TX_TYPE_HASH,
2640};
2641
2642enum netdev_lag_hash {
2643 NETDEV_LAG_HASH_NONE,
2644 NETDEV_LAG_HASH_L2,
2645 NETDEV_LAG_HASH_L34,
2646 NETDEV_LAG_HASH_L23,
2647 NETDEV_LAG_HASH_E23,
2648 NETDEV_LAG_HASH_E34,
2649 NETDEV_LAG_HASH_VLAN_SRCMAC,
2650 NETDEV_LAG_HASH_UNKNOWN,
2651};
2652
2653struct netdev_lag_upper_info {
2654 enum netdev_lag_tx_type tx_type;
2655 enum netdev_lag_hash hash_type;
2656};
2657
2658struct netdev_lag_lower_state_info {
2659 u8 link_up : 1,
2660 tx_enabled : 1;
2661};
2662
2663#include <linux/notifier.h>
2664
2665/* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2666 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2667 * adding new types.
2668 */
2669enum netdev_cmd {
2670 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2671 NETDEV_DOWN,
2672 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2673 detected a hardware crash and restarted
2674 - we can use this eg to kick tcp sessions
2675 once done */
2676 NETDEV_CHANGE, /* Notify device state change */
2677 NETDEV_REGISTER,
2678 NETDEV_UNREGISTER,
2679 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2680 NETDEV_CHANGEADDR, /* notify after the address change */
2681 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2682 NETDEV_GOING_DOWN,
2683 NETDEV_CHANGENAME,
2684 NETDEV_FEAT_CHANGE,
2685 NETDEV_BONDING_FAILOVER,
2686 NETDEV_PRE_UP,
2687 NETDEV_PRE_TYPE_CHANGE,
2688 NETDEV_POST_TYPE_CHANGE,
2689 NETDEV_POST_INIT,
2690 NETDEV_RELEASE,
2691 NETDEV_NOTIFY_PEERS,
2692 NETDEV_JOIN,
2693 NETDEV_CHANGEUPPER,
2694 NETDEV_RESEND_IGMP,
2695 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2696 NETDEV_CHANGEINFODATA,
2697 NETDEV_BONDING_INFO,
2698 NETDEV_PRECHANGEUPPER,
2699 NETDEV_CHANGELOWERSTATE,
2700 NETDEV_UDP_TUNNEL_PUSH_INFO,
2701 NETDEV_UDP_TUNNEL_DROP_INFO,
2702 NETDEV_CHANGE_TX_QUEUE_LEN,
2703 NETDEV_CVLAN_FILTER_PUSH_INFO,
2704 NETDEV_CVLAN_FILTER_DROP_INFO,
2705 NETDEV_SVLAN_FILTER_PUSH_INFO,
2706 NETDEV_SVLAN_FILTER_DROP_INFO,
2707};
2708const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2709
2710int register_netdevice_notifier(struct notifier_block *nb);
2711int unregister_netdevice_notifier(struct notifier_block *nb);
2712int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2713int unregister_netdevice_notifier_net(struct net *net,
2714 struct notifier_block *nb);
2715int register_netdevice_notifier_dev_net(struct net_device *dev,
2716 struct notifier_block *nb,
2717 struct netdev_net_notifier *nn);
2718int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2719 struct notifier_block *nb,
2720 struct netdev_net_notifier *nn);
2721
2722struct netdev_notifier_info {
2723 struct net_device *dev;
2724 struct netlink_ext_ack *extack;
2725};
2726
2727struct netdev_notifier_info_ext {
2728 struct netdev_notifier_info info; /* must be first */
2729 union {
2730 u32 mtu;
2731 } ext;
2732};
2733
2734struct netdev_notifier_change_info {
2735 struct netdev_notifier_info info; /* must be first */
2736 unsigned int flags_changed;
2737};
2738
2739struct netdev_notifier_changeupper_info {
2740 struct netdev_notifier_info info; /* must be first */
2741 struct net_device *upper_dev; /* new upper dev */
2742 bool master; /* is upper dev master */
2743 bool linking; /* is the notification for link or unlink */
2744 void *upper_info; /* upper dev info */
2745};
2746
2747struct netdev_notifier_changelowerstate_info {
2748 struct netdev_notifier_info info; /* must be first */
2749 void *lower_state_info; /* is lower dev state */
2750};
2751
2752struct netdev_notifier_pre_changeaddr_info {
2753 struct netdev_notifier_info info; /* must be first */
2754 const unsigned char *dev_addr;
2755};
2756
2757static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2758 struct net_device *dev)
2759{
2760 info->dev = dev;
2761 info->extack = NULL;
2762}
2763
2764static inline struct net_device *
2765netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2766{
2767 return info->dev;
2768}
2769
2770static inline struct netlink_ext_ack *
2771netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2772{
2773 return info->extack;
2774}
2775
2776int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2777
2778
2779extern rwlock_t dev_base_lock; /* Device list lock */
2780
2781#define for_each_netdev(net, d) \
2782 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2783#define for_each_netdev_reverse(net, d) \
2784 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2785#define for_each_netdev_rcu(net, d) \
2786 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2787#define for_each_netdev_safe(net, d, n) \
2788 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2789#define for_each_netdev_continue(net, d) \
2790 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2791#define for_each_netdev_continue_reverse(net, d) \
2792 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2793 dev_list)
2794#define for_each_netdev_continue_rcu(net, d) \
2795 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2796#define for_each_netdev_in_bond_rcu(bond, slave) \
2797 for_each_netdev_rcu(&init_net, slave) \
2798 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2799#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2800
2801static inline struct net_device *next_net_device(struct net_device *dev)
2802{
2803 struct list_head *lh;
2804 struct net *net;
2805
2806 net = dev_net(dev);
2807 lh = dev->dev_list.next;
2808 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2809}
2810
2811static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2812{
2813 struct list_head *lh;
2814 struct net *net;
2815
2816 net = dev_net(dev);
2817 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2818 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2819}
2820
2821static inline struct net_device *first_net_device(struct net *net)
2822{
2823 return list_empty(&net->dev_base_head) ? NULL :
2824 net_device_entry(net->dev_base_head.next);
2825}
2826
2827static inline struct net_device *first_net_device_rcu(struct net *net)
2828{
2829 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2830
2831 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2832}
2833
2834int netdev_boot_setup_check(struct net_device *dev);
2835unsigned long netdev_boot_base(const char *prefix, int unit);
2836struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2837 const char *hwaddr);
2838struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2839void dev_add_pack(struct packet_type *pt);
2840void dev_remove_pack(struct packet_type *pt);
2841void __dev_remove_pack(struct packet_type *pt);
2842void dev_add_offload(struct packet_offload *po);
2843void dev_remove_offload(struct packet_offload *po);
2844
2845int dev_get_iflink(const struct net_device *dev);
2846int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2847struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2848 unsigned short mask);
2849struct net_device *dev_get_by_name(struct net *net, const char *name);
2850struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2851struct net_device *__dev_get_by_name(struct net *net, const char *name);
2852int dev_alloc_name(struct net_device *dev, const char *name);
2853int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2854void dev_close(struct net_device *dev);
2855void dev_close_many(struct list_head *head, bool unlink);
2856void dev_disable_lro(struct net_device *dev);
2857int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2858u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2859 struct net_device *sb_dev);
2860u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2861 struct net_device *sb_dev);
2862
2863int dev_queue_xmit(struct sk_buff *skb);
2864int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2865int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2866
2867static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2868{
2869 int ret;
2870
2871 ret = __dev_direct_xmit(skb, queue_id);
2872 if (!dev_xmit_complete(ret))
2873 kfree_skb(skb);
2874 return ret;
2875}
2876
2877int register_netdevice(struct net_device *dev);
2878void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2879void unregister_netdevice_many(struct list_head *head);
2880static inline void unregister_netdevice(struct net_device *dev)
2881{
2882 unregister_netdevice_queue(dev, NULL);
2883}
2884
2885int netdev_refcnt_read(const struct net_device *dev);
2886void free_netdev(struct net_device *dev);
2887void netdev_freemem(struct net_device *dev);
2888int init_dummy_netdev(struct net_device *dev);
2889
2890struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2891 struct sk_buff *skb,
2892 bool all_slaves);
2893struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2894 struct sock *sk);
2895struct net_device *dev_get_by_index(struct net *net, int ifindex);
2896struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2897struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2898struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2899int netdev_get_name(struct net *net, char *name, int ifindex);
2900int dev_restart(struct net_device *dev);
2901int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2902int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2903
2904static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2905{
2906 return NAPI_GRO_CB(skb)->data_offset;
2907}
2908
2909static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2910{
2911 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2912}
2913
2914static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2915{
2916 NAPI_GRO_CB(skb)->data_offset += len;
2917}
2918
2919static inline void *skb_gro_header_fast(struct sk_buff *skb,
2920 unsigned int offset)
2921{
2922 return NAPI_GRO_CB(skb)->frag0 + offset;
2923}
2924
2925static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2926{
2927 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2928}
2929
2930static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2931{
2932 NAPI_GRO_CB(skb)->frag0 = NULL;
2933 NAPI_GRO_CB(skb)->frag0_len = 0;
2934}
2935
2936static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2937 unsigned int offset)
2938{
2939 if (!pskb_may_pull(skb, hlen))
2940 return NULL;
2941
2942 skb_gro_frag0_invalidate(skb);
2943 return skb->data + offset;
2944}
2945
2946static inline void *skb_gro_network_header(struct sk_buff *skb)
2947{
2948 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2949 skb_network_offset(skb);
2950}
2951
2952static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2953 const void *start, unsigned int len)
2954{
2955 if (NAPI_GRO_CB(skb)->csum_valid)
2956 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2957 csum_partial(start, len, 0));
2958}
2959
2960/* GRO checksum functions. These are logical equivalents of the normal
2961 * checksum functions (in skbuff.h) except that they operate on the GRO
2962 * offsets and fields in sk_buff.
2963 */
2964
2965__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2966
2967static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2968{
2969 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2970}
2971
2972static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2973 bool zero_okay,
2974 __sum16 check)
2975{
2976 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2977 skb_checksum_start_offset(skb) <
2978 skb_gro_offset(skb)) &&
2979 !skb_at_gro_remcsum_start(skb) &&
2980 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2981 (!zero_okay || check));
2982}
2983
2984static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2985 __wsum psum)
2986{
2987 if (NAPI_GRO_CB(skb)->csum_valid &&
2988 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2989 return 0;
2990
2991 NAPI_GRO_CB(skb)->csum = psum;
2992
2993 return __skb_gro_checksum_complete(skb);
2994}
2995
2996static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2997{
2998 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2999 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3000 NAPI_GRO_CB(skb)->csum_cnt--;
3001 } else {
3002 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3003 * verified a new top level checksum or an encapsulated one
3004 * during GRO. This saves work if we fallback to normal path.
3005 */
3006 __skb_incr_checksum_unnecessary(skb);
3007 }
3008}
3009
3010#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3011 compute_pseudo) \
3012({ \
3013 __sum16 __ret = 0; \
3014 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3015 __ret = __skb_gro_checksum_validate_complete(skb, \
3016 compute_pseudo(skb, proto)); \
3017 if (!__ret) \
3018 skb_gro_incr_csum_unnecessary(skb); \
3019 __ret; \
3020})
3021
3022#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3023 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3024
3025#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3026 compute_pseudo) \
3027 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3028
3029#define skb_gro_checksum_simple_validate(skb) \
3030 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3031
3032static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3033{
3034 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3035 !NAPI_GRO_CB(skb)->csum_valid);
3036}
3037
3038static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3039 __wsum pseudo)
3040{
3041 NAPI_GRO_CB(skb)->csum = ~pseudo;
3042 NAPI_GRO_CB(skb)->csum_valid = 1;
3043}
3044
3045#define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3046do { \
3047 if (__skb_gro_checksum_convert_check(skb)) \
3048 __skb_gro_checksum_convert(skb, \
3049 compute_pseudo(skb, proto)); \
3050} while (0)
3051
3052struct gro_remcsum {
3053 int offset;
3054 __wsum delta;
3055};
3056
3057static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3058{
3059 grc->offset = 0;
3060 grc->delta = 0;
3061}
3062
3063static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3064 unsigned int off, size_t hdrlen,
3065 int start, int offset,
3066 struct gro_remcsum *grc,
3067 bool nopartial)
3068{
3069 __wsum delta;
3070 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3071
3072 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3073
3074 if (!nopartial) {
3075 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3076 return ptr;
3077 }
3078
3079 ptr = skb_gro_header_fast(skb, off);
3080 if (skb_gro_header_hard(skb, off + plen)) {
3081 ptr = skb_gro_header_slow(skb, off + plen, off);
3082 if (!ptr)
3083 return NULL;
3084 }
3085
3086 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3087 start, offset);
3088
3089 /* Adjust skb->csum since we changed the packet */
3090 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3091
3092 grc->offset = off + hdrlen + offset;
3093 grc->delta = delta;
3094
3095 return ptr;
3096}
3097
3098static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3099 struct gro_remcsum *grc)
3100{
3101 void *ptr;
3102 size_t plen = grc->offset + sizeof(u16);
3103
3104 if (!grc->delta)
3105 return;
3106
3107 ptr = skb_gro_header_fast(skb, grc->offset);
3108 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3109 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3110 if (!ptr)
3111 return;
3112 }
3113
3114 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3115}
3116
3117#ifdef CONFIG_XFRM_OFFLOAD
3118static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3119{
3120 if (PTR_ERR(pp) != -EINPROGRESS)
3121 NAPI_GRO_CB(skb)->flush |= flush;
3122}
3123static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3124 struct sk_buff *pp,
3125 int flush,
3126 struct gro_remcsum *grc)
3127{
3128 if (PTR_ERR(pp) != -EINPROGRESS) {
3129 NAPI_GRO_CB(skb)->flush |= flush;
3130 skb_gro_remcsum_cleanup(skb, grc);
3131 skb->remcsum_offload = 0;
3132 }
3133}
3134#else
3135static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3136{
3137 NAPI_GRO_CB(skb)->flush |= flush;
3138}
3139static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3140 struct sk_buff *pp,
3141 int flush,
3142 struct gro_remcsum *grc)
3143{
3144 NAPI_GRO_CB(skb)->flush |= flush;
3145 skb_gro_remcsum_cleanup(skb, grc);
3146 skb->remcsum_offload = 0;
3147}
3148#endif
3149
3150static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3151 unsigned short type,
3152 const void *daddr, const void *saddr,
3153 unsigned int len)
3154{
3155 if (!dev->header_ops || !dev->header_ops->create)
3156 return 0;
3157
3158 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3159}
3160
3161static inline int dev_parse_header(const struct sk_buff *skb,
3162 unsigned char *haddr)
3163{
3164 const struct net_device *dev = skb->dev;
3165
3166 if (!dev->header_ops || !dev->header_ops->parse)
3167 return 0;
3168 return dev->header_ops->parse(skb, haddr);
3169}
3170
3171static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3172{
3173 const struct net_device *dev = skb->dev;
3174
3175 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3176 return 0;
3177 return dev->header_ops->parse_protocol(skb);
3178}
3179
3180/* ll_header must have at least hard_header_len allocated */
3181static inline bool dev_validate_header(const struct net_device *dev,
3182 char *ll_header, int len)
3183{
3184 if (likely(len >= dev->hard_header_len))
3185 return true;
3186 if (len < dev->min_header_len)
3187 return false;
3188
3189 if (capable(CAP_SYS_RAWIO)) {
3190 memset(ll_header + len, 0, dev->hard_header_len - len);
3191 return true;
3192 }
3193
3194 if (dev->header_ops && dev->header_ops->validate)
3195 return dev->header_ops->validate(ll_header, len);
3196
3197 return false;
3198}
3199
3200static inline bool dev_has_header(const struct net_device *dev)
3201{
3202 return dev->header_ops && dev->header_ops->create;
3203}
3204
3205typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3206 int len, int size);
3207int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3208static inline int unregister_gifconf(unsigned int family)
3209{
3210 return register_gifconf(family, NULL);
3211}
3212
3213#ifdef CONFIG_NET_FLOW_LIMIT
3214#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3215struct sd_flow_limit {
3216 u64 count;
3217 unsigned int num_buckets;
3218 unsigned int history_head;
3219 u16 history[FLOW_LIMIT_HISTORY];
3220 u8 buckets[];
3221};
3222
3223extern int netdev_flow_limit_table_len;
3224#endif /* CONFIG_NET_FLOW_LIMIT */
3225
3226/*
3227 * Incoming packets are placed on per-CPU queues
3228 */
3229struct softnet_data {
3230 struct list_head poll_list;
3231 struct sk_buff_head process_queue;
3232
3233 /* stats */
3234 unsigned int processed;
3235 unsigned int time_squeeze;
3236 unsigned int received_rps;
3237#ifdef CONFIG_RPS
3238 struct softnet_data *rps_ipi_list;
3239#endif
3240#ifdef CONFIG_NET_FLOW_LIMIT
3241 struct sd_flow_limit __rcu *flow_limit;
3242#endif
3243 struct Qdisc *output_queue;
3244 struct Qdisc **output_queue_tailp;
3245 struct sk_buff *completion_queue;
3246#ifdef CONFIG_XFRM_OFFLOAD
3247 struct sk_buff_head xfrm_backlog;
3248#endif
3249 /* written and read only by owning cpu: */
3250 struct {
3251 u16 recursion;
3252 u8 more;
3253 } xmit;
3254#ifdef CONFIG_RPS
3255 /* input_queue_head should be written by cpu owning this struct,
3256 * and only read by other cpus. Worth using a cache line.
3257 */
3258 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3259
3260 /* Elements below can be accessed between CPUs for RPS/RFS */
3261 call_single_data_t csd ____cacheline_aligned_in_smp;
3262 struct softnet_data *rps_ipi_next;
3263 unsigned int cpu;
3264 unsigned int input_queue_tail;
3265#endif
3266 unsigned int dropped;
3267 struct sk_buff_head input_pkt_queue;
3268 struct napi_struct backlog;
3269
3270};
3271
3272static inline void input_queue_head_incr(struct softnet_data *sd)
3273{
3274#ifdef CONFIG_RPS
3275 sd->input_queue_head++;
3276#endif
3277}
3278
3279static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3280 unsigned int *qtail)
3281{
3282#ifdef CONFIG_RPS
3283 *qtail = ++sd->input_queue_tail;
3284#endif
3285}
3286
3287DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3288
3289static inline int dev_recursion_level(void)
3290{
3291 return this_cpu_read(softnet_data.xmit.recursion);
3292}
3293
3294#define XMIT_RECURSION_LIMIT 8
3295static inline bool dev_xmit_recursion(void)
3296{
3297 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3298 XMIT_RECURSION_LIMIT);
3299}
3300
3301static inline void dev_xmit_recursion_inc(void)
3302{
3303 __this_cpu_inc(softnet_data.xmit.recursion);
3304}
3305
3306static inline void dev_xmit_recursion_dec(void)
3307{
3308 __this_cpu_dec(softnet_data.xmit.recursion);
3309}
3310
3311void __netif_schedule(struct Qdisc *q);
3312void netif_schedule_queue(struct netdev_queue *txq);
3313
3314static inline void netif_tx_schedule_all(struct net_device *dev)
3315{
3316 unsigned int i;
3317
3318 for (i = 0; i < dev->num_tx_queues; i++)
3319 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3320}
3321
3322static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3323{
3324 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3325}
3326
3327/**
3328 * netif_start_queue - allow transmit
3329 * @dev: network device
3330 *
3331 * Allow upper layers to call the device hard_start_xmit routine.
3332 */
3333static inline void netif_start_queue(struct net_device *dev)
3334{
3335 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3336}
3337
3338static inline void netif_tx_start_all_queues(struct net_device *dev)
3339{
3340 unsigned int i;
3341
3342 for (i = 0; i < dev->num_tx_queues; i++) {
3343 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3344 netif_tx_start_queue(txq);
3345 }
3346}
3347
3348void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3349
3350/**
3351 * netif_wake_queue - restart transmit
3352 * @dev: network device
3353 *
3354 * Allow upper layers to call the device hard_start_xmit routine.
3355 * Used for flow control when transmit resources are available.
3356 */
3357static inline void netif_wake_queue(struct net_device *dev)
3358{
3359 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3360}
3361
3362static inline void netif_tx_wake_all_queues(struct net_device *dev)
3363{
3364 unsigned int i;
3365
3366 for (i = 0; i < dev->num_tx_queues; i++) {
3367 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3368 netif_tx_wake_queue(txq);
3369 }
3370}
3371
3372static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3373{
3374 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3375}
3376
3377/**
3378 * netif_stop_queue - stop transmitted packets
3379 * @dev: network device
3380 *
3381 * Stop upper layers calling the device hard_start_xmit routine.
3382 * Used for flow control when transmit resources are unavailable.
3383 */
3384static inline void netif_stop_queue(struct net_device *dev)
3385{
3386 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3387}
3388
3389void netif_tx_stop_all_queues(struct net_device *dev);
3390
3391static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3392{
3393 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3394}
3395
3396/**
3397 * netif_queue_stopped - test if transmit queue is flowblocked
3398 * @dev: network device
3399 *
3400 * Test if transmit queue on device is currently unable to send.
3401 */
3402static inline bool netif_queue_stopped(const struct net_device *dev)
3403{
3404 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3405}
3406
3407static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3408{
3409 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3410}
3411
3412static inline bool
3413netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3414{
3415 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3416}
3417
3418static inline bool
3419netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3420{
3421 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3422}
3423
3424/**
3425 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3426 * @dev_queue: pointer to transmit queue
3427 *
3428 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3429 * to give appropriate hint to the CPU.
3430 */
3431static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3432{
3433#ifdef CONFIG_BQL
3434 prefetchw(&dev_queue->dql.num_queued);
3435#endif
3436}
3437
3438/**
3439 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3440 * @dev_queue: pointer to transmit queue
3441 *
3442 * BQL enabled drivers might use this helper in their TX completion path,
3443 * to give appropriate hint to the CPU.
3444 */
3445static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3446{
3447#ifdef CONFIG_BQL
3448 prefetchw(&dev_queue->dql.limit);
3449#endif
3450}
3451
3452static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3453 unsigned int bytes)
3454{
3455#ifdef CONFIG_BQL
3456 dql_queued(&dev_queue->dql, bytes);
3457
3458 if (likely(dql_avail(&dev_queue->dql) >= 0))
3459 return;
3460
3461 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3462
3463 /*
3464 * The XOFF flag must be set before checking the dql_avail below,
3465 * because in netdev_tx_completed_queue we update the dql_completed
3466 * before checking the XOFF flag.
3467 */
3468 smp_mb();
3469
3470 /* check again in case another CPU has just made room avail */
3471 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3472 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3473#endif
3474}
3475
3476/* Variant of netdev_tx_sent_queue() for drivers that are aware
3477 * that they should not test BQL status themselves.
3478 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3479 * skb of a batch.
3480 * Returns true if the doorbell must be used to kick the NIC.
3481 */
3482static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3483 unsigned int bytes,
3484 bool xmit_more)
3485{
3486 if (xmit_more) {
3487#ifdef CONFIG_BQL
3488 dql_queued(&dev_queue->dql, bytes);
3489#endif
3490 return netif_tx_queue_stopped(dev_queue);
3491 }
3492 netdev_tx_sent_queue(dev_queue, bytes);
3493 return true;
3494}
3495
3496/**
3497 * netdev_sent_queue - report the number of bytes queued to hardware
3498 * @dev: network device
3499 * @bytes: number of bytes queued to the hardware device queue
3500 *
3501 * Report the number of bytes queued for sending/completion to the network
3502 * device hardware queue. @bytes should be a good approximation and should
3503 * exactly match netdev_completed_queue() @bytes
3504 */
3505static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3506{
3507 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3508}
3509
3510static inline bool __netdev_sent_queue(struct net_device *dev,
3511 unsigned int bytes,
3512 bool xmit_more)
3513{
3514 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3515 xmit_more);
3516}
3517
3518static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3519 unsigned int pkts, unsigned int bytes)
3520{
3521#ifdef CONFIG_BQL
3522 if (unlikely(!bytes))
3523 return;
3524
3525 dql_completed(&dev_queue->dql, bytes);
3526
3527 /*
3528 * Without the memory barrier there is a small possiblity that
3529 * netdev_tx_sent_queue will miss the update and cause the queue to
3530 * be stopped forever
3531 */
3532 smp_mb();
3533
3534 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3535 return;
3536
3537 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3538 netif_schedule_queue(dev_queue);
3539#endif
3540}
3541
3542/**
3543 * netdev_completed_queue - report bytes and packets completed by device
3544 * @dev: network device
3545 * @pkts: actual number of packets sent over the medium
3546 * @bytes: actual number of bytes sent over the medium
3547 *
3548 * Report the number of bytes and packets transmitted by the network device
3549 * hardware queue over the physical medium, @bytes must exactly match the
3550 * @bytes amount passed to netdev_sent_queue()
3551 */
3552static inline void netdev_completed_queue(struct net_device *dev,
3553 unsigned int pkts, unsigned int bytes)
3554{
3555 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3556}
3557
3558static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3559{
3560#ifdef CONFIG_BQL
3561 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3562 dql_reset(&q->dql);
3563#endif
3564}
3565
3566/**
3567 * netdev_reset_queue - reset the packets and bytes count of a network device
3568 * @dev_queue: network device
3569 *
3570 * Reset the bytes and packet count of a network device and clear the
3571 * software flow control OFF bit for this network device
3572 */
3573static inline void netdev_reset_queue(struct net_device *dev_queue)
3574{
3575 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3576}
3577
3578/**
3579 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3580 * @dev: network device
3581 * @queue_index: given tx queue index
3582 *
3583 * Returns 0 if given tx queue index >= number of device tx queues,
3584 * otherwise returns the originally passed tx queue index.
3585 */
3586static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3587{
3588 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3589 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3590 dev->name, queue_index,
3591 dev->real_num_tx_queues);
3592 return 0;
3593 }
3594
3595 return queue_index;
3596}
3597
3598/**
3599 * netif_running - test if up
3600 * @dev: network device
3601 *
3602 * Test if the device has been brought up.
3603 */
3604static inline bool netif_running(const struct net_device *dev)
3605{
3606 return test_bit(__LINK_STATE_START, &dev->state);
3607}
3608
3609/*
3610 * Routines to manage the subqueues on a device. We only need start,
3611 * stop, and a check if it's stopped. All other device management is
3612 * done at the overall netdevice level.
3613 * Also test the device if we're multiqueue.
3614 */
3615
3616/**
3617 * netif_start_subqueue - allow sending packets on subqueue
3618 * @dev: network device
3619 * @queue_index: sub queue index
3620 *
3621 * Start individual transmit queue of a device with multiple transmit queues.
3622 */
3623static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3624{
3625 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3626
3627 netif_tx_start_queue(txq);
3628}
3629
3630/**
3631 * netif_stop_subqueue - stop sending packets on subqueue
3632 * @dev: network device
3633 * @queue_index: sub queue index
3634 *
3635 * Stop individual transmit queue of a device with multiple transmit queues.
3636 */
3637static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3638{
3639 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3640 netif_tx_stop_queue(txq);
3641}
3642
3643/**
3644 * __netif_subqueue_stopped - test status of subqueue
3645 * @dev: network device
3646 * @queue_index: sub queue index
3647 *
3648 * Check individual transmit queue of a device with multiple transmit queues.
3649 */
3650static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3651 u16 queue_index)
3652{
3653 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3654
3655 return netif_tx_queue_stopped(txq);
3656}
3657
3658/**
3659 * netif_subqueue_stopped - test status of subqueue
3660 * @dev: network device
3661 * @skb: sub queue buffer pointer
3662 *
3663 * Check individual transmit queue of a device with multiple transmit queues.
3664 */
3665static inline bool netif_subqueue_stopped(const struct net_device *dev,
3666 struct sk_buff *skb)
3667{
3668 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3669}
3670
3671/**
3672 * netif_wake_subqueue - allow sending packets on subqueue
3673 * @dev: network device
3674 * @queue_index: sub queue index
3675 *
3676 * Resume individual transmit queue of a device with multiple transmit queues.
3677 */
3678static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3679{
3680 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3681
3682 netif_tx_wake_queue(txq);
3683}
3684
3685#ifdef CONFIG_XPS
3686int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3687 u16 index);
3688int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3689 u16 index, bool is_rxqs_map);
3690
3691/**
3692 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3693 * @j: CPU/Rx queue index
3694 * @mask: bitmask of all cpus/rx queues
3695 * @nr_bits: number of bits in the bitmask
3696 *
3697 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3698 */
3699static inline bool netif_attr_test_mask(unsigned long j,
3700 const unsigned long *mask,
3701 unsigned int nr_bits)
3702{
3703 cpu_max_bits_warn(j, nr_bits);
3704 return test_bit(j, mask);
3705}
3706
3707/**
3708 * netif_attr_test_online - Test for online CPU/Rx queue
3709 * @j: CPU/Rx queue index
3710 * @online_mask: bitmask for CPUs/Rx queues that are online
3711 * @nr_bits: number of bits in the bitmask
3712 *
3713 * Returns true if a CPU/Rx queue is online.
3714 */
3715static inline bool netif_attr_test_online(unsigned long j,
3716 const unsigned long *online_mask,
3717 unsigned int nr_bits)
3718{
3719 cpu_max_bits_warn(j, nr_bits);
3720
3721 if (online_mask)
3722 return test_bit(j, online_mask);
3723
3724 return (j < nr_bits);
3725}
3726
3727/**
3728 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3729 * @n: CPU/Rx queue index
3730 * @srcp: the cpumask/Rx queue mask pointer
3731 * @nr_bits: number of bits in the bitmask
3732 *
3733 * Returns >= nr_bits if no further CPUs/Rx queues set.
3734 */
3735static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3736 unsigned int nr_bits)
3737{
3738 /* -1 is a legal arg here. */
3739 if (n != -1)
3740 cpu_max_bits_warn(n, nr_bits);
3741
3742 if (srcp)
3743 return find_next_bit(srcp, nr_bits, n + 1);
3744
3745 return n + 1;
3746}
3747
3748/**
3749 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3750 * @n: CPU/Rx queue index
3751 * @src1p: the first CPUs/Rx queues mask pointer
3752 * @src2p: the second CPUs/Rx queues mask pointer
3753 * @nr_bits: number of bits in the bitmask
3754 *
3755 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3756 */
3757static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3758 const unsigned long *src2p,
3759 unsigned int nr_bits)
3760{
3761 /* -1 is a legal arg here. */
3762 if (n != -1)
3763 cpu_max_bits_warn(n, nr_bits);
3764
3765 if (src1p && src2p)
3766 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3767 else if (src1p)
3768 return find_next_bit(src1p, nr_bits, n + 1);
3769 else if (src2p)
3770 return find_next_bit(src2p, nr_bits, n + 1);
3771
3772 return n + 1;
3773}
3774#else
3775static inline int netif_set_xps_queue(struct net_device *dev,
3776 const struct cpumask *mask,
3777 u16 index)
3778{
3779 return 0;
3780}
3781
3782static inline int __netif_set_xps_queue(struct net_device *dev,
3783 const unsigned long *mask,
3784 u16 index, bool is_rxqs_map)
3785{
3786 return 0;
3787}
3788#endif
3789
3790/**
3791 * netif_is_multiqueue - test if device has multiple transmit queues
3792 * @dev: network device
3793 *
3794 * Check if device has multiple transmit queues
3795 */
3796static inline bool netif_is_multiqueue(const struct net_device *dev)
3797{
3798 return dev->num_tx_queues > 1;
3799}
3800
3801int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3802
3803#ifdef CONFIG_SYSFS
3804int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3805#else
3806static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3807 unsigned int rxqs)
3808{
3809 dev->real_num_rx_queues = rxqs;
3810 return 0;
3811}
3812#endif
3813
3814static inline struct netdev_rx_queue *
3815__netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3816{
3817 return dev->_rx + rxq;
3818}
3819
3820#ifdef CONFIG_SYSFS
3821static inline unsigned int get_netdev_rx_queue_index(
3822 struct netdev_rx_queue *queue)
3823{
3824 struct net_device *dev = queue->dev;
3825 int index = queue - dev->_rx;
3826
3827 BUG_ON(index >= dev->num_rx_queues);
3828 return index;
3829}
3830#endif
3831
3832#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3833int netif_get_num_default_rss_queues(void);
3834
3835enum skb_free_reason {
3836 SKB_REASON_CONSUMED,
3837 SKB_REASON_DROPPED,
3838};
3839
3840void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3841void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3842
3843/*
3844 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3845 * interrupt context or with hardware interrupts being disabled.
3846 * (in_irq() || irqs_disabled())
3847 *
3848 * We provide four helpers that can be used in following contexts :
3849 *
3850 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3851 * replacing kfree_skb(skb)
3852 *
3853 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3854 * Typically used in place of consume_skb(skb) in TX completion path
3855 *
3856 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3857 * replacing kfree_skb(skb)
3858 *
3859 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3860 * and consumed a packet. Used in place of consume_skb(skb)
3861 */
3862static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3863{
3864 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3865}
3866
3867static inline void dev_consume_skb_irq(struct sk_buff *skb)
3868{
3869 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3870}
3871
3872static inline void dev_kfree_skb_any(struct sk_buff *skb)
3873{
3874 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3875}
3876
3877static inline void dev_consume_skb_any(struct sk_buff *skb)
3878{
3879 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3880}
3881
3882void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3883int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3884int netif_rx(struct sk_buff *skb);
3885int netif_rx_ni(struct sk_buff *skb);
3886int netif_rx_any_context(struct sk_buff *skb);
3887int netif_receive_skb(struct sk_buff *skb);
3888int netif_receive_skb_core(struct sk_buff *skb);
3889void netif_receive_skb_list(struct list_head *head);
3890gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3891void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3892struct sk_buff *napi_get_frags(struct napi_struct *napi);
3893gro_result_t napi_gro_frags(struct napi_struct *napi);
3894struct packet_offload *gro_find_receive_by_type(__be16 type);
3895struct packet_offload *gro_find_complete_by_type(__be16 type);
3896
3897static inline void napi_free_frags(struct napi_struct *napi)
3898{
3899 kfree_skb(napi->skb);
3900 napi->skb = NULL;
3901}
3902
3903bool netdev_is_rx_handler_busy(struct net_device *dev);
3904int netdev_rx_handler_register(struct net_device *dev,
3905 rx_handler_func_t *rx_handler,
3906 void *rx_handler_data);
3907void netdev_rx_handler_unregister(struct net_device *dev);
3908
3909bool dev_valid_name(const char *name);
3910int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3911 bool *need_copyout);
3912int dev_ifconf(struct net *net, struct ifconf *, int);
3913int dev_ethtool(struct net *net, struct ifreq *);
3914unsigned int dev_get_flags(const struct net_device *);
3915int __dev_change_flags(struct net_device *dev, unsigned int flags,
3916 struct netlink_ext_ack *extack);
3917int dev_change_flags(struct net_device *dev, unsigned int flags,
3918 struct netlink_ext_ack *extack);
3919void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3920 unsigned int gchanges);
3921int dev_change_name(struct net_device *, const char *);
3922int dev_set_alias(struct net_device *, const char *, size_t);
3923int dev_get_alias(const struct net_device *, char *, size_t);
3924int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3925int __dev_set_mtu(struct net_device *, int);
3926int dev_validate_mtu(struct net_device *dev, int mtu,
3927 struct netlink_ext_ack *extack);
3928int dev_set_mtu_ext(struct net_device *dev, int mtu,
3929 struct netlink_ext_ack *extack);
3930int dev_set_mtu(struct net_device *, int);
3931int dev_change_tx_queue_len(struct net_device *, unsigned long);
3932void dev_set_group(struct net_device *, int);
3933int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3934 struct netlink_ext_ack *extack);
3935int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3936 struct netlink_ext_ack *extack);
3937int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3938 struct netlink_ext_ack *extack);
3939int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3940int dev_change_carrier(struct net_device *, bool new_carrier);
3941int dev_get_phys_port_id(struct net_device *dev,
3942 struct netdev_phys_item_id *ppid);
3943int dev_get_phys_port_name(struct net_device *dev,
3944 char *name, size_t len);
3945int dev_get_port_parent_id(struct net_device *dev,
3946 struct netdev_phys_item_id *ppid, bool recurse);
3947bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3948int dev_change_proto_down(struct net_device *dev, bool proto_down);
3949int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3950void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3951 u32 value);
3952struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3953struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3954 struct netdev_queue *txq, int *ret);
3955
3956typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3957int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3958 int fd, int expected_fd, u32 flags);
3959int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3960u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3961
3962int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3963int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3964int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3965bool is_skb_forwardable(const struct net_device *dev,
3966 const struct sk_buff *skb);
3967
3968static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3969 const struct sk_buff *skb,
3970 const bool check_mtu)
3971{
3972 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3973 unsigned int len;
3974
3975 if (!(dev->flags & IFF_UP))
3976 return false;
3977
3978 if (!check_mtu)
3979 return true;
3980
3981 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3982 if (skb->len <= len)
3983 return true;
3984
3985 /* if TSO is enabled, we don't care about the length as the packet
3986 * could be forwarded without being segmented before
3987 */
3988 if (skb_is_gso(skb))
3989 return true;
3990
3991 return false;
3992}
3993
3994static __always_inline int ____dev_forward_skb(struct net_device *dev,
3995 struct sk_buff *skb,
3996 const bool check_mtu)
3997{
3998 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3999 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4000 atomic_long_inc(&dev->rx_dropped);
4001 kfree_skb(skb);
4002 return NET_RX_DROP;
4003 }
4004
4005 skb_scrub_packet(skb, true);
4006 skb->priority = 0;
4007 return 0;
4008}
4009
4010bool dev_nit_active(struct net_device *dev);
4011void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4012
4013extern int netdev_budget;
4014extern unsigned int netdev_budget_usecs;
4015
4016/* Called by rtnetlink.c:rtnl_unlock() */
4017void netdev_run_todo(void);
4018
4019/**
4020 * dev_put - release reference to device
4021 * @dev: network device
4022 *
4023 * Release reference to device to allow it to be freed.
4024 */
4025static inline void dev_put(struct net_device *dev)
4026{
4027 this_cpu_dec(*dev->pcpu_refcnt);
4028}
4029
4030/**
4031 * dev_hold - get reference to device
4032 * @dev: network device
4033 *
4034 * Hold reference to device to keep it from being freed.
4035 */
4036static inline void dev_hold(struct net_device *dev)
4037{
4038 this_cpu_inc(*dev->pcpu_refcnt);
4039}
4040
4041/* Carrier loss detection, dial on demand. The functions netif_carrier_on
4042 * and _off may be called from IRQ context, but it is caller
4043 * who is responsible for serialization of these calls.
4044 *
4045 * The name carrier is inappropriate, these functions should really be
4046 * called netif_lowerlayer_*() because they represent the state of any
4047 * kind of lower layer not just hardware media.
4048 */
4049
4050void linkwatch_init_dev(struct net_device *dev);
4051void linkwatch_fire_event(struct net_device *dev);
4052void linkwatch_forget_dev(struct net_device *dev);
4053
4054/**
4055 * netif_carrier_ok - test if carrier present
4056 * @dev: network device
4057 *
4058 * Check if carrier is present on device
4059 */
4060static inline bool netif_carrier_ok(const struct net_device *dev)
4061{
4062 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4063}
4064
4065unsigned long dev_trans_start(struct net_device *dev);
4066
4067void __netdev_watchdog_up(struct net_device *dev);
4068
4069void netif_carrier_on(struct net_device *dev);
4070
4071void netif_carrier_off(struct net_device *dev);
4072
4073/**
4074 * netif_dormant_on - mark device as dormant.
4075 * @dev: network device
4076 *
4077 * Mark device as dormant (as per RFC2863).
4078 *
4079 * The dormant state indicates that the relevant interface is not
4080 * actually in a condition to pass packets (i.e., it is not 'up') but is
4081 * in a "pending" state, waiting for some external event. For "on-
4082 * demand" interfaces, this new state identifies the situation where the
4083 * interface is waiting for events to place it in the up state.
4084 */
4085static inline void netif_dormant_on(struct net_device *dev)
4086{
4087 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4088 linkwatch_fire_event(dev);
4089}
4090
4091/**
4092 * netif_dormant_off - set device as not dormant.
4093 * @dev: network device
4094 *
4095 * Device is not in dormant state.
4096 */
4097static inline void netif_dormant_off(struct net_device *dev)
4098{
4099 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4100 linkwatch_fire_event(dev);
4101}
4102
4103/**
4104 * netif_dormant - test if device is dormant
4105 * @dev: network device
4106 *
4107 * Check if device is dormant.
4108 */
4109static inline bool netif_dormant(const struct net_device *dev)
4110{
4111 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4112}
4113
4114
4115/**
4116 * netif_testing_on - mark device as under test.
4117 * @dev: network device
4118 *
4119 * Mark device as under test (as per RFC2863).
4120 *
4121 * The testing state indicates that some test(s) must be performed on
4122 * the interface. After completion, of the test, the interface state
4123 * will change to up, dormant, or down, as appropriate.
4124 */
4125static inline void netif_testing_on(struct net_device *dev)
4126{
4127 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4128 linkwatch_fire_event(dev);
4129}
4130
4131/**
4132 * netif_testing_off - set device as not under test.
4133 * @dev: network device
4134 *
4135 * Device is not in testing state.
4136 */
4137static inline void netif_testing_off(struct net_device *dev)
4138{
4139 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4140 linkwatch_fire_event(dev);
4141}
4142
4143/**
4144 * netif_testing - test if device is under test
4145 * @dev: network device
4146 *
4147 * Check if device is under test
4148 */
4149static inline bool netif_testing(const struct net_device *dev)
4150{
4151 return test_bit(__LINK_STATE_TESTING, &dev->state);
4152}
4153
4154
4155/**
4156 * netif_oper_up - test if device is operational
4157 * @dev: network device
4158 *
4159 * Check if carrier is operational
4160 */
4161static inline bool netif_oper_up(const struct net_device *dev)
4162{
4163 return (dev->operstate == IF_OPER_UP ||
4164 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4165}
4166
4167/**
4168 * netif_device_present - is device available or removed
4169 * @dev: network device
4170 *
4171 * Check if device has not been removed from system.
4172 */
4173static inline bool netif_device_present(struct net_device *dev)
4174{
4175 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4176}
4177
4178void netif_device_detach(struct net_device *dev);
4179
4180void netif_device_attach(struct net_device *dev);
4181
4182/*
4183 * Network interface message level settings
4184 */
4185
4186enum {
4187 NETIF_MSG_DRV_BIT,
4188 NETIF_MSG_PROBE_BIT,
4189 NETIF_MSG_LINK_BIT,
4190 NETIF_MSG_TIMER_BIT,
4191 NETIF_MSG_IFDOWN_BIT,
4192 NETIF_MSG_IFUP_BIT,
4193 NETIF_MSG_RX_ERR_BIT,
4194 NETIF_MSG_TX_ERR_BIT,
4195 NETIF_MSG_TX_QUEUED_BIT,
4196 NETIF_MSG_INTR_BIT,
4197 NETIF_MSG_TX_DONE_BIT,
4198 NETIF_MSG_RX_STATUS_BIT,
4199 NETIF_MSG_PKTDATA_BIT,
4200 NETIF_MSG_HW_BIT,
4201 NETIF_MSG_WOL_BIT,
4202
4203 /* When you add a new bit above, update netif_msg_class_names array
4204 * in net/ethtool/common.c
4205 */
4206 NETIF_MSG_CLASS_COUNT,
4207};
4208/* Both ethtool_ops interface and internal driver implementation use u32 */
4209static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4210
4211#define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4212#define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4213
4214#define NETIF_MSG_DRV __NETIF_MSG(DRV)
4215#define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4216#define NETIF_MSG_LINK __NETIF_MSG(LINK)
4217#define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4218#define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4219#define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4220#define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4221#define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4222#define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4223#define NETIF_MSG_INTR __NETIF_MSG(INTR)
4224#define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4225#define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4226#define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4227#define NETIF_MSG_HW __NETIF_MSG(HW)
4228#define NETIF_MSG_WOL __NETIF_MSG(WOL)
4229
4230#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4231#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4232#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4233#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4234#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4235#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4236#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4237#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4238#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4239#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4240#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4241#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4242#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4243#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4244#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4245
4246static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4247{
4248 /* use default */
4249 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4250 return default_msg_enable_bits;
4251 if (debug_value == 0) /* no output */
4252 return 0;
4253 /* set low N bits */
4254 return (1U << debug_value) - 1;
4255}
4256
4257static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4258{
4259 spin_lock(&txq->_xmit_lock);
4260 txq->xmit_lock_owner = cpu;
4261}
4262
4263static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4264{
4265 __acquire(&txq->_xmit_lock);
4266 return true;
4267}
4268
4269static inline void __netif_tx_release(struct netdev_queue *txq)
4270{
4271 __release(&txq->_xmit_lock);
4272}
4273
4274static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4275{
4276 spin_lock_bh(&txq->_xmit_lock);
4277 txq->xmit_lock_owner = smp_processor_id();
4278}
4279
4280static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4281{
4282 bool ok = spin_trylock(&txq->_xmit_lock);
4283 if (likely(ok))
4284 txq->xmit_lock_owner = smp_processor_id();
4285 return ok;
4286}
4287
4288static inline void __netif_tx_unlock(struct netdev_queue *txq)
4289{
4290 txq->xmit_lock_owner = -1;
4291 spin_unlock(&txq->_xmit_lock);
4292}
4293
4294static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4295{
4296 txq->xmit_lock_owner = -1;
4297 spin_unlock_bh(&txq->_xmit_lock);
4298}
4299
4300static inline void txq_trans_update(struct netdev_queue *txq)
4301{
4302 if (txq->xmit_lock_owner != -1)
4303 txq->trans_start = jiffies;
4304}
4305
4306/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4307static inline void netif_trans_update(struct net_device *dev)
4308{
4309 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4310
4311 if (txq->trans_start != jiffies)
4312 txq->trans_start = jiffies;
4313}
4314
4315/**
4316 * netif_tx_lock - grab network device transmit lock
4317 * @dev: network device
4318 *
4319 * Get network device transmit lock
4320 */
4321static inline void netif_tx_lock(struct net_device *dev)
4322{
4323 unsigned int i;
4324 int cpu;
4325
4326 spin_lock(&dev->tx_global_lock);
4327 cpu = smp_processor_id();
4328 for (i = 0; i < dev->num_tx_queues; i++) {
4329 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4330
4331 /* We are the only thread of execution doing a
4332 * freeze, but we have to grab the _xmit_lock in
4333 * order to synchronize with threads which are in
4334 * the ->hard_start_xmit() handler and already
4335 * checked the frozen bit.
4336 */
4337 __netif_tx_lock(txq, cpu);
4338 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4339 __netif_tx_unlock(txq);
4340 }
4341}
4342
4343static inline void netif_tx_lock_bh(struct net_device *dev)
4344{
4345 local_bh_disable();
4346 netif_tx_lock(dev);
4347}
4348
4349static inline void netif_tx_unlock(struct net_device *dev)
4350{
4351 unsigned int i;
4352
4353 for (i = 0; i < dev->num_tx_queues; i++) {
4354 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4355
4356 /* No need to grab the _xmit_lock here. If the
4357 * queue is not stopped for another reason, we
4358 * force a schedule.
4359 */
4360 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4361 netif_schedule_queue(txq);
4362 }
4363 spin_unlock(&dev->tx_global_lock);
4364}
4365
4366static inline void netif_tx_unlock_bh(struct net_device *dev)
4367{
4368 netif_tx_unlock(dev);
4369 local_bh_enable();
4370}
4371
4372#define HARD_TX_LOCK(dev, txq, cpu) { \
4373 if ((dev->features & NETIF_F_LLTX) == 0) { \
4374 __netif_tx_lock(txq, cpu); \
4375 } else { \
4376 __netif_tx_acquire(txq); \
4377 } \
4378}
4379
4380#define HARD_TX_TRYLOCK(dev, txq) \
4381 (((dev->features & NETIF_F_LLTX) == 0) ? \
4382 __netif_tx_trylock(txq) : \
4383 __netif_tx_acquire(txq))
4384
4385#define HARD_TX_UNLOCK(dev, txq) { \
4386 if ((dev->features & NETIF_F_LLTX) == 0) { \
4387 __netif_tx_unlock(txq); \
4388 } else { \
4389 __netif_tx_release(txq); \
4390 } \
4391}
4392
4393static inline void netif_tx_disable(struct net_device *dev)
4394{
4395 unsigned int i;
4396 int cpu;
4397
4398 local_bh_disable();
4399 cpu = smp_processor_id();
4400 spin_lock(&dev->tx_global_lock);
4401 for (i = 0; i < dev->num_tx_queues; i++) {
4402 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4403
4404 __netif_tx_lock(txq, cpu);
4405 netif_tx_stop_queue(txq);
4406 __netif_tx_unlock(txq);
4407 }
4408 spin_unlock(&dev->tx_global_lock);
4409 local_bh_enable();
4410}
4411
4412static inline void netif_addr_lock(struct net_device *dev)
4413{
4414 unsigned char nest_level = 0;
4415
4416#ifdef CONFIG_LOCKDEP
4417 nest_level = dev->nested_level;
4418#endif
4419 spin_lock_nested(&dev->addr_list_lock, nest_level);
4420}
4421
4422static inline void netif_addr_lock_bh(struct net_device *dev)
4423{
4424 unsigned char nest_level = 0;
4425
4426#ifdef CONFIG_LOCKDEP
4427 nest_level = dev->nested_level;
4428#endif
4429 local_bh_disable();
4430 spin_lock_nested(&dev->addr_list_lock, nest_level);
4431}
4432
4433static inline void netif_addr_unlock(struct net_device *dev)
4434{
4435 spin_unlock(&dev->addr_list_lock);
4436}
4437
4438static inline void netif_addr_unlock_bh(struct net_device *dev)
4439{
4440 spin_unlock_bh(&dev->addr_list_lock);
4441}
4442
4443/*
4444 * dev_addrs walker. Should be used only for read access. Call with
4445 * rcu_read_lock held.
4446 */
4447#define for_each_dev_addr(dev, ha) \
4448 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4449
4450/* These functions live elsewhere (drivers/net/net_init.c, but related) */
4451
4452void ether_setup(struct net_device *dev);
4453
4454/* Support for loadable net-drivers */
4455struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4456 unsigned char name_assign_type,
4457 void (*setup)(struct net_device *),
4458 unsigned int txqs, unsigned int rxqs);
4459#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4460 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4461
4462#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4463 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4464 count)
4465
4466int register_netdev(struct net_device *dev);
4467void unregister_netdev(struct net_device *dev);
4468
4469int devm_register_netdev(struct device *dev, struct net_device *ndev);
4470
4471/* General hardware address lists handling functions */
4472int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4473 struct netdev_hw_addr_list *from_list, int addr_len);
4474void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4475 struct netdev_hw_addr_list *from_list, int addr_len);
4476int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4477 struct net_device *dev,
4478 int (*sync)(struct net_device *, const unsigned char *),
4479 int (*unsync)(struct net_device *,
4480 const unsigned char *));
4481int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4482 struct net_device *dev,
4483 int (*sync)(struct net_device *,
4484 const unsigned char *, int),
4485 int (*unsync)(struct net_device *,
4486 const unsigned char *, int));
4487void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4488 struct net_device *dev,
4489 int (*unsync)(struct net_device *,
4490 const unsigned char *, int));
4491void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4492 struct net_device *dev,
4493 int (*unsync)(struct net_device *,
4494 const unsigned char *));
4495void __hw_addr_init(struct netdev_hw_addr_list *list);
4496
4497/* Functions used for device addresses handling */
4498int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4499 unsigned char addr_type);
4500int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4501 unsigned char addr_type);
4502void dev_addr_flush(struct net_device *dev);
4503int dev_addr_init(struct net_device *dev);
4504
4505/* Functions used for unicast addresses handling */
4506int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4507int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4508int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4509int dev_uc_sync(struct net_device *to, struct net_device *from);
4510int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4511void dev_uc_unsync(struct net_device *to, struct net_device *from);
4512void dev_uc_flush(struct net_device *dev);
4513void dev_uc_init(struct net_device *dev);
4514
4515/**
4516 * __dev_uc_sync - Synchonize device's unicast list
4517 * @dev: device to sync
4518 * @sync: function to call if address should be added
4519 * @unsync: function to call if address should be removed
4520 *
4521 * Add newly added addresses to the interface, and release
4522 * addresses that have been deleted.
4523 */
4524static inline int __dev_uc_sync(struct net_device *dev,
4525 int (*sync)(struct net_device *,
4526 const unsigned char *),
4527 int (*unsync)(struct net_device *,
4528 const unsigned char *))
4529{
4530 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4531}
4532
4533/**
4534 * __dev_uc_unsync - Remove synchronized addresses from device
4535 * @dev: device to sync
4536 * @unsync: function to call if address should be removed
4537 *
4538 * Remove all addresses that were added to the device by dev_uc_sync().
4539 */
4540static inline void __dev_uc_unsync(struct net_device *dev,
4541 int (*unsync)(struct net_device *,
4542 const unsigned char *))
4543{
4544 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4545}
4546
4547/* Functions used for multicast addresses handling */
4548int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4549int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4550int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4551int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4552int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4553int dev_mc_sync(struct net_device *to, struct net_device *from);
4554int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4555void dev_mc_unsync(struct net_device *to, struct net_device *from);
4556void dev_mc_flush(struct net_device *dev);
4557void dev_mc_init(struct net_device *dev);
4558
4559/**
4560 * __dev_mc_sync - Synchonize device's multicast list
4561 * @dev: device to sync
4562 * @sync: function to call if address should be added
4563 * @unsync: function to call if address should be removed
4564 *
4565 * Add newly added addresses to the interface, and release
4566 * addresses that have been deleted.
4567 */
4568static inline int __dev_mc_sync(struct net_device *dev,
4569 int (*sync)(struct net_device *,
4570 const unsigned char *),
4571 int (*unsync)(struct net_device *,
4572 const unsigned char *))
4573{
4574 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4575}
4576
4577/**
4578 * __dev_mc_unsync - Remove synchronized addresses from device
4579 * @dev: device to sync
4580 * @unsync: function to call if address should be removed
4581 *
4582 * Remove all addresses that were added to the device by dev_mc_sync().
4583 */
4584static inline void __dev_mc_unsync(struct net_device *dev,
4585 int (*unsync)(struct net_device *,
4586 const unsigned char *))
4587{
4588 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4589}
4590
4591/* Functions used for secondary unicast and multicast support */
4592void dev_set_rx_mode(struct net_device *dev);
4593void __dev_set_rx_mode(struct net_device *dev);
4594int dev_set_promiscuity(struct net_device *dev, int inc);
4595int dev_set_allmulti(struct net_device *dev, int inc);
4596void netdev_state_change(struct net_device *dev);
4597void __netdev_notify_peers(struct net_device *dev);
4598void netdev_notify_peers(struct net_device *dev);
4599void netdev_features_change(struct net_device *dev);
4600/* Load a device via the kmod */
4601void dev_load(struct net *net, const char *name);
4602struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4603 struct rtnl_link_stats64 *storage);
4604void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4605 const struct net_device_stats *netdev_stats);
4606void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4607 const struct pcpu_sw_netstats __percpu *netstats);
4608void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4609
4610extern int netdev_max_backlog;
4611extern int netdev_tstamp_prequeue;
4612extern int weight_p;
4613extern int dev_weight_rx_bias;
4614extern int dev_weight_tx_bias;
4615extern int dev_rx_weight;
4616extern int dev_tx_weight;
4617extern int gro_normal_batch;
4618
4619enum {
4620 NESTED_SYNC_IMM_BIT,
4621 NESTED_SYNC_TODO_BIT,
4622};
4623
4624#define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4625#define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4626
4627#define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4628#define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4629
4630struct netdev_nested_priv {
4631 unsigned char flags;
4632 void *data;
4633};
4634
4635bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4636struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4637 struct list_head **iter);
4638struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4639 struct list_head **iter);
4640
4641#ifdef CONFIG_LOCKDEP
4642static LIST_HEAD(net_unlink_list);
4643
4644static inline void net_unlink_todo(struct net_device *dev)
4645{
4646 if (list_empty(&dev->unlink_list))
4647 list_add_tail(&dev->unlink_list, &net_unlink_list);
4648}
4649#endif
4650
4651/* iterate through upper list, must be called under RCU read lock */
4652#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4653 for (iter = &(dev)->adj_list.upper, \
4654 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4655 updev; \
4656 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4657
4658int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4659 int (*fn)(struct net_device *upper_dev,
4660 struct netdev_nested_priv *priv),
4661 struct netdev_nested_priv *priv);
4662
4663bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4664 struct net_device *upper_dev);
4665
4666bool netdev_has_any_upper_dev(struct net_device *dev);
4667
4668void *netdev_lower_get_next_private(struct net_device *dev,
4669 struct list_head **iter);
4670void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4671 struct list_head **iter);
4672
4673#define netdev_for_each_lower_private(dev, priv, iter) \
4674 for (iter = (dev)->adj_list.lower.next, \
4675 priv = netdev_lower_get_next_private(dev, &(iter)); \
4676 priv; \
4677 priv = netdev_lower_get_next_private(dev, &(iter)))
4678
4679#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4680 for (iter = &(dev)->adj_list.lower, \
4681 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4682 priv; \
4683 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4684
4685void *netdev_lower_get_next(struct net_device *dev,
4686 struct list_head **iter);
4687
4688#define netdev_for_each_lower_dev(dev, ldev, iter) \
4689 for (iter = (dev)->adj_list.lower.next, \
4690 ldev = netdev_lower_get_next(dev, &(iter)); \
4691 ldev; \
4692 ldev = netdev_lower_get_next(dev, &(iter)))
4693
4694struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4695 struct list_head **iter);
4696int netdev_walk_all_lower_dev(struct net_device *dev,
4697 int (*fn)(struct net_device *lower_dev,
4698 struct netdev_nested_priv *priv),
4699 struct netdev_nested_priv *priv);
4700int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4701 int (*fn)(struct net_device *lower_dev,
4702 struct netdev_nested_priv *priv),
4703 struct netdev_nested_priv *priv);
4704
4705void *netdev_adjacent_get_private(struct list_head *adj_list);
4706void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4707struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4708struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4709int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4710 struct netlink_ext_ack *extack);
4711int netdev_master_upper_dev_link(struct net_device *dev,
4712 struct net_device *upper_dev,
4713 void *upper_priv, void *upper_info,
4714 struct netlink_ext_ack *extack);
4715void netdev_upper_dev_unlink(struct net_device *dev,
4716 struct net_device *upper_dev);
4717int netdev_adjacent_change_prepare(struct net_device *old_dev,
4718 struct net_device *new_dev,
4719 struct net_device *dev,
4720 struct netlink_ext_ack *extack);
4721void netdev_adjacent_change_commit(struct net_device *old_dev,
4722 struct net_device *new_dev,
4723 struct net_device *dev);
4724void netdev_adjacent_change_abort(struct net_device *old_dev,
4725 struct net_device *new_dev,
4726 struct net_device *dev);
4727void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4728void *netdev_lower_dev_get_private(struct net_device *dev,
4729 struct net_device *lower_dev);
4730void netdev_lower_state_changed(struct net_device *lower_dev,
4731 void *lower_state_info);
4732
4733/* RSS keys are 40 or 52 bytes long */
4734#define NETDEV_RSS_KEY_LEN 52
4735extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4736void netdev_rss_key_fill(void *buffer, size_t len);
4737
4738int skb_checksum_help(struct sk_buff *skb);
4739int skb_crc32c_csum_help(struct sk_buff *skb);
4740int skb_csum_hwoffload_help(struct sk_buff *skb,
4741 const netdev_features_t features);
4742
4743struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4744 netdev_features_t features, bool tx_path);
4745struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4746 netdev_features_t features);
4747
4748struct netdev_bonding_info {
4749 ifslave slave;
4750 ifbond master;
4751};
4752
4753struct netdev_notifier_bonding_info {
4754 struct netdev_notifier_info info; /* must be first */
4755 struct netdev_bonding_info bonding_info;
4756};
4757
4758void netdev_bonding_info_change(struct net_device *dev,
4759 struct netdev_bonding_info *bonding_info);
4760
4761#if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4762void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4763#else
4764static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4765 const void *data)
4766{
4767}
4768#endif
4769
4770static inline
4771struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4772{
4773 return __skb_gso_segment(skb, features, true);
4774}
4775__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4776
4777static inline bool can_checksum_protocol(netdev_features_t features,
4778 __be16 protocol)
4779{
4780 if (protocol == htons(ETH_P_FCOE))
4781 return !!(features & NETIF_F_FCOE_CRC);
4782
4783 /* Assume this is an IP checksum (not SCTP CRC) */
4784
4785 if (features & NETIF_F_HW_CSUM) {
4786 /* Can checksum everything */
4787 return true;
4788 }
4789
4790 switch (protocol) {
4791 case htons(ETH_P_IP):
4792 return !!(features & NETIF_F_IP_CSUM);
4793 case htons(ETH_P_IPV6):
4794 return !!(features & NETIF_F_IPV6_CSUM);
4795 default:
4796 return false;
4797 }
4798}
4799
4800#ifdef CONFIG_BUG
4801void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4802#else
4803static inline void netdev_rx_csum_fault(struct net_device *dev,
4804 struct sk_buff *skb)
4805{
4806}
4807#endif
4808/* rx skb timestamps */
4809void net_enable_timestamp(void);
4810void net_disable_timestamp(void);
4811
4812#ifdef CONFIG_PROC_FS
4813int __init dev_proc_init(void);
4814#else
4815#define dev_proc_init() 0
4816#endif
4817
4818static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4819 struct sk_buff *skb, struct net_device *dev,
4820 bool more)
4821{
4822 __this_cpu_write(softnet_data.xmit.more, more);
4823 return ops->ndo_start_xmit(skb, dev);
4824}
4825
4826static inline bool netdev_xmit_more(void)
4827{
4828 return __this_cpu_read(softnet_data.xmit.more);
4829}
4830
4831static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4832 struct netdev_queue *txq, bool more)
4833{
4834 const struct net_device_ops *ops = dev->netdev_ops;
4835 netdev_tx_t rc;
4836
4837 rc = __netdev_start_xmit(ops, skb, dev, more);
4838 if (rc == NETDEV_TX_OK)
4839 txq_trans_update(txq);
4840
4841 return rc;
4842}
4843
4844int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4845 const void *ns);
4846void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4847 const void *ns);
4848
4849extern const struct kobj_ns_type_operations net_ns_type_operations;
4850
4851const char *netdev_drivername(const struct net_device *dev);
4852
4853void linkwatch_run_queue(void);
4854
4855static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4856 netdev_features_t f2)
4857{
4858 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4859 if (f1 & NETIF_F_HW_CSUM)
4860 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4861 else
4862 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4863 }
4864
4865 return f1 & f2;
4866}
4867
4868static inline netdev_features_t netdev_get_wanted_features(
4869 struct net_device *dev)
4870{
4871 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4872}
4873netdev_features_t netdev_increment_features(netdev_features_t all,
4874 netdev_features_t one, netdev_features_t mask);
4875
4876/* Allow TSO being used on stacked device :
4877 * Performing the GSO segmentation before last device
4878 * is a performance improvement.
4879 */
4880static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4881 netdev_features_t mask)
4882{
4883 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4884}
4885
4886int __netdev_update_features(struct net_device *dev);
4887void netdev_update_features(struct net_device *dev);
4888void netdev_change_features(struct net_device *dev);
4889
4890void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4891 struct net_device *dev);
4892
4893netdev_features_t passthru_features_check(struct sk_buff *skb,
4894 struct net_device *dev,
4895 netdev_features_t features);
4896netdev_features_t netif_skb_features(struct sk_buff *skb);
4897
4898static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4899{
4900 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4901
4902 /* check flags correspondence */
4903 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4904 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4905 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4906 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4907 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4908 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4909 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4910 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4911 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4912 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4913 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4914 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4915 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4916 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4917 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4918 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4919 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4920 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4921 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4922
4923 return (features & feature) == feature;
4924}
4925
4926static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4927{
4928 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4929 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4930}
4931
4932static inline bool netif_needs_gso(struct sk_buff *skb,
4933 netdev_features_t features)
4934{
4935 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4936 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4937 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4938}
4939
4940static inline void netif_set_gso_max_size(struct net_device *dev,
4941 unsigned int size)
4942{
4943 dev->gso_max_size = size;
4944}
4945
4946static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4947 int pulled_hlen, u16 mac_offset,
4948 int mac_len)
4949{
4950 skb->protocol = protocol;
4951 skb->encapsulation = 1;
4952 skb_push(skb, pulled_hlen);
4953 skb_reset_transport_header(skb);
4954 skb->mac_header = mac_offset;
4955 skb->network_header = skb->mac_header + mac_len;
4956 skb->mac_len = mac_len;
4957}
4958
4959static inline bool netif_is_macsec(const struct net_device *dev)
4960{
4961 return dev->priv_flags & IFF_MACSEC;
4962}
4963
4964static inline bool netif_is_macvlan(const struct net_device *dev)
4965{
4966 return dev->priv_flags & IFF_MACVLAN;
4967}
4968
4969static inline bool netif_is_macvlan_port(const struct net_device *dev)
4970{
4971 return dev->priv_flags & IFF_MACVLAN_PORT;
4972}
4973
4974static inline bool netif_is_bond_master(const struct net_device *dev)
4975{
4976 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4977}
4978
4979static inline bool netif_is_bond_slave(const struct net_device *dev)
4980{
4981 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4982}
4983
4984static inline bool netif_supports_nofcs(struct net_device *dev)
4985{
4986 return dev->priv_flags & IFF_SUPP_NOFCS;
4987}
4988
4989static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4990{
4991 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4992}
4993
4994static inline bool netif_is_l3_master(const struct net_device *dev)
4995{
4996 return dev->priv_flags & IFF_L3MDEV_MASTER;
4997}
4998
4999static inline bool netif_is_l3_slave(const struct net_device *dev)
5000{
5001 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5002}
5003
5004static inline bool netif_is_bridge_master(const struct net_device *dev)
5005{
5006 return dev->priv_flags & IFF_EBRIDGE;
5007}
5008
5009static inline bool netif_is_bridge_port(const struct net_device *dev)
5010{
5011 return dev->priv_flags & IFF_BRIDGE_PORT;
5012}
5013
5014static inline bool netif_is_ovs_master(const struct net_device *dev)
5015{
5016 return dev->priv_flags & IFF_OPENVSWITCH;
5017}
5018
5019static inline bool netif_is_ovs_port(const struct net_device *dev)
5020{
5021 return dev->priv_flags & IFF_OVS_DATAPATH;
5022}
5023
5024static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5025{
5026 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5027}
5028
5029static inline bool netif_is_team_master(const struct net_device *dev)
5030{
5031 return dev->priv_flags & IFF_TEAM;
5032}
5033
5034static inline bool netif_is_team_port(const struct net_device *dev)
5035{
5036 return dev->priv_flags & IFF_TEAM_PORT;
5037}
5038
5039static inline bool netif_is_lag_master(const struct net_device *dev)
5040{
5041 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5042}
5043
5044static inline bool netif_is_lag_port(const struct net_device *dev)
5045{
5046 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5047}
5048
5049static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5050{
5051 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5052}
5053
5054static inline bool netif_is_failover(const struct net_device *dev)
5055{
5056 return dev->priv_flags & IFF_FAILOVER;
5057}
5058
5059static inline bool netif_is_failover_slave(const struct net_device *dev)
5060{
5061 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5062}
5063
5064/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5065static inline void netif_keep_dst(struct net_device *dev)
5066{
5067 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5068}
5069
5070/* return true if dev can't cope with mtu frames that need vlan tag insertion */
5071static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5072{
5073 /* TODO: reserve and use an additional IFF bit, if we get more users */
5074 return dev->priv_flags & IFF_MACSEC;
5075}
5076
5077extern struct pernet_operations __net_initdata loopback_net_ops;
5078
5079/* Logging, debugging and troubleshooting/diagnostic helpers. */
5080
5081/* netdev_printk helpers, similar to dev_printk */
5082
5083static inline const char *netdev_name(const struct net_device *dev)
5084{
5085 if (!dev->name[0] || strchr(dev->name, '%'))
5086 return "(unnamed net_device)";
5087 return dev->name;
5088}
5089
5090static inline bool netdev_unregistering(const struct net_device *dev)
5091{
5092 return dev->reg_state == NETREG_UNREGISTERING;
5093}
5094
5095static inline const char *netdev_reg_state(const struct net_device *dev)
5096{
5097 switch (dev->reg_state) {
5098 case NETREG_UNINITIALIZED: return " (uninitialized)";
5099 case NETREG_REGISTERED: return "";
5100 case NETREG_UNREGISTERING: return " (unregistering)";
5101 case NETREG_UNREGISTERED: return " (unregistered)";
5102 case NETREG_RELEASED: return " (released)";
5103 case NETREG_DUMMY: return " (dummy)";
5104 }
5105
5106 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5107 return " (unknown)";
5108}
5109
5110__printf(3, 4) __cold
5111void netdev_printk(const char *level, const struct net_device *dev,
5112 const char *format, ...);
5113__printf(2, 3) __cold
5114void netdev_emerg(const struct net_device *dev, const char *format, ...);
5115__printf(2, 3) __cold
5116void netdev_alert(const struct net_device *dev, const char *format, ...);
5117__printf(2, 3) __cold
5118void netdev_crit(const struct net_device *dev, const char *format, ...);
5119__printf(2, 3) __cold
5120void netdev_err(const struct net_device *dev, const char *format, ...);
5121__printf(2, 3) __cold
5122void netdev_warn(const struct net_device *dev, const char *format, ...);
5123__printf(2, 3) __cold
5124void netdev_notice(const struct net_device *dev, const char *format, ...);
5125__printf(2, 3) __cold
5126void netdev_info(const struct net_device *dev, const char *format, ...);
5127
5128#define netdev_level_once(level, dev, fmt, ...) \
5129do { \
5130 static bool __print_once __read_mostly; \
5131 \
5132 if (!__print_once) { \
5133 __print_once = true; \
5134 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5135 } \
5136} while (0)
5137
5138#define netdev_emerg_once(dev, fmt, ...) \
5139 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5140#define netdev_alert_once(dev, fmt, ...) \
5141 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5142#define netdev_crit_once(dev, fmt, ...) \
5143 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5144#define netdev_err_once(dev, fmt, ...) \
5145 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5146#define netdev_warn_once(dev, fmt, ...) \
5147 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5148#define netdev_notice_once(dev, fmt, ...) \
5149 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5150#define netdev_info_once(dev, fmt, ...) \
5151 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5152
5153#define MODULE_ALIAS_NETDEV(device) \
5154 MODULE_ALIAS("netdev-" device)
5155
5156#if defined(CONFIG_DYNAMIC_DEBUG) || \
5157 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5158#define netdev_dbg(__dev, format, args...) \
5159do { \
5160 dynamic_netdev_dbg(__dev, format, ##args); \
5161} while (0)
5162#elif defined(DEBUG)
5163#define netdev_dbg(__dev, format, args...) \
5164 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5165#else
5166#define netdev_dbg(__dev, format, args...) \
5167({ \
5168 if (0) \
5169 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5170})
5171#endif
5172
5173#if defined(VERBOSE_DEBUG)
5174#define netdev_vdbg netdev_dbg
5175#else
5176
5177#define netdev_vdbg(dev, format, args...) \
5178({ \
5179 if (0) \
5180 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5181 0; \
5182})
5183#endif
5184
5185/*
5186 * netdev_WARN() acts like dev_printk(), but with the key difference
5187 * of using a WARN/WARN_ON to get the message out, including the
5188 * file/line information and a backtrace.
5189 */
5190#define netdev_WARN(dev, format, args...) \
5191 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5192 netdev_reg_state(dev), ##args)
5193
5194#define netdev_WARN_ONCE(dev, format, args...) \
5195 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5196 netdev_reg_state(dev), ##args)
5197
5198/* netif printk helpers, similar to netdev_printk */
5199
5200#define netif_printk(priv, type, level, dev, fmt, args...) \
5201do { \
5202 if (netif_msg_##type(priv)) \
5203 netdev_printk(level, (dev), fmt, ##args); \
5204} while (0)
5205
5206#define netif_level(level, priv, type, dev, fmt, args...) \
5207do { \
5208 if (netif_msg_##type(priv)) \
5209 netdev_##level(dev, fmt, ##args); \
5210} while (0)
5211
5212#define netif_emerg(priv, type, dev, fmt, args...) \
5213 netif_level(emerg, priv, type, dev, fmt, ##args)
5214#define netif_alert(priv, type, dev, fmt, args...) \
5215 netif_level(alert, priv, type, dev, fmt, ##args)
5216#define netif_crit(priv, type, dev, fmt, args...) \
5217 netif_level(crit, priv, type, dev, fmt, ##args)
5218#define netif_err(priv, type, dev, fmt, args...) \
5219 netif_level(err, priv, type, dev, fmt, ##args)
5220#define netif_warn(priv, type, dev, fmt, args...) \
5221 netif_level(warn, priv, type, dev, fmt, ##args)
5222#define netif_notice(priv, type, dev, fmt, args...) \
5223 netif_level(notice, priv, type, dev, fmt, ##args)
5224#define netif_info(priv, type, dev, fmt, args...) \
5225 netif_level(info, priv, type, dev, fmt, ##args)
5226
5227#if defined(CONFIG_DYNAMIC_DEBUG) || \
5228 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5229#define netif_dbg(priv, type, netdev, format, args...) \
5230do { \
5231 if (netif_msg_##type(priv)) \
5232 dynamic_netdev_dbg(netdev, format, ##args); \
5233} while (0)
5234#elif defined(DEBUG)
5235#define netif_dbg(priv, type, dev, format, args...) \
5236 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5237#else
5238#define netif_dbg(priv, type, dev, format, args...) \
5239({ \
5240 if (0) \
5241 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5242 0; \
5243})
5244#endif
5245
5246/* if @cond then downgrade to debug, else print at @level */
5247#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5248 do { \
5249 if (cond) \
5250 netif_dbg(priv, type, netdev, fmt, ##args); \
5251 else \
5252 netif_ ## level(priv, type, netdev, fmt, ##args); \
5253 } while (0)
5254
5255#if defined(VERBOSE_DEBUG)
5256#define netif_vdbg netif_dbg
5257#else
5258#define netif_vdbg(priv, type, dev, format, args...) \
5259({ \
5260 if (0) \
5261 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5262 0; \
5263})
5264#endif
5265
5266/*
5267 * The list of packet types we will receive (as opposed to discard)
5268 * and the routines to invoke.
5269 *
5270 * Why 16. Because with 16 the only overlap we get on a hash of the
5271 * low nibble of the protocol value is RARP/SNAP/X.25.
5272 *
5273 * 0800 IP
5274 * 0001 802.3
5275 * 0002 AX.25
5276 * 0004 802.2
5277 * 8035 RARP
5278 * 0005 SNAP
5279 * 0805 X.25
5280 * 0806 ARP
5281 * 8137 IPX
5282 * 0009 Localtalk
5283 * 86DD IPv6
5284 */
5285#define PTYPE_HASH_SIZE (16)
5286#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5287
5288extern struct net_device *blackhole_netdev;
5289
5290#endif /* _LINUX_NETDEVICE_H */