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;
85 typedef u32 xdp_features_t;
87 void synchronize_net(void);
88 void netdev_set_default_ethtool_ops(struct net_device *dev,
89 const struct ethtool_ops *ops);
90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
92 /* Backlog congestion levels */
93 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
94 #define NET_RX_DROP 1 /* packet dropped */
96 #define MAX_NEST_DEV 8
99 * Transmit return codes: transmit return codes originate from three different
102 * - qdisc return codes
103 * - driver transmit return codes
106 * Drivers are allowed to return any one of those in their hard_start_xmit()
107 * function. Real network devices commonly used with qdiscs should only return
108 * the driver transmit return codes though - when qdiscs are used, the actual
109 * transmission happens asynchronously, so the value is not propagated to
110 * higher layers. Virtual network devices transmit synchronously; in this case
111 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
112 * others are propagated to higher layers.
115 /* qdisc ->enqueue() return codes. */
116 #define NET_XMIT_SUCCESS 0x00
117 #define NET_XMIT_DROP 0x01 /* skb dropped */
118 #define NET_XMIT_CN 0x02 /* congestion notification */
119 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
122 * indicates that the device will soon be dropping packets, or already drops
123 * some packets of the same priority; prompting us to send less aggressively. */
124 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
125 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
127 /* Driver transmit return codes */
128 #define NETDEV_TX_MASK 0xf0
131 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
132 NETDEV_TX_OK = 0x00, /* driver took care of packet */
133 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
135 typedef enum netdev_tx netdev_tx_t;
138 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
139 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
141 static inline bool dev_xmit_complete(int rc)
144 * Positive cases with an skb consumed by a driver:
145 * - successful transmission (rc == NETDEV_TX_OK)
146 * - error while transmitting (rc < 0)
147 * - error while queueing to a different device (rc & NET_XMIT_MASK)
149 if (likely(rc < NET_XMIT_MASK))
156 * Compute the worst-case header length according to the protocols
160 #if defined(CONFIG_HYPERV_NET)
161 # define LL_MAX_HEADER 128
162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
163 # if defined(CONFIG_MAC80211_MESH)
164 # define LL_MAX_HEADER 128
166 # define LL_MAX_HEADER 96
169 # define LL_MAX_HEADER 32
172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
173 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
174 #define MAX_HEADER LL_MAX_HEADER
176 #define MAX_HEADER (LL_MAX_HEADER + 48)
180 * Old network device statistics. Fields are native words
181 * (unsigned long) so they can be read and written atomically.
184 #define NET_DEV_STAT(FIELD) \
186 unsigned long FIELD; \
187 atomic_long_t __##FIELD; \
190 struct net_device_stats {
191 NET_DEV_STAT(rx_packets);
192 NET_DEV_STAT(tx_packets);
193 NET_DEV_STAT(rx_bytes);
194 NET_DEV_STAT(tx_bytes);
195 NET_DEV_STAT(rx_errors);
196 NET_DEV_STAT(tx_errors);
197 NET_DEV_STAT(rx_dropped);
198 NET_DEV_STAT(tx_dropped);
199 NET_DEV_STAT(multicast);
200 NET_DEV_STAT(collisions);
201 NET_DEV_STAT(rx_length_errors);
202 NET_DEV_STAT(rx_over_errors);
203 NET_DEV_STAT(rx_crc_errors);
204 NET_DEV_STAT(rx_frame_errors);
205 NET_DEV_STAT(rx_fifo_errors);
206 NET_DEV_STAT(rx_missed_errors);
207 NET_DEV_STAT(tx_aborted_errors);
208 NET_DEV_STAT(tx_carrier_errors);
209 NET_DEV_STAT(tx_fifo_errors);
210 NET_DEV_STAT(tx_heartbeat_errors);
211 NET_DEV_STAT(tx_window_errors);
212 NET_DEV_STAT(rx_compressed);
213 NET_DEV_STAT(tx_compressed);
217 /* per-cpu stats, allocated on demand.
218 * Try to fit them in a single cache line, for dev_get_stats() sake.
220 struct net_device_core_stats {
221 unsigned long rx_dropped;
222 unsigned long tx_dropped;
223 unsigned long rx_nohandler;
224 unsigned long rx_otherhost_dropped;
225 } __aligned(4 * sizeof(unsigned long));
227 #include <linux/cache.h>
228 #include <linux/skbuff.h>
231 #include <linux/static_key.h>
232 extern struct static_key_false rps_needed;
233 extern struct static_key_false rfs_needed;
240 struct netdev_hw_addr {
241 struct list_head list;
243 unsigned char addr[MAX_ADDR_LEN];
245 #define NETDEV_HW_ADDR_T_LAN 1
246 #define NETDEV_HW_ADDR_T_SAN 2
247 #define NETDEV_HW_ADDR_T_UNICAST 3
248 #define NETDEV_HW_ADDR_T_MULTICAST 4
253 struct rcu_head rcu_head;
256 struct netdev_hw_addr_list {
257 struct list_head list;
260 /* Auxiliary tree for faster lookup on addition and deletion */
264 #define netdev_hw_addr_list_count(l) ((l)->count)
265 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
266 #define netdev_hw_addr_list_for_each(ha, l) \
267 list_for_each_entry(ha, &(l)->list, list)
269 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
270 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
271 #define netdev_for_each_uc_addr(ha, dev) \
272 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
273 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
274 netdev_for_each_uc_addr((_ha), (_dev)) \
277 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
278 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
279 #define netdev_for_each_mc_addr(ha, dev) \
280 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
281 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
282 netdev_for_each_mc_addr((_ha), (_dev)) \
289 /* cached hardware header; allow for machine alignment needs. */
290 #define HH_DATA_MOD 16
291 #define HH_DATA_OFF(__len) \
292 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
293 #define HH_DATA_ALIGN(__len) \
294 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
295 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
298 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
300 * dev->hard_header_len ? (dev->hard_header_len +
301 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
303 * We could use other alignment values, but we must maintain the
304 * relationship HH alignment <= LL alignment.
306 #define LL_RESERVED_SPACE(dev) \
307 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
308 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
309 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
310 ((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
311 & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
314 int (*create) (struct sk_buff *skb, struct net_device *dev,
315 unsigned short type, const void *daddr,
316 const void *saddr, unsigned int len);
317 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
318 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
319 void (*cache_update)(struct hh_cache *hh,
320 const struct net_device *dev,
321 const unsigned char *haddr);
322 bool (*validate)(const char *ll_header, unsigned int len);
323 __be16 (*parse_protocol)(const struct sk_buff *skb);
326 /* These flag bits are private to the generic network queueing
327 * layer; they may not be explicitly referenced by any other
331 enum netdev_state_t {
333 __LINK_STATE_PRESENT,
334 __LINK_STATE_NOCARRIER,
335 __LINK_STATE_LINKWATCH_PENDING,
336 __LINK_STATE_DORMANT,
337 __LINK_STATE_TESTING,
341 struct list_head list;
346 * size of gro hash buckets, must less than bit number of
347 * napi_struct::gro_bitmask
349 #define GRO_HASH_BUCKETS 8
352 * Structure for NAPI scheduling similar to tasklet but with weighting
355 /* The poll_list must only be managed by the entity which
356 * changes the state of the NAPI_STATE_SCHED bit. This means
357 * whoever atomically sets that bit can add this napi_struct
358 * to the per-CPU poll_list, and whoever clears that bit
359 * can remove from the list right before clearing the bit.
361 struct list_head poll_list;
365 int defer_hard_irqs_count;
366 unsigned long gro_bitmask;
367 int (*poll)(struct napi_struct *, int);
368 #ifdef CONFIG_NETPOLL
369 /* CPU actively polling if netpoll is configured */
372 /* CPU on which NAPI has been scheduled for processing */
374 struct net_device *dev;
375 struct gro_list gro_hash[GRO_HASH_BUCKETS];
377 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
378 int rx_count; /* length of rx_list */
379 unsigned int napi_id;
380 struct hrtimer timer;
381 struct task_struct *thread;
382 /* control-path-only fields follow */
383 struct list_head dev_list;
384 struct hlist_node napi_hash_node;
389 NAPI_STATE_SCHED, /* Poll is scheduled */
390 NAPI_STATE_MISSED, /* reschedule a napi */
391 NAPI_STATE_DISABLE, /* Disable pending */
392 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
393 NAPI_STATE_LISTED, /* NAPI added to system lists */
394 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
395 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
396 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
397 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
398 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
402 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
403 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
404 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
405 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
406 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
407 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
408 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
409 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
410 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
411 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
421 typedef enum gro_result gro_result_t;
424 * enum rx_handler_result - Possible return values for rx_handlers.
425 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
427 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
428 * case skb->dev was changed by rx_handler.
429 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
430 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
432 * rx_handlers are functions called from inside __netif_receive_skb(), to do
433 * special processing of the skb, prior to delivery to protocol handlers.
435 * Currently, a net_device can only have a single rx_handler registered. Trying
436 * to register a second rx_handler will return -EBUSY.
438 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
439 * To unregister a rx_handler on a net_device, use
440 * netdev_rx_handler_unregister().
442 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
445 * If the rx_handler consumed the skb in some way, it should return
446 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
447 * the skb to be delivered in some other way.
449 * If the rx_handler changed skb->dev, to divert the skb to another
450 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
451 * new device will be called if it exists.
453 * If the rx_handler decides the skb should be ignored, it should return
454 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
455 * are registered on exact device (ptype->dev == skb->dev).
457 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
458 * delivered, it should return RX_HANDLER_PASS.
460 * A device without a registered rx_handler will behave as if rx_handler
461 * returned RX_HANDLER_PASS.
464 enum rx_handler_result {
470 typedef enum rx_handler_result rx_handler_result_t;
471 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
473 void __napi_schedule(struct napi_struct *n);
474 void __napi_schedule_irqoff(struct napi_struct *n);
476 static inline bool napi_disable_pending(struct napi_struct *n)
478 return test_bit(NAPI_STATE_DISABLE, &n->state);
481 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
483 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
487 * napi_is_scheduled - test if NAPI is scheduled
490 * This check is "best-effort". With no locking implemented,
491 * a NAPI can be scheduled or terminate right after this check
492 * and produce not precise results.
494 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
495 * should not be used normally and napi_schedule should be
498 * Use only if the driver really needs to check if a NAPI
499 * is scheduled for example in the context of delayed timer
500 * that can be skipped if a NAPI is already scheduled.
502 * Return True if NAPI is scheduled, False otherwise.
504 static inline bool napi_is_scheduled(struct napi_struct *n)
506 return test_bit(NAPI_STATE_SCHED, &n->state);
509 bool napi_schedule_prep(struct napi_struct *n);
512 * napi_schedule - schedule NAPI poll
515 * Schedule NAPI poll routine to be called if it is not already
517 * Return true if we schedule a NAPI or false if not.
518 * Refer to napi_schedule_prep() for additional reason on why
519 * a NAPI might not be scheduled.
521 static inline bool napi_schedule(struct napi_struct *n)
523 if (napi_schedule_prep(n)) {
532 * napi_schedule_irqoff - schedule NAPI poll
535 * Variant of napi_schedule(), assuming hard irqs are masked.
537 static inline void napi_schedule_irqoff(struct napi_struct *n)
539 if (napi_schedule_prep(n))
540 __napi_schedule_irqoff(n);
544 * napi_complete_done - NAPI processing complete
546 * @work_done: number of packets processed
548 * Mark NAPI processing as complete. Should only be called if poll budget
549 * has not been completely consumed.
550 * Prefer over napi_complete().
551 * Return false if device should avoid rearming interrupts.
553 bool napi_complete_done(struct napi_struct *n, int work_done);
555 static inline bool napi_complete(struct napi_struct *n)
557 return napi_complete_done(n, 0);
560 int dev_set_threaded(struct net_device *dev, bool threaded);
563 * napi_disable - prevent NAPI from scheduling
566 * Stop NAPI from being scheduled on this context.
567 * Waits till any outstanding processing completes.
569 void napi_disable(struct napi_struct *n);
571 void napi_enable(struct napi_struct *n);
574 * napi_synchronize - wait until NAPI is not running
577 * Wait until NAPI is done being scheduled on this context.
578 * Waits till any outstanding processing completes but
579 * does not disable future activations.
581 static inline void napi_synchronize(const struct napi_struct *n)
583 if (IS_ENABLED(CONFIG_SMP))
584 while (test_bit(NAPI_STATE_SCHED, &n->state))
591 * napi_if_scheduled_mark_missed - if napi is running, set the
595 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
598 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
600 unsigned long val, new;
602 val = READ_ONCE(n->state);
604 if (val & NAPIF_STATE_DISABLE)
607 if (!(val & NAPIF_STATE_SCHED))
610 new = val | NAPIF_STATE_MISSED;
611 } while (!try_cmpxchg(&n->state, &val, new));
616 enum netdev_queue_state_t {
617 __QUEUE_STATE_DRV_XOFF,
618 __QUEUE_STATE_STACK_XOFF,
619 __QUEUE_STATE_FROZEN,
622 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
623 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
624 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
626 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
627 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
629 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
633 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
634 * netif_tx_* functions below are used to manipulate this flag. The
635 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
636 * queue independently. The netif_xmit_*stopped functions below are called
637 * to check if the queue has been stopped by the driver or stack (either
638 * of the XOFF bits are set in the state). Drivers should not need to call
639 * netif_xmit*stopped functions, they should only be using netif_tx_*.
642 struct netdev_queue {
646 struct net_device *dev;
647 netdevice_tracker dev_tracker;
649 struct Qdisc __rcu *qdisc;
650 struct Qdisc __rcu *qdisc_sleeping;
654 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 unsigned long tx_maxrate;
659 * Number of TX timeouts for this queue
660 * (/sys/class/net/DEV/Q/trans_timeout)
662 atomic_long_t trans_timeout;
664 /* Subordinate device that the queue has been assigned to */
665 struct net_device *sb_dev;
666 #ifdef CONFIG_XDP_SOCKETS
667 struct xsk_buff_pool *pool;
669 /* NAPI instance for the queue
670 * Readers and writers must hold RTNL
672 struct napi_struct *napi;
676 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
679 * Time (in jiffies) of last Tx
681 unsigned long trans_start;
688 } ____cacheline_aligned_in_smp;
690 extern int sysctl_fb_tunnels_only_for_init_net;
691 extern int sysctl_devconf_inherit_init_net;
694 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
695 * == 1 : For initns only
698 static inline bool net_has_fallback_tunnels(const struct net *net)
700 #if IS_ENABLED(CONFIG_SYSCTL)
701 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
703 return !fb_tunnels_only_for_init_net ||
704 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
710 static inline int net_inherit_devconf(void)
712 #if IS_ENABLED(CONFIG_SYSCTL)
713 return READ_ONCE(sysctl_devconf_inherit_init_net);
719 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
721 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
728 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
730 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
737 * This structure holds an RPS map which can be of variable length. The
738 * map is an array of CPUs.
745 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
748 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
749 * tail pointer for that CPU's input queue at the time of last enqueue, and
750 * a hardware filter index.
752 struct rps_dev_flow {
755 unsigned int last_qtail;
757 #define RPS_NO_FILTER 0xffff
760 * The rps_dev_flow_table structure contains a table of flow mappings.
762 struct rps_dev_flow_table {
765 struct rps_dev_flow flows[];
767 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
768 ((_num) * sizeof(struct rps_dev_flow)))
771 * The rps_sock_flow_table contains mappings of flows to the last CPU
772 * on which they were processed by the application (set in recvmsg).
773 * Each entry is a 32bit value. Upper part is the high-order bits
774 * of flow hash, lower part is CPU number.
775 * rps_cpu_mask is used to partition the space, depending on number of
776 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
777 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
778 * meaning we use 32-6=26 bits for the hash.
780 struct rps_sock_flow_table {
783 u32 ents[] ____cacheline_aligned_in_smp;
785 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
787 #define RPS_NO_CPU 0xffff
789 extern u32 rps_cpu_mask;
790 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
792 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
796 unsigned int index = hash & table->mask;
797 u32 val = hash & ~rps_cpu_mask;
799 /* We only give a hint, preemption can change CPU under us */
800 val |= raw_smp_processor_id();
802 /* The following WRITE_ONCE() is paired with the READ_ONCE()
803 * here, and another one in get_rps_cpu().
805 if (READ_ONCE(table->ents[index]) != val)
806 WRITE_ONCE(table->ents[index], val);
810 #ifdef CONFIG_RFS_ACCEL
811 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
814 #endif /* CONFIG_RPS */
816 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
825 * This structure holds an XPS map which can be of variable length. The
826 * map is an array of queues.
830 unsigned int alloc_len;
834 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
835 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
836 - sizeof(struct xps_map)) / sizeof(u16))
839 * This structure holds all XPS maps for device. Maps are indexed by CPU.
841 * We keep track of the number of cpus/rxqs used when the struct is allocated,
842 * in nr_ids. This will help not accessing out-of-bound memory.
844 * We keep track of the number of traffic classes used when the struct is
845 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
846 * not crossing its upper bound, as the original dev->num_tc can be updated in
849 struct xps_dev_maps {
853 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
856 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
857 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
859 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
860 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
862 #endif /* CONFIG_XPS */
864 #define TC_MAX_QUEUE 16
865 #define TC_BITMASK 15
866 /* HW offloaded queuing disciplines txq count and offset maps */
867 struct netdev_tc_txq {
872 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
874 * This structure is to hold information about the device
875 * configured to run FCoE protocol stack.
877 struct netdev_fcoe_hbainfo {
878 char manufacturer[64];
879 char serial_number[64];
880 char hardware_version[64];
881 char driver_version[64];
882 char optionrom_version[64];
883 char firmware_version[64];
885 char model_description[256];
889 #define MAX_PHYS_ITEM_ID_LEN 32
891 /* This structure holds a unique identifier to identify some
892 * physical item (port for example) used by a netdevice.
894 struct netdev_phys_item_id {
895 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
896 unsigned char id_len;
899 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
900 struct netdev_phys_item_id *b)
902 return a->id_len == b->id_len &&
903 memcmp(a->id, b->id, a->id_len) == 0;
906 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
908 struct net_device *sb_dev);
910 enum net_device_path_type {
911 DEV_PATH_ETHERNET = 0,
919 struct net_device_path {
920 enum net_device_path_type type;
921 const struct net_device *dev;
930 DEV_PATH_BR_VLAN_KEEP,
931 DEV_PATH_BR_VLAN_TAG,
932 DEV_PATH_BR_VLAN_UNTAG,
933 DEV_PATH_BR_VLAN_UNTAG_HW,
952 #define NET_DEVICE_PATH_STACK_MAX 5
953 #define NET_DEVICE_PATH_VLAN_MAX 2
955 struct net_device_path_stack {
957 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
960 struct net_device_path_ctx {
961 const struct net_device *dev;
968 } vlan[NET_DEVICE_PATH_VLAN_MAX];
973 TC_SETUP_QDISC_MQPRIO,
976 TC_SETUP_CLSMATCHALL,
986 TC_SETUP_QDISC_TAPRIO,
995 /* These structures hold the attributes of bpf state that are being passed
996 * to the netdevice through the bpf op.
998 enum bpf_netdev_command {
999 /* Set or clear a bpf program used in the earliest stages of packet
1000 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
1001 * is responsible for calling bpf_prog_put on any old progs that are
1002 * stored. In case of error, the callee need not release the new prog
1003 * reference, but on success it takes ownership and must bpf_prog_put
1004 * when it is no longer used.
1008 /* BPF program for offload callbacks, invoked at program load time. */
1009 BPF_OFFLOAD_MAP_ALLOC,
1010 BPF_OFFLOAD_MAP_FREE,
1014 struct bpf_prog_offload_ops;
1015 struct netlink_ext_ack;
1017 struct xdp_dev_bulk_queue;
1018 struct bpf_xdp_link;
1027 struct bpf_xdp_entity {
1028 struct bpf_prog *prog;
1029 struct bpf_xdp_link *link;
1033 enum bpf_netdev_command command;
1035 /* XDP_SETUP_PROG */
1038 struct bpf_prog *prog;
1039 struct netlink_ext_ack *extack;
1041 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1043 struct bpf_offloaded_map *offmap;
1045 /* XDP_SETUP_XSK_POOL */
1047 struct xsk_buff_pool *pool;
1053 /* Flags for ndo_xsk_wakeup. */
1054 #define XDP_WAKEUP_RX (1 << 0)
1055 #define XDP_WAKEUP_TX (1 << 1)
1057 #ifdef CONFIG_XFRM_OFFLOAD
1058 struct xfrmdev_ops {
1059 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1060 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1061 void (*xdo_dev_state_free) (struct xfrm_state *x);
1062 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1063 struct xfrm_state *x);
1064 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1065 void (*xdo_dev_state_update_stats) (struct xfrm_state *x);
1066 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1067 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1068 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1072 struct dev_ifalias {
1073 struct rcu_head rcuhead;
1080 struct netdev_net_notifier {
1081 struct list_head list;
1082 struct notifier_block *nb;
1086 * This structure defines the management hooks for network devices.
1087 * The following hooks can be defined; unless noted otherwise, they are
1088 * optional and can be filled with a null pointer.
1090 * int (*ndo_init)(struct net_device *dev);
1091 * This function is called once when a network device is registered.
1092 * The network device can use this for any late stage initialization
1093 * or semantic validation. It can fail with an error code which will
1094 * be propagated back to register_netdev.
1096 * void (*ndo_uninit)(struct net_device *dev);
1097 * This function is called when device is unregistered or when registration
1098 * fails. It is not called if init fails.
1100 * int (*ndo_open)(struct net_device *dev);
1101 * This function is called when a network device transitions to the up
1104 * int (*ndo_stop)(struct net_device *dev);
1105 * This function is called when a network device transitions to the down
1108 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1109 * struct net_device *dev);
1110 * Called when a packet needs to be transmitted.
1111 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1112 * the queue before that can happen; it's for obsolete devices and weird
1113 * corner cases, but the stack really does a non-trivial amount
1114 * of useless work if you return NETDEV_TX_BUSY.
1115 * Required; cannot be NULL.
1117 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1118 * struct net_device *dev
1119 * netdev_features_t features);
1120 * Called by core transmit path to determine if device is capable of
1121 * performing offload operations on a given packet. This is to give
1122 * the device an opportunity to implement any restrictions that cannot
1123 * be otherwise expressed by feature flags. The check is called with
1124 * the set of features that the stack has calculated and it returns
1125 * those the driver believes to be appropriate.
1127 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1128 * struct net_device *sb_dev);
1129 * Called to decide which queue to use when device supports multiple
1132 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1133 * This function is called to allow device receiver to make
1134 * changes to configuration when multicast or promiscuous is enabled.
1136 * void (*ndo_set_rx_mode)(struct net_device *dev);
1137 * This function is called device changes address list filtering.
1138 * If driver handles unicast address filtering, it should set
1139 * IFF_UNICAST_FLT in its priv_flags.
1141 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1142 * This function is called when the Media Access Control address
1143 * needs to be changed. If this interface is not defined, the
1144 * MAC address can not be changed.
1146 * int (*ndo_validate_addr)(struct net_device *dev);
1147 * Test if Media Access Control address is valid for the device.
1149 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1150 * Old-style ioctl entry point. This is used internally by the
1151 * appletalk and ieee802154 subsystems but is no longer called by
1152 * the device ioctl handler.
1154 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1155 * Used by the bonding driver for its device specific ioctls:
1156 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1157 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1159 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1160 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1161 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1163 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1164 * Used to set network devices bus interface parameters. This interface
1165 * is retained for legacy reasons; new devices should use the bus
1166 * interface (PCI) for low level management.
1168 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1169 * Called when a user wants to change the Maximum Transfer Unit
1172 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1173 * Callback used when the transmitter has not made any progress
1174 * for dev->watchdog ticks.
1176 * void (*ndo_get_stats64)(struct net_device *dev,
1177 * struct rtnl_link_stats64 *storage);
1178 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1179 * Called when a user wants to get the network device usage
1180 * statistics. Drivers must do one of the following:
1181 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1182 * rtnl_link_stats64 structure passed by the caller.
1183 * 2. Define @ndo_get_stats to update a net_device_stats structure
1184 * (which should normally be dev->stats) and return a pointer to
1185 * it. The structure may be changed asynchronously only if each
1186 * field is written atomically.
1187 * 3. Update dev->stats asynchronously and atomically, and define
1188 * neither operation.
1190 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1191 * Return true if this device supports offload stats of this attr_id.
1193 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1195 * Get statistics for offload operations by attr_id. Write it into the
1196 * attr_data pointer.
1198 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1199 * If device supports VLAN filtering this function is called when a
1200 * VLAN id is registered.
1202 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1203 * If device supports VLAN filtering this function is called when a
1204 * VLAN id is unregistered.
1206 * void (*ndo_poll_controller)(struct net_device *dev);
1208 * SR-IOV management functions.
1209 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1210 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1211 * u8 qos, __be16 proto);
1212 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1214 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1215 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1216 * int (*ndo_get_vf_config)(struct net_device *dev,
1217 * int vf, struct ifla_vf_info *ivf);
1218 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1219 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1220 * struct nlattr *port[]);
1222 * Enable or disable the VF ability to query its RSS Redirection Table and
1223 * Hash Key. This is needed since on some devices VF share this information
1224 * with PF and querying it may introduce a theoretical security risk.
1225 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1226 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1227 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1229 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1230 * This is always called from the stack with the rtnl lock held and netif
1231 * tx queues stopped. This allows the netdevice to perform queue
1232 * management safely.
1234 * Fiber Channel over Ethernet (FCoE) offload functions.
1235 * int (*ndo_fcoe_enable)(struct net_device *dev);
1236 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1237 * so the underlying device can perform whatever needed configuration or
1238 * initialization to support acceleration of FCoE traffic.
1240 * int (*ndo_fcoe_disable)(struct net_device *dev);
1241 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1242 * so the underlying device can perform whatever needed clean-ups to
1243 * stop supporting acceleration of FCoE traffic.
1245 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1246 * struct scatterlist *sgl, unsigned int sgc);
1247 * Called when the FCoE Initiator wants to initialize an I/O that
1248 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1249 * perform necessary setup and returns 1 to indicate the device is set up
1250 * successfully to perform DDP on this I/O, otherwise this returns 0.
1252 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1253 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1254 * indicated by the FC exchange id 'xid', so the underlying device can
1255 * clean up and reuse resources for later DDP requests.
1257 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1258 * struct scatterlist *sgl, unsigned int sgc);
1259 * Called when the FCoE Target wants to initialize an I/O that
1260 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1261 * perform necessary setup and returns 1 to indicate the device is set up
1262 * successfully to perform DDP on this I/O, otherwise this returns 0.
1264 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1265 * struct netdev_fcoe_hbainfo *hbainfo);
1266 * Called when the FCoE Protocol stack wants information on the underlying
1267 * device. This information is utilized by the FCoE protocol stack to
1268 * register attributes with Fiber Channel management service as per the
1269 * FC-GS Fabric Device Management Information(FDMI) specification.
1271 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1272 * Called when the underlying device wants to override default World Wide
1273 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1274 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1275 * protocol stack to use.
1278 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1279 * u16 rxq_index, u32 flow_id);
1280 * Set hardware filter for RFS. rxq_index is the target queue index;
1281 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1282 * Return the filter ID on success, or a negative error code.
1284 * Slave management functions (for bridge, bonding, etc).
1285 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1286 * Called to make another netdev an underling.
1288 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1289 * Called to release previously enslaved netdev.
1291 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1292 * struct sk_buff *skb,
1294 * Get the xmit slave of master device. If all_slaves is true, function
1295 * assume all the slaves can transmit.
1297 * Feature/offload setting functions.
1298 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1299 * netdev_features_t features);
1300 * Adjusts the requested feature flags according to device-specific
1301 * constraints, and returns the resulting flags. Must not modify
1304 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1305 * Called to update device configuration to new features. Passed
1306 * feature set might be less than what was returned by ndo_fix_features()).
1307 * Must return >0 or -errno if it changed dev->features itself.
1309 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1310 * struct net_device *dev,
1311 * const unsigned char *addr, u16 vid, u16 flags,
1312 * struct netlink_ext_ack *extack);
1313 * Adds an FDB entry to dev for addr.
1314 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1315 * struct net_device *dev,
1316 * const unsigned char *addr, u16 vid)
1317 * Deletes the FDB entry from dev coresponding to addr.
1318 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1319 * struct netlink_ext_ack *extack);
1320 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1321 * struct net_device *dev, struct net_device *filter_dev,
1323 * Used to add FDB entries to dump requests. Implementers should add
1324 * entries to skb and update idx with the number of entries.
1326 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1327 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1328 * Adds an MDB entry to dev.
1329 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1330 * struct netlink_ext_ack *extack);
1331 * Deletes the MDB entry from dev.
1332 * int (*ndo_mdb_del_bulk)(struct net_device *dev, struct nlattr *tb[],
1333 * struct netlink_ext_ack *extack);
1334 * Bulk deletes MDB entries from dev.
1335 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1336 * struct netlink_callback *cb);
1337 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1338 * callback is used by core rtnetlink code.
1340 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1341 * u16 flags, struct netlink_ext_ack *extack)
1342 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1343 * struct net_device *dev, u32 filter_mask,
1345 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1348 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1349 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1350 * which do not represent real hardware may define this to allow their
1351 * userspace components to manage their virtual carrier state. Devices
1352 * that determine carrier state from physical hardware properties (eg
1353 * network cables) or protocol-dependent mechanisms (eg
1354 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1356 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1357 * struct netdev_phys_item_id *ppid);
1358 * Called to get ID of physical port of this device. If driver does
1359 * not implement this, it is assumed that the hw is not able to have
1360 * multiple net devices on single physical port.
1362 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1363 * struct netdev_phys_item_id *ppid)
1364 * Called to get the parent ID of the physical port of this device.
1366 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1367 * struct net_device *dev)
1368 * Called by upper layer devices to accelerate switching or other
1369 * station functionality into hardware. 'pdev is the lowerdev
1370 * to use for the offload and 'dev' is the net device that will
1371 * back the offload. Returns a pointer to the private structure
1372 * the upper layer will maintain.
1373 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1374 * Called by upper layer device to delete the station created
1375 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1376 * the station and priv is the structure returned by the add
1378 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1379 * int queue_index, u32 maxrate);
1380 * Called when a user wants to set a max-rate limitation of specific
1382 * int (*ndo_get_iflink)(const struct net_device *dev);
1383 * Called to get the iflink value of this device.
1384 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1385 * This function is used to get egress tunnel information for given skb.
1386 * This is useful for retrieving outer tunnel header parameters while
1388 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1389 * This function is used to specify the headroom that the skb must
1390 * consider when allocation skb during packet reception. Setting
1391 * appropriate rx headroom value allows avoiding skb head copy on
1392 * forward. Setting a negative value resets the rx headroom to the
1394 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1395 * This function is used to set or query state related to XDP on the
1396 * netdevice and manage BPF offload. See definition of
1397 * enum bpf_netdev_command for details.
1398 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1400 * This function is used to submit @n XDP packets for transmit on a
1401 * netdevice. Returns number of frames successfully transmitted, frames
1402 * that got dropped are freed/returned via xdp_return_frame().
1403 * Returns negative number, means general error invoking ndo, meaning
1404 * no frames were xmit'ed and core-caller will free all frames.
1405 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1406 * struct xdp_buff *xdp);
1407 * Get the xmit slave of master device based on the xdp_buff.
1408 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1409 * This function is used to wake up the softirq, ksoftirqd or kthread
1410 * responsible for sending and/or receiving packets on a specific
1411 * queue id bound to an AF_XDP socket. The flags field specifies if
1412 * only RX, only Tx, or both should be woken up using the flags
1413 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1414 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1416 * Add, change, delete or get information on an IPv4 tunnel.
1417 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1418 * If a device is paired with a peer device, return the peer instance.
1419 * The caller must be under RCU read context.
1420 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1421 * Get the forwarding path to reach the real device from the HW destination address
1422 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1423 * const struct skb_shared_hwtstamps *hwtstamps,
1425 * Get hardware timestamp based on normal/adjustable time or free running
1426 * cycle counter. This function is required if physical clock supports a
1427 * free running cycle counter.
1429 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1430 * struct kernel_hwtstamp_config *kernel_config);
1431 * Get the currently configured hardware timestamping parameters for the
1434 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1435 * struct kernel_hwtstamp_config *kernel_config,
1436 * struct netlink_ext_ack *extack);
1437 * Change the hardware timestamping parameters for NIC device.
1439 struct net_device_ops {
1440 int (*ndo_init)(struct net_device *dev);
1441 void (*ndo_uninit)(struct net_device *dev);
1442 int (*ndo_open)(struct net_device *dev);
1443 int (*ndo_stop)(struct net_device *dev);
1444 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1445 struct net_device *dev);
1446 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1447 struct net_device *dev,
1448 netdev_features_t features);
1449 u16 (*ndo_select_queue)(struct net_device *dev,
1450 struct sk_buff *skb,
1451 struct net_device *sb_dev);
1452 void (*ndo_change_rx_flags)(struct net_device *dev,
1454 void (*ndo_set_rx_mode)(struct net_device *dev);
1455 int (*ndo_set_mac_address)(struct net_device *dev,
1457 int (*ndo_validate_addr)(struct net_device *dev);
1458 int (*ndo_do_ioctl)(struct net_device *dev,
1459 struct ifreq *ifr, int cmd);
1460 int (*ndo_eth_ioctl)(struct net_device *dev,
1461 struct ifreq *ifr, int cmd);
1462 int (*ndo_siocbond)(struct net_device *dev,
1463 struct ifreq *ifr, int cmd);
1464 int (*ndo_siocwandev)(struct net_device *dev,
1465 struct if_settings *ifs);
1466 int (*ndo_siocdevprivate)(struct net_device *dev,
1468 void __user *data, int cmd);
1469 int (*ndo_set_config)(struct net_device *dev,
1471 int (*ndo_change_mtu)(struct net_device *dev,
1473 int (*ndo_neigh_setup)(struct net_device *dev,
1474 struct neigh_parms *);
1475 void (*ndo_tx_timeout) (struct net_device *dev,
1476 unsigned int txqueue);
1478 void (*ndo_get_stats64)(struct net_device *dev,
1479 struct rtnl_link_stats64 *storage);
1480 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1481 int (*ndo_get_offload_stats)(int attr_id,
1482 const struct net_device *dev,
1484 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1486 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1487 __be16 proto, u16 vid);
1488 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1489 __be16 proto, u16 vid);
1490 #ifdef CONFIG_NET_POLL_CONTROLLER
1491 void (*ndo_poll_controller)(struct net_device *dev);
1492 int (*ndo_netpoll_setup)(struct net_device *dev,
1493 struct netpoll_info *info);
1494 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1496 int (*ndo_set_vf_mac)(struct net_device *dev,
1497 int queue, u8 *mac);
1498 int (*ndo_set_vf_vlan)(struct net_device *dev,
1499 int queue, u16 vlan,
1500 u8 qos, __be16 proto);
1501 int (*ndo_set_vf_rate)(struct net_device *dev,
1502 int vf, int min_tx_rate,
1504 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1505 int vf, bool setting);
1506 int (*ndo_set_vf_trust)(struct net_device *dev,
1507 int vf, bool setting);
1508 int (*ndo_get_vf_config)(struct net_device *dev,
1510 struct ifla_vf_info *ivf);
1511 int (*ndo_set_vf_link_state)(struct net_device *dev,
1512 int vf, int link_state);
1513 int (*ndo_get_vf_stats)(struct net_device *dev,
1515 struct ifla_vf_stats
1517 int (*ndo_set_vf_port)(struct net_device *dev,
1519 struct nlattr *port[]);
1520 int (*ndo_get_vf_port)(struct net_device *dev,
1521 int vf, struct sk_buff *skb);
1522 int (*ndo_get_vf_guid)(struct net_device *dev,
1524 struct ifla_vf_guid *node_guid,
1525 struct ifla_vf_guid *port_guid);
1526 int (*ndo_set_vf_guid)(struct net_device *dev,
1529 int (*ndo_set_vf_rss_query_en)(
1530 struct net_device *dev,
1531 int vf, bool setting);
1532 int (*ndo_setup_tc)(struct net_device *dev,
1533 enum tc_setup_type type,
1535 #if IS_ENABLED(CONFIG_FCOE)
1536 int (*ndo_fcoe_enable)(struct net_device *dev);
1537 int (*ndo_fcoe_disable)(struct net_device *dev);
1538 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1540 struct scatterlist *sgl,
1542 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1544 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1546 struct scatterlist *sgl,
1548 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1549 struct netdev_fcoe_hbainfo *hbainfo);
1552 #if IS_ENABLED(CONFIG_LIBFCOE)
1553 #define NETDEV_FCOE_WWNN 0
1554 #define NETDEV_FCOE_WWPN 1
1555 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1556 u64 *wwn, int type);
1559 #ifdef CONFIG_RFS_ACCEL
1560 int (*ndo_rx_flow_steer)(struct net_device *dev,
1561 const struct sk_buff *skb,
1565 int (*ndo_add_slave)(struct net_device *dev,
1566 struct net_device *slave_dev,
1567 struct netlink_ext_ack *extack);
1568 int (*ndo_del_slave)(struct net_device *dev,
1569 struct net_device *slave_dev);
1570 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1571 struct sk_buff *skb,
1573 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1575 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1576 netdev_features_t features);
1577 int (*ndo_set_features)(struct net_device *dev,
1578 netdev_features_t features);
1579 int (*ndo_neigh_construct)(struct net_device *dev,
1580 struct neighbour *n);
1581 void (*ndo_neigh_destroy)(struct net_device *dev,
1582 struct neighbour *n);
1584 int (*ndo_fdb_add)(struct ndmsg *ndm,
1585 struct nlattr *tb[],
1586 struct net_device *dev,
1587 const unsigned char *addr,
1590 struct netlink_ext_ack *extack);
1591 int (*ndo_fdb_del)(struct ndmsg *ndm,
1592 struct nlattr *tb[],
1593 struct net_device *dev,
1594 const unsigned char *addr,
1595 u16 vid, struct netlink_ext_ack *extack);
1596 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1597 struct net_device *dev,
1598 struct netlink_ext_ack *extack);
1599 int (*ndo_fdb_dump)(struct sk_buff *skb,
1600 struct netlink_callback *cb,
1601 struct net_device *dev,
1602 struct net_device *filter_dev,
1604 int (*ndo_fdb_get)(struct sk_buff *skb,
1605 struct nlattr *tb[],
1606 struct net_device *dev,
1607 const unsigned char *addr,
1608 u16 vid, u32 portid, u32 seq,
1609 struct netlink_ext_ack *extack);
1610 int (*ndo_mdb_add)(struct net_device *dev,
1611 struct nlattr *tb[],
1613 struct netlink_ext_ack *extack);
1614 int (*ndo_mdb_del)(struct net_device *dev,
1615 struct nlattr *tb[],
1616 struct netlink_ext_ack *extack);
1617 int (*ndo_mdb_del_bulk)(struct net_device *dev,
1618 struct nlattr *tb[],
1619 struct netlink_ext_ack *extack);
1620 int (*ndo_mdb_dump)(struct net_device *dev,
1621 struct sk_buff *skb,
1622 struct netlink_callback *cb);
1623 int (*ndo_mdb_get)(struct net_device *dev,
1624 struct nlattr *tb[], u32 portid,
1626 struct netlink_ext_ack *extack);
1627 int (*ndo_bridge_setlink)(struct net_device *dev,
1628 struct nlmsghdr *nlh,
1630 struct netlink_ext_ack *extack);
1631 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1633 struct net_device *dev,
1636 int (*ndo_bridge_dellink)(struct net_device *dev,
1637 struct nlmsghdr *nlh,
1639 int (*ndo_change_carrier)(struct net_device *dev,
1641 int (*ndo_get_phys_port_id)(struct net_device *dev,
1642 struct netdev_phys_item_id *ppid);
1643 int (*ndo_get_port_parent_id)(struct net_device *dev,
1644 struct netdev_phys_item_id *ppid);
1645 int (*ndo_get_phys_port_name)(struct net_device *dev,
1646 char *name, size_t len);
1647 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1648 struct net_device *dev);
1649 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1652 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1655 int (*ndo_get_iflink)(const struct net_device *dev);
1656 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1657 struct sk_buff *skb);
1658 void (*ndo_set_rx_headroom)(struct net_device *dev,
1659 int needed_headroom);
1660 int (*ndo_bpf)(struct net_device *dev,
1661 struct netdev_bpf *bpf);
1662 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1663 struct xdp_frame **xdp,
1665 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1666 struct xdp_buff *xdp);
1667 int (*ndo_xsk_wakeup)(struct net_device *dev,
1668 u32 queue_id, u32 flags);
1669 int (*ndo_tunnel_ctl)(struct net_device *dev,
1670 struct ip_tunnel_parm *p, int cmd);
1671 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1672 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1673 struct net_device_path *path);
1674 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1675 const struct skb_shared_hwtstamps *hwtstamps,
1677 int (*ndo_hwtstamp_get)(struct net_device *dev,
1678 struct kernel_hwtstamp_config *kernel_config);
1679 int (*ndo_hwtstamp_set)(struct net_device *dev,
1680 struct kernel_hwtstamp_config *kernel_config,
1681 struct netlink_ext_ack *extack);
1685 * enum netdev_priv_flags - &struct net_device priv_flags
1687 * These are the &struct net_device, they are only set internally
1688 * by drivers and used in the kernel. These flags are invisible to
1689 * userspace; this means that the order of these flags can change
1690 * during any kernel release.
1692 * You should have a pretty good reason to be extending these flags.
1694 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1695 * @IFF_EBRIDGE: Ethernet bridging device
1696 * @IFF_BONDING: bonding master or slave
1697 * @IFF_ISATAP: ISATAP interface (RFC4214)
1698 * @IFF_WAN_HDLC: WAN HDLC device
1699 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1701 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1702 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1703 * @IFF_MACVLAN_PORT: device used as macvlan port
1704 * @IFF_BRIDGE_PORT: device used as bridge port
1705 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1706 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1707 * @IFF_UNICAST_FLT: Supports unicast filtering
1708 * @IFF_TEAM_PORT: device used as team port
1709 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1710 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1711 * change when it's running
1712 * @IFF_MACVLAN: Macvlan device
1713 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1714 * underlying stacked devices
1715 * @IFF_L3MDEV_MASTER: device is an L3 master device
1716 * @IFF_NO_QUEUE: device can run without qdisc attached
1717 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1718 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1719 * @IFF_TEAM: device is a team device
1720 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1721 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1722 * entity (i.e. the master device for bridged veth)
1723 * @IFF_MACSEC: device is a MACsec device
1724 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1725 * @IFF_FAILOVER: device is a failover master device
1726 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1727 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1728 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1729 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1730 * skb_headlen(skb) == 0 (data starts from frag0)
1731 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1732 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1733 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1734 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1736 enum netdev_priv_flags {
1737 IFF_802_1Q_VLAN = 1<<0,
1741 IFF_WAN_HDLC = 1<<4,
1742 IFF_XMIT_DST_RELEASE = 1<<5,
1743 IFF_DONT_BRIDGE = 1<<6,
1744 IFF_DISABLE_NETPOLL = 1<<7,
1745 IFF_MACVLAN_PORT = 1<<8,
1746 IFF_BRIDGE_PORT = 1<<9,
1747 IFF_OVS_DATAPATH = 1<<10,
1748 IFF_TX_SKB_SHARING = 1<<11,
1749 IFF_UNICAST_FLT = 1<<12,
1750 IFF_TEAM_PORT = 1<<13,
1751 IFF_SUPP_NOFCS = 1<<14,
1752 IFF_LIVE_ADDR_CHANGE = 1<<15,
1753 IFF_MACVLAN = 1<<16,
1754 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1755 IFF_L3MDEV_MASTER = 1<<18,
1756 IFF_NO_QUEUE = 1<<19,
1757 IFF_OPENVSWITCH = 1<<20,
1758 IFF_L3MDEV_SLAVE = 1<<21,
1760 IFF_RXFH_CONFIGURED = 1<<23,
1761 IFF_PHONY_HEADROOM = 1<<24,
1763 IFF_NO_RX_HANDLER = 1<<26,
1764 IFF_FAILOVER = 1<<27,
1765 IFF_FAILOVER_SLAVE = 1<<28,
1766 IFF_L3MDEV_RX_HANDLER = 1<<29,
1767 IFF_NO_ADDRCONF = BIT_ULL(30),
1768 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1769 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1770 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1773 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1774 #define IFF_EBRIDGE IFF_EBRIDGE
1775 #define IFF_BONDING IFF_BONDING
1776 #define IFF_ISATAP IFF_ISATAP
1777 #define IFF_WAN_HDLC IFF_WAN_HDLC
1778 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1779 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1780 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1781 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1782 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1783 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1784 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1785 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1786 #define IFF_TEAM_PORT IFF_TEAM_PORT
1787 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1788 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1789 #define IFF_MACVLAN IFF_MACVLAN
1790 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1791 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1792 #define IFF_NO_QUEUE IFF_NO_QUEUE
1793 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1794 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1795 #define IFF_TEAM IFF_TEAM
1796 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1797 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1798 #define IFF_MACSEC IFF_MACSEC
1799 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1800 #define IFF_FAILOVER IFF_FAILOVER
1801 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1802 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1803 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1805 /* Specifies the type of the struct net_device::ml_priv pointer */
1806 enum netdev_ml_priv_type {
1811 enum netdev_stat_type {
1812 NETDEV_PCPU_STAT_NONE,
1813 NETDEV_PCPU_STAT_LSTATS, /* struct pcpu_lstats */
1814 NETDEV_PCPU_STAT_TSTATS, /* struct pcpu_sw_netstats */
1815 NETDEV_PCPU_STAT_DSTATS, /* struct pcpu_dstats */
1818 enum netdev_reg_state {
1819 NETREG_UNINITIALIZED = 0,
1820 NETREG_REGISTERED, /* completed register_netdevice */
1821 NETREG_UNREGISTERING, /* called unregister_netdevice */
1822 NETREG_UNREGISTERED, /* completed unregister todo */
1823 NETREG_RELEASED, /* called free_netdev */
1824 NETREG_DUMMY, /* dummy device for NAPI poll */
1828 * struct net_device - The DEVICE structure.
1830 * Actually, this whole structure is a big mistake. It mixes I/O
1831 * data with strictly "high-level" data, and it has to know about
1832 * almost every data structure used in the INET module.
1834 * @name: This is the first field of the "visible" part of this structure
1835 * (i.e. as seen by users in the "Space.c" file). It is the name
1838 * @name_node: Name hashlist node
1839 * @ifalias: SNMP alias
1840 * @mem_end: Shared memory end
1841 * @mem_start: Shared memory start
1842 * @base_addr: Device I/O address
1843 * @irq: Device IRQ number
1845 * @state: Generic network queuing layer state, see netdev_state_t
1846 * @dev_list: The global list of network devices
1847 * @napi_list: List entry used for polling NAPI devices
1848 * @unreg_list: List entry when we are unregistering the
1849 * device; see the function unregister_netdev
1850 * @close_list: List entry used when we are closing the device
1851 * @ptype_all: Device-specific packet handlers for all protocols
1852 * @ptype_specific: Device-specific, protocol-specific packet handlers
1854 * @adj_list: Directly linked devices, like slaves for bonding
1855 * @features: Currently active device features
1856 * @hw_features: User-changeable features
1858 * @wanted_features: User-requested features
1859 * @vlan_features: Mask of features inheritable by VLAN devices
1861 * @hw_enc_features: Mask of features inherited by encapsulating devices
1862 * This field indicates what encapsulation
1863 * offloads the hardware is capable of doing,
1864 * and drivers will need to set them appropriately.
1866 * @mpls_features: Mask of features inheritable by MPLS
1867 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1869 * @ifindex: interface index
1870 * @group: The group the device belongs to
1872 * @stats: Statistics struct, which was left as a legacy, use
1873 * rtnl_link_stats64 instead
1875 * @core_stats: core networking counters,
1876 * do not use this in drivers
1877 * @carrier_up_count: Number of times the carrier has been up
1878 * @carrier_down_count: Number of times the carrier has been down
1880 * @wireless_handlers: List of functions to handle Wireless Extensions,
1882 * see <net/iw_handler.h> for details.
1883 * @wireless_data: Instance data managed by the core of wireless extensions
1885 * @netdev_ops: Includes several pointers to callbacks,
1886 * if one wants to override the ndo_*() functions
1887 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1888 * @xsk_tx_metadata_ops: Includes pointers to AF_XDP TX metadata callbacks.
1889 * @ethtool_ops: Management operations
1890 * @l3mdev_ops: Layer 3 master device operations
1891 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1892 * discovery handling. Necessary for e.g. 6LoWPAN.
1893 * @xfrmdev_ops: Transformation offload operations
1894 * @tlsdev_ops: Transport Layer Security offload operations
1895 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1896 * of Layer 2 headers.
1898 * @flags: Interface flags (a la BSD)
1899 * @xdp_features: XDP capability supported by the device
1900 * @priv_flags: Like 'flags' but invisible to userspace,
1901 * see if.h for the definitions
1902 * @gflags: Global flags ( kept as legacy )
1903 * @padded: How much padding added by alloc_netdev()
1904 * @operstate: RFC2863 operstate
1905 * @link_mode: Mapping policy to operstate
1906 * @if_port: Selectable AUI, TP, ...
1908 * @mtu: Interface MTU value
1909 * @min_mtu: Interface Minimum MTU value
1910 * @max_mtu: Interface Maximum MTU value
1911 * @type: Interface hardware type
1912 * @hard_header_len: Maximum hardware header length.
1913 * @min_header_len: Minimum hardware header length
1915 * @needed_headroom: Extra headroom the hardware may need, but not in all
1916 * cases can this be guaranteed
1917 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1918 * cases can this be guaranteed. Some cases also use
1919 * LL_MAX_HEADER instead to allocate the skb
1921 * interface address info:
1923 * @perm_addr: Permanent hw address
1924 * @addr_assign_type: Hw address assignment type
1925 * @addr_len: Hardware address length
1926 * @upper_level: Maximum depth level of upper devices.
1927 * @lower_level: Maximum depth level of lower devices.
1928 * @neigh_priv_len: Used in neigh_alloc()
1929 * @dev_id: Used to differentiate devices that share
1930 * the same link layer address
1931 * @dev_port: Used to differentiate devices that share
1933 * @addr_list_lock: XXX: need comments on this one
1934 * @name_assign_type: network interface name assignment type
1935 * @uc_promisc: Counter that indicates promiscuous mode
1936 * has been enabled due to the need to listen to
1937 * additional unicast addresses in a device that
1938 * does not implement ndo_set_rx_mode()
1939 * @uc: unicast mac addresses
1940 * @mc: multicast mac addresses
1941 * @dev_addrs: list of device hw addresses
1942 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1943 * @promiscuity: Number of times the NIC is told to work in
1944 * promiscuous mode; if it becomes 0 the NIC will
1945 * exit promiscuous mode
1946 * @allmulti: Counter, enables or disables allmulticast mode
1948 * @vlan_info: VLAN info
1949 * @dsa_ptr: dsa specific data
1950 * @tipc_ptr: TIPC specific data
1951 * @atalk_ptr: AppleTalk link
1952 * @ip_ptr: IPv4 specific data
1953 * @ip6_ptr: IPv6 specific data
1954 * @ax25_ptr: AX.25 specific data
1955 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1956 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1958 * @mpls_ptr: mpls_dev struct pointer
1959 * @mctp_ptr: MCTP specific data
1961 * @dev_addr: Hw address (before bcast,
1962 * because most packets are unicast)
1964 * @_rx: Array of RX queues
1965 * @num_rx_queues: Number of RX queues
1966 * allocated at register_netdev() time
1967 * @real_num_rx_queues: Number of RX queues currently active in device
1968 * @xdp_prog: XDP sockets filter program pointer
1969 * @gro_flush_timeout: timeout for GRO layer in NAPI
1970 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1971 * allow to avoid NIC hard IRQ, on busy queues.
1973 * @rx_handler: handler for received packets
1974 * @rx_handler_data: XXX: need comments on this one
1975 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1976 * @ingress_queue: XXX: need comments on this one
1977 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1978 * @broadcast: hw bcast address
1980 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1981 * indexed by RX queue number. Assigned by driver.
1982 * This must only be set if the ndo_rx_flow_steer
1983 * operation is defined
1984 * @index_hlist: Device index hash chain
1986 * @_tx: Array of TX queues
1987 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1988 * @real_num_tx_queues: Number of TX queues currently active in device
1989 * @qdisc: Root qdisc from userspace point of view
1990 * @tx_queue_len: Max frames per queue allowed
1991 * @tx_global_lock: XXX: need comments on this one
1992 * @xdp_bulkq: XDP device bulk queue
1993 * @xps_maps: all CPUs/RXQs maps for XPS device
1995 * @xps_maps: XXX: need comments on this one
1996 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1997 * @nf_hooks_egress: netfilter hooks executed for egress packets
1998 * @qdisc_hash: qdisc hash table
1999 * @watchdog_timeo: Represents the timeout that is used by
2000 * the watchdog (see dev_watchdog())
2001 * @watchdog_timer: List of timers
2003 * @proto_down_reason: reason a netdev interface is held down
2004 * @pcpu_refcnt: Number of references to this device
2005 * @dev_refcnt: Number of references to this device
2006 * @refcnt_tracker: Tracker directory for tracked references to this device
2007 * @todo_list: Delayed register/unregister
2008 * @link_watch_list: XXX: need comments on this one
2010 * @reg_state: Register/unregister state machine
2011 * @dismantle: Device is going to be freed
2012 * @rtnl_link_state: This enum represents the phases of creating
2015 * @needs_free_netdev: Should unregister perform free_netdev?
2016 * @priv_destructor: Called from unregister
2017 * @npinfo: XXX: need comments on this one
2018 * @nd_net: Network namespace this network device is inside
2020 * @ml_priv: Mid-layer private
2021 * @ml_priv_type: Mid-layer private type
2023 * @pcpu_stat_type: Type of device statistics which the core should
2024 * allocate/free: none, lstats, tstats, dstats. none
2025 * means the driver is handling statistics allocation/
2026 * freeing internally.
2027 * @lstats: Loopback statistics: packets, bytes
2028 * @tstats: Tunnel statistics: RX/TX packets, RX/TX bytes
2029 * @dstats: Dummy statistics: RX/TX/drop packets, RX/TX bytes
2034 * @dm_private: Drop monitor private
2036 * @dev: Class/net/name entry
2037 * @sysfs_groups: Space for optional device, statistics and wireless
2040 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2041 * @rtnl_link_ops: Rtnl_link_ops
2043 * @gso_max_size: Maximum size of generic segmentation offload
2044 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2045 * @gso_max_segs: Maximum number of segments that can be passed to the
2047 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2048 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2051 * @dcbnl_ops: Data Center Bridging netlink ops
2052 * @num_tc: Number of traffic classes in the net device
2053 * @tc_to_txq: XXX: need comments on this one
2054 * @prio_tc_map: XXX: need comments on this one
2056 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2058 * @priomap: XXX: need comments on this one
2059 * @phydev: Physical device may attach itself
2060 * for hardware timestamping
2061 * @sfp_bus: attached &struct sfp_bus structure.
2063 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2065 * @proto_down: protocol port state information can be sent to the
2066 * switch driver and used to set the phys state of the
2069 * @wol_enabled: Wake-on-LAN is enabled
2071 * @threaded: napi threaded mode is enabled
2073 * @net_notifier_list: List of per-net netdev notifier block
2074 * that follow this device when it is moved
2075 * to another network namespace.
2077 * @macsec_ops: MACsec offloading ops
2079 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2080 * offload capabilities of the device
2081 * @udp_tunnel_nic: UDP tunnel offload state
2082 * @xdp_state: stores info on attached XDP BPF programs
2084 * @nested_level: Used as a parameter of spin_lock_nested() of
2085 * dev->addr_list_lock.
2086 * @unlink_list: As netif_addr_lock() can be called recursively,
2087 * keep a list of interfaces to be deleted.
2088 * @gro_max_size: Maximum size of aggregated packet in generic
2089 * receive offload (GRO)
2090 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2091 * receive offload (GRO), for IPv4.
2092 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2095 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2096 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2097 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2098 * @dev_registered_tracker: tracker for reference held while
2100 * @offload_xstats_l3: L3 HW stats for this netdevice.
2102 * @devlink_port: Pointer to related devlink port structure.
2103 * Assigned by a driver before netdev registration using
2104 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2105 * during the time netdevice is registered.
2107 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2108 * where the clock is recovered.
2110 * FIXME: cleanup struct net_device such that network protocol info
2115 /* Cacheline organization can be found documented in
2116 * Documentation/networking/net_cachelines/net_device.rst.
2117 * Please update the document when adding new fields.
2120 /* TX read-mostly hotpath */
2121 __cacheline_group_begin(net_device_read_tx);
2122 unsigned long long priv_flags;
2123 const struct net_device_ops *netdev_ops;
2124 const struct header_ops *header_ops;
2125 struct netdev_queue *_tx;
2126 netdev_features_t gso_partial_features;
2127 unsigned int real_num_tx_queues;
2128 unsigned int gso_max_size;
2129 unsigned int gso_ipv4_max_size;
2132 /* Note : dev->mtu is often read without holding a lock.
2133 * Writers usually hold RTNL.
2134 * It is recommended to use READ_ONCE() to annotate the reads,
2135 * and to use WRITE_ONCE() to annotate the writes.
2138 unsigned short needed_headroom;
2139 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2141 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2143 #ifdef CONFIG_NETFILTER_EGRESS
2144 struct nf_hook_entries __rcu *nf_hooks_egress;
2146 #ifdef CONFIG_NET_XGRESS
2147 struct bpf_mprog_entry __rcu *tcx_egress;
2149 __cacheline_group_end(net_device_read_tx);
2151 /* TXRX read-mostly hotpath */
2152 __cacheline_group_begin(net_device_read_txrx);
2154 struct pcpu_lstats __percpu *lstats;
2155 struct pcpu_sw_netstats __percpu *tstats;
2156 struct pcpu_dstats __percpu *dstats;
2159 unsigned short hard_header_len;
2160 netdev_features_t features;
2161 struct inet6_dev __rcu *ip6_ptr;
2162 __cacheline_group_end(net_device_read_txrx);
2164 /* RX read-mostly hotpath */
2165 __cacheline_group_begin(net_device_read_rx);
2166 struct bpf_prog __rcu *xdp_prog;
2167 struct list_head ptype_specific;
2169 unsigned int real_num_rx_queues;
2170 struct netdev_rx_queue *_rx;
2171 unsigned long gro_flush_timeout;
2172 int napi_defer_hard_irqs;
2173 unsigned int gro_max_size;
2174 unsigned int gro_ipv4_max_size;
2175 rx_handler_func_t __rcu *rx_handler;
2176 void __rcu *rx_handler_data;
2177 possible_net_t nd_net;
2178 #ifdef CONFIG_NETPOLL
2179 struct netpoll_info __rcu *npinfo;
2181 #ifdef CONFIG_NET_XGRESS
2182 struct bpf_mprog_entry __rcu *tcx_ingress;
2184 __cacheline_group_end(net_device_read_rx);
2186 char name[IFNAMSIZ];
2187 struct netdev_name_node *name_node;
2188 struct dev_ifalias __rcu *ifalias;
2190 * I/O specific fields
2191 * FIXME: Merge these and struct ifmap into one
2193 unsigned long mem_end;
2194 unsigned long mem_start;
2195 unsigned long base_addr;
2198 * Some hardware also needs these fields (state,dev_list,
2199 * napi_list,unreg_list,close_list) but they are not
2200 * part of the usual set specified in Space.c.
2203 unsigned long state;
2205 struct list_head dev_list;
2206 struct list_head napi_list;
2207 struct list_head unreg_list;
2208 struct list_head close_list;
2209 struct list_head ptype_all;
2212 struct list_head upper;
2213 struct list_head lower;
2216 /* Read-mostly cache-line for fast-path access */
2217 xdp_features_t xdp_features;
2218 const struct xdp_metadata_ops *xdp_metadata_ops;
2219 const struct xsk_tx_metadata_ops *xsk_tx_metadata_ops;
2220 unsigned short gflags;
2222 unsigned short needed_tailroom;
2224 netdev_features_t hw_features;
2225 netdev_features_t wanted_features;
2226 netdev_features_t vlan_features;
2227 netdev_features_t hw_enc_features;
2228 netdev_features_t mpls_features;
2230 unsigned int min_mtu;
2231 unsigned int max_mtu;
2232 unsigned short type;
2233 unsigned char min_header_len;
2234 unsigned char name_assign_type;
2238 struct net_device_stats stats; /* not used by modern drivers */
2240 struct net_device_core_stats __percpu *core_stats;
2242 /* Stats to monitor link on/off, flapping */
2243 atomic_t carrier_up_count;
2244 atomic_t carrier_down_count;
2246 #ifdef CONFIG_WIRELESS_EXT
2247 const struct iw_handler_def *wireless_handlers;
2248 struct iw_public_data *wireless_data;
2250 const struct ethtool_ops *ethtool_ops;
2251 #ifdef CONFIG_NET_L3_MASTER_DEV
2252 const struct l3mdev_ops *l3mdev_ops;
2254 #if IS_ENABLED(CONFIG_IPV6)
2255 const struct ndisc_ops *ndisc_ops;
2258 #ifdef CONFIG_XFRM_OFFLOAD
2259 const struct xfrmdev_ops *xfrmdev_ops;
2262 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2263 const struct tlsdev_ops *tlsdev_ops;
2266 unsigned int operstate;
2267 unsigned char link_mode;
2269 unsigned char if_port;
2272 /* Interface address info. */
2273 unsigned char perm_addr[MAX_ADDR_LEN];
2274 unsigned char addr_assign_type;
2275 unsigned char addr_len;
2276 unsigned char upper_level;
2277 unsigned char lower_level;
2279 unsigned short neigh_priv_len;
2280 unsigned short dev_id;
2281 unsigned short dev_port;
2282 unsigned short padded;
2284 spinlock_t addr_list_lock;
2287 struct netdev_hw_addr_list uc;
2288 struct netdev_hw_addr_list mc;
2289 struct netdev_hw_addr_list dev_addrs;
2292 struct kset *queues_kset;
2294 #ifdef CONFIG_LOCKDEP
2295 struct list_head unlink_list;
2297 unsigned int promiscuity;
2298 unsigned int allmulti;
2300 #ifdef CONFIG_LOCKDEP
2301 unsigned char nested_level;
2305 /* Protocol-specific pointers */
2306 struct in_device __rcu *ip_ptr;
2307 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2308 struct vlan_info __rcu *vlan_info;
2310 #if IS_ENABLED(CONFIG_NET_DSA)
2311 struct dsa_port *dsa_ptr;
2313 #if IS_ENABLED(CONFIG_TIPC)
2314 struct tipc_bearer __rcu *tipc_ptr;
2316 #if IS_ENABLED(CONFIG_ATALK)
2319 #if IS_ENABLED(CONFIG_AX25)
2322 #if IS_ENABLED(CONFIG_CFG80211)
2323 struct wireless_dev *ieee80211_ptr;
2325 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2326 struct wpan_dev *ieee802154_ptr;
2328 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2329 struct mpls_dev __rcu *mpls_ptr;
2331 #if IS_ENABLED(CONFIG_MCTP)
2332 struct mctp_dev __rcu *mctp_ptr;
2336 * Cache lines mostly used on receive path (including eth_type_trans())
2338 /* Interface address info used in eth_type_trans() */
2339 const unsigned char *dev_addr;
2341 unsigned int num_rx_queues;
2342 #define GRO_LEGACY_MAX_SIZE 65536u
2343 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2344 * and shinfo->gso_segs is a 16bit field.
2346 #define GRO_MAX_SIZE (8 * 65535u)
2347 unsigned int xdp_zc_max_segs;
2348 struct netdev_queue __rcu *ingress_queue;
2349 #ifdef CONFIG_NETFILTER_INGRESS
2350 struct nf_hook_entries __rcu *nf_hooks_ingress;
2353 unsigned char broadcast[MAX_ADDR_LEN];
2354 #ifdef CONFIG_RFS_ACCEL
2355 struct cpu_rmap *rx_cpu_rmap;
2357 struct hlist_node index_hlist;
2360 * Cache lines mostly used on transmit path
2362 unsigned int num_tx_queues;
2363 struct Qdisc __rcu *qdisc;
2364 unsigned int tx_queue_len;
2365 spinlock_t tx_global_lock;
2367 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2369 #ifdef CONFIG_NET_SCHED
2370 DECLARE_HASHTABLE (qdisc_hash, 4);
2372 /* These may be needed for future network-power-down code. */
2373 struct timer_list watchdog_timer;
2376 u32 proto_down_reason;
2378 struct list_head todo_list;
2380 #ifdef CONFIG_PCPU_DEV_REFCNT
2381 int __percpu *pcpu_refcnt;
2383 refcount_t dev_refcnt;
2385 struct ref_tracker_dir refcnt_tracker;
2387 struct list_head link_watch_list;
2394 RTNL_LINK_INITIALIZED,
2395 RTNL_LINK_INITIALIZING,
2396 } rtnl_link_state:16;
2398 bool needs_free_netdev;
2399 void (*priv_destructor)(struct net_device *dev);
2401 /* mid-layer private */
2403 enum netdev_ml_priv_type ml_priv_type;
2405 enum netdev_stat_type pcpu_stat_type:8;
2407 #if IS_ENABLED(CONFIG_GARP)
2408 struct garp_port __rcu *garp_port;
2410 #if IS_ENABLED(CONFIG_MRP)
2411 struct mrp_port __rcu *mrp_port;
2413 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2414 struct dm_hw_stat_delta __rcu *dm_private;
2417 const struct attribute_group *sysfs_groups[4];
2418 const struct attribute_group *sysfs_rx_queue_group;
2420 const struct rtnl_link_ops *rtnl_link_ops;
2422 /* for setting kernel sock attribute on TCP connection setup */
2423 #define GSO_MAX_SEGS 65535u
2424 #define GSO_LEGACY_MAX_SIZE 65536u
2425 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2426 * and shinfo->gso_segs is a 16bit field.
2428 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2430 #define TSO_LEGACY_MAX_SIZE 65536
2431 #define TSO_MAX_SIZE UINT_MAX
2432 unsigned int tso_max_size;
2433 #define TSO_MAX_SEGS U16_MAX
2437 const struct dcbnl_rtnl_ops *dcbnl_ops;
2439 u8 prio_tc_map[TC_BITMASK + 1];
2441 #if IS_ENABLED(CONFIG_FCOE)
2442 unsigned int fcoe_ddp_xid;
2444 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2445 struct netprio_map __rcu *priomap;
2447 struct phy_device *phydev;
2448 struct sfp_bus *sfp_bus;
2449 struct lock_class_key *qdisc_tx_busylock;
2451 unsigned wol_enabled:1;
2452 unsigned threaded:1;
2454 struct list_head net_notifier_list;
2456 #if IS_ENABLED(CONFIG_MACSEC)
2457 /* MACsec management functions */
2458 const struct macsec_ops *macsec_ops;
2460 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2461 struct udp_tunnel_nic *udp_tunnel_nic;
2463 /* protected by rtnl_lock */
2464 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2466 u8 dev_addr_shadow[MAX_ADDR_LEN];
2467 netdevice_tracker linkwatch_dev_tracker;
2468 netdevice_tracker watchdog_dev_tracker;
2469 netdevice_tracker dev_registered_tracker;
2470 struct rtnl_hw_stats64 *offload_xstats_l3;
2472 struct devlink_port *devlink_port;
2474 #if IS_ENABLED(CONFIG_DPLL)
2475 struct dpll_pin *dpll_pin;
2477 #if IS_ENABLED(CONFIG_PAGE_POOL)
2478 /** @page_pools: page pools created for this netdevice */
2479 struct hlist_head page_pools;
2482 #define to_net_dev(d) container_of(d, struct net_device, dev)
2485 * Driver should use this to assign devlink port instance to a netdevice
2486 * before it registers the netdevice. Therefore devlink_port is static
2487 * during the netdev lifetime after it is registered.
2489 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2491 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2492 ((dev)->devlink_port = (port)); \
2495 static inline bool netif_elide_gro(const struct net_device *dev)
2497 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2502 #define NETDEV_ALIGN 32
2505 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2507 return dev->prio_tc_map[prio & TC_BITMASK];
2511 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2513 if (tc >= dev->num_tc)
2516 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2520 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2521 void netdev_reset_tc(struct net_device *dev);
2522 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2523 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2526 int netdev_get_num_tc(struct net_device *dev)
2531 static inline void net_prefetch(void *p)
2534 #if L1_CACHE_BYTES < 128
2535 prefetch((u8 *)p + L1_CACHE_BYTES);
2539 static inline void net_prefetchw(void *p)
2542 #if L1_CACHE_BYTES < 128
2543 prefetchw((u8 *)p + L1_CACHE_BYTES);
2547 void netdev_unbind_sb_channel(struct net_device *dev,
2548 struct net_device *sb_dev);
2549 int netdev_bind_sb_channel_queue(struct net_device *dev,
2550 struct net_device *sb_dev,
2551 u8 tc, u16 count, u16 offset);
2552 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2553 static inline int netdev_get_sb_channel(struct net_device *dev)
2555 return max_t(int, -dev->num_tc, 0);
2559 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2562 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2563 return &dev->_tx[index];
2566 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2567 const struct sk_buff *skb)
2569 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2572 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2573 void (*f)(struct net_device *,
2574 struct netdev_queue *,
2580 for (i = 0; i < dev->num_tx_queues; i++)
2581 f(dev, &dev->_tx[i], arg);
2584 #define netdev_lockdep_set_classes(dev) \
2586 static struct lock_class_key qdisc_tx_busylock_key; \
2587 static struct lock_class_key qdisc_xmit_lock_key; \
2588 static struct lock_class_key dev_addr_list_lock_key; \
2591 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2592 lockdep_set_class(&(dev)->addr_list_lock, \
2593 &dev_addr_list_lock_key); \
2594 for (i = 0; i < (dev)->num_tx_queues; i++) \
2595 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2596 &qdisc_xmit_lock_key); \
2599 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2600 struct net_device *sb_dev);
2601 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2602 struct sk_buff *skb,
2603 struct net_device *sb_dev);
2605 /* returns the headroom that the master device needs to take in account
2606 * when forwarding to this dev
2608 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2610 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2613 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2615 if (dev->netdev_ops->ndo_set_rx_headroom)
2616 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2619 /* set the device rx headroom to the dev's default */
2620 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2622 netdev_set_rx_headroom(dev, -1);
2625 static inline void *netdev_get_ml_priv(struct net_device *dev,
2626 enum netdev_ml_priv_type type)
2628 if (dev->ml_priv_type != type)
2631 return dev->ml_priv;
2634 static inline void netdev_set_ml_priv(struct net_device *dev,
2636 enum netdev_ml_priv_type type)
2638 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2639 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2640 dev->ml_priv_type, type);
2641 WARN(!dev->ml_priv_type && dev->ml_priv,
2642 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2644 dev->ml_priv = ml_priv;
2645 dev->ml_priv_type = type;
2649 * Net namespace inlines
2652 struct net *dev_net(const struct net_device *dev)
2654 return read_pnet(&dev->nd_net);
2658 void dev_net_set(struct net_device *dev, struct net *net)
2660 write_pnet(&dev->nd_net, net);
2664 * netdev_priv - access network device private data
2665 * @dev: network device
2667 * Get network device private data
2669 static inline void *netdev_priv(const struct net_device *dev)
2671 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2674 /* Set the sysfs physical device reference for the network logical device
2675 * if set prior to registration will cause a symlink during initialization.
2677 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2679 /* Set the sysfs device type for the network logical device to allow
2680 * fine-grained identification of different network device types. For
2681 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2683 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2685 void netif_queue_set_napi(struct net_device *dev, unsigned int queue_index,
2686 enum netdev_queue_type type,
2687 struct napi_struct *napi);
2689 static inline void netif_napi_set_irq(struct napi_struct *napi, int irq)
2694 /* Default NAPI poll() weight
2695 * Device drivers are strongly advised to not use bigger value
2697 #define NAPI_POLL_WEIGHT 64
2699 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2700 int (*poll)(struct napi_struct *, int), int weight);
2703 * netif_napi_add() - initialize a NAPI context
2704 * @dev: network device
2705 * @napi: NAPI context
2706 * @poll: polling function
2708 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2709 * *any* of the other NAPI-related functions.
2712 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2713 int (*poll)(struct napi_struct *, int))
2715 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2719 netif_napi_add_tx_weight(struct net_device *dev,
2720 struct napi_struct *napi,
2721 int (*poll)(struct napi_struct *, int),
2724 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2725 netif_napi_add_weight(dev, napi, poll, weight);
2729 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2730 * @dev: network device
2731 * @napi: NAPI context
2732 * @poll: polling function
2734 * This variant of netif_napi_add() should be used from drivers using NAPI
2735 * to exclusively poll a TX queue.
2736 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2738 static inline void netif_napi_add_tx(struct net_device *dev,
2739 struct napi_struct *napi,
2740 int (*poll)(struct napi_struct *, int))
2742 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2746 * __netif_napi_del - remove a NAPI context
2747 * @napi: NAPI context
2749 * Warning: caller must observe RCU grace period before freeing memory
2750 * containing @napi. Drivers might want to call this helper to combine
2751 * all the needed RCU grace periods into a single one.
2753 void __netif_napi_del(struct napi_struct *napi);
2756 * netif_napi_del - remove a NAPI context
2757 * @napi: NAPI context
2759 * netif_napi_del() removes a NAPI context from the network device NAPI list
2761 static inline void netif_napi_del(struct napi_struct *napi)
2763 __netif_napi_del(napi);
2767 struct packet_type {
2768 __be16 type; /* This is really htons(ether_type). */
2769 bool ignore_outgoing;
2770 struct net_device *dev; /* NULL is wildcarded here */
2771 netdevice_tracker dev_tracker;
2772 int (*func) (struct sk_buff *,
2773 struct net_device *,
2774 struct packet_type *,
2775 struct net_device *);
2776 void (*list_func) (struct list_head *,
2777 struct packet_type *,
2778 struct net_device *);
2779 bool (*id_match)(struct packet_type *ptype,
2781 struct net *af_packet_net;
2782 void *af_packet_priv;
2783 struct list_head list;
2786 struct offload_callbacks {
2787 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2788 netdev_features_t features);
2789 struct sk_buff *(*gro_receive)(struct list_head *head,
2790 struct sk_buff *skb);
2791 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2794 struct packet_offload {
2795 __be16 type; /* This is really htons(ether_type). */
2797 struct offload_callbacks callbacks;
2798 struct list_head list;
2801 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2802 struct pcpu_sw_netstats {
2803 u64_stats_t rx_packets;
2804 u64_stats_t rx_bytes;
2805 u64_stats_t tx_packets;
2806 u64_stats_t tx_bytes;
2807 struct u64_stats_sync syncp;
2808 } __aligned(4 * sizeof(u64));
2810 struct pcpu_dstats {
2817 struct u64_stats_sync syncp;
2818 } __aligned(8 * sizeof(u64));
2820 struct pcpu_lstats {
2821 u64_stats_t packets;
2823 struct u64_stats_sync syncp;
2824 } __aligned(2 * sizeof(u64));
2826 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2828 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2830 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2832 u64_stats_update_begin(&tstats->syncp);
2833 u64_stats_add(&tstats->rx_bytes, len);
2834 u64_stats_inc(&tstats->rx_packets);
2835 u64_stats_update_end(&tstats->syncp);
2838 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2839 unsigned int packets,
2842 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2844 u64_stats_update_begin(&tstats->syncp);
2845 u64_stats_add(&tstats->tx_bytes, len);
2846 u64_stats_add(&tstats->tx_packets, packets);
2847 u64_stats_update_end(&tstats->syncp);
2850 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2852 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2854 u64_stats_update_begin(&lstats->syncp);
2855 u64_stats_add(&lstats->bytes, len);
2856 u64_stats_inc(&lstats->packets);
2857 u64_stats_update_end(&lstats->syncp);
2860 #define __netdev_alloc_pcpu_stats(type, gfp) \
2862 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2865 for_each_possible_cpu(__cpu) { \
2866 typeof(type) *stat; \
2867 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2868 u64_stats_init(&stat->syncp); \
2874 #define netdev_alloc_pcpu_stats(type) \
2875 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2877 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2879 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2882 for_each_possible_cpu(__cpu) { \
2883 typeof(type) *stat; \
2884 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2885 u64_stats_init(&stat->syncp); \
2891 enum netdev_lag_tx_type {
2892 NETDEV_LAG_TX_TYPE_UNKNOWN,
2893 NETDEV_LAG_TX_TYPE_RANDOM,
2894 NETDEV_LAG_TX_TYPE_BROADCAST,
2895 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2896 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2897 NETDEV_LAG_TX_TYPE_HASH,
2900 enum netdev_lag_hash {
2901 NETDEV_LAG_HASH_NONE,
2903 NETDEV_LAG_HASH_L34,
2904 NETDEV_LAG_HASH_L23,
2905 NETDEV_LAG_HASH_E23,
2906 NETDEV_LAG_HASH_E34,
2907 NETDEV_LAG_HASH_VLAN_SRCMAC,
2908 NETDEV_LAG_HASH_UNKNOWN,
2911 struct netdev_lag_upper_info {
2912 enum netdev_lag_tx_type tx_type;
2913 enum netdev_lag_hash hash_type;
2916 struct netdev_lag_lower_state_info {
2921 #include <linux/notifier.h>
2923 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2924 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2928 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2930 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2931 detected a hardware crash and restarted
2932 - we can use this eg to kick tcp sessions
2934 NETDEV_CHANGE, /* Notify device state change */
2937 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2938 NETDEV_CHANGEADDR, /* notify after the address change */
2939 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2943 NETDEV_BONDING_FAILOVER,
2945 NETDEV_PRE_TYPE_CHANGE,
2946 NETDEV_POST_TYPE_CHANGE,
2950 NETDEV_NOTIFY_PEERS,
2954 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2955 NETDEV_CHANGEINFODATA,
2956 NETDEV_BONDING_INFO,
2957 NETDEV_PRECHANGEUPPER,
2958 NETDEV_CHANGELOWERSTATE,
2959 NETDEV_UDP_TUNNEL_PUSH_INFO,
2960 NETDEV_UDP_TUNNEL_DROP_INFO,
2961 NETDEV_CHANGE_TX_QUEUE_LEN,
2962 NETDEV_CVLAN_FILTER_PUSH_INFO,
2963 NETDEV_CVLAN_FILTER_DROP_INFO,
2964 NETDEV_SVLAN_FILTER_PUSH_INFO,
2965 NETDEV_SVLAN_FILTER_DROP_INFO,
2966 NETDEV_OFFLOAD_XSTATS_ENABLE,
2967 NETDEV_OFFLOAD_XSTATS_DISABLE,
2968 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2969 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2970 NETDEV_XDP_FEAT_CHANGE,
2972 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2974 int register_netdevice_notifier(struct notifier_block *nb);
2975 int unregister_netdevice_notifier(struct notifier_block *nb);
2976 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2977 int unregister_netdevice_notifier_net(struct net *net,
2978 struct notifier_block *nb);
2979 int register_netdevice_notifier_dev_net(struct net_device *dev,
2980 struct notifier_block *nb,
2981 struct netdev_net_notifier *nn);
2982 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2983 struct notifier_block *nb,
2984 struct netdev_net_notifier *nn);
2986 struct netdev_notifier_info {
2987 struct net_device *dev;
2988 struct netlink_ext_ack *extack;
2991 struct netdev_notifier_info_ext {
2992 struct netdev_notifier_info info; /* must be first */
2998 struct netdev_notifier_change_info {
2999 struct netdev_notifier_info info; /* must be first */
3000 unsigned int flags_changed;
3003 struct netdev_notifier_changeupper_info {
3004 struct netdev_notifier_info info; /* must be first */
3005 struct net_device *upper_dev; /* new upper dev */
3006 bool master; /* is upper dev master */
3007 bool linking; /* is the notification for link or unlink */
3008 void *upper_info; /* upper dev info */
3011 struct netdev_notifier_changelowerstate_info {
3012 struct netdev_notifier_info info; /* must be first */
3013 void *lower_state_info; /* is lower dev state */
3016 struct netdev_notifier_pre_changeaddr_info {
3017 struct netdev_notifier_info info; /* must be first */
3018 const unsigned char *dev_addr;
3021 enum netdev_offload_xstats_type {
3022 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
3025 struct netdev_notifier_offload_xstats_info {
3026 struct netdev_notifier_info info; /* must be first */
3027 enum netdev_offload_xstats_type type;
3030 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
3031 struct netdev_notifier_offload_xstats_rd *report_delta;
3032 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
3033 struct netdev_notifier_offload_xstats_ru *report_used;
3037 int netdev_offload_xstats_enable(struct net_device *dev,
3038 enum netdev_offload_xstats_type type,
3039 struct netlink_ext_ack *extack);
3040 int netdev_offload_xstats_disable(struct net_device *dev,
3041 enum netdev_offload_xstats_type type);
3042 bool netdev_offload_xstats_enabled(const struct net_device *dev,
3043 enum netdev_offload_xstats_type type);
3044 int netdev_offload_xstats_get(struct net_device *dev,
3045 enum netdev_offload_xstats_type type,
3046 struct rtnl_hw_stats64 *stats, bool *used,
3047 struct netlink_ext_ack *extack);
3049 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
3050 const struct rtnl_hw_stats64 *stats);
3052 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
3053 void netdev_offload_xstats_push_delta(struct net_device *dev,
3054 enum netdev_offload_xstats_type type,
3055 const struct rtnl_hw_stats64 *stats);
3057 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
3058 struct net_device *dev)
3061 info->extack = NULL;
3064 static inline struct net_device *
3065 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3070 static inline struct netlink_ext_ack *
3071 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3073 return info->extack;
3076 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3077 int call_netdevice_notifiers_info(unsigned long val,
3078 struct netdev_notifier_info *info);
3080 #define for_each_netdev(net, d) \
3081 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3082 #define for_each_netdev_reverse(net, d) \
3083 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3084 #define for_each_netdev_rcu(net, d) \
3085 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3086 #define for_each_netdev_safe(net, d, n) \
3087 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3088 #define for_each_netdev_continue(net, d) \
3089 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3090 #define for_each_netdev_continue_reverse(net, d) \
3091 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3093 #define for_each_netdev_continue_rcu(net, d) \
3094 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3095 #define for_each_netdev_in_bond_rcu(bond, slave) \
3096 for_each_netdev_rcu(&init_net, slave) \
3097 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3098 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3100 #define for_each_netdev_dump(net, d, ifindex) \
3101 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3103 static inline struct net_device *next_net_device(struct net_device *dev)
3105 struct list_head *lh;
3109 lh = dev->dev_list.next;
3110 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3113 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3115 struct list_head *lh;
3119 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3120 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3123 static inline struct net_device *first_net_device(struct net *net)
3125 return list_empty(&net->dev_base_head) ? NULL :
3126 net_device_entry(net->dev_base_head.next);
3129 static inline struct net_device *first_net_device_rcu(struct net *net)
3131 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3133 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3136 int netdev_boot_setup_check(struct net_device *dev);
3137 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3138 const char *hwaddr);
3139 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3140 void dev_add_pack(struct packet_type *pt);
3141 void dev_remove_pack(struct packet_type *pt);
3142 void __dev_remove_pack(struct packet_type *pt);
3143 void dev_add_offload(struct packet_offload *po);
3144 void dev_remove_offload(struct packet_offload *po);
3146 int dev_get_iflink(const struct net_device *dev);
3147 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3148 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3149 struct net_device_path_stack *stack);
3150 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3151 unsigned short mask);
3152 struct net_device *dev_get_by_name(struct net *net, const char *name);
3153 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3154 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3155 bool netdev_name_in_use(struct net *net, const char *name);
3156 int dev_alloc_name(struct net_device *dev, const char *name);
3157 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3158 void dev_close(struct net_device *dev);
3159 void dev_close_many(struct list_head *head, bool unlink);
3160 void dev_disable_lro(struct net_device *dev);
3161 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3162 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3163 struct net_device *sb_dev);
3164 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3165 struct net_device *sb_dev);
3167 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3168 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3170 static inline int dev_queue_xmit(struct sk_buff *skb)
3172 return __dev_queue_xmit(skb, NULL);
3175 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3176 struct net_device *sb_dev)
3178 return __dev_queue_xmit(skb, sb_dev);
3181 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3185 ret = __dev_direct_xmit(skb, queue_id);
3186 if (!dev_xmit_complete(ret))
3191 int register_netdevice(struct net_device *dev);
3192 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3193 void unregister_netdevice_many(struct list_head *head);
3194 static inline void unregister_netdevice(struct net_device *dev)
3196 unregister_netdevice_queue(dev, NULL);
3199 int netdev_refcnt_read(const struct net_device *dev);
3200 void free_netdev(struct net_device *dev);
3201 void netdev_freemem(struct net_device *dev);
3202 void init_dummy_netdev(struct net_device *dev);
3204 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3205 struct sk_buff *skb,
3207 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3209 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3210 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3211 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3212 netdevice_tracker *tracker, gfp_t gfp);
3213 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3214 netdevice_tracker *tracker, gfp_t gfp);
3215 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3216 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3218 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3219 unsigned short type,
3220 const void *daddr, const void *saddr,
3223 if (!dev->header_ops || !dev->header_ops->create)
3226 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3229 static inline int dev_parse_header(const struct sk_buff *skb,
3230 unsigned char *haddr)
3232 const struct net_device *dev = skb->dev;
3234 if (!dev->header_ops || !dev->header_ops->parse)
3236 return dev->header_ops->parse(skb, haddr);
3239 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3241 const struct net_device *dev = skb->dev;
3243 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3245 return dev->header_ops->parse_protocol(skb);
3248 /* ll_header must have at least hard_header_len allocated */
3249 static inline bool dev_validate_header(const struct net_device *dev,
3250 char *ll_header, int len)
3252 if (likely(len >= dev->hard_header_len))
3254 if (len < dev->min_header_len)
3257 if (capable(CAP_SYS_RAWIO)) {
3258 memset(ll_header + len, 0, dev->hard_header_len - len);
3262 if (dev->header_ops && dev->header_ops->validate)
3263 return dev->header_ops->validate(ll_header, len);
3268 static inline bool dev_has_header(const struct net_device *dev)
3270 return dev->header_ops && dev->header_ops->create;
3274 * Incoming packets are placed on per-CPU queues
3276 struct softnet_data {
3277 struct list_head poll_list;
3278 struct sk_buff_head process_queue;
3281 unsigned int processed;
3282 unsigned int time_squeeze;
3284 struct softnet_data *rps_ipi_list;
3287 bool in_net_rx_action;
3288 bool in_napi_threaded_poll;
3290 #ifdef CONFIG_NET_FLOW_LIMIT
3291 struct sd_flow_limit __rcu *flow_limit;
3293 struct Qdisc *output_queue;
3294 struct Qdisc **output_queue_tailp;
3295 struct sk_buff *completion_queue;
3296 #ifdef CONFIG_XFRM_OFFLOAD
3297 struct sk_buff_head xfrm_backlog;
3299 /* written and read only by owning cpu: */
3303 #ifdef CONFIG_NET_EGRESS
3308 /* input_queue_head should be written by cpu owning this struct,
3309 * and only read by other cpus. Worth using a cache line.
3311 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3313 /* Elements below can be accessed between CPUs for RPS/RFS */
3314 call_single_data_t csd ____cacheline_aligned_in_smp;
3315 struct softnet_data *rps_ipi_next;
3317 unsigned int input_queue_tail;
3319 unsigned int received_rps;
3320 unsigned int dropped;
3321 struct sk_buff_head input_pkt_queue;
3322 struct napi_struct backlog;
3324 /* Another possibly contended cache line */
3325 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3327 int defer_ipi_scheduled;
3328 struct sk_buff *defer_list;
3329 call_single_data_t defer_csd;
3332 static inline void input_queue_head_incr(struct softnet_data *sd)
3335 sd->input_queue_head++;
3339 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3340 unsigned int *qtail)
3343 *qtail = ++sd->input_queue_tail;
3347 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3349 static inline int dev_recursion_level(void)
3351 return this_cpu_read(softnet_data.xmit.recursion);
3354 #define XMIT_RECURSION_LIMIT 8
3355 static inline bool dev_xmit_recursion(void)
3357 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3358 XMIT_RECURSION_LIMIT);
3361 static inline void dev_xmit_recursion_inc(void)
3363 __this_cpu_inc(softnet_data.xmit.recursion);
3366 static inline void dev_xmit_recursion_dec(void)
3368 __this_cpu_dec(softnet_data.xmit.recursion);
3371 void __netif_schedule(struct Qdisc *q);
3372 void netif_schedule_queue(struct netdev_queue *txq);
3374 static inline void netif_tx_schedule_all(struct net_device *dev)
3378 for (i = 0; i < dev->num_tx_queues; i++)
3379 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3382 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3384 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3388 * netif_start_queue - allow transmit
3389 * @dev: network device
3391 * Allow upper layers to call the device hard_start_xmit routine.
3393 static inline void netif_start_queue(struct net_device *dev)
3395 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3398 static inline void netif_tx_start_all_queues(struct net_device *dev)
3402 for (i = 0; i < dev->num_tx_queues; i++) {
3403 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3404 netif_tx_start_queue(txq);
3408 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3411 * netif_wake_queue - restart transmit
3412 * @dev: network device
3414 * Allow upper layers to call the device hard_start_xmit routine.
3415 * Used for flow control when transmit resources are available.
3417 static inline void netif_wake_queue(struct net_device *dev)
3419 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3422 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3426 for (i = 0; i < dev->num_tx_queues; i++) {
3427 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3428 netif_tx_wake_queue(txq);
3432 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3434 /* Must be an atomic op see netif_txq_try_stop() */
3435 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3439 * netif_stop_queue - stop transmitted packets
3440 * @dev: network device
3442 * Stop upper layers calling the device hard_start_xmit routine.
3443 * Used for flow control when transmit resources are unavailable.
3445 static inline void netif_stop_queue(struct net_device *dev)
3447 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3450 void netif_tx_stop_all_queues(struct net_device *dev);
3452 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3454 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3458 * netif_queue_stopped - test if transmit queue is flowblocked
3459 * @dev: network device
3461 * Test if transmit queue on device is currently unable to send.
3463 static inline bool netif_queue_stopped(const struct net_device *dev)
3465 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3468 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3470 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3474 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3476 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3480 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3482 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3486 * netdev_queue_set_dql_min_limit - set dql minimum limit
3487 * @dev_queue: pointer to transmit queue
3488 * @min_limit: dql minimum limit
3490 * Forces xmit_more() to return true until the minimum threshold
3491 * defined by @min_limit is reached (or until the tx queue is
3492 * empty). Warning: to be use with care, misuse will impact the
3495 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3496 unsigned int min_limit)
3499 dev_queue->dql.min_limit = min_limit;
3504 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3505 * @dev_queue: pointer to transmit queue
3507 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3508 * to give appropriate hint to the CPU.
3510 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3513 prefetchw(&dev_queue->dql.num_queued);
3518 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3519 * @dev_queue: pointer to transmit queue
3521 * BQL enabled drivers might use this helper in their TX completion path,
3522 * to give appropriate hint to the CPU.
3524 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3527 prefetchw(&dev_queue->dql.limit);
3532 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3533 * @dev_queue: network device queue
3534 * @bytes: number of bytes queued to the device queue
3536 * Report the number of bytes queued for sending/completion to the network
3537 * device hardware queue. @bytes should be a good approximation and should
3538 * exactly match netdev_completed_queue() @bytes.
3539 * This is typically called once per packet, from ndo_start_xmit().
3541 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3545 dql_queued(&dev_queue->dql, bytes);
3547 if (likely(dql_avail(&dev_queue->dql) >= 0))
3550 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3553 * The XOFF flag must be set before checking the dql_avail below,
3554 * because in netdev_tx_completed_queue we update the dql_completed
3555 * before checking the XOFF flag.
3559 /* check again in case another CPU has just made room avail */
3560 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3561 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3565 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3566 * that they should not test BQL status themselves.
3567 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3569 * Returns true if the doorbell must be used to kick the NIC.
3571 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3577 dql_queued(&dev_queue->dql, bytes);
3579 return netif_tx_queue_stopped(dev_queue);
3581 netdev_tx_sent_queue(dev_queue, bytes);
3586 * netdev_sent_queue - report the number of bytes queued to hardware
3587 * @dev: network device
3588 * @bytes: number of bytes queued to the hardware device queue
3590 * Report the number of bytes queued for sending/completion to the network
3591 * device hardware queue#0. @bytes should be a good approximation and should
3592 * exactly match netdev_completed_queue() @bytes.
3593 * This is typically called once per packet, from ndo_start_xmit().
3595 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3597 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3600 static inline bool __netdev_sent_queue(struct net_device *dev,
3604 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3609 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3610 * @dev_queue: network device queue
3611 * @pkts: number of packets (currently ignored)
3612 * @bytes: number of bytes dequeued from the device queue
3614 * Must be called at most once per TX completion round (and not per
3615 * individual packet), so that BQL can adjust its limits appropriately.
3617 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3618 unsigned int pkts, unsigned int bytes)
3621 if (unlikely(!bytes))
3624 dql_completed(&dev_queue->dql, bytes);
3627 * Without the memory barrier there is a small possiblity that
3628 * netdev_tx_sent_queue will miss the update and cause the queue to
3629 * be stopped forever
3631 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3633 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3636 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3637 netif_schedule_queue(dev_queue);
3642 * netdev_completed_queue - report bytes and packets completed by device
3643 * @dev: network device
3644 * @pkts: actual number of packets sent over the medium
3645 * @bytes: actual number of bytes sent over the medium
3647 * Report the number of bytes and packets transmitted by the network device
3648 * hardware queue over the physical medium, @bytes must exactly match the
3649 * @bytes amount passed to netdev_sent_queue()
3651 static inline void netdev_completed_queue(struct net_device *dev,
3652 unsigned int pkts, unsigned int bytes)
3654 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3657 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3660 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3666 * netdev_reset_queue - reset the packets and bytes count of a network device
3667 * @dev_queue: network device
3669 * Reset the bytes and packet count of a network device and clear the
3670 * software flow control OFF bit for this network device
3672 static inline void netdev_reset_queue(struct net_device *dev_queue)
3674 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3678 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3679 * @dev: network device
3680 * @queue_index: given tx queue index
3682 * Returns 0 if given tx queue index >= number of device tx queues,
3683 * otherwise returns the originally passed tx queue index.
3685 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3687 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3688 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3689 dev->name, queue_index,
3690 dev->real_num_tx_queues);
3698 * netif_running - test if up
3699 * @dev: network device
3701 * Test if the device has been brought up.
3703 static inline bool netif_running(const struct net_device *dev)
3705 return test_bit(__LINK_STATE_START, &dev->state);
3709 * Routines to manage the subqueues on a device. We only need start,
3710 * stop, and a check if it's stopped. All other device management is
3711 * done at the overall netdevice level.
3712 * Also test the device if we're multiqueue.
3716 * netif_start_subqueue - allow sending packets on subqueue
3717 * @dev: network device
3718 * @queue_index: sub queue index
3720 * Start individual transmit queue of a device with multiple transmit queues.
3722 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3724 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3726 netif_tx_start_queue(txq);
3730 * netif_stop_subqueue - stop sending packets on subqueue
3731 * @dev: network device
3732 * @queue_index: sub queue index
3734 * Stop individual transmit queue of a device with multiple transmit queues.
3736 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3738 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3739 netif_tx_stop_queue(txq);
3743 * __netif_subqueue_stopped - test status of subqueue
3744 * @dev: network device
3745 * @queue_index: sub queue index
3747 * Check individual transmit queue of a device with multiple transmit queues.
3749 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3752 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3754 return netif_tx_queue_stopped(txq);
3758 * netif_subqueue_stopped - test status of subqueue
3759 * @dev: network device
3760 * @skb: sub queue buffer pointer
3762 * Check individual transmit queue of a device with multiple transmit queues.
3764 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3765 struct sk_buff *skb)
3767 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3771 * netif_wake_subqueue - allow sending packets on subqueue
3772 * @dev: network device
3773 * @queue_index: sub queue index
3775 * Resume individual transmit queue of a device with multiple transmit queues.
3777 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3779 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3781 netif_tx_wake_queue(txq);
3785 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3787 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3788 u16 index, enum xps_map_type type);
3791 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3792 * @j: CPU/Rx queue index
3793 * @mask: bitmask of all cpus/rx queues
3794 * @nr_bits: number of bits in the bitmask
3796 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3798 static inline bool netif_attr_test_mask(unsigned long j,
3799 const unsigned long *mask,
3800 unsigned int nr_bits)
3802 cpu_max_bits_warn(j, nr_bits);
3803 return test_bit(j, mask);
3807 * netif_attr_test_online - Test for online CPU/Rx queue
3808 * @j: CPU/Rx queue index
3809 * @online_mask: bitmask for CPUs/Rx queues that are online
3810 * @nr_bits: number of bits in the bitmask
3812 * Returns true if a CPU/Rx queue is online.
3814 static inline bool netif_attr_test_online(unsigned long j,
3815 const unsigned long *online_mask,
3816 unsigned int nr_bits)
3818 cpu_max_bits_warn(j, nr_bits);
3821 return test_bit(j, online_mask);
3823 return (j < nr_bits);
3827 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3828 * @n: CPU/Rx queue index
3829 * @srcp: the cpumask/Rx queue mask pointer
3830 * @nr_bits: number of bits in the bitmask
3832 * Returns >= nr_bits if no further CPUs/Rx queues set.
3834 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3835 unsigned int nr_bits)
3837 /* -1 is a legal arg here. */
3839 cpu_max_bits_warn(n, nr_bits);
3842 return find_next_bit(srcp, nr_bits, n + 1);
3848 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3849 * @n: CPU/Rx queue index
3850 * @src1p: the first CPUs/Rx queues mask pointer
3851 * @src2p: the second CPUs/Rx queues mask pointer
3852 * @nr_bits: number of bits in the bitmask
3854 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3856 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3857 const unsigned long *src2p,
3858 unsigned int nr_bits)
3860 /* -1 is a legal arg here. */
3862 cpu_max_bits_warn(n, nr_bits);
3865 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3867 return find_next_bit(src1p, nr_bits, n + 1);
3869 return find_next_bit(src2p, nr_bits, n + 1);
3874 static inline int netif_set_xps_queue(struct net_device *dev,
3875 const struct cpumask *mask,
3881 static inline int __netif_set_xps_queue(struct net_device *dev,
3882 const unsigned long *mask,
3883 u16 index, enum xps_map_type type)
3890 * netif_is_multiqueue - test if device has multiple transmit queues
3891 * @dev: network device
3893 * Check if device has multiple transmit queues
3895 static inline bool netif_is_multiqueue(const struct net_device *dev)
3897 return dev->num_tx_queues > 1;
3900 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3903 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3905 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3908 dev->real_num_rx_queues = rxqs;
3912 int netif_set_real_num_queues(struct net_device *dev,
3913 unsigned int txq, unsigned int rxq);
3915 int netif_get_num_default_rss_queues(void);
3917 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3918 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3921 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3922 * interrupt context or with hardware interrupts being disabled.
3923 * (in_hardirq() || irqs_disabled())
3925 * We provide four helpers that can be used in following contexts :
3927 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3928 * replacing kfree_skb(skb)
3930 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3931 * Typically used in place of consume_skb(skb) in TX completion path
3933 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3934 * replacing kfree_skb(skb)
3936 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3937 * and consumed a packet. Used in place of consume_skb(skb)
3939 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3941 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3944 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3946 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3949 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3951 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3954 static inline void dev_consume_skb_any(struct sk_buff *skb)
3956 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3959 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3960 struct bpf_prog *xdp_prog);
3961 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3962 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff **pskb);
3963 int netif_rx(struct sk_buff *skb);
3964 int __netif_rx(struct sk_buff *skb);
3966 int netif_receive_skb(struct sk_buff *skb);
3967 int netif_receive_skb_core(struct sk_buff *skb);
3968 void netif_receive_skb_list_internal(struct list_head *head);
3969 void netif_receive_skb_list(struct list_head *head);
3970 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3971 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3972 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3973 void napi_get_frags_check(struct napi_struct *napi);
3974 gro_result_t napi_gro_frags(struct napi_struct *napi);
3975 struct packet_offload *gro_find_receive_by_type(__be16 type);
3976 struct packet_offload *gro_find_complete_by_type(__be16 type);
3978 static inline void napi_free_frags(struct napi_struct *napi)
3980 kfree_skb(napi->skb);
3984 bool netdev_is_rx_handler_busy(struct net_device *dev);
3985 int netdev_rx_handler_register(struct net_device *dev,
3986 rx_handler_func_t *rx_handler,
3987 void *rx_handler_data);
3988 void netdev_rx_handler_unregister(struct net_device *dev);
3990 bool dev_valid_name(const char *name);
3991 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3993 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3995 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3996 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3997 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3998 void __user *data, bool *need_copyout);
3999 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4000 int generic_hwtstamp_get_lower(struct net_device *dev,
4001 struct kernel_hwtstamp_config *kernel_cfg);
4002 int generic_hwtstamp_set_lower(struct net_device *dev,
4003 struct kernel_hwtstamp_config *kernel_cfg,
4004 struct netlink_ext_ack *extack);
4005 int dev_set_hwtstamp_phylib(struct net_device *dev,
4006 struct kernel_hwtstamp_config *cfg,
4007 struct netlink_ext_ack *extack);
4008 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4009 unsigned int dev_get_flags(const struct net_device *);
4010 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4011 struct netlink_ext_ack *extack);
4012 int dev_change_flags(struct net_device *dev, unsigned int flags,
4013 struct netlink_ext_ack *extack);
4014 int dev_set_alias(struct net_device *, const char *, size_t);
4015 int dev_get_alias(const struct net_device *, char *, size_t);
4016 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4017 const char *pat, int new_ifindex);
4019 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4022 return __dev_change_net_namespace(dev, net, pat, 0);
4024 int __dev_set_mtu(struct net_device *, int);
4025 int dev_set_mtu(struct net_device *, int);
4026 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4027 struct netlink_ext_ack *extack);
4028 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4029 struct netlink_ext_ack *extack);
4030 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4031 struct netlink_ext_ack *extack);
4032 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4033 int dev_get_port_parent_id(struct net_device *dev,
4034 struct netdev_phys_item_id *ppid, bool recurse);
4035 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4036 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
4037 void netdev_dpll_pin_clear(struct net_device *dev);
4039 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
4041 #if IS_ENABLED(CONFIG_DPLL)
4042 return dev->dpll_pin;
4048 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4049 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4050 struct netdev_queue *txq, int *ret);
4052 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4053 u8 dev_xdp_prog_count(struct net_device *dev);
4054 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4056 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4057 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4058 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4059 bool is_skb_forwardable(const struct net_device *dev,
4060 const struct sk_buff *skb);
4062 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4063 const struct sk_buff *skb,
4064 const bool check_mtu)
4066 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4069 if (!(dev->flags & IFF_UP))
4075 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4076 if (skb->len <= len)
4079 /* if TSO is enabled, we don't care about the length as the packet
4080 * could be forwarded without being segmented before
4082 if (skb_is_gso(skb))
4088 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4090 #define DEV_CORE_STATS_INC(FIELD) \
4091 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4093 netdev_core_stats_inc(dev, \
4094 offsetof(struct net_device_core_stats, FIELD)); \
4096 DEV_CORE_STATS_INC(rx_dropped)
4097 DEV_CORE_STATS_INC(tx_dropped)
4098 DEV_CORE_STATS_INC(rx_nohandler)
4099 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4100 #undef DEV_CORE_STATS_INC
4102 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4103 struct sk_buff *skb,
4104 const bool check_mtu)
4106 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4107 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4108 dev_core_stats_rx_dropped_inc(dev);
4113 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4118 bool dev_nit_active(struct net_device *dev);
4119 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4121 static inline void __dev_put(struct net_device *dev)
4124 #ifdef CONFIG_PCPU_DEV_REFCNT
4125 this_cpu_dec(*dev->pcpu_refcnt);
4127 refcount_dec(&dev->dev_refcnt);
4132 static inline void __dev_hold(struct net_device *dev)
4135 #ifdef CONFIG_PCPU_DEV_REFCNT
4136 this_cpu_inc(*dev->pcpu_refcnt);
4138 refcount_inc(&dev->dev_refcnt);
4143 static inline void __netdev_tracker_alloc(struct net_device *dev,
4144 netdevice_tracker *tracker,
4147 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4148 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4152 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4153 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4155 static inline void netdev_tracker_alloc(struct net_device *dev,
4156 netdevice_tracker *tracker, gfp_t gfp)
4158 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4159 refcount_dec(&dev->refcnt_tracker.no_tracker);
4160 __netdev_tracker_alloc(dev, tracker, gfp);
4164 static inline void netdev_tracker_free(struct net_device *dev,
4165 netdevice_tracker *tracker)
4167 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4168 ref_tracker_free(&dev->refcnt_tracker, tracker);
4172 static inline void netdev_hold(struct net_device *dev,
4173 netdevice_tracker *tracker, gfp_t gfp)
4177 __netdev_tracker_alloc(dev, tracker, gfp);
4181 static inline void netdev_put(struct net_device *dev,
4182 netdevice_tracker *tracker)
4185 netdev_tracker_free(dev, tracker);
4191 * dev_hold - get reference to device
4192 * @dev: network device
4194 * Hold reference to device to keep it from being freed.
4195 * Try using netdev_hold() instead.
4197 static inline void dev_hold(struct net_device *dev)
4199 netdev_hold(dev, NULL, GFP_ATOMIC);
4203 * dev_put - release reference to device
4204 * @dev: network device
4206 * Release reference to device to allow it to be freed.
4207 * Try using netdev_put() instead.
4209 static inline void dev_put(struct net_device *dev)
4211 netdev_put(dev, NULL);
4214 static inline void netdev_ref_replace(struct net_device *odev,
4215 struct net_device *ndev,
4216 netdevice_tracker *tracker,
4220 netdev_tracker_free(odev, tracker);
4226 __netdev_tracker_alloc(ndev, tracker, gfp);
4229 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4230 * and _off may be called from IRQ context, but it is caller
4231 * who is responsible for serialization of these calls.
4233 * The name carrier is inappropriate, these functions should really be
4234 * called netif_lowerlayer_*() because they represent the state of any
4235 * kind of lower layer not just hardware media.
4237 void linkwatch_fire_event(struct net_device *dev);
4240 * linkwatch_sync_dev - sync linkwatch for the given device
4241 * @dev: network device to sync linkwatch for
4243 * Sync linkwatch for the given device, removing it from the
4244 * pending work list (if queued).
4246 void linkwatch_sync_dev(struct net_device *dev);
4249 * netif_carrier_ok - test if carrier present
4250 * @dev: network device
4252 * Check if carrier is present on device
4254 static inline bool netif_carrier_ok(const struct net_device *dev)
4256 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4259 unsigned long dev_trans_start(struct net_device *dev);
4261 void __netdev_watchdog_up(struct net_device *dev);
4263 void netif_carrier_on(struct net_device *dev);
4264 void netif_carrier_off(struct net_device *dev);
4265 void netif_carrier_event(struct net_device *dev);
4268 * netif_dormant_on - mark device as dormant.
4269 * @dev: network device
4271 * Mark device as dormant (as per RFC2863).
4273 * The dormant state indicates that the relevant interface is not
4274 * actually in a condition to pass packets (i.e., it is not 'up') but is
4275 * in a "pending" state, waiting for some external event. For "on-
4276 * demand" interfaces, this new state identifies the situation where the
4277 * interface is waiting for events to place it in the up state.
4279 static inline void netif_dormant_on(struct net_device *dev)
4281 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4282 linkwatch_fire_event(dev);
4286 * netif_dormant_off - set device as not dormant.
4287 * @dev: network device
4289 * Device is not in dormant state.
4291 static inline void netif_dormant_off(struct net_device *dev)
4293 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4294 linkwatch_fire_event(dev);
4298 * netif_dormant - test if device is dormant
4299 * @dev: network device
4301 * Check if device is dormant.
4303 static inline bool netif_dormant(const struct net_device *dev)
4305 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4310 * netif_testing_on - mark device as under test.
4311 * @dev: network device
4313 * Mark device as under test (as per RFC2863).
4315 * The testing state indicates that some test(s) must be performed on
4316 * the interface. After completion, of the test, the interface state
4317 * will change to up, dormant, or down, as appropriate.
4319 static inline void netif_testing_on(struct net_device *dev)
4321 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4322 linkwatch_fire_event(dev);
4326 * netif_testing_off - set device as not under test.
4327 * @dev: network device
4329 * Device is not in testing state.
4331 static inline void netif_testing_off(struct net_device *dev)
4333 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4334 linkwatch_fire_event(dev);
4338 * netif_testing - test if device is under test
4339 * @dev: network device
4341 * Check if device is under test
4343 static inline bool netif_testing(const struct net_device *dev)
4345 return test_bit(__LINK_STATE_TESTING, &dev->state);
4350 * netif_oper_up - test if device is operational
4351 * @dev: network device
4353 * Check if carrier is operational
4355 static inline bool netif_oper_up(const struct net_device *dev)
4357 return (dev->operstate == IF_OPER_UP ||
4358 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4362 * netif_device_present - is device available or removed
4363 * @dev: network device
4365 * Check if device has not been removed from system.
4367 static inline bool netif_device_present(const struct net_device *dev)
4369 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4372 void netif_device_detach(struct net_device *dev);
4374 void netif_device_attach(struct net_device *dev);
4377 * Network interface message level settings
4382 NETIF_MSG_PROBE_BIT,
4384 NETIF_MSG_TIMER_BIT,
4385 NETIF_MSG_IFDOWN_BIT,
4387 NETIF_MSG_RX_ERR_BIT,
4388 NETIF_MSG_TX_ERR_BIT,
4389 NETIF_MSG_TX_QUEUED_BIT,
4391 NETIF_MSG_TX_DONE_BIT,
4392 NETIF_MSG_RX_STATUS_BIT,
4393 NETIF_MSG_PKTDATA_BIT,
4397 /* When you add a new bit above, update netif_msg_class_names array
4398 * in net/ethtool/common.c
4400 NETIF_MSG_CLASS_COUNT,
4402 /* Both ethtool_ops interface and internal driver implementation use u32 */
4403 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4405 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4406 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4408 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4409 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4410 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4411 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4412 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4413 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4414 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4415 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4416 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4417 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4418 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4419 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4420 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4421 #define NETIF_MSG_HW __NETIF_MSG(HW)
4422 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4424 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4425 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4426 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4427 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4428 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4429 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4430 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4431 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4432 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4433 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4434 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4435 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4436 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4437 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4438 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4440 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4443 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4444 return default_msg_enable_bits;
4445 if (debug_value == 0) /* no output */
4447 /* set low N bits */
4448 return (1U << debug_value) - 1;
4451 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4453 spin_lock(&txq->_xmit_lock);
4454 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4455 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4458 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4460 __acquire(&txq->_xmit_lock);
4464 static inline void __netif_tx_release(struct netdev_queue *txq)
4466 __release(&txq->_xmit_lock);
4469 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4471 spin_lock_bh(&txq->_xmit_lock);
4472 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4473 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4476 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4478 bool ok = spin_trylock(&txq->_xmit_lock);
4481 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4482 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4487 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4489 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4490 WRITE_ONCE(txq->xmit_lock_owner, -1);
4491 spin_unlock(&txq->_xmit_lock);
4494 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4496 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4497 WRITE_ONCE(txq->xmit_lock_owner, -1);
4498 spin_unlock_bh(&txq->_xmit_lock);
4502 * txq->trans_start can be read locklessly from dev_watchdog()
4504 static inline void txq_trans_update(struct netdev_queue *txq)
4506 if (txq->xmit_lock_owner != -1)
4507 WRITE_ONCE(txq->trans_start, jiffies);
4510 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4512 unsigned long now = jiffies;
4514 if (READ_ONCE(txq->trans_start) != now)
4515 WRITE_ONCE(txq->trans_start, now);
4518 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4519 static inline void netif_trans_update(struct net_device *dev)
4521 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4523 txq_trans_cond_update(txq);
4527 * netif_tx_lock - grab network device transmit lock
4528 * @dev: network device
4530 * Get network device transmit lock
4532 void netif_tx_lock(struct net_device *dev);
4534 static inline void netif_tx_lock_bh(struct net_device *dev)
4540 void netif_tx_unlock(struct net_device *dev);
4542 static inline void netif_tx_unlock_bh(struct net_device *dev)
4544 netif_tx_unlock(dev);
4548 #define HARD_TX_LOCK(dev, txq, cpu) { \
4549 if ((dev->features & NETIF_F_LLTX) == 0) { \
4550 __netif_tx_lock(txq, cpu); \
4552 __netif_tx_acquire(txq); \
4556 #define HARD_TX_TRYLOCK(dev, txq) \
4557 (((dev->features & NETIF_F_LLTX) == 0) ? \
4558 __netif_tx_trylock(txq) : \
4559 __netif_tx_acquire(txq))
4561 #define HARD_TX_UNLOCK(dev, txq) { \
4562 if ((dev->features & NETIF_F_LLTX) == 0) { \
4563 __netif_tx_unlock(txq); \
4565 __netif_tx_release(txq); \
4569 static inline void netif_tx_disable(struct net_device *dev)
4575 cpu = smp_processor_id();
4576 spin_lock(&dev->tx_global_lock);
4577 for (i = 0; i < dev->num_tx_queues; i++) {
4578 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4580 __netif_tx_lock(txq, cpu);
4581 netif_tx_stop_queue(txq);
4582 __netif_tx_unlock(txq);
4584 spin_unlock(&dev->tx_global_lock);
4588 static inline void netif_addr_lock(struct net_device *dev)
4590 unsigned char nest_level = 0;
4592 #ifdef CONFIG_LOCKDEP
4593 nest_level = dev->nested_level;
4595 spin_lock_nested(&dev->addr_list_lock, nest_level);
4598 static inline void netif_addr_lock_bh(struct net_device *dev)
4600 unsigned char nest_level = 0;
4602 #ifdef CONFIG_LOCKDEP
4603 nest_level = dev->nested_level;
4606 spin_lock_nested(&dev->addr_list_lock, nest_level);
4609 static inline void netif_addr_unlock(struct net_device *dev)
4611 spin_unlock(&dev->addr_list_lock);
4614 static inline void netif_addr_unlock_bh(struct net_device *dev)
4616 spin_unlock_bh(&dev->addr_list_lock);
4620 * dev_addrs walker. Should be used only for read access. Call with
4621 * rcu_read_lock held.
4623 #define for_each_dev_addr(dev, ha) \
4624 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4626 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4628 void ether_setup(struct net_device *dev);
4630 /* Support for loadable net-drivers */
4631 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4632 unsigned char name_assign_type,
4633 void (*setup)(struct net_device *),
4634 unsigned int txqs, unsigned int rxqs);
4635 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4636 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4638 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4639 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4642 int register_netdev(struct net_device *dev);
4643 void unregister_netdev(struct net_device *dev);
4645 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4647 /* General hardware address lists handling functions */
4648 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4649 struct netdev_hw_addr_list *from_list, int addr_len);
4650 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4651 struct netdev_hw_addr_list *from_list, int addr_len);
4652 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4653 struct net_device *dev,
4654 int (*sync)(struct net_device *, const unsigned char *),
4655 int (*unsync)(struct net_device *,
4656 const unsigned char *));
4657 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4658 struct net_device *dev,
4659 int (*sync)(struct net_device *,
4660 const unsigned char *, int),
4661 int (*unsync)(struct net_device *,
4662 const unsigned char *, int));
4663 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4664 struct net_device *dev,
4665 int (*unsync)(struct net_device *,
4666 const unsigned char *, int));
4667 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4668 struct net_device *dev,
4669 int (*unsync)(struct net_device *,
4670 const unsigned char *));
4671 void __hw_addr_init(struct netdev_hw_addr_list *list);
4673 /* Functions used for device addresses handling */
4674 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4675 const void *addr, size_t len);
4678 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4680 dev_addr_mod(dev, 0, addr, len);
4683 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4685 __dev_addr_set(dev, addr, dev->addr_len);
4688 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4689 unsigned char addr_type);
4690 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4691 unsigned char addr_type);
4693 /* Functions used for unicast addresses handling */
4694 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4695 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4696 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4697 int dev_uc_sync(struct net_device *to, struct net_device *from);
4698 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4699 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4700 void dev_uc_flush(struct net_device *dev);
4701 void dev_uc_init(struct net_device *dev);
4704 * __dev_uc_sync - Synchonize device's unicast list
4705 * @dev: device to sync
4706 * @sync: function to call if address should be added
4707 * @unsync: function to call if address should be removed
4709 * Add newly added addresses to the interface, and release
4710 * addresses that have been deleted.
4712 static inline int __dev_uc_sync(struct net_device *dev,
4713 int (*sync)(struct net_device *,
4714 const unsigned char *),
4715 int (*unsync)(struct net_device *,
4716 const unsigned char *))
4718 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4722 * __dev_uc_unsync - Remove synchronized addresses from device
4723 * @dev: device to sync
4724 * @unsync: function to call if address should be removed
4726 * Remove all addresses that were added to the device by dev_uc_sync().
4728 static inline void __dev_uc_unsync(struct net_device *dev,
4729 int (*unsync)(struct net_device *,
4730 const unsigned char *))
4732 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4735 /* Functions used for multicast addresses handling */
4736 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4737 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4738 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4739 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4740 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4741 int dev_mc_sync(struct net_device *to, struct net_device *from);
4742 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4743 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4744 void dev_mc_flush(struct net_device *dev);
4745 void dev_mc_init(struct net_device *dev);
4748 * __dev_mc_sync - Synchonize device's multicast list
4749 * @dev: device to sync
4750 * @sync: function to call if address should be added
4751 * @unsync: function to call if address should be removed
4753 * Add newly added addresses to the interface, and release
4754 * addresses that have been deleted.
4756 static inline int __dev_mc_sync(struct net_device *dev,
4757 int (*sync)(struct net_device *,
4758 const unsigned char *),
4759 int (*unsync)(struct net_device *,
4760 const unsigned char *))
4762 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4766 * __dev_mc_unsync - Remove synchronized addresses from device
4767 * @dev: device to sync
4768 * @unsync: function to call if address should be removed
4770 * Remove all addresses that were added to the device by dev_mc_sync().
4772 static inline void __dev_mc_unsync(struct net_device *dev,
4773 int (*unsync)(struct net_device *,
4774 const unsigned char *))
4776 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4779 /* Functions used for secondary unicast and multicast support */
4780 void dev_set_rx_mode(struct net_device *dev);
4781 int dev_set_promiscuity(struct net_device *dev, int inc);
4782 int dev_set_allmulti(struct net_device *dev, int inc);
4783 void netdev_state_change(struct net_device *dev);
4784 void __netdev_notify_peers(struct net_device *dev);
4785 void netdev_notify_peers(struct net_device *dev);
4786 void netdev_features_change(struct net_device *dev);
4787 /* Load a device via the kmod */
4788 void dev_load(struct net *net, const char *name);
4789 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4790 struct rtnl_link_stats64 *storage);
4791 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4792 const struct net_device_stats *netdev_stats);
4793 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4794 const struct pcpu_sw_netstats __percpu *netstats);
4795 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4797 extern int netdev_max_backlog;
4798 extern int dev_rx_weight;
4799 extern int dev_tx_weight;
4800 extern int gro_normal_batch;
4803 NESTED_SYNC_IMM_BIT,
4804 NESTED_SYNC_TODO_BIT,
4807 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4808 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4810 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4811 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4813 struct netdev_nested_priv {
4814 unsigned char flags;
4818 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4819 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4820 struct list_head **iter);
4822 /* iterate through upper list, must be called under RCU read lock */
4823 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4824 for (iter = &(dev)->adj_list.upper, \
4825 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4827 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4829 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4830 int (*fn)(struct net_device *upper_dev,
4831 struct netdev_nested_priv *priv),
4832 struct netdev_nested_priv *priv);
4834 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4835 struct net_device *upper_dev);
4837 bool netdev_has_any_upper_dev(struct net_device *dev);
4839 void *netdev_lower_get_next_private(struct net_device *dev,
4840 struct list_head **iter);
4841 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4842 struct list_head **iter);
4844 #define netdev_for_each_lower_private(dev, priv, iter) \
4845 for (iter = (dev)->adj_list.lower.next, \
4846 priv = netdev_lower_get_next_private(dev, &(iter)); \
4848 priv = netdev_lower_get_next_private(dev, &(iter)))
4850 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4851 for (iter = &(dev)->adj_list.lower, \
4852 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4854 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4856 void *netdev_lower_get_next(struct net_device *dev,
4857 struct list_head **iter);
4859 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4860 for (iter = (dev)->adj_list.lower.next, \
4861 ldev = netdev_lower_get_next(dev, &(iter)); \
4863 ldev = netdev_lower_get_next(dev, &(iter)))
4865 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4866 struct list_head **iter);
4867 int netdev_walk_all_lower_dev(struct net_device *dev,
4868 int (*fn)(struct net_device *lower_dev,
4869 struct netdev_nested_priv *priv),
4870 struct netdev_nested_priv *priv);
4871 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4872 int (*fn)(struct net_device *lower_dev,
4873 struct netdev_nested_priv *priv),
4874 struct netdev_nested_priv *priv);
4876 void *netdev_adjacent_get_private(struct list_head *adj_list);
4877 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4878 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4879 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4880 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4881 struct netlink_ext_ack *extack);
4882 int netdev_master_upper_dev_link(struct net_device *dev,
4883 struct net_device *upper_dev,
4884 void *upper_priv, void *upper_info,
4885 struct netlink_ext_ack *extack);
4886 void netdev_upper_dev_unlink(struct net_device *dev,
4887 struct net_device *upper_dev);
4888 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4889 struct net_device *new_dev,
4890 struct net_device *dev,
4891 struct netlink_ext_ack *extack);
4892 void netdev_adjacent_change_commit(struct net_device *old_dev,
4893 struct net_device *new_dev,
4894 struct net_device *dev);
4895 void netdev_adjacent_change_abort(struct net_device *old_dev,
4896 struct net_device *new_dev,
4897 struct net_device *dev);
4898 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4899 void *netdev_lower_dev_get_private(struct net_device *dev,
4900 struct net_device *lower_dev);
4901 void netdev_lower_state_changed(struct net_device *lower_dev,
4902 void *lower_state_info);
4904 /* RSS keys are 40 or 52 bytes long */
4905 #define NETDEV_RSS_KEY_LEN 52
4906 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4907 void netdev_rss_key_fill(void *buffer, size_t len);
4909 int skb_checksum_help(struct sk_buff *skb);
4910 int skb_crc32c_csum_help(struct sk_buff *skb);
4911 int skb_csum_hwoffload_help(struct sk_buff *skb,
4912 const netdev_features_t features);
4914 struct netdev_bonding_info {
4919 struct netdev_notifier_bonding_info {
4920 struct netdev_notifier_info info; /* must be first */
4921 struct netdev_bonding_info bonding_info;
4924 void netdev_bonding_info_change(struct net_device *dev,
4925 struct netdev_bonding_info *bonding_info);
4927 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4928 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4930 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4936 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4938 static inline bool can_checksum_protocol(netdev_features_t features,
4941 if (protocol == htons(ETH_P_FCOE))
4942 return !!(features & NETIF_F_FCOE_CRC);
4944 /* Assume this is an IP checksum (not SCTP CRC) */
4946 if (features & NETIF_F_HW_CSUM) {
4947 /* Can checksum everything */
4952 case htons(ETH_P_IP):
4953 return !!(features & NETIF_F_IP_CSUM);
4954 case htons(ETH_P_IPV6):
4955 return !!(features & NETIF_F_IPV6_CSUM);
4962 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4964 static inline void netdev_rx_csum_fault(struct net_device *dev,
4965 struct sk_buff *skb)
4969 /* rx skb timestamps */
4970 void net_enable_timestamp(void);
4971 void net_disable_timestamp(void);
4973 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4974 const struct skb_shared_hwtstamps *hwtstamps,
4977 const struct net_device_ops *ops = dev->netdev_ops;
4979 if (ops->ndo_get_tstamp)
4980 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4982 return hwtstamps->hwtstamp;
4985 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4986 struct sk_buff *skb, struct net_device *dev,
4989 __this_cpu_write(softnet_data.xmit.more, more);
4990 return ops->ndo_start_xmit(skb, dev);
4993 static inline bool netdev_xmit_more(void)
4995 return __this_cpu_read(softnet_data.xmit.more);
4998 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4999 struct netdev_queue *txq, bool more)
5001 const struct net_device_ops *ops = dev->netdev_ops;
5004 rc = __netdev_start_xmit(ops, skb, dev, more);
5005 if (rc == NETDEV_TX_OK)
5006 txq_trans_update(txq);
5011 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
5013 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
5016 extern const struct kobj_ns_type_operations net_ns_type_operations;
5018 const char *netdev_drivername(const struct net_device *dev);
5020 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
5021 netdev_features_t f2)
5023 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
5024 if (f1 & NETIF_F_HW_CSUM)
5025 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5027 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5033 static inline netdev_features_t netdev_get_wanted_features(
5034 struct net_device *dev)
5036 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5038 netdev_features_t netdev_increment_features(netdev_features_t all,
5039 netdev_features_t one, netdev_features_t mask);
5041 /* Allow TSO being used on stacked device :
5042 * Performing the GSO segmentation before last device
5043 * is a performance improvement.
5045 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5046 netdev_features_t mask)
5048 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5051 int __netdev_update_features(struct net_device *dev);
5052 void netdev_update_features(struct net_device *dev);
5053 void netdev_change_features(struct net_device *dev);
5055 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5056 struct net_device *dev);
5058 netdev_features_t passthru_features_check(struct sk_buff *skb,
5059 struct net_device *dev,
5060 netdev_features_t features);
5061 netdev_features_t netif_skb_features(struct sk_buff *skb);
5062 void skb_warn_bad_offload(const struct sk_buff *skb);
5064 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5066 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5068 /* check flags correspondence */
5069 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5070 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5071 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5072 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5073 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5074 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5075 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5076 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5077 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5078 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5079 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5080 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5081 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5082 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5083 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5084 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5085 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5086 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5087 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5089 return (features & feature) == feature;
5092 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5094 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5095 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5098 static inline bool netif_needs_gso(struct sk_buff *skb,
5099 netdev_features_t features)
5101 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5102 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5103 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5106 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5107 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5108 void netif_inherit_tso_max(struct net_device *to,
5109 const struct net_device *from);
5111 static inline bool netif_is_macsec(const struct net_device *dev)
5113 return dev->priv_flags & IFF_MACSEC;
5116 static inline bool netif_is_macvlan(const struct net_device *dev)
5118 return dev->priv_flags & IFF_MACVLAN;
5121 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5123 return dev->priv_flags & IFF_MACVLAN_PORT;
5126 static inline bool netif_is_bond_master(const struct net_device *dev)
5128 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5131 static inline bool netif_is_bond_slave(const struct net_device *dev)
5133 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5136 static inline bool netif_supports_nofcs(struct net_device *dev)
5138 return dev->priv_flags & IFF_SUPP_NOFCS;
5141 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5143 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5146 static inline bool netif_is_l3_master(const struct net_device *dev)
5148 return dev->priv_flags & IFF_L3MDEV_MASTER;
5151 static inline bool netif_is_l3_slave(const struct net_device *dev)
5153 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5156 static inline int dev_sdif(const struct net_device *dev)
5158 #ifdef CONFIG_NET_L3_MASTER_DEV
5159 if (netif_is_l3_slave(dev))
5160 return dev->ifindex;
5165 static inline bool netif_is_bridge_master(const struct net_device *dev)
5167 return dev->priv_flags & IFF_EBRIDGE;
5170 static inline bool netif_is_bridge_port(const struct net_device *dev)
5172 return dev->priv_flags & IFF_BRIDGE_PORT;
5175 static inline bool netif_is_ovs_master(const struct net_device *dev)
5177 return dev->priv_flags & IFF_OPENVSWITCH;
5180 static inline bool netif_is_ovs_port(const struct net_device *dev)
5182 return dev->priv_flags & IFF_OVS_DATAPATH;
5185 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5187 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5190 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5192 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5195 static inline bool netif_is_team_master(const struct net_device *dev)
5197 return dev->priv_flags & IFF_TEAM;
5200 static inline bool netif_is_team_port(const struct net_device *dev)
5202 return dev->priv_flags & IFF_TEAM_PORT;
5205 static inline bool netif_is_lag_master(const struct net_device *dev)
5207 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5210 static inline bool netif_is_lag_port(const struct net_device *dev)
5212 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5215 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5217 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5220 static inline bool netif_is_failover(const struct net_device *dev)
5222 return dev->priv_flags & IFF_FAILOVER;
5225 static inline bool netif_is_failover_slave(const struct net_device *dev)
5227 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5230 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5231 static inline void netif_keep_dst(struct net_device *dev)
5233 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5236 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5237 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5239 /* TODO: reserve and use an additional IFF bit, if we get more users */
5240 return netif_is_macsec(dev);
5243 extern struct pernet_operations __net_initdata loopback_net_ops;
5245 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5247 /* netdev_printk helpers, similar to dev_printk */
5249 static inline const char *netdev_name(const struct net_device *dev)
5251 if (!dev->name[0] || strchr(dev->name, '%'))
5252 return "(unnamed net_device)";
5256 static inline const char *netdev_reg_state(const struct net_device *dev)
5258 u8 reg_state = READ_ONCE(dev->reg_state);
5260 switch (reg_state) {
5261 case NETREG_UNINITIALIZED: return " (uninitialized)";
5262 case NETREG_REGISTERED: return "";
5263 case NETREG_UNREGISTERING: return " (unregistering)";
5264 case NETREG_UNREGISTERED: return " (unregistered)";
5265 case NETREG_RELEASED: return " (released)";
5266 case NETREG_DUMMY: return " (dummy)";
5269 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, reg_state);
5270 return " (unknown)";
5273 #define MODULE_ALIAS_NETDEV(device) \
5274 MODULE_ALIAS("netdev-" device)
5277 * netdev_WARN() acts like dev_printk(), but with the key difference
5278 * of using a WARN/WARN_ON to get the message out, including the
5279 * file/line information and a backtrace.
5281 #define netdev_WARN(dev, format, args...) \
5282 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5283 netdev_reg_state(dev), ##args)
5285 #define netdev_WARN_ONCE(dev, format, args...) \
5286 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5287 netdev_reg_state(dev), ##args)
5290 * The list of packet types we will receive (as opposed to discard)
5291 * and the routines to invoke.
5293 * Why 16. Because with 16 the only overlap we get on a hash of the
5294 * low nibble of the protocol value is RARP/SNAP/X.25.
5308 #define PTYPE_HASH_SIZE (16)
5309 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5311 extern struct list_head ptype_all __read_mostly;
5312 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5314 extern struct net_device *blackhole_netdev;
5316 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5317 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5318 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5319 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5320 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5322 #endif /* _LINUX_NETDEVICE_H */