1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
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>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
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 #include <asm/local.h>
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
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 <uapi/linux/netdev.h>
50 #include <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
54 #include <net/dropreason-core.h>
59 struct kernel_hwtstamp_config;
62 struct ip_tunnel_parm;
63 struct macsec_context;
65 struct netdev_name_node;
70 /* 802.15.4 specific */
73 /* UDP Tunnel offloads */
74 struct udp_tunnel_info;
75 struct udp_tunnel_nic_info;
76 struct udp_tunnel_nic;
80 struct xdp_metadata_ops;
83 typedef u32 xdp_features_t;
85 void synchronize_net(void);
86 void netdev_set_default_ethtool_ops(struct net_device *dev,
87 const struct ethtool_ops *ops);
88 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
90 /* Backlog congestion levels */
91 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
92 #define NET_RX_DROP 1 /* packet dropped */
94 #define MAX_NEST_DEV 8
97 * Transmit return codes: transmit return codes originate from three different
100 * - qdisc return codes
101 * - driver transmit return codes
104 * Drivers are allowed to return any one of those in their hard_start_xmit()
105 * function. Real network devices commonly used with qdiscs should only return
106 * the driver transmit return codes though - when qdiscs are used, the actual
107 * transmission happens asynchronously, so the value is not propagated to
108 * higher layers. Virtual network devices transmit synchronously; in this case
109 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
110 * others are propagated to higher layers.
113 /* qdisc ->enqueue() return codes. */
114 #define NET_XMIT_SUCCESS 0x00
115 #define NET_XMIT_DROP 0x01 /* skb dropped */
116 #define NET_XMIT_CN 0x02 /* congestion notification */
117 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
119 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
120 * indicates that the device will soon be dropping packets, or already drops
121 * some packets of the same priority; prompting us to send less aggressively. */
122 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
123 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
125 /* Driver transmit return codes */
126 #define NETDEV_TX_MASK 0xf0
129 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
130 NETDEV_TX_OK = 0x00, /* driver took care of packet */
131 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
133 typedef enum netdev_tx netdev_tx_t;
136 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
137 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
139 static inline bool dev_xmit_complete(int rc)
142 * Positive cases with an skb consumed by a driver:
143 * - successful transmission (rc == NETDEV_TX_OK)
144 * - error while transmitting (rc < 0)
145 * - error while queueing to a different device (rc & NET_XMIT_MASK)
147 if (likely(rc < NET_XMIT_MASK))
154 * Compute the worst-case header length according to the protocols
158 #if defined(CONFIG_HYPERV_NET)
159 # define LL_MAX_HEADER 128
160 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
161 # if defined(CONFIG_MAC80211_MESH)
162 # define LL_MAX_HEADER 128
164 # define LL_MAX_HEADER 96
167 # define LL_MAX_HEADER 32
170 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
171 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
172 #define MAX_HEADER LL_MAX_HEADER
174 #define MAX_HEADER (LL_MAX_HEADER + 48)
178 * Old network device statistics. Fields are native words
179 * (unsigned long) so they can be read and written atomically.
182 #define NET_DEV_STAT(FIELD) \
184 unsigned long FIELD; \
185 atomic_long_t __##FIELD; \
188 struct net_device_stats {
189 NET_DEV_STAT(rx_packets);
190 NET_DEV_STAT(tx_packets);
191 NET_DEV_STAT(rx_bytes);
192 NET_DEV_STAT(tx_bytes);
193 NET_DEV_STAT(rx_errors);
194 NET_DEV_STAT(tx_errors);
195 NET_DEV_STAT(rx_dropped);
196 NET_DEV_STAT(tx_dropped);
197 NET_DEV_STAT(multicast);
198 NET_DEV_STAT(collisions);
199 NET_DEV_STAT(rx_length_errors);
200 NET_DEV_STAT(rx_over_errors);
201 NET_DEV_STAT(rx_crc_errors);
202 NET_DEV_STAT(rx_frame_errors);
203 NET_DEV_STAT(rx_fifo_errors);
204 NET_DEV_STAT(rx_missed_errors);
205 NET_DEV_STAT(tx_aborted_errors);
206 NET_DEV_STAT(tx_carrier_errors);
207 NET_DEV_STAT(tx_fifo_errors);
208 NET_DEV_STAT(tx_heartbeat_errors);
209 NET_DEV_STAT(tx_window_errors);
210 NET_DEV_STAT(rx_compressed);
211 NET_DEV_STAT(tx_compressed);
215 /* per-cpu stats, allocated on demand.
216 * Try to fit them in a single cache line, for dev_get_stats() sake.
218 struct net_device_core_stats {
219 unsigned long rx_dropped;
220 unsigned long tx_dropped;
221 unsigned long rx_nohandler;
222 unsigned long rx_otherhost_dropped;
223 } __aligned(4 * sizeof(unsigned long));
225 #include <linux/cache.h>
226 #include <linux/skbuff.h>
229 #include <linux/static_key.h>
230 extern struct static_key_false rps_needed;
231 extern struct static_key_false rfs_needed;
238 struct netdev_hw_addr {
239 struct list_head list;
241 unsigned char addr[MAX_ADDR_LEN];
243 #define NETDEV_HW_ADDR_T_LAN 1
244 #define NETDEV_HW_ADDR_T_SAN 2
245 #define NETDEV_HW_ADDR_T_UNICAST 3
246 #define NETDEV_HW_ADDR_T_MULTICAST 4
251 struct rcu_head rcu_head;
254 struct netdev_hw_addr_list {
255 struct list_head list;
258 /* Auxiliary tree for faster lookup on addition and deletion */
262 #define netdev_hw_addr_list_count(l) ((l)->count)
263 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
264 #define netdev_hw_addr_list_for_each(ha, l) \
265 list_for_each_entry(ha, &(l)->list, list)
267 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
268 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
269 #define netdev_for_each_uc_addr(ha, dev) \
270 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
271 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
272 netdev_for_each_uc_addr((_ha), (_dev)) \
275 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
276 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
277 #define netdev_for_each_mc_addr(ha, dev) \
278 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
279 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
280 netdev_for_each_mc_addr((_ha), (_dev)) \
287 /* cached hardware header; allow for machine alignment needs. */
288 #define HH_DATA_MOD 16
289 #define HH_DATA_OFF(__len) \
290 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
291 #define HH_DATA_ALIGN(__len) \
292 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
293 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
296 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
298 * dev->hard_header_len ? (dev->hard_header_len +
299 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
301 * We could use other alignment values, but we must maintain the
302 * relationship HH alignment <= LL alignment.
304 #define LL_RESERVED_SPACE(dev) \
305 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
306 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
307 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
308 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
309 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
312 int (*create) (struct sk_buff *skb, struct net_device *dev,
313 unsigned short type, const void *daddr,
314 const void *saddr, unsigned int len);
315 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
316 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
317 void (*cache_update)(struct hh_cache *hh,
318 const struct net_device *dev,
319 const unsigned char *haddr);
320 bool (*validate)(const char *ll_header, unsigned int len);
321 __be16 (*parse_protocol)(const struct sk_buff *skb);
324 /* These flag bits are private to the generic network queueing
325 * layer; they may not be explicitly referenced by any other
329 enum netdev_state_t {
331 __LINK_STATE_PRESENT,
332 __LINK_STATE_NOCARRIER,
333 __LINK_STATE_LINKWATCH_PENDING,
334 __LINK_STATE_DORMANT,
335 __LINK_STATE_TESTING,
339 struct list_head list;
344 * size of gro hash buckets, must less than bit number of
345 * napi_struct::gro_bitmask
347 #define GRO_HASH_BUCKETS 8
350 * Structure for NAPI scheduling similar to tasklet but with weighting
353 /* The poll_list must only be managed by the entity which
354 * changes the state of the NAPI_STATE_SCHED bit. This means
355 * whoever atomically sets that bit can add this napi_struct
356 * to the per-CPU poll_list, and whoever clears that bit
357 * can remove from the list right before clearing the bit.
359 struct list_head poll_list;
363 int defer_hard_irqs_count;
364 unsigned long gro_bitmask;
365 int (*poll)(struct napi_struct *, int);
366 #ifdef CONFIG_NETPOLL
367 /* CPU actively polling if netpoll is configured */
370 /* CPU on which NAPI has been scheduled for processing */
372 struct net_device *dev;
373 struct gro_list gro_hash[GRO_HASH_BUCKETS];
375 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
376 int rx_count; /* length of rx_list */
377 unsigned int napi_id;
378 struct hrtimer timer;
379 struct task_struct *thread;
380 /* control-path-only fields follow */
381 struct list_head dev_list;
382 struct hlist_node napi_hash_node;
387 NAPI_STATE_SCHED, /* Poll is scheduled */
388 NAPI_STATE_MISSED, /* reschedule a napi */
389 NAPI_STATE_DISABLE, /* Disable pending */
390 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
391 NAPI_STATE_LISTED, /* NAPI added to system lists */
392 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
393 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
394 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
395 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
396 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
400 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
401 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
402 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
403 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
404 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
405 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
406 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
407 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
408 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
409 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
419 typedef enum gro_result gro_result_t;
422 * enum rx_handler_result - Possible return values for rx_handlers.
423 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
425 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
426 * case skb->dev was changed by rx_handler.
427 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
428 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
430 * rx_handlers are functions called from inside __netif_receive_skb(), to do
431 * special processing of the skb, prior to delivery to protocol handlers.
433 * Currently, a net_device can only have a single rx_handler registered. Trying
434 * to register a second rx_handler will return -EBUSY.
436 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
437 * To unregister a rx_handler on a net_device, use
438 * netdev_rx_handler_unregister().
440 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
443 * If the rx_handler consumed the skb in some way, it should return
444 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
445 * the skb to be delivered in some other way.
447 * If the rx_handler changed skb->dev, to divert the skb to another
448 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
449 * new device will be called if it exists.
451 * If the rx_handler decides the skb should be ignored, it should return
452 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
453 * are registered on exact device (ptype->dev == skb->dev).
455 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
456 * delivered, it should return RX_HANDLER_PASS.
458 * A device without a registered rx_handler will behave as if rx_handler
459 * returned RX_HANDLER_PASS.
462 enum rx_handler_result {
468 typedef enum rx_handler_result rx_handler_result_t;
469 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
471 void __napi_schedule(struct napi_struct *n);
472 void __napi_schedule_irqoff(struct napi_struct *n);
474 static inline bool napi_disable_pending(struct napi_struct *n)
476 return test_bit(NAPI_STATE_DISABLE, &n->state);
479 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
481 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
485 * napi_is_scheduled - test if NAPI is scheduled
488 * This check is "best-effort". With no locking implemented,
489 * a NAPI can be scheduled or terminate right after this check
490 * and produce not precise results.
492 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
493 * should not be used normally and napi_schedule should be
496 * Use only if the driver really needs to check if a NAPI
497 * is scheduled for example in the context of delayed timer
498 * that can be skipped if a NAPI is already scheduled.
500 * Return True if NAPI is scheduled, False otherwise.
502 static inline bool napi_is_scheduled(struct napi_struct *n)
504 return test_bit(NAPI_STATE_SCHED, &n->state);
507 bool napi_schedule_prep(struct napi_struct *n);
510 * napi_schedule - schedule NAPI poll
513 * Schedule NAPI poll routine to be called if it is not already
515 * Return true if we schedule a NAPI or false if not.
516 * Refer to napi_schedule_prep() for additional reason on why
517 * a NAPI might not be scheduled.
519 static inline bool napi_schedule(struct napi_struct *n)
521 if (napi_schedule_prep(n)) {
530 * napi_schedule_irqoff - schedule NAPI poll
533 * Variant of napi_schedule(), assuming hard irqs are masked.
535 static inline void napi_schedule_irqoff(struct napi_struct *n)
537 if (napi_schedule_prep(n))
538 __napi_schedule_irqoff(n);
542 * napi_complete_done - NAPI processing complete
544 * @work_done: number of packets processed
546 * Mark NAPI processing as complete. Should only be called if poll budget
547 * has not been completely consumed.
548 * Prefer over napi_complete().
549 * Return false if device should avoid rearming interrupts.
551 bool napi_complete_done(struct napi_struct *n, int work_done);
553 static inline bool napi_complete(struct napi_struct *n)
555 return napi_complete_done(n, 0);
558 int dev_set_threaded(struct net_device *dev, bool threaded);
561 * napi_disable - prevent NAPI from scheduling
564 * Stop NAPI from being scheduled on this context.
565 * Waits till any outstanding processing completes.
567 void napi_disable(struct napi_struct *n);
569 void napi_enable(struct napi_struct *n);
572 * napi_synchronize - wait until NAPI is not running
575 * Wait until NAPI is done being scheduled on this context.
576 * Waits till any outstanding processing completes but
577 * does not disable future activations.
579 static inline void napi_synchronize(const struct napi_struct *n)
581 if (IS_ENABLED(CONFIG_SMP))
582 while (test_bit(NAPI_STATE_SCHED, &n->state))
589 * napi_if_scheduled_mark_missed - if napi is running, set the
593 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
596 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
598 unsigned long val, new;
600 val = READ_ONCE(n->state);
602 if (val & NAPIF_STATE_DISABLE)
605 if (!(val & NAPIF_STATE_SCHED))
608 new = val | NAPIF_STATE_MISSED;
609 } while (!try_cmpxchg(&n->state, &val, new));
614 enum netdev_queue_state_t {
615 __QUEUE_STATE_DRV_XOFF,
616 __QUEUE_STATE_STACK_XOFF,
617 __QUEUE_STATE_FROZEN,
620 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
621 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
622 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
624 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
625 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
627 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
631 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
632 * netif_tx_* functions below are used to manipulate this flag. The
633 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
634 * queue independently. The netif_xmit_*stopped functions below are called
635 * to check if the queue has been stopped by the driver or stack (either
636 * of the XOFF bits are set in the state). Drivers should not need to call
637 * netif_xmit*stopped functions, they should only be using netif_tx_*.
640 struct netdev_queue {
644 struct net_device *dev;
645 netdevice_tracker dev_tracker;
647 struct Qdisc __rcu *qdisc;
648 struct Qdisc __rcu *qdisc_sleeping;
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
655 unsigned long tx_maxrate;
657 * Number of TX timeouts for this queue
658 * (/sys/class/net/DEV/Q/trans_timeout)
660 atomic_long_t trans_timeout;
662 /* Subordinate device that the queue has been assigned to */
663 struct net_device *sb_dev;
664 #ifdef CONFIG_XDP_SOCKETS
665 struct xsk_buff_pool *pool;
667 /* NAPI instance for the queue
668 * Readers and writers must hold RTNL
670 struct napi_struct *napi;
674 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
677 * Time (in jiffies) of last Tx
679 unsigned long trans_start;
686 } ____cacheline_aligned_in_smp;
688 extern int sysctl_fb_tunnels_only_for_init_net;
689 extern int sysctl_devconf_inherit_init_net;
692 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
693 * == 1 : For initns only
696 static inline bool net_has_fallback_tunnels(const struct net *net)
698 #if IS_ENABLED(CONFIG_SYSCTL)
699 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
701 return !fb_tunnels_only_for_init_net ||
702 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
708 static inline int net_inherit_devconf(void)
710 #if IS_ENABLED(CONFIG_SYSCTL)
711 return READ_ONCE(sysctl_devconf_inherit_init_net);
717 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
719 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
726 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
728 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
735 * This structure holds an RPS map which can be of variable length. The
736 * map is an array of CPUs.
743 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
746 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
747 * tail pointer for that CPU's input queue at the time of last enqueue, and
748 * a hardware filter index.
750 struct rps_dev_flow {
753 unsigned int last_qtail;
755 #define RPS_NO_FILTER 0xffff
758 * The rps_dev_flow_table structure contains a table of flow mappings.
760 struct rps_dev_flow_table {
763 struct rps_dev_flow flows[];
765 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
766 ((_num) * sizeof(struct rps_dev_flow)))
769 * The rps_sock_flow_table contains mappings of flows to the last CPU
770 * on which they were processed by the application (set in recvmsg).
771 * Each entry is a 32bit value. Upper part is the high-order bits
772 * of flow hash, lower part is CPU number.
773 * rps_cpu_mask is used to partition the space, depending on number of
774 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
775 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
776 * meaning we use 32-6=26 bits for the hash.
778 struct rps_sock_flow_table {
781 u32 ents[] ____cacheline_aligned_in_smp;
783 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
785 #define RPS_NO_CPU 0xffff
787 extern u32 rps_cpu_mask;
788 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
790 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
794 unsigned int index = hash & table->mask;
795 u32 val = hash & ~rps_cpu_mask;
797 /* We only give a hint, preemption can change CPU under us */
798 val |= raw_smp_processor_id();
800 /* The following WRITE_ONCE() is paired with the READ_ONCE()
801 * here, and another one in get_rps_cpu().
803 if (READ_ONCE(table->ents[index]) != val)
804 WRITE_ONCE(table->ents[index], val);
808 #ifdef CONFIG_RFS_ACCEL
809 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
812 #endif /* CONFIG_RPS */
814 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
823 * This structure holds an XPS map which can be of variable length. The
824 * map is an array of queues.
828 unsigned int alloc_len;
832 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
833 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
834 - sizeof(struct xps_map)) / sizeof(u16))
837 * This structure holds all XPS maps for device. Maps are indexed by CPU.
839 * We keep track of the number of cpus/rxqs used when the struct is allocated,
840 * in nr_ids. This will help not accessing out-of-bound memory.
842 * We keep track of the number of traffic classes used when the struct is
843 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
844 * not crossing its upper bound, as the original dev->num_tc can be updated in
847 struct xps_dev_maps {
851 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
854 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
855 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
857 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
858 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
860 #endif /* CONFIG_XPS */
862 #define TC_MAX_QUEUE 16
863 #define TC_BITMASK 15
864 /* HW offloaded queuing disciplines txq count and offset maps */
865 struct netdev_tc_txq {
870 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
872 * This structure is to hold information about the device
873 * configured to run FCoE protocol stack.
875 struct netdev_fcoe_hbainfo {
876 char manufacturer[64];
877 char serial_number[64];
878 char hardware_version[64];
879 char driver_version[64];
880 char optionrom_version[64];
881 char firmware_version[64];
883 char model_description[256];
887 #define MAX_PHYS_ITEM_ID_LEN 32
889 /* This structure holds a unique identifier to identify some
890 * physical item (port for example) used by a netdevice.
892 struct netdev_phys_item_id {
893 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
894 unsigned char id_len;
897 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
898 struct netdev_phys_item_id *b)
900 return a->id_len == b->id_len &&
901 memcmp(a->id, b->id, a->id_len) == 0;
904 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
906 struct net_device *sb_dev);
908 enum net_device_path_type {
909 DEV_PATH_ETHERNET = 0,
917 struct net_device_path {
918 enum net_device_path_type type;
919 const struct net_device *dev;
928 DEV_PATH_BR_VLAN_KEEP,
929 DEV_PATH_BR_VLAN_TAG,
930 DEV_PATH_BR_VLAN_UNTAG,
931 DEV_PATH_BR_VLAN_UNTAG_HW,
950 #define NET_DEVICE_PATH_STACK_MAX 5
951 #define NET_DEVICE_PATH_VLAN_MAX 2
953 struct net_device_path_stack {
955 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
958 struct net_device_path_ctx {
959 const struct net_device *dev;
966 } vlan[NET_DEVICE_PATH_VLAN_MAX];
971 TC_SETUP_QDISC_MQPRIO,
974 TC_SETUP_CLSMATCHALL,
984 TC_SETUP_QDISC_TAPRIO,
993 /* These structures hold the attributes of bpf state that are being passed
994 * to the netdevice through the bpf op.
996 enum bpf_netdev_command {
997 /* Set or clear a bpf program used in the earliest stages of packet
998 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
999 * is responsible for calling bpf_prog_put on any old progs that are
1000 * stored. In case of error, the callee need not release the new prog
1001 * reference, but on success it takes ownership and must bpf_prog_put
1002 * when it is no longer used.
1006 /* BPF program for offload callbacks, invoked at program load time. */
1007 BPF_OFFLOAD_MAP_ALLOC,
1008 BPF_OFFLOAD_MAP_FREE,
1012 struct bpf_prog_offload_ops;
1013 struct netlink_ext_ack;
1015 struct xdp_dev_bulk_queue;
1016 struct bpf_xdp_link;
1025 struct bpf_xdp_entity {
1026 struct bpf_prog *prog;
1027 struct bpf_xdp_link *link;
1031 enum bpf_netdev_command command;
1033 /* XDP_SETUP_PROG */
1036 struct bpf_prog *prog;
1037 struct netlink_ext_ack *extack;
1039 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1041 struct bpf_offloaded_map *offmap;
1043 /* XDP_SETUP_XSK_POOL */
1045 struct xsk_buff_pool *pool;
1051 /* Flags for ndo_xsk_wakeup. */
1052 #define XDP_WAKEUP_RX (1 << 0)
1053 #define XDP_WAKEUP_TX (1 << 1)
1055 #ifdef CONFIG_XFRM_OFFLOAD
1056 struct xfrmdev_ops {
1057 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1058 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1059 void (*xdo_dev_state_free) (struct xfrm_state *x);
1060 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1061 struct xfrm_state *x);
1062 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1063 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1064 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1065 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1066 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1070 struct dev_ifalias {
1071 struct rcu_head rcuhead;
1078 struct netdev_net_notifier {
1079 struct list_head list;
1080 struct notifier_block *nb;
1084 * This structure defines the management hooks for network devices.
1085 * The following hooks can be defined; unless noted otherwise, they are
1086 * optional and can be filled with a null pointer.
1088 * int (*ndo_init)(struct net_device *dev);
1089 * This function is called once when a network device is registered.
1090 * The network device can use this for any late stage initialization
1091 * or semantic validation. It can fail with an error code which will
1092 * be propagated back to register_netdev.
1094 * void (*ndo_uninit)(struct net_device *dev);
1095 * This function is called when device is unregistered or when registration
1096 * fails. It is not called if init fails.
1098 * int (*ndo_open)(struct net_device *dev);
1099 * This function is called when a network device transitions to the up
1102 * int (*ndo_stop)(struct net_device *dev);
1103 * This function is called when a network device transitions to the down
1106 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1107 * struct net_device *dev);
1108 * Called when a packet needs to be transmitted.
1109 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1110 * the queue before that can happen; it's for obsolete devices and weird
1111 * corner cases, but the stack really does a non-trivial amount
1112 * of useless work if you return NETDEV_TX_BUSY.
1113 * Required; cannot be NULL.
1115 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1116 * struct net_device *dev
1117 * netdev_features_t features);
1118 * Called by core transmit path to determine if device is capable of
1119 * performing offload operations on a given packet. This is to give
1120 * the device an opportunity to implement any restrictions that cannot
1121 * be otherwise expressed by feature flags. The check is called with
1122 * the set of features that the stack has calculated and it returns
1123 * those the driver believes to be appropriate.
1125 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1126 * struct net_device *sb_dev);
1127 * Called to decide which queue to use when device supports multiple
1130 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1131 * This function is called to allow device receiver to make
1132 * changes to configuration when multicast or promiscuous is enabled.
1134 * void (*ndo_set_rx_mode)(struct net_device *dev);
1135 * This function is called device changes address list filtering.
1136 * If driver handles unicast address filtering, it should set
1137 * IFF_UNICAST_FLT in its priv_flags.
1139 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1140 * This function is called when the Media Access Control address
1141 * needs to be changed. If this interface is not defined, the
1142 * MAC address can not be changed.
1144 * int (*ndo_validate_addr)(struct net_device *dev);
1145 * Test if Media Access Control address is valid for the device.
1147 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1148 * Old-style ioctl entry point. This is used internally by the
1149 * appletalk and ieee802154 subsystems but is no longer called by
1150 * the device ioctl handler.
1152 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1153 * Used by the bonding driver for its device specific ioctls:
1154 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1155 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1157 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1158 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1159 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1161 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1162 * Used to set network devices bus interface parameters. This interface
1163 * is retained for legacy reasons; new devices should use the bus
1164 * interface (PCI) for low level management.
1166 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1167 * Called when a user wants to change the Maximum Transfer Unit
1170 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1171 * Callback used when the transmitter has not made any progress
1172 * for dev->watchdog ticks.
1174 * void (*ndo_get_stats64)(struct net_device *dev,
1175 * struct rtnl_link_stats64 *storage);
1176 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1177 * Called when a user wants to get the network device usage
1178 * statistics. Drivers must do one of the following:
1179 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1180 * rtnl_link_stats64 structure passed by the caller.
1181 * 2. Define @ndo_get_stats to update a net_device_stats structure
1182 * (which should normally be dev->stats) and return a pointer to
1183 * it. The structure may be changed asynchronously only if each
1184 * field is written atomically.
1185 * 3. Update dev->stats asynchronously and atomically, and define
1186 * neither operation.
1188 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1189 * Return true if this device supports offload stats of this attr_id.
1191 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1193 * Get statistics for offload operations by attr_id. Write it into the
1194 * attr_data pointer.
1196 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1197 * If device supports VLAN filtering this function is called when a
1198 * VLAN id is registered.
1200 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1201 * If device supports VLAN filtering this function is called when a
1202 * VLAN id is unregistered.
1204 * void (*ndo_poll_controller)(struct net_device *dev);
1206 * SR-IOV management functions.
1207 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1208 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1209 * u8 qos, __be16 proto);
1210 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1212 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1213 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1214 * int (*ndo_get_vf_config)(struct net_device *dev,
1215 * int vf, struct ifla_vf_info *ivf);
1216 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1217 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1218 * struct nlattr *port[]);
1220 * Enable or disable the VF ability to query its RSS Redirection Table and
1221 * Hash Key. This is needed since on some devices VF share this information
1222 * with PF and querying it may introduce a theoretical security risk.
1223 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1224 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1225 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1227 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1228 * This is always called from the stack with the rtnl lock held and netif
1229 * tx queues stopped. This allows the netdevice to perform queue
1230 * management safely.
1232 * Fiber Channel over Ethernet (FCoE) offload functions.
1233 * int (*ndo_fcoe_enable)(struct net_device *dev);
1234 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1235 * so the underlying device can perform whatever needed configuration or
1236 * initialization to support acceleration of FCoE traffic.
1238 * int (*ndo_fcoe_disable)(struct net_device *dev);
1239 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1240 * so the underlying device can perform whatever needed clean-ups to
1241 * stop supporting acceleration of FCoE traffic.
1243 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1244 * struct scatterlist *sgl, unsigned int sgc);
1245 * Called when the FCoE Initiator wants to initialize an I/O that
1246 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1247 * perform necessary setup and returns 1 to indicate the device is set up
1248 * successfully to perform DDP on this I/O, otherwise this returns 0.
1250 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1251 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1252 * indicated by the FC exchange id 'xid', so the underlying device can
1253 * clean up and reuse resources for later DDP requests.
1255 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1256 * struct scatterlist *sgl, unsigned int sgc);
1257 * Called when the FCoE Target wants to initialize an I/O that
1258 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1259 * perform necessary setup and returns 1 to indicate the device is set up
1260 * successfully to perform DDP on this I/O, otherwise this returns 0.
1262 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1263 * struct netdev_fcoe_hbainfo *hbainfo);
1264 * Called when the FCoE Protocol stack wants information on the underlying
1265 * device. This information is utilized by the FCoE protocol stack to
1266 * register attributes with Fiber Channel management service as per the
1267 * FC-GS Fabric Device Management Information(FDMI) specification.
1269 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1270 * Called when the underlying device wants to override default World Wide
1271 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1272 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1273 * protocol stack to use.
1276 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1277 * u16 rxq_index, u32 flow_id);
1278 * Set hardware filter for RFS. rxq_index is the target queue index;
1279 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1280 * Return the filter ID on success, or a negative error code.
1282 * Slave management functions (for bridge, bonding, etc).
1283 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1284 * Called to make another netdev an underling.
1286 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1287 * Called to release previously enslaved netdev.
1289 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1290 * struct sk_buff *skb,
1292 * Get the xmit slave of master device. If all_slaves is true, function
1293 * assume all the slaves can transmit.
1295 * Feature/offload setting functions.
1296 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1297 * netdev_features_t features);
1298 * Adjusts the requested feature flags according to device-specific
1299 * constraints, and returns the resulting flags. Must not modify
1302 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1303 * Called to update device configuration to new features. Passed
1304 * feature set might be less than what was returned by ndo_fix_features()).
1305 * Must return >0 or -errno if it changed dev->features itself.
1307 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1308 * struct net_device *dev,
1309 * const unsigned char *addr, u16 vid, u16 flags,
1310 * struct netlink_ext_ack *extack);
1311 * Adds an FDB entry to dev for addr.
1312 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1313 * struct net_device *dev,
1314 * const unsigned char *addr, u16 vid)
1315 * Deletes the FDB entry from dev coresponding to addr.
1316 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1317 * struct netlink_ext_ack *extack);
1318 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1319 * struct net_device *dev, struct net_device *filter_dev,
1321 * Used to add FDB entries to dump requests. Implementers should add
1322 * entries to skb and update idx with the number of entries.
1324 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1325 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1326 * Adds an MDB entry to dev.
1327 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1328 * struct netlink_ext_ack *extack);
1329 * Deletes the MDB entry from dev.
1330 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1331 * struct netlink_ext_ack *extack);
1332 * Bulk deletes MDB entries from dev.
1333 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1334 * struct netlink_callback *cb);
1335 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1336 * callback is used by core rtnetlink code.
1338 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1339 * u16 flags, struct netlink_ext_ack *extack)
1340 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1341 * struct net_device *dev, u32 filter_mask,
1343 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1346 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1347 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1348 * which do not represent real hardware may define this to allow their
1349 * userspace components to manage their virtual carrier state. Devices
1350 * that determine carrier state from physical hardware properties (eg
1351 * network cables) or protocol-dependent mechanisms (eg
1352 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1354 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1355 * struct netdev_phys_item_id *ppid);
1356 * Called to get ID of physical port of this device. If driver does
1357 * not implement this, it is assumed that the hw is not able to have
1358 * multiple net devices on single physical port.
1360 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1361 * struct netdev_phys_item_id *ppid)
1362 * Called to get the parent ID of the physical port of this device.
1364 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1365 * struct net_device *dev)
1366 * Called by upper layer devices to accelerate switching or other
1367 * station functionality into hardware. 'pdev is the lowerdev
1368 * to use for the offload and 'dev' is the net device that will
1369 * back the offload. Returns a pointer to the private structure
1370 * the upper layer will maintain.
1371 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1372 * Called by upper layer device to delete the station created
1373 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1374 * the station and priv is the structure returned by the add
1376 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1377 * int queue_index, u32 maxrate);
1378 * Called when a user wants to set a max-rate limitation of specific
1380 * int (*ndo_get_iflink)(const struct net_device *dev);
1381 * Called to get the iflink value of this device.
1382 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1383 * This function is used to get egress tunnel information for given skb.
1384 * This is useful for retrieving outer tunnel header parameters while
1386 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1387 * This function is used to specify the headroom that the skb must
1388 * consider when allocation skb during packet reception. Setting
1389 * appropriate rx headroom value allows avoiding skb head copy on
1390 * forward. Setting a negative value resets the rx headroom to the
1392 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1393 * This function is used to set or query state related to XDP on the
1394 * netdevice and manage BPF offload. See definition of
1395 * enum bpf_netdev_command for details.
1396 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1398 * This function is used to submit @n XDP packets for transmit on a
1399 * netdevice. Returns number of frames successfully transmitted, frames
1400 * that got dropped are freed/returned via xdp_return_frame().
1401 * Returns negative number, means general error invoking ndo, meaning
1402 * no frames were xmit'ed and core-caller will free all frames.
1403 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1404 * struct xdp_buff *xdp);
1405 * Get the xmit slave of master device based on the xdp_buff.
1406 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1407 * This function is used to wake up the softirq, ksoftirqd or kthread
1408 * responsible for sending and/or receiving packets on a specific
1409 * queue id bound to an AF_XDP socket. The flags field specifies if
1410 * only RX, only Tx, or both should be woken up using the flags
1411 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1412 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1414 * Add, change, delete or get information on an IPv4 tunnel.
1415 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1416 * If a device is paired with a peer device, return the peer instance.
1417 * The caller must be under RCU read context.
1418 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1419 * Get the forwarding path to reach the real device from the HW destination address
1420 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1421 * const struct skb_shared_hwtstamps *hwtstamps,
1423 * Get hardware timestamp based on normal/adjustable time or free running
1424 * cycle counter. This function is required if physical clock supports a
1425 * free running cycle counter.
1427 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1428 * struct kernel_hwtstamp_config *kernel_config);
1429 * Get the currently configured hardware timestamping parameters for the
1432 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1433 * struct kernel_hwtstamp_config *kernel_config,
1434 * struct netlink_ext_ack *extack);
1435 * Change the hardware timestamping parameters for NIC device.
1437 struct net_device_ops {
1438 int (*ndo_init)(struct net_device *dev);
1439 void (*ndo_uninit)(struct net_device *dev);
1440 int (*ndo_open)(struct net_device *dev);
1441 int (*ndo_stop)(struct net_device *dev);
1442 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1443 struct net_device *dev);
1444 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1445 struct net_device *dev,
1446 netdev_features_t features);
1447 u16 (*ndo_select_queue)(struct net_device *dev,
1448 struct sk_buff *skb,
1449 struct net_device *sb_dev);
1450 void (*ndo_change_rx_flags)(struct net_device *dev,
1452 void (*ndo_set_rx_mode)(struct net_device *dev);
1453 int (*ndo_set_mac_address)(struct net_device *dev,
1455 int (*ndo_validate_addr)(struct net_device *dev);
1456 int (*ndo_do_ioctl)(struct net_device *dev,
1457 struct ifreq *ifr, int cmd);
1458 int (*ndo_eth_ioctl)(struct net_device *dev,
1459 struct ifreq *ifr, int cmd);
1460 int (*ndo_siocbond)(struct net_device *dev,
1461 struct ifreq *ifr, int cmd);
1462 int (*ndo_siocwandev)(struct net_device *dev,
1463 struct if_settings *ifs);
1464 int (*ndo_siocdevprivate)(struct net_device *dev,
1466 void __user *data, int cmd);
1467 int (*ndo_set_config)(struct net_device *dev,
1469 int (*ndo_change_mtu)(struct net_device *dev,
1471 int (*ndo_neigh_setup)(struct net_device *dev,
1472 struct neigh_parms *);
1473 void (*ndo_tx_timeout) (struct net_device *dev,
1474 unsigned int txqueue);
1476 void (*ndo_get_stats64)(struct net_device *dev,
1477 struct rtnl_link_stats64 *storage);
1478 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1479 int (*ndo_get_offload_stats)(int attr_id,
1480 const struct net_device *dev,
1482 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1484 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1485 __be16 proto, u16 vid);
1486 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1487 __be16 proto, u16 vid);
1488 #ifdef CONFIG_NET_POLL_CONTROLLER
1489 void (*ndo_poll_controller)(struct net_device *dev);
1490 int (*ndo_netpoll_setup)(struct net_device *dev,
1491 struct netpoll_info *info);
1492 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1494 int (*ndo_set_vf_mac)(struct net_device *dev,
1495 int queue, u8 *mac);
1496 int (*ndo_set_vf_vlan)(struct net_device *dev,
1497 int queue, u16 vlan,
1498 u8 qos, __be16 proto);
1499 int (*ndo_set_vf_rate)(struct net_device *dev,
1500 int vf, int min_tx_rate,
1502 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1503 int vf, bool setting);
1504 int (*ndo_set_vf_trust)(struct net_device *dev,
1505 int vf, bool setting);
1506 int (*ndo_get_vf_config)(struct net_device *dev,
1508 struct ifla_vf_info *ivf);
1509 int (*ndo_set_vf_link_state)(struct net_device *dev,
1510 int vf, int link_state);
1511 int (*ndo_get_vf_stats)(struct net_device *dev,
1513 struct ifla_vf_stats
1515 int (*ndo_set_vf_port)(struct net_device *dev,
1517 struct nlattr *port[]);
1518 int (*ndo_get_vf_port)(struct net_device *dev,
1519 int vf, struct sk_buff *skb);
1520 int (*ndo_get_vf_guid)(struct net_device *dev,
1522 struct ifla_vf_guid *node_guid,
1523 struct ifla_vf_guid *port_guid);
1524 int (*ndo_set_vf_guid)(struct net_device *dev,
1527 int (*ndo_set_vf_rss_query_en)(
1528 struct net_device *dev,
1529 int vf, bool setting);
1530 int (*ndo_setup_tc)(struct net_device *dev,
1531 enum tc_setup_type type,
1533 #if IS_ENABLED(CONFIG_FCOE)
1534 int (*ndo_fcoe_enable)(struct net_device *dev);
1535 int (*ndo_fcoe_disable)(struct net_device *dev);
1536 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1538 struct scatterlist *sgl,
1540 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1542 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1544 struct scatterlist *sgl,
1546 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1547 struct netdev_fcoe_hbainfo *hbainfo);
1550 #if IS_ENABLED(CONFIG_LIBFCOE)
1551 #define NETDEV_FCOE_WWNN 0
1552 #define NETDEV_FCOE_WWPN 1
1553 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1554 u64 *wwn, int type);
1557 #ifdef CONFIG_RFS_ACCEL
1558 int (*ndo_rx_flow_steer)(struct net_device *dev,
1559 const struct sk_buff *skb,
1563 int (*ndo_add_slave)(struct net_device *dev,
1564 struct net_device *slave_dev,
1565 struct netlink_ext_ack *extack);
1566 int (*ndo_del_slave)(struct net_device *dev,
1567 struct net_device *slave_dev);
1568 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1569 struct sk_buff *skb,
1571 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1573 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1574 netdev_features_t features);
1575 int (*ndo_set_features)(struct net_device *dev,
1576 netdev_features_t features);
1577 int (*ndo_neigh_construct)(struct net_device *dev,
1578 struct neighbour *n);
1579 void (*ndo_neigh_destroy)(struct net_device *dev,
1580 struct neighbour *n);
1582 int (*ndo_fdb_add)(struct ndmsg *ndm,
1583 struct nlattr *tb[],
1584 struct net_device *dev,
1585 const unsigned char *addr,
1588 struct netlink_ext_ack *extack);
1589 int (*ndo_fdb_del)(struct ndmsg *ndm,
1590 struct nlattr *tb[],
1591 struct net_device *dev,
1592 const unsigned char *addr,
1593 u16 vid, struct netlink_ext_ack *extack);
1594 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1595 struct net_device *dev,
1596 struct netlink_ext_ack *extack);
1597 int (*ndo_fdb_dump)(struct sk_buff *skb,
1598 struct netlink_callback *cb,
1599 struct net_device *dev,
1600 struct net_device *filter_dev,
1602 int (*ndo_fdb_get)(struct sk_buff *skb,
1603 struct nlattr *tb[],
1604 struct net_device *dev,
1605 const unsigned char *addr,
1606 u16 vid, u32 portid, u32 seq,
1607 struct netlink_ext_ack *extack);
1608 int (*ndo_mdb_add)(struct net_device *dev,
1609 struct nlattr *tb[],
1611 struct netlink_ext_ack *extack);
1612 int (*ndo_mdb_del)(struct net_device *dev,
1613 struct nlattr *tb[],
1614 struct netlink_ext_ack *extack);
1615 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1616 struct nlattr *tb[],
1617 struct netlink_ext_ack *extack);
1618 int (*ndo_mdb_dump)(struct net_device *dev,
1619 struct sk_buff *skb,
1620 struct netlink_callback *cb);
1621 int (*ndo_mdb_get)(struct net_device *dev,
1622 struct nlattr *tb[], u32 portid,
1624 struct netlink_ext_ack *extack);
1625 int (*ndo_bridge_setlink)(struct net_device *dev,
1626 struct nlmsghdr *nlh,
1628 struct netlink_ext_ack *extack);
1629 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1631 struct net_device *dev,
1634 int (*ndo_bridge_dellink)(struct net_device *dev,
1635 struct nlmsghdr *nlh,
1637 int (*ndo_change_carrier)(struct net_device *dev,
1639 int (*ndo_get_phys_port_id)(struct net_device *dev,
1640 struct netdev_phys_item_id *ppid);
1641 int (*ndo_get_port_parent_id)(struct net_device *dev,
1642 struct netdev_phys_item_id *ppid);
1643 int (*ndo_get_phys_port_name)(struct net_device *dev,
1644 char *name, size_t len);
1645 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1646 struct net_device *dev);
1647 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1650 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1653 int (*ndo_get_iflink)(const struct net_device *dev);
1654 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1655 struct sk_buff *skb);
1656 void (*ndo_set_rx_headroom)(struct net_device *dev,
1657 int needed_headroom);
1658 int (*ndo_bpf)(struct net_device *dev,
1659 struct netdev_bpf *bpf);
1660 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1661 struct xdp_frame **xdp,
1663 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1664 struct xdp_buff *xdp);
1665 int (*ndo_xsk_wakeup)(struct net_device *dev,
1666 u32 queue_id, u32 flags);
1667 int (*ndo_tunnel_ctl)(struct net_device *dev,
1668 struct ip_tunnel_parm *p, int cmd);
1669 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1670 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1671 struct net_device_path *path);
1672 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1673 const struct skb_shared_hwtstamps *hwtstamps,
1675 int (*ndo_hwtstamp_get)(struct net_device *dev,
1676 struct kernel_hwtstamp_config *kernel_config);
1677 int (*ndo_hwtstamp_set)(struct net_device *dev,
1678 struct kernel_hwtstamp_config *kernel_config,
1679 struct netlink_ext_ack *extack);
1683 * enum netdev_priv_flags - &struct net_device priv_flags
1685 * These are the &struct net_device, they are only set internally
1686 * by drivers and used in the kernel. These flags are invisible to
1687 * userspace; this means that the order of these flags can change
1688 * during any kernel release.
1690 * You should have a pretty good reason to be extending these flags.
1692 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1693 * @IFF_EBRIDGE: Ethernet bridging device
1694 * @IFF_BONDING: bonding master or slave
1695 * @IFF_ISATAP: ISATAP interface (RFC4214)
1696 * @IFF_WAN_HDLC: WAN HDLC device
1697 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1699 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1700 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1701 * @IFF_MACVLAN_PORT: device used as macvlan port
1702 * @IFF_BRIDGE_PORT: device used as bridge port
1703 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1704 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1705 * @IFF_UNICAST_FLT: Supports unicast filtering
1706 * @IFF_TEAM_PORT: device used as team port
1707 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1708 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1709 * change when it's running
1710 * @IFF_MACVLAN: Macvlan device
1711 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1712 * underlying stacked devices
1713 * @IFF_L3MDEV_MASTER: device is an L3 master device
1714 * @IFF_NO_QUEUE: device can run without qdisc attached
1715 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1716 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1717 * @IFF_TEAM: device is a team device
1718 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1719 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1720 * entity (i.e. the master device for bridged veth)
1721 * @IFF_MACSEC: device is a MACsec device
1722 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1723 * @IFF_FAILOVER: device is a failover master device
1724 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1725 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1726 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1727 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1728 * skb_headlen(skb) == 0 (data starts from frag0)
1729 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1730 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1731 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1732 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1734 enum netdev_priv_flags {
1735 IFF_802_1Q_VLAN = 1<<0,
1739 IFF_WAN_HDLC = 1<<4,
1740 IFF_XMIT_DST_RELEASE = 1<<5,
1741 IFF_DONT_BRIDGE = 1<<6,
1742 IFF_DISABLE_NETPOLL = 1<<7,
1743 IFF_MACVLAN_PORT = 1<<8,
1744 IFF_BRIDGE_PORT = 1<<9,
1745 IFF_OVS_DATAPATH = 1<<10,
1746 IFF_TX_SKB_SHARING = 1<<11,
1747 IFF_UNICAST_FLT = 1<<12,
1748 IFF_TEAM_PORT = 1<<13,
1749 IFF_SUPP_NOFCS = 1<<14,
1750 IFF_LIVE_ADDR_CHANGE = 1<<15,
1751 IFF_MACVLAN = 1<<16,
1752 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1753 IFF_L3MDEV_MASTER = 1<<18,
1754 IFF_NO_QUEUE = 1<<19,
1755 IFF_OPENVSWITCH = 1<<20,
1756 IFF_L3MDEV_SLAVE = 1<<21,
1758 IFF_RXFH_CONFIGURED = 1<<23,
1759 IFF_PHONY_HEADROOM = 1<<24,
1761 IFF_NO_RX_HANDLER = 1<<26,
1762 IFF_FAILOVER = 1<<27,
1763 IFF_FAILOVER_SLAVE = 1<<28,
1764 IFF_L3MDEV_RX_HANDLER = 1<<29,
1765 IFF_NO_ADDRCONF = BIT_ULL(30),
1766 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1767 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1768 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1771 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1772 #define IFF_EBRIDGE IFF_EBRIDGE
1773 #define IFF_BONDING IFF_BONDING
1774 #define IFF_ISATAP IFF_ISATAP
1775 #define IFF_WAN_HDLC IFF_WAN_HDLC
1776 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1777 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1778 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1779 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1780 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1781 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1782 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1783 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1784 #define IFF_TEAM_PORT IFF_TEAM_PORT
1785 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1786 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1787 #define IFF_MACVLAN IFF_MACVLAN
1788 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1789 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1790 #define IFF_NO_QUEUE IFF_NO_QUEUE
1791 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1792 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1793 #define IFF_TEAM IFF_TEAM
1794 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1795 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1796 #define IFF_MACSEC IFF_MACSEC
1797 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1798 #define IFF_FAILOVER IFF_FAILOVER
1799 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1800 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1801 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1803 /* Specifies the type of the struct net_device::ml_priv pointer */
1804 enum netdev_ml_priv_type {
1809 enum netdev_stat_type {
1810 NETDEV_PCPU_STAT_NONE,
1811 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1812 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1813 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1817 * struct net_device - The DEVICE structure.
1819 * Actually, this whole structure is a big mistake. It mixes I/O
1820 * data with strictly "high-level" data, and it has to know about
1821 * almost every data structure used in the INET module.
1823 * @name: This is the first field of the "visible" part of this structure
1824 * (i.e. as seen by users in the "Space.c" file). It is the name
1827 * @name_node: Name hashlist node
1828 * @ifalias: SNMP alias
1829 * @mem_end: Shared memory end
1830 * @mem_start: Shared memory start
1831 * @base_addr: Device I/O address
1832 * @irq: Device IRQ number
1834 * @state: Generic network queuing layer state, see netdev_state_t
1835 * @dev_list: The global list of network devices
1836 * @napi_list: List entry used for polling NAPI devices
1837 * @unreg_list: List entry when we are unregistering the
1838 * device; see the function unregister_netdev
1839 * @close_list: List entry used when we are closing the device
1840 * @ptype_all: Device-specific packet handlers for all protocols
1841 * @ptype_specific: Device-specific, protocol-specific packet handlers
1843 * @adj_list: Directly linked devices, like slaves for bonding
1844 * @features: Currently active device features
1845 * @hw_features: User-changeable features
1847 * @wanted_features: User-requested features
1848 * @vlan_features: Mask of features inheritable by VLAN devices
1850 * @hw_enc_features: Mask of features inherited by encapsulating devices
1851 * This field indicates what encapsulation
1852 * offloads the hardware is capable of doing,
1853 * and drivers will need to set them appropriately.
1855 * @mpls_features: Mask of features inheritable by MPLS
1856 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1858 * @ifindex: interface index
1859 * @group: The group the device belongs to
1861 * @stats: Statistics struct, which was left as a legacy, use
1862 * rtnl_link_stats64 instead
1864 * @core_stats: core networking counters,
1865 * do not use this in drivers
1866 * @carrier_up_count: Number of times the carrier has been up
1867 * @carrier_down_count: Number of times the carrier has been down
1869 * @wireless_handlers: List of functions to handle Wireless Extensions,
1871 * see <net/iw_handler.h> for details.
1872 * @wireless_data: Instance data managed by the core of wireless extensions
1874 * @netdev_ops: Includes several pointers to callbacks,
1875 * if one wants to override the ndo_*() functions
1876 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1877 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1878 * @ethtool_ops: Management operations
1879 * @l3mdev_ops: Layer 3 master device operations
1880 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1881 * discovery handling. Necessary for e.g. 6LoWPAN.
1882 * @xfrmdev_ops: Transformation offload operations
1883 * @tlsdev_ops: Transport Layer Security offload operations
1884 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1885 * of Layer 2 headers.
1887 * @flags: Interface flags (a la BSD)
1888 * @xdp_features: XDP capability supported by the device
1889 * @priv_flags: Like 'flags' but invisible to userspace,
1890 * see if.h for the definitions
1891 * @gflags: Global flags ( kept as legacy )
1892 * @padded: How much padding added by alloc_netdev()
1893 * @operstate: RFC2863 operstate
1894 * @link_mode: Mapping policy to operstate
1895 * @if_port: Selectable AUI, TP, ...
1897 * @mtu: Interface MTU value
1898 * @min_mtu: Interface Minimum MTU value
1899 * @max_mtu: Interface Maximum MTU value
1900 * @type: Interface hardware type
1901 * @hard_header_len: Maximum hardware header length.
1902 * @min_header_len: Minimum hardware header length
1904 * @needed_headroom: Extra headroom the hardware may need, but not in all
1905 * cases can this be guaranteed
1906 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1907 * cases can this be guaranteed. Some cases also use
1908 * LL_MAX_HEADER instead to allocate the skb
1910 * interface address info:
1912 * @perm_addr: Permanent hw address
1913 * @addr_assign_type: Hw address assignment type
1914 * @addr_len: Hardware address length
1915 * @upper_level: Maximum depth level of upper devices.
1916 * @lower_level: Maximum depth level of lower devices.
1917 * @neigh_priv_len: Used in neigh_alloc()
1918 * @dev_id: Used to differentiate devices that share
1919 * the same link layer address
1920 * @dev_port: Used to differentiate devices that share
1922 * @addr_list_lock: XXX: need comments on this one
1923 * @name_assign_type: network interface name assignment type
1924 * @uc_promisc: Counter that indicates promiscuous mode
1925 * has been enabled due to the need to listen to
1926 * additional unicast addresses in a device that
1927 * does not implement ndo_set_rx_mode()
1928 * @uc: unicast mac addresses
1929 * @mc: multicast mac addresses
1930 * @dev_addrs: list of device hw addresses
1931 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1932 * @promiscuity: Number of times the NIC is told to work in
1933 * promiscuous mode; if it becomes 0 the NIC will
1934 * exit promiscuous mode
1935 * @allmulti: Counter, enables or disables allmulticast mode
1937 * @vlan_info: VLAN info
1938 * @dsa_ptr: dsa specific data
1939 * @tipc_ptr: TIPC specific data
1940 * @atalk_ptr: AppleTalk link
1941 * @ip_ptr: IPv4 specific data
1942 * @ip6_ptr: IPv6 specific data
1943 * @ax25_ptr: AX.25 specific data
1944 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1945 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1947 * @mpls_ptr: mpls_dev struct pointer
1948 * @mctp_ptr: MCTP specific data
1950 * @dev_addr: Hw address (before bcast,
1951 * because most packets are unicast)
1953 * @_rx: Array of RX queues
1954 * @num_rx_queues: Number of RX queues
1955 * allocated at register_netdev() time
1956 * @real_num_rx_queues: Number of RX queues currently active in device
1957 * @xdp_prog: XDP sockets filter program pointer
1958 * @gro_flush_timeout: timeout for GRO layer in NAPI
1959 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1960 * allow to avoid NIC hard IRQ, on busy queues.
1962 * @rx_handler: handler for received packets
1963 * @rx_handler_data: XXX: need comments on this one
1964 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1965 * @ingress_queue: XXX: need comments on this one
1966 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1967 * @broadcast: hw bcast address
1969 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1970 * indexed by RX queue number. Assigned by driver.
1971 * This must only be set if the ndo_rx_flow_steer
1972 * operation is defined
1973 * @index_hlist: Device index hash chain
1975 * @_tx: Array of TX queues
1976 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1977 * @real_num_tx_queues: Number of TX queues currently active in device
1978 * @qdisc: Root qdisc from userspace point of view
1979 * @tx_queue_len: Max frames per queue allowed
1980 * @tx_global_lock: XXX: need comments on this one
1981 * @xdp_bulkq: XDP device bulk queue
1982 * @xps_maps: all CPUs/RXQs maps for XPS device
1984 * @xps_maps: XXX: need comments on this one
1985 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1986 * @nf_hooks_egress: netfilter hooks executed for egress packets
1987 * @qdisc_hash: qdisc hash table
1988 * @watchdog_timeo: Represents the timeout that is used by
1989 * the watchdog (see dev_watchdog())
1990 * @watchdog_timer: List of timers
1992 * @proto_down_reason: reason a netdev interface is held down
1993 * @pcpu_refcnt: Number of references to this device
1994 * @dev_refcnt: Number of references to this device
1995 * @refcnt_tracker: Tracker directory for tracked references to this device
1996 * @todo_list: Delayed register/unregister
1997 * @link_watch_list: XXX: need comments on this one
1999 * @reg_state: Register/unregister state machine
2000 * @dismantle: Device is going to be freed
2001 * @rtnl_link_state: This enum represents the phases of creating
2004 * @needs_free_netdev: Should unregister perform free_netdev?
2005 * @priv_destructor: Called from unregister
2006 * @npinfo: XXX: need comments on this one
2007 * @nd_net: Network namespace this network device is inside
2009 * @ml_priv: Mid-layer private
2010 * @ml_priv_type: Mid-layer private type
2012 * @pcpu_stat_type: Type of device statistics which the core should
2013 * allocate/free: none, lstats, tstats, dstats. none
2014 * means the driver is handling statistics allocation/
2015 * freeing internally.
2016 * @lstats: Loopback statistics: packets, bytes
2017 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
2018 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
2023 * @dm_private: Drop monitor private
2025 * @dev: Class/net/name entry
2026 * @sysfs_groups: Space for optional device, statistics and wireless
2029 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2030 * @rtnl_link_ops: Rtnl_link_ops
2032 * @gso_max_size: Maximum size of generic segmentation offload
2033 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2034 * @gso_max_segs: Maximum number of segments that can be passed to the
2036 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2037 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2040 * @dcbnl_ops: Data Center Bridging netlink ops
2041 * @num_tc: Number of traffic classes in the net device
2042 * @tc_to_txq: XXX: need comments on this one
2043 * @prio_tc_map: XXX: need comments on this one
2045 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2047 * @priomap: XXX: need comments on this one
2048 * @phydev: Physical device may attach itself
2049 * for hardware timestamping
2050 * @sfp_bus: attached &struct sfp_bus structure.
2052 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2054 * @proto_down: protocol port state information can be sent to the
2055 * switch driver and used to set the phys state of the
2058 * @wol_enabled: Wake-on-LAN is enabled
2060 * @threaded: napi threaded mode is enabled
2062 * @net_notifier_list: List of per-net netdev notifier block
2063 * that follow this device when it is moved
2064 * to another network namespace.
2066 * @macsec_ops: MACsec offloading ops
2068 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2069 * offload capabilities of the device
2070 * @udp_tunnel_nic: UDP tunnel offload state
2071 * @xdp_state: stores info on attached XDP BPF programs
2073 * @nested_level: Used as a parameter of spin_lock_nested() of
2074 * dev->addr_list_lock.
2075 * @unlink_list: As netif_addr_lock() can be called recursively,
2076 * keep a list of interfaces to be deleted.
2077 * @gro_max_size: Maximum size of aggregated packet in generic
2078 * receive offload (GRO)
2079 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2080 * receive offload (GRO), for IPv4.
2081 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2084 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2085 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2086 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2087 * @dev_registered_tracker: tracker for reference held while
2089 * @offload_xstats_l3: L3 HW stats for this netdevice.
2091 * @devlink_port: Pointer to related devlink port structure.
2092 * Assigned by a driver before netdev registration using
2093 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2094 * during the time netdevice is registered.
2096 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2097 * where the clock is recovered.
2099 * FIXME: cleanup struct net_device such that network protocol info
2104 /* Cacheline organization can be found documented in
2105 * Documentation/networking/net_cachelines/net_device.rst.
2106 * Please update the document when adding new fields.
2109 /* TX read-mostly hotpath */
2110 __cacheline_group_begin(net_device_read_tx);
2111 unsigned long long priv_flags;
2112 const struct net_device_ops *netdev_ops;
2113 const struct header_ops *header_ops;
2114 struct netdev_queue *_tx;
2115 netdev_features_t gso_partial_features;
2116 unsigned int real_num_tx_queues;
2117 unsigned int gso_max_size;
2118 unsigned int gso_ipv4_max_size;
2121 /* Note : dev->mtu is often read without holding a lock.
2122 * Writers usually hold RTNL.
2123 * It is recommended to use READ_ONCE() to annotate the reads,
2124 * and to use WRITE_ONCE() to annotate the writes.
2127 unsigned short needed_headroom;
2128 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2130 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2132 #ifdef CONFIG_NETFILTER_EGRESS
2133 struct nf_hook_entries __rcu *nf_hooks_egress;
2135 #ifdef CONFIG_NET_XGRESS
2136 struct bpf_mprog_entry __rcu *tcx_egress;
2138 __cacheline_group_end(net_device_read_tx);
2140 /* TXRX read-mostly hotpath */
2141 __cacheline_group_begin(net_device_read_txrx);
2143 struct pcpu_lstats __percpu *lstats;
2144 struct pcpu_sw_netstats __percpu *tstats;
2145 struct pcpu_dstats __percpu *dstats;
2148 unsigned short hard_header_len;
2149 netdev_features_t features;
2150 struct inet6_dev __rcu *ip6_ptr;
2151 __cacheline_group_end(net_device_read_txrx);
2153 /* RX read-mostly hotpath */
2154 __cacheline_group_begin(net_device_read_rx);
2155 struct bpf_prog __rcu *xdp_prog;
2156 struct list_head ptype_specific;
2158 unsigned int real_num_rx_queues;
2159 struct netdev_rx_queue *_rx;
2160 unsigned long gro_flush_timeout;
2161 int napi_defer_hard_irqs;
2162 unsigned int gro_max_size;
2163 unsigned int gro_ipv4_max_size;
2164 rx_handler_func_t __rcu *rx_handler;
2165 void __rcu *rx_handler_data;
2166 possible_net_t nd_net;
2167 #ifdef CONFIG_NETPOLL
2168 struct netpoll_info __rcu *npinfo;
2170 #ifdef CONFIG_NET_XGRESS
2171 struct bpf_mprog_entry __rcu *tcx_ingress;
2173 __cacheline_group_end(net_device_read_rx);
2175 char name[IFNAMSIZ];
2176 struct netdev_name_node *name_node;
2177 struct dev_ifalias __rcu *ifalias;
2179 * I/O specific fields
2180 * FIXME: Merge these and struct ifmap into one
2182 unsigned long mem_end;
2183 unsigned long mem_start;
2184 unsigned long base_addr;
2187 * Some hardware also needs these fields (state,dev_list,
2188 * napi_list,unreg_list,close_list) but they are not
2189 * part of the usual set specified in Space.c.
2192 unsigned long state;
2194 struct list_head dev_list;
2195 struct list_head napi_list;
2196 struct list_head unreg_list;
2197 struct list_head close_list;
2198 struct list_head ptype_all;
2201 struct list_head upper;
2202 struct list_head lower;
2205 /* Read-mostly cache-line for fast-path access */
2206 xdp_features_t xdp_features;
2207 const struct xdp_metadata_ops *xdp_metadata_ops;
2208 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2209 unsigned short gflags;
2211 unsigned short needed_tailroom;
2213 netdev_features_t hw_features;
2214 netdev_features_t wanted_features;
2215 netdev_features_t vlan_features;
2216 netdev_features_t hw_enc_features;
2217 netdev_features_t mpls_features;
2219 unsigned int min_mtu;
2220 unsigned int max_mtu;
2221 unsigned short type;
2222 unsigned char min_header_len;
2223 unsigned char name_assign_type;
2227 struct net_device_stats stats; /* not used by modern drivers */
2229 struct net_device_core_stats __percpu *core_stats;
2231 /* Stats to monitor link on/off, flapping */
2232 atomic_t carrier_up_count;
2233 atomic_t carrier_down_count;
2235 #ifdef CONFIG_WIRELESS_EXT
2236 const struct iw_handler_def *wireless_handlers;
2237 struct iw_public_data *wireless_data;
2239 const struct ethtool_ops *ethtool_ops;
2240 #ifdef CONFIG_NET_L3_MASTER_DEV
2241 const struct l3mdev_ops *l3mdev_ops;
2243 #if IS_ENABLED(CONFIG_IPV6)
2244 const struct ndisc_ops *ndisc_ops;
2247 #ifdef CONFIG_XFRM_OFFLOAD
2248 const struct xfrmdev_ops *xfrmdev_ops;
2251 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2252 const struct tlsdev_ops *tlsdev_ops;
2255 unsigned char operstate;
2256 unsigned char link_mode;
2258 unsigned char if_port;
2261 /* Interface address info. */
2262 unsigned char perm_addr[MAX_ADDR_LEN];
2263 unsigned char addr_assign_type;
2264 unsigned char addr_len;
2265 unsigned char upper_level;
2266 unsigned char lower_level;
2268 unsigned short neigh_priv_len;
2269 unsigned short dev_id;
2270 unsigned short dev_port;
2271 unsigned short padded;
2273 spinlock_t addr_list_lock;
2276 struct netdev_hw_addr_list uc;
2277 struct netdev_hw_addr_list mc;
2278 struct netdev_hw_addr_list dev_addrs;
2281 struct kset *queues_kset;
2283 #ifdef CONFIG_LOCKDEP
2284 struct list_head unlink_list;
2286 unsigned int promiscuity;
2287 unsigned int allmulti;
2289 #ifdef CONFIG_LOCKDEP
2290 unsigned char nested_level;
2294 /* Protocol-specific pointers */
2295 struct in_device __rcu *ip_ptr;
2296 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2297 struct vlan_info __rcu *vlan_info;
2299 #if IS_ENABLED(CONFIG_NET_DSA)
2300 struct dsa_port *dsa_ptr;
2302 #if IS_ENABLED(CONFIG_TIPC)
2303 struct tipc_bearer __rcu *tipc_ptr;
2305 #if IS_ENABLED(CONFIG_ATALK)
2308 #if IS_ENABLED(CONFIG_AX25)
2311 #if IS_ENABLED(CONFIG_CFG80211)
2312 struct wireless_dev *ieee80211_ptr;
2314 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2315 struct wpan_dev *ieee802154_ptr;
2317 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2318 struct mpls_dev __rcu *mpls_ptr;
2320 #if IS_ENABLED(CONFIG_MCTP)
2321 struct mctp_dev __rcu *mctp_ptr;
2325 * Cache lines mostly used on receive path (including eth_type_trans())
2327 /* Interface address info used in eth_type_trans() */
2328 const unsigned char *dev_addr;
2330 unsigned int num_rx_queues;
2331 #define GRO_LEGACY_MAX_SIZE 65536u
2332 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2333 * and shinfo->gso_segs is a 16bit field.
2335 #define GRO_MAX_SIZE (8 * 65535u)
2336 unsigned int xdp_zc_max_segs;
2337 struct netdev_queue __rcu *ingress_queue;
2338 #ifdef CONFIG_NETFILTER_INGRESS
2339 struct nf_hook_entries __rcu *nf_hooks_ingress;
2342 unsigned char broadcast[MAX_ADDR_LEN];
2343 #ifdef CONFIG_RFS_ACCEL
2344 struct cpu_rmap *rx_cpu_rmap;
2346 struct hlist_node index_hlist;
2349 * Cache lines mostly used on transmit path
2351 unsigned int num_tx_queues;
2352 struct Qdisc __rcu *qdisc;
2353 unsigned int tx_queue_len;
2354 spinlock_t tx_global_lock;
2356 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2358 #ifdef CONFIG_NET_SCHED
2359 DECLARE_HASHTABLE (qdisc_hash, 4);
2361 /* These may be needed for future network-power-down code. */
2362 struct timer_list watchdog_timer;
2365 u32 proto_down_reason;
2367 struct list_head todo_list;
2369 #ifdef CONFIG_PCPU_DEV_REFCNT
2370 int __percpu *pcpu_refcnt;
2372 refcount_t dev_refcnt;
2374 struct ref_tracker_dir refcnt_tracker;
2376 struct list_head link_watch_list;
2378 enum { NETREG_UNINITIALIZED=0,
2379 NETREG_REGISTERED, /* completed register_netdevice */
2380 NETREG_UNREGISTERING, /* called unregister_netdevice */
2381 NETREG_UNREGISTERED, /* completed unregister todo */
2382 NETREG_RELEASED, /* called free_netdev */
2383 NETREG_DUMMY, /* dummy device for NAPI poll */
2389 RTNL_LINK_INITIALIZED,
2390 RTNL_LINK_INITIALIZING,
2391 } rtnl_link_state:16;
2393 bool needs_free_netdev;
2394 void (*priv_destructor)(struct net_device *dev);
2396 /* mid-layer private */
2398 enum netdev_ml_priv_type ml_priv_type;
2400 enum netdev_stat_type pcpu_stat_type:8;
2402 #if IS_ENABLED(CONFIG_GARP)
2403 struct garp_port __rcu *garp_port;
2405 #if IS_ENABLED(CONFIG_MRP)
2406 struct mrp_port __rcu *mrp_port;
2408 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2409 struct dm_hw_stat_delta __rcu *dm_private;
2412 const struct attribute_group *sysfs_groups[4];
2413 const struct attribute_group *sysfs_rx_queue_group;
2415 const struct rtnl_link_ops *rtnl_link_ops;
2417 /* for setting kernel sock attribute on TCP connection setup */
2418 #define GSO_MAX_SEGS 65535u
2419 #define GSO_LEGACY_MAX_SIZE 65536u
2420 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2421 * and shinfo->gso_segs is a 16bit field.
2423 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2425 #define TSO_LEGACY_MAX_SIZE 65536
2426 #define TSO_MAX_SIZE UINT_MAX
2427 unsigned int tso_max_size;
2428 #define TSO_MAX_SEGS U16_MAX
2432 const struct dcbnl_rtnl_ops *dcbnl_ops;
2434 u8 prio_tc_map[TC_BITMASK + 1];
2436 #if IS_ENABLED(CONFIG_FCOE)
2437 unsigned int fcoe_ddp_xid;
2439 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2440 struct netprio_map __rcu *priomap;
2442 struct phy_device *phydev;
2443 struct sfp_bus *sfp_bus;
2444 struct lock_class_key *qdisc_tx_busylock;
2446 unsigned wol_enabled:1;
2447 unsigned threaded:1;
2449 struct list_head net_notifier_list;
2451 #if IS_ENABLED(CONFIG_MACSEC)
2452 /* MACsec management functions */
2453 const struct macsec_ops *macsec_ops;
2455 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2456 struct udp_tunnel_nic *udp_tunnel_nic;
2458 /* protected by rtnl_lock */
2459 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2461 u8 dev_addr_shadow[MAX_ADDR_LEN];
2462 netdevice_tracker linkwatch_dev_tracker;
2463 netdevice_tracker watchdog_dev_tracker;
2464 netdevice_tracker dev_registered_tracker;
2465 struct rtnl_hw_stats64 *offload_xstats_l3;
2467 struct devlink_port *devlink_port;
2469 #if IS_ENABLED(CONFIG_DPLL)
2470 struct dpll_pin __rcu *dpll_pin;
2472 #if IS_ENABLED(CONFIG_PAGE_POOL)
2473 /** @page_pools: page pools created for this netdevice */
2474 struct hlist_head page_pools;
2477 #define to_net_dev(d) container_of(d, struct net_device, dev)
2480 * Driver should use this to assign devlink port instance to a netdevice
2481 * before it registers the netdevice. Therefore devlink_port is static
2482 * during the netdev lifetime after it is registered.
2484 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2486 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2487 ((dev)->devlink_port = (port)); \
2490 static inline bool netif_elide_gro(const struct net_device *dev)
2492 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2497 #define NETDEV_ALIGN 32
2500 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2502 return dev->prio_tc_map[prio & TC_BITMASK];
2506 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2508 if (tc >= dev->num_tc)
2511 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2515 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2516 void netdev_reset_tc(struct net_device *dev);
2517 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2518 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2521 int netdev_get_num_tc(struct net_device *dev)
2526 static inline void net_prefetch(void *p)
2529 #if L1_CACHE_BYTES < 128
2530 prefetch((u8 *)p + L1_CACHE_BYTES);
2534 static inline void net_prefetchw(void *p)
2537 #if L1_CACHE_BYTES < 128
2538 prefetchw((u8 *)p + L1_CACHE_BYTES);
2542 void netdev_unbind_sb_channel(struct net_device *dev,
2543 struct net_device *sb_dev);
2544 int netdev_bind_sb_channel_queue(struct net_device *dev,
2545 struct net_device *sb_dev,
2546 u8 tc, u16 count, u16 offset);
2547 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2548 static inline int netdev_get_sb_channel(struct net_device *dev)
2550 return max_t(int, -dev->num_tc, 0);
2554 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2557 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2558 return &dev->_tx[index];
2561 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2562 const struct sk_buff *skb)
2564 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2567 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2568 void (*f)(struct net_device *,
2569 struct netdev_queue *,
2575 for (i = 0; i < dev->num_tx_queues; i++)
2576 f(dev, &dev->_tx[i], arg);
2579 #define netdev_lockdep_set_classes(dev) \
2581 static struct lock_class_key qdisc_tx_busylock_key; \
2582 static struct lock_class_key qdisc_xmit_lock_key; \
2583 static struct lock_class_key dev_addr_list_lock_key; \
2586 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2587 lockdep_set_class(&(dev)->addr_list_lock, \
2588 &dev_addr_list_lock_key); \
2589 for (i = 0; i < (dev)->num_tx_queues; i++) \
2590 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2591 &qdisc_xmit_lock_key); \
2594 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2595 struct net_device *sb_dev);
2596 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2597 struct sk_buff *skb,
2598 struct net_device *sb_dev);
2600 /* returns the headroom that the master device needs to take in account
2601 * when forwarding to this dev
2603 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2605 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2608 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2610 if (dev->netdev_ops->ndo_set_rx_headroom)
2611 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2614 /* set the device rx headroom to the dev's default */
2615 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2617 netdev_set_rx_headroom(dev, -1);
2620 static inline void *netdev_get_ml_priv(struct net_device *dev,
2621 enum netdev_ml_priv_type type)
2623 if (dev->ml_priv_type != type)
2626 return dev->ml_priv;
2629 static inline void netdev_set_ml_priv(struct net_device *dev,
2631 enum netdev_ml_priv_type type)
2633 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2634 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2635 dev->ml_priv_type, type);
2636 WARN(!dev->ml_priv_type && dev->ml_priv,
2637 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2639 dev->ml_priv = ml_priv;
2640 dev->ml_priv_type = type;
2644 * Net namespace inlines
2647 struct net *dev_net(const struct net_device *dev)
2649 return read_pnet(&dev->nd_net);
2653 void dev_net_set(struct net_device *dev, struct net *net)
2655 write_pnet(&dev->nd_net, net);
2659 * netdev_priv - access network device private data
2660 * @dev: network device
2662 * Get network device private data
2664 static inline void *netdev_priv(const struct net_device *dev)
2666 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2669 /* Set the sysfs physical device reference for the network logical device
2670 * if set prior to registration will cause a symlink during initialization.
2672 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2674 /* Set the sysfs device type for the network logical device to allow
2675 * fine-grained identification of different network device types. For
2676 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2678 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2680 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2681 enum netdev_queue_type type,
2682 struct napi_struct *napi);
2684 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2689 /* Default NAPI poll() weight
2690 * Device drivers are strongly advised to not use bigger value
2692 #define NAPI_POLL_WEIGHT 64
2694 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2695 int (*poll)(struct napi_struct *, int), int weight);
2698 * netif_napi_add() - initialize a NAPI context
2699 * @dev: network device
2700 * @napi: NAPI context
2701 * @poll: polling function
2703 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2704 * *any* of the other NAPI-related functions.
2707 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2708 int (*poll)(struct napi_struct *, int))
2710 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2714 netif_napi_add_tx_weight(struct net_device *dev,
2715 struct napi_struct *napi,
2716 int (*poll)(struct napi_struct *, int),
2719 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2720 netif_napi_add_weight(dev, napi, poll, weight);
2724 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2725 * @dev: network device
2726 * @napi: NAPI context
2727 * @poll: polling function
2729 * This variant of netif_napi_add() should be used from drivers using NAPI
2730 * to exclusively poll a TX queue.
2731 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2733 static inline void netif_napi_add_tx(struct net_device *dev,
2734 struct napi_struct *napi,
2735 int (*poll)(struct napi_struct *, int))
2737 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2741 * __netif_napi_del - remove a NAPI context
2742 * @napi: NAPI context
2744 * Warning: caller must observe RCU grace period before freeing memory
2745 * containing @napi. Drivers might want to call this helper to combine
2746 * all the needed RCU grace periods into a single one.
2748 void __netif_napi_del(struct napi_struct *napi);
2751 * netif_napi_del - remove a NAPI context
2752 * @napi: NAPI context
2754 * netif_napi_del() removes a NAPI context from the network device NAPI list
2756 static inline void netif_napi_del(struct napi_struct *napi)
2758 __netif_napi_del(napi);
2762 struct packet_type {
2763 __be16 type; /* This is really htons(ether_type). */
2764 bool ignore_outgoing;
2765 struct net_device *dev; /* NULL is wildcarded here */
2766 netdevice_tracker dev_tracker;
2767 int (*func) (struct sk_buff *,
2768 struct net_device *,
2769 struct packet_type *,
2770 struct net_device *);
2771 void (*list_func) (struct list_head *,
2772 struct packet_type *,
2773 struct net_device *);
2774 bool (*id_match)(struct packet_type *ptype,
2776 struct net *af_packet_net;
2777 void *af_packet_priv;
2778 struct list_head list;
2781 struct offload_callbacks {
2782 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2783 netdev_features_t features);
2784 struct sk_buff *(*gro_receive)(struct list_head *head,
2785 struct sk_buff *skb);
2786 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2789 struct packet_offload {
2790 __be16 type; /* This is really htons(ether_type). */
2792 struct offload_callbacks callbacks;
2793 struct list_head list;
2796 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2797 struct pcpu_sw_netstats {
2798 u64_stats_t rx_packets;
2799 u64_stats_t rx_bytes;
2800 u64_stats_t tx_packets;
2801 u64_stats_t tx_bytes;
2802 struct u64_stats_sync syncp;
2803 } __aligned(4 * sizeof(u64));
2805 struct pcpu_dstats {
2812 struct u64_stats_sync syncp;
2813 } __aligned(8 * sizeof(u64));
2815 struct pcpu_lstats {
2816 u64_stats_t packets;
2818 struct u64_stats_sync syncp;
2819 } __aligned(2 * sizeof(u64));
2821 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2823 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2825 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2827 u64_stats_update_begin(&tstats->syncp);
2828 u64_stats_add(&tstats->rx_bytes, len);
2829 u64_stats_inc(&tstats->rx_packets);
2830 u64_stats_update_end(&tstats->syncp);
2833 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2834 unsigned int packets,
2837 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2839 u64_stats_update_begin(&tstats->syncp);
2840 u64_stats_add(&tstats->tx_bytes, len);
2841 u64_stats_add(&tstats->tx_packets, packets);
2842 u64_stats_update_end(&tstats->syncp);
2845 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2847 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2849 u64_stats_update_begin(&lstats->syncp);
2850 u64_stats_add(&lstats->bytes, len);
2851 u64_stats_inc(&lstats->packets);
2852 u64_stats_update_end(&lstats->syncp);
2855 #define __netdev_alloc_pcpu_stats(type, gfp) \
2857 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2860 for_each_possible_cpu(__cpu) { \
2861 typeof(type) *stat; \
2862 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2863 u64_stats_init(&stat->syncp); \
2869 #define netdev_alloc_pcpu_stats(type) \
2870 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2872 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2874 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2877 for_each_possible_cpu(__cpu) { \
2878 typeof(type) *stat; \
2879 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2880 u64_stats_init(&stat->syncp); \
2886 enum netdev_lag_tx_type {
2887 NETDEV_LAG_TX_TYPE_UNKNOWN,
2888 NETDEV_LAG_TX_TYPE_RANDOM,
2889 NETDEV_LAG_TX_TYPE_BROADCAST,
2890 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2891 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2892 NETDEV_LAG_TX_TYPE_HASH,
2895 enum netdev_lag_hash {
2896 NETDEV_LAG_HASH_NONE,
2898 NETDEV_LAG_HASH_L34,
2899 NETDEV_LAG_HASH_L23,
2900 NETDEV_LAG_HASH_E23,
2901 NETDEV_LAG_HASH_E34,
2902 NETDEV_LAG_HASH_VLAN_SRCMAC,
2903 NETDEV_LAG_HASH_UNKNOWN,
2906 struct netdev_lag_upper_info {
2907 enum netdev_lag_tx_type tx_type;
2908 enum netdev_lag_hash hash_type;
2911 struct netdev_lag_lower_state_info {
2916 #include <linux/notifier.h>
2918 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2919 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2923 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2925 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2926 detected a hardware crash and restarted
2927 - we can use this eg to kick tcp sessions
2929 NETDEV_CHANGE, /* Notify device state change */
2932 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2933 NETDEV_CHANGEADDR, /* notify after the address change */
2934 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2938 NETDEV_BONDING_FAILOVER,
2940 NETDEV_PRE_TYPE_CHANGE,
2941 NETDEV_POST_TYPE_CHANGE,
2945 NETDEV_NOTIFY_PEERS,
2949 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2950 NETDEV_CHANGEINFODATA,
2951 NETDEV_BONDING_INFO,
2952 NETDEV_PRECHANGEUPPER,
2953 NETDEV_CHANGELOWERSTATE,
2954 NETDEV_UDP_TUNNEL_PUSH_INFO,
2955 NETDEV_UDP_TUNNEL_DROP_INFO,
2956 NETDEV_CHANGE_TX_QUEUE_LEN,
2957 NETDEV_CVLAN_FILTER_PUSH_INFO,
2958 NETDEV_CVLAN_FILTER_DROP_INFO,
2959 NETDEV_SVLAN_FILTER_PUSH_INFO,
2960 NETDEV_SVLAN_FILTER_DROP_INFO,
2961 NETDEV_OFFLOAD_XSTATS_ENABLE,
2962 NETDEV_OFFLOAD_XSTATS_DISABLE,
2963 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2964 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2965 NETDEV_XDP_FEAT_CHANGE,
2967 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2969 int register_netdevice_notifier(struct notifier_block *nb);
2970 int unregister_netdevice_notifier(struct notifier_block *nb);
2971 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2972 int unregister_netdevice_notifier_net(struct net *net,
2973 struct notifier_block *nb);
2974 int register_netdevice_notifier_dev_net(struct net_device *dev,
2975 struct notifier_block *nb,
2976 struct netdev_net_notifier *nn);
2977 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2978 struct notifier_block *nb,
2979 struct netdev_net_notifier *nn);
2981 struct netdev_notifier_info {
2982 struct net_device *dev;
2983 struct netlink_ext_ack *extack;
2986 struct netdev_notifier_info_ext {
2987 struct netdev_notifier_info info; /* must be first */
2993 struct netdev_notifier_change_info {
2994 struct netdev_notifier_info info; /* must be first */
2995 unsigned int flags_changed;
2998 struct netdev_notifier_changeupper_info {
2999 struct netdev_notifier_info info; /* must be first */
3000 struct net_device *upper_dev; /* new upper dev */
3001 bool master; /* is upper dev master */
3002 bool linking; /* is the notification for link or unlink */
3003 void *upper_info; /* upper dev info */
3006 struct netdev_notifier_changelowerstate_info {
3007 struct netdev_notifier_info info; /* must be first */
3008 void *lower_state_info; /* is lower dev state */
3011 struct netdev_notifier_pre_changeaddr_info {
3012 struct netdev_notifier_info info; /* must be first */
3013 const unsigned char *dev_addr;
3016 enum netdev_offload_xstats_type {
3017 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3020 struct netdev_notifier_offload_xstats_info {
3021 struct netdev_notifier_info info; /* must be first */
3022 enum netdev_offload_xstats_type type;
3025 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3026 struct netdev_notifier_offload_xstats_rd *report_delta;
3027 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3028 struct netdev_notifier_offload_xstats_ru *report_used;
3032 int netdev_offload_xstats_enable(struct net_device *dev,
3033 enum netdev_offload_xstats_type type,
3034 struct netlink_ext_ack *extack);
3035 int netdev_offload_xstats_disable(struct net_device *dev,
3036 enum netdev_offload_xstats_type type);
3037 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3038 enum netdev_offload_xstats_type type);
3039 int netdev_offload_xstats_get(struct net_device *dev,
3040 enum netdev_offload_xstats_type type,
3041 struct rtnl_hw_stats64 *stats, bool *used,
3042 struct netlink_ext_ack *extack);
3044 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3045 const struct rtnl_hw_stats64 *stats);
3047 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3048 void netdev_offload_xstats_push_delta(struct net_device *dev,
3049 enum netdev_offload_xstats_type type,
3050 const struct rtnl_hw_stats64 *stats);
3052 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3053 struct net_device *dev)
3056 info->extack = NULL;
3059 static inline struct net_device *
3060 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3065 static inline struct netlink_ext_ack *
3066 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3068 return info->extack;
3071 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3072 int call_netdevice_notifiers_info(unsigned long val,
3073 struct netdev_notifier_info *info);
3075 extern rwlock_t dev_base_lock; /* Device list lock */
3077 #define for_each_netdev(net, d) \
3078 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3079 #define for_each_netdev_reverse(net, d) \
3080 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3081 #define for_each_netdev_rcu(net, d) \
3082 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3083 #define for_each_netdev_safe(net, d, n) \
3084 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3085 #define for_each_netdev_continue(net, d) \
3086 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3087 #define for_each_netdev_continue_reverse(net, d) \
3088 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3090 #define for_each_netdev_continue_rcu(net, d) \
3091 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3092 #define for_each_netdev_in_bond_rcu(bond, slave) \
3093 for_each_netdev_rcu(&init_net, slave) \
3094 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3095 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3097 #define for_each_netdev_dump(net, d, ifindex) \
3098 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3100 static inline struct net_device *next_net_device(struct net_device *dev)
3102 struct list_head *lh;
3106 lh = dev->dev_list.next;
3107 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3110 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3112 struct list_head *lh;
3116 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3117 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3120 static inline struct net_device *first_net_device(struct net *net)
3122 return list_empty(&net->dev_base_head) ? NULL :
3123 net_device_entry(net->dev_base_head.next);
3126 static inline struct net_device *first_net_device_rcu(struct net *net)
3128 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3130 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3133 int netdev_boot_setup_check(struct net_device *dev);
3134 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3135 const char *hwaddr);
3136 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3137 void dev_add_pack(struct packet_type *pt);
3138 void dev_remove_pack(struct packet_type *pt);
3139 void __dev_remove_pack(struct packet_type *pt);
3140 void dev_add_offload(struct packet_offload *po);
3141 void dev_remove_offload(struct packet_offload *po);
3143 int dev_get_iflink(const struct net_device *dev);
3144 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3145 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3146 struct net_device_path_stack *stack);
3147 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3148 unsigned short mask);
3149 struct net_device *dev_get_by_name(struct net *net, const char *name);
3150 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3151 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3152 bool netdev_name_in_use(struct net *net, const char *name);
3153 int dev_alloc_name(struct net_device *dev, const char *name);
3154 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3155 void dev_close(struct net_device *dev);
3156 void dev_close_many(struct list_head *head, bool unlink);
3157 void dev_disable_lro(struct net_device *dev);
3158 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3159 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3160 struct net_device *sb_dev);
3161 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3162 struct net_device *sb_dev);
3164 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3165 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3167 static inline int dev_queue_xmit(struct sk_buff *skb)
3169 return __dev_queue_xmit(skb, NULL);
3172 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3173 struct net_device *sb_dev)
3175 return __dev_queue_xmit(skb, sb_dev);
3178 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3182 ret = __dev_direct_xmit(skb, queue_id);
3183 if (!dev_xmit_complete(ret))
3188 int register_netdevice(struct net_device *dev);
3189 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3190 void unregister_netdevice_many(struct list_head *head);
3191 static inline void unregister_netdevice(struct net_device *dev)
3193 unregister_netdevice_queue(dev, NULL);
3196 int netdev_refcnt_read(const struct net_device *dev);
3197 void free_netdev(struct net_device *dev);
3198 void netdev_freemem(struct net_device *dev);
3199 int init_dummy_netdev(struct net_device *dev);
3201 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3202 struct sk_buff *skb,
3204 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3206 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3207 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3208 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3209 netdevice_tracker *tracker, gfp_t gfp);
3210 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3211 netdevice_tracker *tracker, gfp_t gfp);
3212 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3213 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3215 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3216 unsigned short type,
3217 const void *daddr, const void *saddr,
3220 if (!dev->header_ops || !dev->header_ops->create)
3223 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3226 static inline int dev_parse_header(const struct sk_buff *skb,
3227 unsigned char *haddr)
3229 const struct net_device *dev = skb->dev;
3231 if (!dev->header_ops || !dev->header_ops->parse)
3233 return dev->header_ops->parse(skb, haddr);
3236 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3238 const struct net_device *dev = skb->dev;
3240 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3242 return dev->header_ops->parse_protocol(skb);
3245 /* ll_header must have at least hard_header_len allocated */
3246 static inline bool dev_validate_header(const struct net_device *dev,
3247 char *ll_header, int len)
3249 if (likely(len >= dev->hard_header_len))
3251 if (len < dev->min_header_len)
3254 if (capable(CAP_SYS_RAWIO)) {
3255 memset(ll_header + len, 0, dev->hard_header_len - len);
3259 if (dev->header_ops && dev->header_ops->validate)
3260 return dev->header_ops->validate(ll_header, len);
3265 static inline bool dev_has_header(const struct net_device *dev)
3267 return dev->header_ops && dev->header_ops->create;
3271 * Incoming packets are placed on per-CPU queues
3273 struct softnet_data {
3274 struct list_head poll_list;
3275 struct sk_buff_head process_queue;
3278 unsigned int processed;
3279 unsigned int time_squeeze;
3281 struct softnet_data *rps_ipi_list;
3284 bool in_net_rx_action;
3285 bool in_napi_threaded_poll;
3287 #ifdef CONFIG_NET_FLOW_LIMIT
3288 struct sd_flow_limit __rcu *flow_limit;
3290 struct Qdisc *output_queue;
3291 struct Qdisc **output_queue_tailp;
3292 struct sk_buff *completion_queue;
3293 #ifdef CONFIG_XFRM_OFFLOAD
3294 struct sk_buff_head xfrm_backlog;
3296 /* written and read only by owning cpu: */
3300 #ifdef CONFIG_NET_EGRESS
3305 /* input_queue_head should be written by cpu owning this struct,
3306 * and only read by other cpus. Worth using a cache line.
3308 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3310 /* Elements below can be accessed between CPUs for RPS/RFS */
3311 call_single_data_t csd ____cacheline_aligned_in_smp;
3312 struct softnet_data *rps_ipi_next;
3314 unsigned int input_queue_tail;
3316 unsigned int received_rps;
3317 unsigned int dropped;
3318 struct sk_buff_head input_pkt_queue;
3319 struct napi_struct backlog;
3321 /* Another possibly contended cache line */
3322 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3324 int defer_ipi_scheduled;
3325 struct sk_buff *defer_list;
3326 call_single_data_t defer_csd;
3329 static inline void input_queue_head_incr(struct softnet_data *sd)
3332 sd->input_queue_head++;
3336 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3337 unsigned int *qtail)
3340 *qtail = ++sd->input_queue_tail;
3344 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3346 static inline int dev_recursion_level(void)
3348 return this_cpu_read(softnet_data.xmit.recursion);
3351 #define XMIT_RECURSION_LIMIT 8
3352 static inline bool dev_xmit_recursion(void)
3354 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3355 XMIT_RECURSION_LIMIT);
3358 static inline void dev_xmit_recursion_inc(void)
3360 __this_cpu_inc(softnet_data.xmit.recursion);
3363 static inline void dev_xmit_recursion_dec(void)
3365 __this_cpu_dec(softnet_data.xmit.recursion);
3368 void __netif_schedule(struct Qdisc *q);
3369 void netif_schedule_queue(struct netdev_queue *txq);
3371 static inline void netif_tx_schedule_all(struct net_device *dev)
3375 for (i = 0; i < dev->num_tx_queues; i++)
3376 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3379 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3381 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3385 * netif_start_queue - allow transmit
3386 * @dev: network device
3388 * Allow upper layers to call the device hard_start_xmit routine.
3390 static inline void netif_start_queue(struct net_device *dev)
3392 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3395 static inline void netif_tx_start_all_queues(struct net_device *dev)
3399 for (i = 0; i < dev->num_tx_queues; i++) {
3400 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3401 netif_tx_start_queue(txq);
3405 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3408 * netif_wake_queue - restart transmit
3409 * @dev: network device
3411 * Allow upper layers to call the device hard_start_xmit routine.
3412 * Used for flow control when transmit resources are available.
3414 static inline void netif_wake_queue(struct net_device *dev)
3416 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3419 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3423 for (i = 0; i < dev->num_tx_queues; i++) {
3424 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3425 netif_tx_wake_queue(txq);
3429 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3431 /* Must be an atomic op see netif_txq_try_stop() */
3432 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3436 * netif_stop_queue - stop transmitted packets
3437 * @dev: network device
3439 * Stop upper layers calling the device hard_start_xmit routine.
3440 * Used for flow control when transmit resources are unavailable.
3442 static inline void netif_stop_queue(struct net_device *dev)
3444 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3447 void netif_tx_stop_all_queues(struct net_device *dev);
3449 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3451 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3455 * netif_queue_stopped - test if transmit queue is flowblocked
3456 * @dev: network device
3458 * Test if transmit queue on device is currently unable to send.
3460 static inline bool netif_queue_stopped(const struct net_device *dev)
3462 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3465 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3467 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3471 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3473 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3477 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3479 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3483 * netdev_queue_set_dql_min_limit - set dql minimum limit
3484 * @dev_queue: pointer to transmit queue
3485 * @min_limit: dql minimum limit
3487 * Forces xmit_more() to return true until the minimum threshold
3488 * defined by @min_limit is reached (or until the tx queue is
3489 * empty). Warning: to be use with care, misuse will impact the
3492 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3493 unsigned int min_limit)
3496 dev_queue->dql.min_limit = min_limit;
3500 static inline int netdev_queue_dql_avail(const struct netdev_queue *txq)
3503 /* Non-BQL migrated drivers will return 0, too. */
3504 return dql_avail(&txq->dql);
3511 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3512 * @dev_queue: pointer to transmit queue
3514 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3515 * to give appropriate hint to the CPU.
3517 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3520 prefetchw(&dev_queue->dql.num_queued);
3525 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3526 * @dev_queue: pointer to transmit queue
3528 * BQL enabled drivers might use this helper in their TX completion path,
3529 * to give appropriate hint to the CPU.
3531 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3534 prefetchw(&dev_queue->dql.limit);
3539 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3540 * @dev_queue: network device queue
3541 * @bytes: number of bytes queued to the device queue
3543 * Report the number of bytes queued for sending/completion to the network
3544 * device hardware queue. @bytes should be a good approximation and should
3545 * exactly match netdev_completed_queue() @bytes.
3546 * This is typically called once per packet, from ndo_start_xmit().
3548 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3552 dql_queued(&dev_queue->dql, bytes);
3554 if (likely(dql_avail(&dev_queue->dql) >= 0))
3557 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3560 * The XOFF flag must be set before checking the dql_avail below,
3561 * because in netdev_tx_completed_queue we update the dql_completed
3562 * before checking the XOFF flag.
3566 /* check again in case another CPU has just made room avail */
3567 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3568 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3572 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3573 * that they should not test BQL status themselves.
3574 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3576 * Returns true if the doorbell must be used to kick the NIC.
3578 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3584 dql_queued(&dev_queue->dql, bytes);
3586 return netif_tx_queue_stopped(dev_queue);
3588 netdev_tx_sent_queue(dev_queue, bytes);
3593 * netdev_sent_queue - report the number of bytes queued to hardware
3594 * @dev: network device
3595 * @bytes: number of bytes queued to the hardware device queue
3597 * Report the number of bytes queued for sending/completion to the network
3598 * device hardware queue#0. @bytes should be a good approximation and should
3599 * exactly match netdev_completed_queue() @bytes.
3600 * This is typically called once per packet, from ndo_start_xmit().
3602 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3604 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3607 static inline bool __netdev_sent_queue(struct net_device *dev,
3611 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3616 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3617 * @dev_queue: network device queue
3618 * @pkts: number of packets (currently ignored)
3619 * @bytes: number of bytes dequeued from the device queue
3621 * Must be called at most once per TX completion round (and not per
3622 * individual packet), so that BQL can adjust its limits appropriately.
3624 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3625 unsigned int pkts, unsigned int bytes)
3628 if (unlikely(!bytes))
3631 dql_completed(&dev_queue->dql, bytes);
3634 * Without the memory barrier there is a small possiblity that
3635 * netdev_tx_sent_queue will miss the update and cause the queue to
3636 * be stopped forever
3638 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3640 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3643 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3644 netif_schedule_queue(dev_queue);
3649 * netdev_completed_queue - report bytes and packets completed by device
3650 * @dev: network device
3651 * @pkts: actual number of packets sent over the medium
3652 * @bytes: actual number of bytes sent over the medium
3654 * Report the number of bytes and packets transmitted by the network device
3655 * hardware queue over the physical medium, @bytes must exactly match the
3656 * @bytes amount passed to netdev_sent_queue()
3658 static inline void netdev_completed_queue(struct net_device *dev,
3659 unsigned int pkts, unsigned int bytes)
3661 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3664 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3667 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3673 * netdev_reset_queue - reset the packets and bytes count of a network device
3674 * @dev_queue: network device
3676 * Reset the bytes and packet count of a network device and clear the
3677 * software flow control OFF bit for this network device
3679 static inline void netdev_reset_queue(struct net_device *dev_queue)
3681 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3685 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3686 * @dev: network device
3687 * @queue_index: given tx queue index
3689 * Returns 0 if given tx queue index >= number of device tx queues,
3690 * otherwise returns the originally passed tx queue index.
3692 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3694 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3695 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3696 dev->name, queue_index,
3697 dev->real_num_tx_queues);
3705 * netif_running - test if up
3706 * @dev: network device
3708 * Test if the device has been brought up.
3710 static inline bool netif_running(const struct net_device *dev)
3712 return test_bit(__LINK_STATE_START, &dev->state);
3716 * Routines to manage the subqueues on a device. We only need start,
3717 * stop, and a check if it's stopped. All other device management is
3718 * done at the overall netdevice level.
3719 * Also test the device if we're multiqueue.
3723 * netif_start_subqueue - allow sending packets on subqueue
3724 * @dev: network device
3725 * @queue_index: sub queue index
3727 * Start individual transmit queue of a device with multiple transmit queues.
3729 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3731 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3733 netif_tx_start_queue(txq);
3737 * netif_stop_subqueue - stop sending packets on subqueue
3738 * @dev: network device
3739 * @queue_index: sub queue index
3741 * Stop individual transmit queue of a device with multiple transmit queues.
3743 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3745 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3746 netif_tx_stop_queue(txq);
3750 * __netif_subqueue_stopped - test status of subqueue
3751 * @dev: network device
3752 * @queue_index: sub queue index
3754 * Check individual transmit queue of a device with multiple transmit queues.
3756 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3759 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3761 return netif_tx_queue_stopped(txq);
3765 * netif_subqueue_stopped - test status of subqueue
3766 * @dev: network device
3767 * @skb: sub queue buffer pointer
3769 * Check individual transmit queue of a device with multiple transmit queues.
3771 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3772 struct sk_buff *skb)
3774 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3778 * netif_wake_subqueue - allow sending packets on subqueue
3779 * @dev: network device
3780 * @queue_index: sub queue index
3782 * Resume individual transmit queue of a device with multiple transmit queues.
3784 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3786 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3788 netif_tx_wake_queue(txq);
3792 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3794 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3795 u16 index, enum xps_map_type type);
3798 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3799 * @j: CPU/Rx queue index
3800 * @mask: bitmask of all cpus/rx queues
3801 * @nr_bits: number of bits in the bitmask
3803 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3805 static inline bool netif_attr_test_mask(unsigned long j,
3806 const unsigned long *mask,
3807 unsigned int nr_bits)
3809 cpu_max_bits_warn(j, nr_bits);
3810 return test_bit(j, mask);
3814 * netif_attr_test_online - Test for online CPU/Rx queue
3815 * @j: CPU/Rx queue index
3816 * @online_mask: bitmask for CPUs/Rx queues that are online
3817 * @nr_bits: number of bits in the bitmask
3819 * Returns true if a CPU/Rx queue is online.
3821 static inline bool netif_attr_test_online(unsigned long j,
3822 const unsigned long *online_mask,
3823 unsigned int nr_bits)
3825 cpu_max_bits_warn(j, nr_bits);
3828 return test_bit(j, online_mask);
3830 return (j < nr_bits);
3834 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3835 * @n: CPU/Rx queue index
3836 * @srcp: the cpumask/Rx queue mask pointer
3837 * @nr_bits: number of bits in the bitmask
3839 * Returns >= nr_bits if no further CPUs/Rx queues set.
3841 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3842 unsigned int nr_bits)
3844 /* -1 is a legal arg here. */
3846 cpu_max_bits_warn(n, nr_bits);
3849 return find_next_bit(srcp, nr_bits, n + 1);
3855 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3856 * @n: CPU/Rx queue index
3857 * @src1p: the first CPUs/Rx queues mask pointer
3858 * @src2p: the second CPUs/Rx queues mask pointer
3859 * @nr_bits: number of bits in the bitmask
3861 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3863 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3864 const unsigned long *src2p,
3865 unsigned int nr_bits)
3867 /* -1 is a legal arg here. */
3869 cpu_max_bits_warn(n, nr_bits);
3872 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3874 return find_next_bit(src1p, nr_bits, n + 1);
3876 return find_next_bit(src2p, nr_bits, n + 1);
3881 static inline int netif_set_xps_queue(struct net_device *dev,
3882 const struct cpumask *mask,
3888 static inline int __netif_set_xps_queue(struct net_device *dev,
3889 const unsigned long *mask,
3890 u16 index, enum xps_map_type type)
3897 * netif_is_multiqueue - test if device has multiple transmit queues
3898 * @dev: network device
3900 * Check if device has multiple transmit queues
3902 static inline bool netif_is_multiqueue(const struct net_device *dev)
3904 return dev->num_tx_queues > 1;
3907 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3910 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3912 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3915 dev->real_num_rx_queues = rxqs;
3919 int netif_set_real_num_queues(struct net_device *dev,
3920 unsigned int txq, unsigned int rxq);
3922 int netif_get_num_default_rss_queues(void);
3924 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3925 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3928 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3929 * interrupt context or with hardware interrupts being disabled.
3930 * (in_hardirq() || irqs_disabled())
3932 * We provide four helpers that can be used in following contexts :
3934 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3935 * replacing kfree_skb(skb)
3937 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3938 * Typically used in place of consume_skb(skb) in TX completion path
3940 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3941 * replacing kfree_skb(skb)
3943 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3944 * and consumed a packet. Used in place of consume_skb(skb)
3946 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3948 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3951 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3953 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3956 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3958 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3961 static inline void dev_consume_skb_any(struct sk_buff *skb)
3963 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3966 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3967 struct bpf_prog *xdp_prog);
3968 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3969 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3970 int netif_rx(struct sk_buff *skb);
3971 int __netif_rx(struct sk_buff *skb);
3973 int netif_receive_skb(struct sk_buff *skb);
3974 int netif_receive_skb_core(struct sk_buff *skb);
3975 void netif_receive_skb_list_internal(struct list_head *head);
3976 void netif_receive_skb_list(struct list_head *head);
3977 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3978 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3979 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3980 void napi_get_frags_check(struct napi_struct *napi);
3981 gro_result_t napi_gro_frags(struct napi_struct *napi);
3982 struct packet_offload *gro_find_receive_by_type(__be16 type);
3983 struct packet_offload *gro_find_complete_by_type(__be16 type);
3985 static inline void napi_free_frags(struct napi_struct *napi)
3987 kfree_skb(napi->skb);
3991 bool netdev_is_rx_handler_busy(struct net_device *dev);
3992 int netdev_rx_handler_register(struct net_device *dev,
3993 rx_handler_func_t *rx_handler,
3994 void *rx_handler_data);
3995 void netdev_rx_handler_unregister(struct net_device *dev);
3997 bool dev_valid_name(const char *name);
3998 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
4000 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
4002 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4003 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4004 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4005 void __user *data, bool *need_copyout);
4006 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4007 int generic_hwtstamp_get_lower(struct net_device *dev,
4008 struct kernel_hwtstamp_config *kernel_cfg);
4009 int generic_hwtstamp_set_lower(struct net_device *dev,
4010 struct kernel_hwtstamp_config *kernel_cfg,
4011 struct netlink_ext_ack *extack);
4012 int dev_set_hwtstamp_phylib(struct net_device *dev,
4013 struct kernel_hwtstamp_config *cfg,
4014 struct netlink_ext_ack *extack);
4015 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4016 unsigned int dev_get_flags(const struct net_device *);
4017 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4018 struct netlink_ext_ack *extack);
4019 int dev_change_flags(struct net_device *dev, unsigned int flags,
4020 struct netlink_ext_ack *extack);
4021 int dev_set_alias(struct net_device *, const char *, size_t);
4022 int dev_get_alias(const struct net_device *, char *, size_t);
4023 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4024 const char *pat, int new_ifindex);
4026 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4029 return __dev_change_net_namespace(dev, net, pat, 0);
4031 int __dev_set_mtu(struct net_device *, int);
4032 int dev_set_mtu(struct net_device *, int);
4033 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4034 struct netlink_ext_ack *extack);
4035 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4036 struct netlink_ext_ack *extack);
4037 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4038 struct netlink_ext_ack *extack);
4039 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4040 int dev_get_port_parent_id(struct net_device *dev,
4041 struct netdev_phys_item_id *ppid, bool recurse);
4042 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4044 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4045 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4046 struct netdev_queue *txq, int *ret);
4048 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4049 u8 dev_xdp_prog_count(struct net_device *dev);
4050 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4052 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4053 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4054 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4055 bool is_skb_forwardable(const struct net_device *dev,
4056 const struct sk_buff *skb);
4058 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4059 const struct sk_buff *skb,
4060 const bool check_mtu)
4062 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4065 if (!(dev->flags & IFF_UP))
4071 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4072 if (skb->len <= len)
4075 /* if TSO is enabled, we don't care about the length as the packet
4076 * could be forwarded without being segmented before
4078 if (skb_is_gso(skb))
4084 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4086 #define DEV_CORE_STATS_INC(FIELD) \
4087 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4089 netdev_core_stats_inc(dev, \
4090 offsetof(struct net_device_core_stats, FIELD)); \
4092 DEV_CORE_STATS_INC(rx_dropped)
4093 DEV_CORE_STATS_INC(tx_dropped)
4094 DEV_CORE_STATS_INC(rx_nohandler)
4095 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4096 #undef DEV_CORE_STATS_INC
4098 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4099 struct sk_buff *skb,
4100 const bool check_mtu)
4102 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4103 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4104 dev_core_stats_rx_dropped_inc(dev);
4109 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4114 bool dev_nit_active(struct net_device *dev);
4115 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4117 static inline void __dev_put(struct net_device *dev)
4120 #ifdef CONFIG_PCPU_DEV_REFCNT
4121 this_cpu_dec(*dev->pcpu_refcnt);
4123 refcount_dec(&dev->dev_refcnt);
4128 static inline void __dev_hold(struct net_device *dev)
4131 #ifdef CONFIG_PCPU_DEV_REFCNT
4132 this_cpu_inc(*dev->pcpu_refcnt);
4134 refcount_inc(&dev->dev_refcnt);
4139 static inline void __netdev_tracker_alloc(struct net_device *dev,
4140 netdevice_tracker *tracker,
4143 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4144 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4148 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4149 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4151 static inline void netdev_tracker_alloc(struct net_device *dev,
4152 netdevice_tracker *tracker, gfp_t gfp)
4154 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4155 refcount_dec(&dev->refcnt_tracker.no_tracker);
4156 __netdev_tracker_alloc(dev, tracker, gfp);
4160 static inline void netdev_tracker_free(struct net_device *dev,
4161 netdevice_tracker *tracker)
4163 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4164 ref_tracker_free(&dev->refcnt_tracker, tracker);
4168 static inline void netdev_hold(struct net_device *dev,
4169 netdevice_tracker *tracker, gfp_t gfp)
4173 __netdev_tracker_alloc(dev, tracker, gfp);
4177 static inline void netdev_put(struct net_device *dev,
4178 netdevice_tracker *tracker)
4181 netdev_tracker_free(dev, tracker);
4187 * dev_hold - get reference to device
4188 * @dev: network device
4190 * Hold reference to device to keep it from being freed.
4191 * Try using netdev_hold() instead.
4193 static inline void dev_hold(struct net_device *dev)
4195 netdev_hold(dev, NULL, GFP_ATOMIC);
4199 * dev_put - release reference to device
4200 * @dev: network device
4202 * Release reference to device to allow it to be freed.
4203 * Try using netdev_put() instead.
4205 static inline void dev_put(struct net_device *dev)
4207 netdev_put(dev, NULL);
4210 static inline void netdev_ref_replace(struct net_device *odev,
4211 struct net_device *ndev,
4212 netdevice_tracker *tracker,
4216 netdev_tracker_free(odev, tracker);
4222 __netdev_tracker_alloc(ndev, tracker, gfp);
4225 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4226 * and _off may be called from IRQ context, but it is caller
4227 * who is responsible for serialization of these calls.
4229 * The name carrier is inappropriate, these functions should really be
4230 * called netif_lowerlayer_*() because they represent the state of any
4231 * kind of lower layer not just hardware media.
4233 void linkwatch_fire_event(struct net_device *dev);
4236 * linkwatch_sync_dev - sync linkwatch for the given device
4237 * @dev: network device to sync linkwatch for
4239 * Sync linkwatch for the given device, removing it from the
4240 * pending work list (if queued).
4242 void linkwatch_sync_dev(struct net_device *dev);
4245 * netif_carrier_ok - test if carrier present
4246 * @dev: network device
4248 * Check if carrier is present on device
4250 static inline bool netif_carrier_ok(const struct net_device *dev)
4252 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4255 unsigned long dev_trans_start(struct net_device *dev);
4257 void __netdev_watchdog_up(struct net_device *dev);
4259 void netif_carrier_on(struct net_device *dev);
4260 void netif_carrier_off(struct net_device *dev);
4261 void netif_carrier_event(struct net_device *dev);
4264 * netif_dormant_on - mark device as dormant.
4265 * @dev: network device
4267 * Mark device as dormant (as per RFC2863).
4269 * The dormant state indicates that the relevant interface is not
4270 * actually in a condition to pass packets (i.e., it is not 'up') but is
4271 * in a "pending" state, waiting for some external event. For "on-
4272 * demand" interfaces, this new state identifies the situation where the
4273 * interface is waiting for events to place it in the up state.
4275 static inline void netif_dormant_on(struct net_device *dev)
4277 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4278 linkwatch_fire_event(dev);
4282 * netif_dormant_off - set device as not dormant.
4283 * @dev: network device
4285 * Device is not in dormant state.
4287 static inline void netif_dormant_off(struct net_device *dev)
4289 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4290 linkwatch_fire_event(dev);
4294 * netif_dormant - test if device is dormant
4295 * @dev: network device
4297 * Check if device is dormant.
4299 static inline bool netif_dormant(const struct net_device *dev)
4301 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4306 * netif_testing_on - mark device as under test.
4307 * @dev: network device
4309 * Mark device as under test (as per RFC2863).
4311 * The testing state indicates that some test(s) must be performed on
4312 * the interface. After completion, of the test, the interface state
4313 * will change to up, dormant, or down, as appropriate.
4315 static inline void netif_testing_on(struct net_device *dev)
4317 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4318 linkwatch_fire_event(dev);
4322 * netif_testing_off - set device as not under test.
4323 * @dev: network device
4325 * Device is not in testing state.
4327 static inline void netif_testing_off(struct net_device *dev)
4329 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4330 linkwatch_fire_event(dev);
4334 * netif_testing - test if device is under test
4335 * @dev: network device
4337 * Check if device is under test
4339 static inline bool netif_testing(const struct net_device *dev)
4341 return test_bit(__LINK_STATE_TESTING, &dev->state);
4346 * netif_oper_up - test if device is operational
4347 * @dev: network device
4349 * Check if carrier is operational
4351 static inline bool netif_oper_up(const struct net_device *dev)
4353 return (dev->operstate == IF_OPER_UP ||
4354 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4358 * netif_device_present - is device available or removed
4359 * @dev: network device
4361 * Check if device has not been removed from system.
4363 static inline bool netif_device_present(const struct net_device *dev)
4365 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4368 void netif_device_detach(struct net_device *dev);
4370 void netif_device_attach(struct net_device *dev);
4373 * Network interface message level settings
4378 NETIF_MSG_PROBE_BIT,
4380 NETIF_MSG_TIMER_BIT,
4381 NETIF_MSG_IFDOWN_BIT,
4383 NETIF_MSG_RX_ERR_BIT,
4384 NETIF_MSG_TX_ERR_BIT,
4385 NETIF_MSG_TX_QUEUED_BIT,
4387 NETIF_MSG_TX_DONE_BIT,
4388 NETIF_MSG_RX_STATUS_BIT,
4389 NETIF_MSG_PKTDATA_BIT,
4393 /* When you add a new bit above, update netif_msg_class_names array
4394 * in net/ethtool/common.c
4396 NETIF_MSG_CLASS_COUNT,
4398 /* Both ethtool_ops interface and internal driver implementation use u32 */
4399 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4401 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4402 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4404 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4405 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4406 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4407 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4408 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4409 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4410 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4411 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4412 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4413 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4414 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4415 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4416 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4417 #define NETIF_MSG_HW __NETIF_MSG(HW)
4418 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4420 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4421 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4422 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4423 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4424 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4425 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4426 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4427 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4428 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4429 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4430 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4431 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4432 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4433 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4434 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4436 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4439 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4440 return default_msg_enable_bits;
4441 if (debug_value == 0) /* no output */
4443 /* set low N bits */
4444 return (1U << debug_value) - 1;
4447 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4449 spin_lock(&txq->_xmit_lock);
4450 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4451 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4454 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4456 __acquire(&txq->_xmit_lock);
4460 static inline void __netif_tx_release(struct netdev_queue *txq)
4462 __release(&txq->_xmit_lock);
4465 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4467 spin_lock_bh(&txq->_xmit_lock);
4468 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4469 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4472 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4474 bool ok = spin_trylock(&txq->_xmit_lock);
4477 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4478 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4483 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4485 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4486 WRITE_ONCE(txq->xmit_lock_owner, -1);
4487 spin_unlock(&txq->_xmit_lock);
4490 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4492 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4493 WRITE_ONCE(txq->xmit_lock_owner, -1);
4494 spin_unlock_bh(&txq->_xmit_lock);
4498 * txq->trans_start can be read locklessly from dev_watchdog()
4500 static inline void txq_trans_update(struct netdev_queue *txq)
4502 if (txq->xmit_lock_owner != -1)
4503 WRITE_ONCE(txq->trans_start, jiffies);
4506 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4508 unsigned long now = jiffies;
4510 if (READ_ONCE(txq->trans_start) != now)
4511 WRITE_ONCE(txq->trans_start, now);
4514 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4515 static inline void netif_trans_update(struct net_device *dev)
4517 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4519 txq_trans_cond_update(txq);
4523 * netif_tx_lock - grab network device transmit lock
4524 * @dev: network device
4526 * Get network device transmit lock
4528 void netif_tx_lock(struct net_device *dev);
4530 static inline void netif_tx_lock_bh(struct net_device *dev)
4536 void netif_tx_unlock(struct net_device *dev);
4538 static inline void netif_tx_unlock_bh(struct net_device *dev)
4540 netif_tx_unlock(dev);
4544 #define HARD_TX_LOCK(dev, txq, cpu) { \
4545 if ((dev->features & NETIF_F_LLTX) == 0) { \
4546 __netif_tx_lock(txq, cpu); \
4548 __netif_tx_acquire(txq); \
4552 #define HARD_TX_TRYLOCK(dev, txq) \
4553 (((dev->features & NETIF_F_LLTX) == 0) ? \
4554 __netif_tx_trylock(txq) : \
4555 __netif_tx_acquire(txq))
4557 #define HARD_TX_UNLOCK(dev, txq) { \
4558 if ((dev->features & NETIF_F_LLTX) == 0) { \
4559 __netif_tx_unlock(txq); \
4561 __netif_tx_release(txq); \
4565 static inline void netif_tx_disable(struct net_device *dev)
4571 cpu = smp_processor_id();
4572 spin_lock(&dev->tx_global_lock);
4573 for (i = 0; i < dev->num_tx_queues; i++) {
4574 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4576 __netif_tx_lock(txq, cpu);
4577 netif_tx_stop_queue(txq);
4578 __netif_tx_unlock(txq);
4580 spin_unlock(&dev->tx_global_lock);
4584 static inline void netif_addr_lock(struct net_device *dev)
4586 unsigned char nest_level = 0;
4588 #ifdef CONFIG_LOCKDEP
4589 nest_level = dev->nested_level;
4591 spin_lock_nested(&dev->addr_list_lock, nest_level);
4594 static inline void netif_addr_lock_bh(struct net_device *dev)
4596 unsigned char nest_level = 0;
4598 #ifdef CONFIG_LOCKDEP
4599 nest_level = dev->nested_level;
4602 spin_lock_nested(&dev->addr_list_lock, nest_level);
4605 static inline void netif_addr_unlock(struct net_device *dev)
4607 spin_unlock(&dev->addr_list_lock);
4610 static inline void netif_addr_unlock_bh(struct net_device *dev)
4612 spin_unlock_bh(&dev->addr_list_lock);
4616 * dev_addrs walker. Should be used only for read access. Call with
4617 * rcu_read_lock held.
4619 #define for_each_dev_addr(dev, ha) \
4620 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4622 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4624 void ether_setup(struct net_device *dev);
4626 /* Support for loadable net-drivers */
4627 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4628 unsigned char name_assign_type,
4629 void (*setup)(struct net_device *),
4630 unsigned int txqs, unsigned int rxqs);
4631 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4632 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4634 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4635 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4638 int register_netdev(struct net_device *dev);
4639 void unregister_netdev(struct net_device *dev);
4641 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4643 /* General hardware address lists handling functions */
4644 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4645 struct netdev_hw_addr_list *from_list, int addr_len);
4646 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4647 struct netdev_hw_addr_list *from_list, int addr_len);
4648 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4649 struct net_device *dev,
4650 int (*sync)(struct net_device *, const unsigned char *),
4651 int (*unsync)(struct net_device *,
4652 const unsigned char *));
4653 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4654 struct net_device *dev,
4655 int (*sync)(struct net_device *,
4656 const unsigned char *, int),
4657 int (*unsync)(struct net_device *,
4658 const unsigned char *, int));
4659 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4660 struct net_device *dev,
4661 int (*unsync)(struct net_device *,
4662 const unsigned char *, int));
4663 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4664 struct net_device *dev,
4665 int (*unsync)(struct net_device *,
4666 const unsigned char *));
4667 void __hw_addr_init(struct netdev_hw_addr_list *list);
4669 /* Functions used for device addresses handling */
4670 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4671 const void *addr, size_t len);
4674 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4676 dev_addr_mod(dev, 0, addr, len);
4679 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4681 __dev_addr_set(dev, addr, dev->addr_len);
4684 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4685 unsigned char addr_type);
4686 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4687 unsigned char addr_type);
4689 /* Functions used for unicast addresses handling */
4690 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4691 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4692 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4693 int dev_uc_sync(struct net_device *to, struct net_device *from);
4694 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4695 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4696 void dev_uc_flush(struct net_device *dev);
4697 void dev_uc_init(struct net_device *dev);
4700 * __dev_uc_sync - Synchonize device's unicast list
4701 * @dev: device to sync
4702 * @sync: function to call if address should be added
4703 * @unsync: function to call if address should be removed
4705 * Add newly added addresses to the interface, and release
4706 * addresses that have been deleted.
4708 static inline int __dev_uc_sync(struct net_device *dev,
4709 int (*sync)(struct net_device *,
4710 const unsigned char *),
4711 int (*unsync)(struct net_device *,
4712 const unsigned char *))
4714 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4718 * __dev_uc_unsync - Remove synchronized addresses from device
4719 * @dev: device to sync
4720 * @unsync: function to call if address should be removed
4722 * Remove all addresses that were added to the device by dev_uc_sync().
4724 static inline void __dev_uc_unsync(struct net_device *dev,
4725 int (*unsync)(struct net_device *,
4726 const unsigned char *))
4728 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4731 /* Functions used for multicast addresses handling */
4732 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4733 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4734 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4735 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4736 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4737 int dev_mc_sync(struct net_device *to, struct net_device *from);
4738 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4739 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4740 void dev_mc_flush(struct net_device *dev);
4741 void dev_mc_init(struct net_device *dev);
4744 * __dev_mc_sync - Synchonize device's multicast list
4745 * @dev: device to sync
4746 * @sync: function to call if address should be added
4747 * @unsync: function to call if address should be removed
4749 * Add newly added addresses to the interface, and release
4750 * addresses that have been deleted.
4752 static inline int __dev_mc_sync(struct net_device *dev,
4753 int (*sync)(struct net_device *,
4754 const unsigned char *),
4755 int (*unsync)(struct net_device *,
4756 const unsigned char *))
4758 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4762 * __dev_mc_unsync - Remove synchronized addresses from device
4763 * @dev: device to sync
4764 * @unsync: function to call if address should be removed
4766 * Remove all addresses that were added to the device by dev_mc_sync().
4768 static inline void __dev_mc_unsync(struct net_device *dev,
4769 int (*unsync)(struct net_device *,
4770 const unsigned char *))
4772 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4775 /* Functions used for secondary unicast and multicast support */
4776 void dev_set_rx_mode(struct net_device *dev);
4777 int dev_set_promiscuity(struct net_device *dev, int inc);
4778 int dev_set_allmulti(struct net_device *dev, int inc);
4779 void netdev_state_change(struct net_device *dev);
4780 void __netdev_notify_peers(struct net_device *dev);
4781 void netdev_notify_peers(struct net_device *dev);
4782 void netdev_features_change(struct net_device *dev);
4783 /* Load a device via the kmod */
4784 void dev_load(struct net *net, const char *name);
4785 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4786 struct rtnl_link_stats64 *storage);
4787 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4788 const struct net_device_stats *netdev_stats);
4789 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4790 const struct pcpu_sw_netstats __percpu *netstats);
4791 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4793 extern int netdev_max_backlog;
4794 extern int dev_rx_weight;
4795 extern int dev_tx_weight;
4796 extern int gro_normal_batch;
4799 NESTED_SYNC_IMM_BIT,
4800 NESTED_SYNC_TODO_BIT,
4803 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4804 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4806 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4807 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4809 struct netdev_nested_priv {
4810 unsigned char flags;
4814 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4815 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4816 struct list_head **iter);
4818 /* iterate through upper list, must be called under RCU read lock */
4819 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4820 for (iter = &(dev)->adj_list.upper, \
4821 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4823 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4825 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4826 int (*fn)(struct net_device *upper_dev,
4827 struct netdev_nested_priv *priv),
4828 struct netdev_nested_priv *priv);
4830 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4831 struct net_device *upper_dev);
4833 bool netdev_has_any_upper_dev(struct net_device *dev);
4835 void *netdev_lower_get_next_private(struct net_device *dev,
4836 struct list_head **iter);
4837 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4838 struct list_head **iter);
4840 #define netdev_for_each_lower_private(dev, priv, iter) \
4841 for (iter = (dev)->adj_list.lower.next, \
4842 priv = netdev_lower_get_next_private(dev, &(iter)); \
4844 priv = netdev_lower_get_next_private(dev, &(iter)))
4846 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4847 for (iter = &(dev)->adj_list.lower, \
4848 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4850 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4852 void *netdev_lower_get_next(struct net_device *dev,
4853 struct list_head **iter);
4855 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4856 for (iter = (dev)->adj_list.lower.next, \
4857 ldev = netdev_lower_get_next(dev, &(iter)); \
4859 ldev = netdev_lower_get_next(dev, &(iter)))
4861 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4862 struct list_head **iter);
4863 int netdev_walk_all_lower_dev(struct net_device *dev,
4864 int (*fn)(struct net_device *lower_dev,
4865 struct netdev_nested_priv *priv),
4866 struct netdev_nested_priv *priv);
4867 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4868 int (*fn)(struct net_device *lower_dev,
4869 struct netdev_nested_priv *priv),
4870 struct netdev_nested_priv *priv);
4872 void *netdev_adjacent_get_private(struct list_head *adj_list);
4873 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4874 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4875 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4876 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4877 struct netlink_ext_ack *extack);
4878 int netdev_master_upper_dev_link(struct net_device *dev,
4879 struct net_device *upper_dev,
4880 void *upper_priv, void *upper_info,
4881 struct netlink_ext_ack *extack);
4882 void netdev_upper_dev_unlink(struct net_device *dev,
4883 struct net_device *upper_dev);
4884 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4885 struct net_device *new_dev,
4886 struct net_device *dev,
4887 struct netlink_ext_ack *extack);
4888 void netdev_adjacent_change_commit(struct net_device *old_dev,
4889 struct net_device *new_dev,
4890 struct net_device *dev);
4891 void netdev_adjacent_change_abort(struct net_device *old_dev,
4892 struct net_device *new_dev,
4893 struct net_device *dev);
4894 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4895 void *netdev_lower_dev_get_private(struct net_device *dev,
4896 struct net_device *lower_dev);
4897 void netdev_lower_state_changed(struct net_device *lower_dev,
4898 void *lower_state_info);
4900 /* RSS keys are 40 or 52 bytes long */
4901 #define NETDEV_RSS_KEY_LEN 52
4902 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4903 void netdev_rss_key_fill(void *buffer, size_t len);
4905 int skb_checksum_help(struct sk_buff *skb);
4906 int skb_crc32c_csum_help(struct sk_buff *skb);
4907 int skb_csum_hwoffload_help(struct sk_buff *skb,
4908 const netdev_features_t features);
4910 struct netdev_bonding_info {
4915 struct netdev_notifier_bonding_info {
4916 struct netdev_notifier_info info; /* must be first */
4917 struct netdev_bonding_info bonding_info;
4920 void netdev_bonding_info_change(struct net_device *dev,
4921 struct netdev_bonding_info *bonding_info);
4923 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4924 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4926 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4932 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4934 static inline bool can_checksum_protocol(netdev_features_t features,
4937 if (protocol == htons(ETH_P_FCOE))
4938 return !!(features & NETIF_F_FCOE_CRC);
4940 /* Assume this is an IP checksum (not SCTP CRC) */
4942 if (features & NETIF_F_HW_CSUM) {
4943 /* Can checksum everything */
4948 case htons(ETH_P_IP):
4949 return !!(features & NETIF_F_IP_CSUM);
4950 case htons(ETH_P_IPV6):
4951 return !!(features & NETIF_F_IPV6_CSUM);
4958 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4960 static inline void netdev_rx_csum_fault(struct net_device *dev,
4961 struct sk_buff *skb)
4965 /* rx skb timestamps */
4966 void net_enable_timestamp(void);
4967 void net_disable_timestamp(void);
4969 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4970 const struct skb_shared_hwtstamps *hwtstamps,
4973 const struct net_device_ops *ops = dev->netdev_ops;
4975 if (ops->ndo_get_tstamp)
4976 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4978 return hwtstamps->hwtstamp;
4981 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4982 struct sk_buff *skb, struct net_device *dev,
4985 __this_cpu_write(softnet_data.xmit.more, more);
4986 return ops->ndo_start_xmit(skb, dev);
4989 static inline bool netdev_xmit_more(void)
4991 return __this_cpu_read(softnet_data.xmit.more);
4994 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4995 struct netdev_queue *txq, bool more)
4997 const struct net_device_ops *ops = dev->netdev_ops;
5000 rc = __netdev_start_xmit(ops, skb, dev, more);
5001 if (rc == NETDEV_TX_OK)
5002 txq_trans_update(txq);
5007 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5009 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5012 extern const struct kobj_ns_type_operations net_ns_type_operations;
5014 const char *netdev_drivername(const struct net_device *dev);
5016 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5017 netdev_features_t f2)
5019 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5020 if (f1 & NETIF_F_HW_CSUM)
5021 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5023 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5029 static inline netdev_features_t netdev_get_wanted_features(
5030 struct net_device *dev)
5032 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5034 netdev_features_t netdev_increment_features(netdev_features_t all,
5035 netdev_features_t one, netdev_features_t mask);
5037 /* Allow TSO being used on stacked device :
5038 * Performing the GSO segmentation before last device
5039 * is a performance improvement.
5041 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5042 netdev_features_t mask)
5044 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5047 int __netdev_update_features(struct net_device *dev);
5048 void netdev_update_features(struct net_device *dev);
5049 void netdev_change_features(struct net_device *dev);
5051 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5052 struct net_device *dev);
5054 netdev_features_t passthru_features_check(struct sk_buff *skb,
5055 struct net_device *dev,
5056 netdev_features_t features);
5057 netdev_features_t netif_skb_features(struct sk_buff *skb);
5058 void skb_warn_bad_offload(const struct sk_buff *skb);
5060 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5062 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5064 /* check flags correspondence */
5065 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5066 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5067 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5068 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5069 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5070 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5071 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5072 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5073 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5074 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5075 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5076 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5077 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5078 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5079 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5080 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5081 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5082 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5083 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5085 return (features & feature) == feature;
5088 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5090 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5091 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5094 static inline bool netif_needs_gso(struct sk_buff *skb,
5095 netdev_features_t features)
5097 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5098 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5099 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5102 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5103 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5104 void netif_inherit_tso_max(struct net_device *to,
5105 const struct net_device *from);
5107 static inline bool netif_is_macsec(const struct net_device *dev)
5109 return dev->priv_flags & IFF_MACSEC;
5112 static inline bool netif_is_macvlan(const struct net_device *dev)
5114 return dev->priv_flags & IFF_MACVLAN;
5117 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5119 return dev->priv_flags & IFF_MACVLAN_PORT;
5122 static inline bool netif_is_bond_master(const struct net_device *dev)
5124 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5127 static inline bool netif_is_bond_slave(const struct net_device *dev)
5129 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5132 static inline bool netif_supports_nofcs(struct net_device *dev)
5134 return dev->priv_flags & IFF_SUPP_NOFCS;
5137 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5139 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5142 static inline bool netif_is_l3_master(const struct net_device *dev)
5144 return dev->priv_flags & IFF_L3MDEV_MASTER;
5147 static inline bool netif_is_l3_slave(const struct net_device *dev)
5149 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5152 static inline int dev_sdif(const struct net_device *dev)
5154 #ifdef CONFIG_NET_L3_MASTER_DEV
5155 if (netif_is_l3_slave(dev))
5156 return dev->ifindex;
5161 static inline bool netif_is_bridge_master(const struct net_device *dev)
5163 return dev->priv_flags & IFF_EBRIDGE;
5166 static inline bool netif_is_bridge_port(const struct net_device *dev)
5168 return dev->priv_flags & IFF_BRIDGE_PORT;
5171 static inline bool netif_is_ovs_master(const struct net_device *dev)
5173 return dev->priv_flags & IFF_OPENVSWITCH;
5176 static inline bool netif_is_ovs_port(const struct net_device *dev)
5178 return dev->priv_flags & IFF_OVS_DATAPATH;
5181 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5183 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5186 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5188 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5191 static inline bool netif_is_team_master(const struct net_device *dev)
5193 return dev->priv_flags & IFF_TEAM;
5196 static inline bool netif_is_team_port(const struct net_device *dev)
5198 return dev->priv_flags & IFF_TEAM_PORT;
5201 static inline bool netif_is_lag_master(const struct net_device *dev)
5203 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5206 static inline bool netif_is_lag_port(const struct net_device *dev)
5208 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5211 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5213 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5216 static inline bool netif_is_failover(const struct net_device *dev)
5218 return dev->priv_flags & IFF_FAILOVER;
5221 static inline bool netif_is_failover_slave(const struct net_device *dev)
5223 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5226 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5227 static inline void netif_keep_dst(struct net_device *dev)
5229 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5232 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5233 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5235 /* TODO: reserve and use an additional IFF bit, if we get more users */
5236 return netif_is_macsec(dev);
5239 extern struct pernet_operations __net_initdata loopback_net_ops;
5241 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5243 /* netdev_printk helpers, similar to dev_printk */
5245 static inline const char *netdev_name(const struct net_device *dev)
5247 if (!dev->name[0] || strchr(dev->name, '%'))
5248 return "(unnamed net_device)";
5252 static inline const char *netdev_reg_state(const struct net_device *dev)
5254 switch (dev->reg_state) {
5255 case NETREG_UNINITIALIZED: return " (uninitialized)";
5256 case NETREG_REGISTERED: return "";
5257 case NETREG_UNREGISTERING: return " (unregistering)";
5258 case NETREG_UNREGISTERED: return " (unregistered)";
5259 case NETREG_RELEASED: return " (released)";
5260 case NETREG_DUMMY: return " (dummy)";
5263 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5264 return " (unknown)";
5267 #define MODULE_ALIAS_NETDEV(device) \
5268 MODULE_ALIAS("netdev-" device)
5271 * netdev_WARN() acts like dev_printk(), but with the key difference
5272 * of using a WARN/WARN_ON to get the message out, including the
5273 * file/line information and a backtrace.
5275 #define netdev_WARN(dev, format, args...) \
5276 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5277 netdev_reg_state(dev), ##args)
5279 #define netdev_WARN_ONCE(dev, format, args...) \
5280 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5281 netdev_reg_state(dev), ##args)
5284 * The list of packet types we will receive (as opposed to discard)
5285 * and the routines to invoke.
5287 * Why 16. Because with 16 the only overlap we get on a hash of the
5288 * low nibble of the protocol value is RARP/SNAP/X.25.
5302 #define PTYPE_HASH_SIZE (16)
5303 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5305 extern struct list_head ptype_all __read_mostly;
5306 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5308 extern struct net_device *blackhole_netdev;
5310 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5311 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5312 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5313 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5314 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5316 #endif /* _LINUX_NETDEVICE_H */