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
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;
388 NAPI_STATE_SCHED, /* Poll is scheduled */
389 NAPI_STATE_MISSED, /* reschedule a napi */
390 NAPI_STATE_DISABLE, /* Disable pending */
391 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
392 NAPI_STATE_LISTED, /* NAPI added to system lists */
393 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
394 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
395 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
396 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
397 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
401 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
402 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
403 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
404 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
405 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
406 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
407 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
408 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
409 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
410 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
420 typedef enum gro_result gro_result_t;
423 * enum rx_handler_result - Possible return values for rx_handlers.
424 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
426 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
427 * case skb->dev was changed by rx_handler.
428 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
429 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
431 * rx_handlers are functions called from inside __netif_receive_skb(), to do
432 * special processing of the skb, prior to delivery to protocol handlers.
434 * Currently, a net_device can only have a single rx_handler registered. Trying
435 * to register a second rx_handler will return -EBUSY.
437 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
438 * To unregister a rx_handler on a net_device, use
439 * netdev_rx_handler_unregister().
441 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
444 * If the rx_handler consumed the skb in some way, it should return
445 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
446 * the skb to be delivered in some other way.
448 * If the rx_handler changed skb->dev, to divert the skb to another
449 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
450 * new device will be called if it exists.
452 * If the rx_handler decides the skb should be ignored, it should return
453 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
454 * are registered on exact device (ptype->dev == skb->dev).
456 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
457 * delivered, it should return RX_HANDLER_PASS.
459 * A device without a registered rx_handler will behave as if rx_handler
460 * returned RX_HANDLER_PASS.
463 enum rx_handler_result {
469 typedef enum rx_handler_result rx_handler_result_t;
470 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
472 void __napi_schedule(struct napi_struct *n);
473 void __napi_schedule_irqoff(struct napi_struct *n);
475 static inline bool napi_disable_pending(struct napi_struct *n)
477 return test_bit(NAPI_STATE_DISABLE, &n->state);
480 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
482 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
486 * napi_is_scheduled - test if NAPI is scheduled
489 * This check is "best-effort". With no locking implemented,
490 * a NAPI can be scheduled or terminate right after this check
491 * and produce not precise results.
493 * NAPI_STATE_SCHED is an internal state, napi_is_scheduled
494 * should not be used normally and napi_schedule should be
497 * Use only if the driver really needs to check if a NAPI
498 * is scheduled for example in the context of delayed timer
499 * that can be skipped if a NAPI is already scheduled.
501 * Return True if NAPI is scheduled, False otherwise.
503 static inline bool napi_is_scheduled(struct napi_struct *n)
505 return test_bit(NAPI_STATE_SCHED, &n->state);
508 bool napi_schedule_prep(struct napi_struct *n);
511 * napi_schedule - schedule NAPI poll
514 * Schedule NAPI poll routine to be called if it is not already
516 * Return true if we schedule a NAPI or false if not.
517 * Refer to napi_schedule_prep() for additional reason on why
518 * a NAPI might not be scheduled.
520 static inline bool napi_schedule(struct napi_struct *n)
522 if (napi_schedule_prep(n)) {
531 * napi_schedule_irqoff - schedule NAPI poll
534 * Variant of napi_schedule(), assuming hard irqs are masked.
536 static inline void napi_schedule_irqoff(struct napi_struct *n)
538 if (napi_schedule_prep(n))
539 __napi_schedule_irqoff(n);
543 * napi_complete_done - NAPI processing complete
545 * @work_done: number of packets processed
547 * Mark NAPI processing as complete. Should only be called if poll budget
548 * has not been completely consumed.
549 * Prefer over napi_complete().
550 * Return false if device should avoid rearming interrupts.
552 bool napi_complete_done(struct napi_struct *n, int work_done);
554 static inline bool napi_complete(struct napi_struct *n)
556 return napi_complete_done(n, 0);
559 int dev_set_threaded(struct net_device *dev, bool threaded);
562 * napi_disable - prevent NAPI from scheduling
565 * Stop NAPI from being scheduled on this context.
566 * Waits till any outstanding processing completes.
568 void napi_disable(struct napi_struct *n);
570 void napi_enable(struct napi_struct *n);
573 * napi_synchronize - wait until NAPI is not running
576 * Wait until NAPI is done being scheduled on this context.
577 * Waits till any outstanding processing completes but
578 * does not disable future activations.
580 static inline void napi_synchronize(const struct napi_struct *n)
582 if (IS_ENABLED(CONFIG_SMP))
583 while (test_bit(NAPI_STATE_SCHED, &n->state))
590 * napi_if_scheduled_mark_missed - if napi is running, set the
594 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
597 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
599 unsigned long val, new;
601 val = READ_ONCE(n->state);
603 if (val & NAPIF_STATE_DISABLE)
606 if (!(val & NAPIF_STATE_SCHED))
609 new = val | NAPIF_STATE_MISSED;
610 } while (!try_cmpxchg(&n->state, &val, new));
615 enum netdev_queue_state_t {
616 __QUEUE_STATE_DRV_XOFF,
617 __QUEUE_STATE_STACK_XOFF,
618 __QUEUE_STATE_FROZEN,
621 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
622 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
623 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
625 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
626 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
628 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
632 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
633 * netif_tx_* functions below are used to manipulate this flag. The
634 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
635 * queue independently. The netif_xmit_*stopped functions below are called
636 * to check if the queue has been stopped by the driver or stack (either
637 * of the XOFF bits are set in the state). Drivers should not need to call
638 * netif_xmit*stopped functions, they should only be using netif_tx_*.
641 struct netdev_queue {
645 struct net_device *dev;
646 netdevice_tracker dev_tracker;
648 struct Qdisc __rcu *qdisc;
649 struct Qdisc __rcu *qdisc_sleeping;
653 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
656 unsigned long tx_maxrate;
658 * Number of TX timeouts for this queue
659 * (/sys/class/net/DEV/Q/trans_timeout)
661 atomic_long_t trans_timeout;
663 /* Subordinate device that the queue has been assigned to */
664 struct net_device *sb_dev;
665 #ifdef CONFIG_XDP_SOCKETS
666 struct xsk_buff_pool *pool;
671 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
674 * Time (in jiffies) of last Tx
676 unsigned long trans_start;
683 } ____cacheline_aligned_in_smp;
685 extern int sysctl_fb_tunnels_only_for_init_net;
686 extern int sysctl_devconf_inherit_init_net;
689 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
690 * == 1 : For initns only
693 static inline bool net_has_fallback_tunnels(const struct net *net)
695 #if IS_ENABLED(CONFIG_SYSCTL)
696 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
698 return !fb_tunnels_only_for_init_net ||
699 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
705 static inline int net_inherit_devconf(void)
707 #if IS_ENABLED(CONFIG_SYSCTL)
708 return READ_ONCE(sysctl_devconf_inherit_init_net);
714 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
716 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
723 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
725 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
732 * This structure holds an RPS map which can be of variable length. The
733 * map is an array of CPUs.
740 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
743 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
744 * tail pointer for that CPU's input queue at the time of last enqueue, and
745 * a hardware filter index.
747 struct rps_dev_flow {
750 unsigned int last_qtail;
752 #define RPS_NO_FILTER 0xffff
755 * The rps_dev_flow_table structure contains a table of flow mappings.
757 struct rps_dev_flow_table {
760 struct rps_dev_flow flows[];
762 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
763 ((_num) * sizeof(struct rps_dev_flow)))
766 * The rps_sock_flow_table contains mappings of flows to the last CPU
767 * on which they were processed by the application (set in recvmsg).
768 * Each entry is a 32bit value. Upper part is the high-order bits
769 * of flow hash, lower part is CPU number.
770 * rps_cpu_mask is used to partition the space, depending on number of
771 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
772 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
773 * meaning we use 32-6=26 bits for the hash.
775 struct rps_sock_flow_table {
778 u32 ents[] ____cacheline_aligned_in_smp;
780 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
782 #define RPS_NO_CPU 0xffff
784 extern u32 rps_cpu_mask;
785 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
787 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
791 unsigned int index = hash & table->mask;
792 u32 val = hash & ~rps_cpu_mask;
794 /* We only give a hint, preemption can change CPU under us */
795 val |= raw_smp_processor_id();
797 /* The following WRITE_ONCE() is paired with the READ_ONCE()
798 * here, and another one in get_rps_cpu().
800 if (READ_ONCE(table->ents[index]) != val)
801 WRITE_ONCE(table->ents[index], val);
805 #ifdef CONFIG_RFS_ACCEL
806 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
809 #endif /* CONFIG_RPS */
811 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
820 * This structure holds an XPS map which can be of variable length. The
821 * map is an array of queues.
825 unsigned int alloc_len;
829 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
830 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
831 - sizeof(struct xps_map)) / sizeof(u16))
834 * This structure holds all XPS maps for device. Maps are indexed by CPU.
836 * We keep track of the number of cpus/rxqs used when the struct is allocated,
837 * in nr_ids. This will help not accessing out-of-bound memory.
839 * We keep track of the number of traffic classes used when the struct is
840 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
841 * not crossing its upper bound, as the original dev->num_tc can be updated in
844 struct xps_dev_maps {
848 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
851 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
852 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
854 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
855 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
857 #endif /* CONFIG_XPS */
859 #define TC_MAX_QUEUE 16
860 #define TC_BITMASK 15
861 /* HW offloaded queuing disciplines txq count and offset maps */
862 struct netdev_tc_txq {
867 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
869 * This structure is to hold information about the device
870 * configured to run FCoE protocol stack.
872 struct netdev_fcoe_hbainfo {
873 char manufacturer[64];
874 char serial_number[64];
875 char hardware_version[64];
876 char driver_version[64];
877 char optionrom_version[64];
878 char firmware_version[64];
880 char model_description[256];
884 #define MAX_PHYS_ITEM_ID_LEN 32
886 /* This structure holds a unique identifier to identify some
887 * physical item (port for example) used by a netdevice.
889 struct netdev_phys_item_id {
890 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
891 unsigned char id_len;
894 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
895 struct netdev_phys_item_id *b)
897 return a->id_len == b->id_len &&
898 memcmp(a->id, b->id, a->id_len) == 0;
901 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
903 struct net_device *sb_dev);
905 enum net_device_path_type {
906 DEV_PATH_ETHERNET = 0,
914 struct net_device_path {
915 enum net_device_path_type type;
916 const struct net_device *dev;
925 DEV_PATH_BR_VLAN_KEEP,
926 DEV_PATH_BR_VLAN_TAG,
927 DEV_PATH_BR_VLAN_UNTAG,
928 DEV_PATH_BR_VLAN_UNTAG_HW,
947 #define NET_DEVICE_PATH_STACK_MAX 5
948 #define NET_DEVICE_PATH_VLAN_MAX 2
950 struct net_device_path_stack {
952 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
955 struct net_device_path_ctx {
956 const struct net_device *dev;
963 } vlan[NET_DEVICE_PATH_VLAN_MAX];
968 TC_SETUP_QDISC_MQPRIO,
971 TC_SETUP_CLSMATCHALL,
981 TC_SETUP_QDISC_TAPRIO,
990 /* These structures hold the attributes of bpf state that are being passed
991 * to the netdevice through the bpf op.
993 enum bpf_netdev_command {
994 /* Set or clear a bpf program used in the earliest stages of packet
995 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
996 * is responsible for calling bpf_prog_put on any old progs that are
997 * stored. In case of error, the callee need not release the new prog
998 * reference, but on success it takes ownership and must bpf_prog_put
999 * when it is no longer used.
1003 /* BPF program for offload callbacks, invoked at program load time. */
1004 BPF_OFFLOAD_MAP_ALLOC,
1005 BPF_OFFLOAD_MAP_FREE,
1009 struct bpf_prog_offload_ops;
1010 struct netlink_ext_ack;
1012 struct xdp_dev_bulk_queue;
1013 struct bpf_xdp_link;
1022 struct bpf_xdp_entity {
1023 struct bpf_prog *prog;
1024 struct bpf_xdp_link *link;
1028 enum bpf_netdev_command command;
1030 /* XDP_SETUP_PROG */
1033 struct bpf_prog *prog;
1034 struct netlink_ext_ack *extack;
1036 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1038 struct bpf_offloaded_map *offmap;
1040 /* XDP_SETUP_XSK_POOL */
1042 struct xsk_buff_pool *pool;
1048 /* Flags for ndo_xsk_wakeup. */
1049 #define XDP_WAKEUP_RX (1 << 0)
1050 #define XDP_WAKEUP_TX (1 << 1)
1052 #ifdef CONFIG_XFRM_OFFLOAD
1053 struct xfrmdev_ops {
1054 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1055 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1056 void (*xdo_dev_state_free) (struct xfrm_state *x);
1057 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1058 struct xfrm_state *x);
1059 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1060 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1061 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1062 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1063 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1067 struct dev_ifalias {
1068 struct rcu_head rcuhead;
1075 struct netdev_net_notifier {
1076 struct list_head list;
1077 struct notifier_block *nb;
1081 * This structure defines the management hooks for network devices.
1082 * The following hooks can be defined; unless noted otherwise, they are
1083 * optional and can be filled with a null pointer.
1085 * int (*ndo_init)(struct net_device *dev);
1086 * This function is called once when a network device is registered.
1087 * The network device can use this for any late stage initialization
1088 * or semantic validation. It can fail with an error code which will
1089 * be propagated back to register_netdev.
1091 * void (*ndo_uninit)(struct net_device *dev);
1092 * This function is called when device is unregistered or when registration
1093 * fails. It is not called if init fails.
1095 * int (*ndo_open)(struct net_device *dev);
1096 * This function is called when a network device transitions to the up
1099 * int (*ndo_stop)(struct net_device *dev);
1100 * This function is called when a network device transitions to the down
1103 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1104 * struct net_device *dev);
1105 * Called when a packet needs to be transmitted.
1106 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1107 * the queue before that can happen; it's for obsolete devices and weird
1108 * corner cases, but the stack really does a non-trivial amount
1109 * of useless work if you return NETDEV_TX_BUSY.
1110 * Required; cannot be NULL.
1112 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1113 * struct net_device *dev
1114 * netdev_features_t features);
1115 * Called by core transmit path to determine if device is capable of
1116 * performing offload operations on a given packet. This is to give
1117 * the device an opportunity to implement any restrictions that cannot
1118 * be otherwise expressed by feature flags. The check is called with
1119 * the set of features that the stack has calculated and it returns
1120 * those the driver believes to be appropriate.
1122 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1123 * struct net_device *sb_dev);
1124 * Called to decide which queue to use when device supports multiple
1127 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1128 * This function is called to allow device receiver to make
1129 * changes to configuration when multicast or promiscuous is enabled.
1131 * void (*ndo_set_rx_mode)(struct net_device *dev);
1132 * This function is called device changes address list filtering.
1133 * If driver handles unicast address filtering, it should set
1134 * IFF_UNICAST_FLT in its priv_flags.
1136 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1137 * This function is called when the Media Access Control address
1138 * needs to be changed. If this interface is not defined, the
1139 * MAC address can not be changed.
1141 * int (*ndo_validate_addr)(struct net_device *dev);
1142 * Test if Media Access Control address is valid for the device.
1144 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1145 * Old-style ioctl entry point. This is used internally by the
1146 * appletalk and ieee802154 subsystems but is no longer called by
1147 * the device ioctl handler.
1149 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1150 * Used by the bonding driver for its device specific ioctls:
1151 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1152 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1154 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1155 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1156 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1158 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1159 * Used to set network devices bus interface parameters. This interface
1160 * is retained for legacy reasons; new devices should use the bus
1161 * interface (PCI) for low level management.
1163 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1164 * Called when a user wants to change the Maximum Transfer Unit
1167 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1168 * Callback used when the transmitter has not made any progress
1169 * for dev->watchdog ticks.
1171 * void (*ndo_get_stats64)(struct net_device *dev,
1172 * struct rtnl_link_stats64 *storage);
1173 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1174 * Called when a user wants to get the network device usage
1175 * statistics. Drivers must do one of the following:
1176 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1177 * rtnl_link_stats64 structure passed by the caller.
1178 * 2. Define @ndo_get_stats to update a net_device_stats structure
1179 * (which should normally be dev->stats) and return a pointer to
1180 * it. The structure may be changed asynchronously only if each
1181 * field is written atomically.
1182 * 3. Update dev->stats asynchronously and atomically, and define
1183 * neither operation.
1185 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1186 * Return true if this device supports offload stats of this attr_id.
1188 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1190 * Get statistics for offload operations by attr_id. Write it into the
1191 * attr_data pointer.
1193 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1194 * If device supports VLAN filtering this function is called when a
1195 * VLAN id is registered.
1197 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1198 * If device supports VLAN filtering this function is called when a
1199 * VLAN id is unregistered.
1201 * void (*ndo_poll_controller)(struct net_device *dev);
1203 * SR-IOV management functions.
1204 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1205 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1206 * u8 qos, __be16 proto);
1207 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1209 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1210 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1211 * int (*ndo_get_vf_config)(struct net_device *dev,
1212 * int vf, struct ifla_vf_info *ivf);
1213 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1214 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1215 * struct nlattr *port[]);
1217 * Enable or disable the VF ability to query its RSS Redirection Table and
1218 * Hash Key. This is needed since on some devices VF share this information
1219 * with PF and querying it may introduce a theoretical security risk.
1220 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1221 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1222 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1224 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1225 * This is always called from the stack with the rtnl lock held and netif
1226 * tx queues stopped. This allows the netdevice to perform queue
1227 * management safely.
1229 * Fiber Channel over Ethernet (FCoE) offload functions.
1230 * int (*ndo_fcoe_enable)(struct net_device *dev);
1231 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1232 * so the underlying device can perform whatever needed configuration or
1233 * initialization to support acceleration of FCoE traffic.
1235 * int (*ndo_fcoe_disable)(struct net_device *dev);
1236 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1237 * so the underlying device can perform whatever needed clean-ups to
1238 * stop supporting acceleration of FCoE traffic.
1240 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1241 * struct scatterlist *sgl, unsigned int sgc);
1242 * Called when the FCoE Initiator wants to initialize an I/O that
1243 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1244 * perform necessary setup and returns 1 to indicate the device is set up
1245 * successfully to perform DDP on this I/O, otherwise this returns 0.
1247 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1248 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1249 * indicated by the FC exchange id 'xid', so the underlying device can
1250 * clean up and reuse resources for later DDP requests.
1252 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1253 * struct scatterlist *sgl, unsigned int sgc);
1254 * Called when the FCoE Target wants to initialize an I/O that
1255 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1256 * perform necessary setup and returns 1 to indicate the device is set up
1257 * successfully to perform DDP on this I/O, otherwise this returns 0.
1259 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1260 * struct netdev_fcoe_hbainfo *hbainfo);
1261 * Called when the FCoE Protocol stack wants information on the underlying
1262 * device. This information is utilized by the FCoE protocol stack to
1263 * register attributes with Fiber Channel management service as per the
1264 * FC-GS Fabric Device Management Information(FDMI) specification.
1266 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1267 * Called when the underlying device wants to override default World Wide
1268 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1269 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1270 * protocol stack to use.
1273 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1274 * u16 rxq_index, u32 flow_id);
1275 * Set hardware filter for RFS. rxq_index is the target queue index;
1276 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1277 * Return the filter ID on success, or a negative error code.
1279 * Slave management functions (for bridge, bonding, etc).
1280 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1281 * Called to make another netdev an underling.
1283 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1284 * Called to release previously enslaved netdev.
1286 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1287 * struct sk_buff *skb,
1289 * Get the xmit slave of master device. If all_slaves is true, function
1290 * assume all the slaves can transmit.
1292 * Feature/offload setting functions.
1293 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1294 * netdev_features_t features);
1295 * Adjusts the requested feature flags according to device-specific
1296 * constraints, and returns the resulting flags. Must not modify
1299 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1300 * Called to update device configuration to new features. Passed
1301 * feature set might be less than what was returned by ndo_fix_features()).
1302 * Must return >0 or -errno if it changed dev->features itself.
1304 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1305 * struct net_device *dev,
1306 * const unsigned char *addr, u16 vid, u16 flags,
1307 * struct netlink_ext_ack *extack);
1308 * Adds an FDB entry to dev for addr.
1309 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1310 * struct net_device *dev,
1311 * const unsigned char *addr, u16 vid)
1312 * Deletes the FDB entry from dev coresponding to addr.
1313 * int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh, struct net_device *dev,
1314 * struct netlink_ext_ack *extack);
1315 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1316 * struct net_device *dev, struct net_device *filter_dev,
1318 * Used to add FDB entries to dump requests. Implementers should add
1319 * entries to skb and update idx with the number of entries.
1321 * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1322 * u16 nlmsg_flags, struct netlink_ext_ack *extack);
1323 * Adds an MDB entry to dev.
1324 * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1325 * struct netlink_ext_ack *extack);
1326 * Deletes the MDB entry from dev.
1327 * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1328 * struct netlink_callback *cb);
1329 * Dumps MDB entries from dev. The first argument (marker) in the netlink
1330 * callback is used by core rtnetlink code.
1332 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1333 * u16 flags, struct netlink_ext_ack *extack)
1334 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1335 * struct net_device *dev, u32 filter_mask,
1337 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1340 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1341 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1342 * which do not represent real hardware may define this to allow their
1343 * userspace components to manage their virtual carrier state. Devices
1344 * that determine carrier state from physical hardware properties (eg
1345 * network cables) or protocol-dependent mechanisms (eg
1346 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1348 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1349 * struct netdev_phys_item_id *ppid);
1350 * Called to get ID of physical port of this device. If driver does
1351 * not implement this, it is assumed that the hw is not able to have
1352 * multiple net devices on single physical port.
1354 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1355 * struct netdev_phys_item_id *ppid)
1356 * Called to get the parent ID of the physical port of this device.
1358 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1359 * struct net_device *dev)
1360 * Called by upper layer devices to accelerate switching or other
1361 * station functionality into hardware. 'pdev is the lowerdev
1362 * to use for the offload and 'dev' is the net device that will
1363 * back the offload. Returns a pointer to the private structure
1364 * the upper layer will maintain.
1365 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1366 * Called by upper layer device to delete the station created
1367 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1368 * the station and priv is the structure returned by the add
1370 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1371 * int queue_index, u32 maxrate);
1372 * Called when a user wants to set a max-rate limitation of specific
1374 * int (*ndo_get_iflink)(const struct net_device *dev);
1375 * Called to get the iflink value of this device.
1376 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1377 * This function is used to get egress tunnel information for given skb.
1378 * This is useful for retrieving outer tunnel header parameters while
1380 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1381 * This function is used to specify the headroom that the skb must
1382 * consider when allocation skb during packet reception. Setting
1383 * appropriate rx headroom value allows avoiding skb head copy on
1384 * forward. Setting a negative value resets the rx headroom to the
1386 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1387 * This function is used to set or query state related to XDP on the
1388 * netdevice and manage BPF offload. See definition of
1389 * enum bpf_netdev_command for details.
1390 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1392 * This function is used to submit @n XDP packets for transmit on a
1393 * netdevice. Returns number of frames successfully transmitted, frames
1394 * that got dropped are freed/returned via xdp_return_frame().
1395 * Returns negative number, means general error invoking ndo, meaning
1396 * no frames were xmit'ed and core-caller will free all frames.
1397 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1398 * struct xdp_buff *xdp);
1399 * Get the xmit slave of master device based on the xdp_buff.
1400 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1401 * This function is used to wake up the softirq, ksoftirqd or kthread
1402 * responsible for sending and/or receiving packets on a specific
1403 * queue id bound to an AF_XDP socket. The flags field specifies if
1404 * only RX, only Tx, or both should be woken up using the flags
1405 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1406 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1408 * Add, change, delete or get information on an IPv4 tunnel.
1409 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1410 * If a device is paired with a peer device, return the peer instance.
1411 * The caller must be under RCU read context.
1412 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1413 * Get the forwarding path to reach the real device from the HW destination address
1414 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1415 * const struct skb_shared_hwtstamps *hwtstamps,
1417 * Get hardware timestamp based on normal/adjustable time or free running
1418 * cycle counter. This function is required if physical clock supports a
1419 * free running cycle counter.
1421 * int (*ndo_hwtstamp_get)(struct net_device *dev,
1422 * struct kernel_hwtstamp_config *kernel_config);
1423 * Get the currently configured hardware timestamping parameters for the
1426 * int (*ndo_hwtstamp_set)(struct net_device *dev,
1427 * struct kernel_hwtstamp_config *kernel_config,
1428 * struct netlink_ext_ack *extack);
1429 * Change the hardware timestamping parameters for NIC device.
1431 struct net_device_ops {
1432 int (*ndo_init)(struct net_device *dev);
1433 void (*ndo_uninit)(struct net_device *dev);
1434 int (*ndo_open)(struct net_device *dev);
1435 int (*ndo_stop)(struct net_device *dev);
1436 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1437 struct net_device *dev);
1438 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1439 struct net_device *dev,
1440 netdev_features_t features);
1441 u16 (*ndo_select_queue)(struct net_device *dev,
1442 struct sk_buff *skb,
1443 struct net_device *sb_dev);
1444 void (*ndo_change_rx_flags)(struct net_device *dev,
1446 void (*ndo_set_rx_mode)(struct net_device *dev);
1447 int (*ndo_set_mac_address)(struct net_device *dev,
1449 int (*ndo_validate_addr)(struct net_device *dev);
1450 int (*ndo_do_ioctl)(struct net_device *dev,
1451 struct ifreq *ifr, int cmd);
1452 int (*ndo_eth_ioctl)(struct net_device *dev,
1453 struct ifreq *ifr, int cmd);
1454 int (*ndo_siocbond)(struct net_device *dev,
1455 struct ifreq *ifr, int cmd);
1456 int (*ndo_siocwandev)(struct net_device *dev,
1457 struct if_settings *ifs);
1458 int (*ndo_siocdevprivate)(struct net_device *dev,
1460 void __user *data, int cmd);
1461 int (*ndo_set_config)(struct net_device *dev,
1463 int (*ndo_change_mtu)(struct net_device *dev,
1465 int (*ndo_neigh_setup)(struct net_device *dev,
1466 struct neigh_parms *);
1467 void (*ndo_tx_timeout) (struct net_device *dev,
1468 unsigned int txqueue);
1470 void (*ndo_get_stats64)(struct net_device *dev,
1471 struct rtnl_link_stats64 *storage);
1472 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1473 int (*ndo_get_offload_stats)(int attr_id,
1474 const struct net_device *dev,
1476 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1478 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1479 __be16 proto, u16 vid);
1480 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1481 __be16 proto, u16 vid);
1482 #ifdef CONFIG_NET_POLL_CONTROLLER
1483 void (*ndo_poll_controller)(struct net_device *dev);
1484 int (*ndo_netpoll_setup)(struct net_device *dev,
1485 struct netpoll_info *info);
1486 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1488 int (*ndo_set_vf_mac)(struct net_device *dev,
1489 int queue, u8 *mac);
1490 int (*ndo_set_vf_vlan)(struct net_device *dev,
1491 int queue, u16 vlan,
1492 u8 qos, __be16 proto);
1493 int (*ndo_set_vf_rate)(struct net_device *dev,
1494 int vf, int min_tx_rate,
1496 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1497 int vf, bool setting);
1498 int (*ndo_set_vf_trust)(struct net_device *dev,
1499 int vf, bool setting);
1500 int (*ndo_get_vf_config)(struct net_device *dev,
1502 struct ifla_vf_info *ivf);
1503 int (*ndo_set_vf_link_state)(struct net_device *dev,
1504 int vf, int link_state);
1505 int (*ndo_get_vf_stats)(struct net_device *dev,
1507 struct ifla_vf_stats
1509 int (*ndo_set_vf_port)(struct net_device *dev,
1511 struct nlattr *port[]);
1512 int (*ndo_get_vf_port)(struct net_device *dev,
1513 int vf, struct sk_buff *skb);
1514 int (*ndo_get_vf_guid)(struct net_device *dev,
1516 struct ifla_vf_guid *node_guid,
1517 struct ifla_vf_guid *port_guid);
1518 int (*ndo_set_vf_guid)(struct net_device *dev,
1521 int (*ndo_set_vf_rss_query_en)(
1522 struct net_device *dev,
1523 int vf, bool setting);
1524 int (*ndo_setup_tc)(struct net_device *dev,
1525 enum tc_setup_type type,
1527 #if IS_ENABLED(CONFIG_FCOE)
1528 int (*ndo_fcoe_enable)(struct net_device *dev);
1529 int (*ndo_fcoe_disable)(struct net_device *dev);
1530 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1532 struct scatterlist *sgl,
1534 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1536 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1538 struct scatterlist *sgl,
1540 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1541 struct netdev_fcoe_hbainfo *hbainfo);
1544 #if IS_ENABLED(CONFIG_LIBFCOE)
1545 #define NETDEV_FCOE_WWNN 0
1546 #define NETDEV_FCOE_WWPN 1
1547 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1548 u64 *wwn, int type);
1551 #ifdef CONFIG_RFS_ACCEL
1552 int (*ndo_rx_flow_steer)(struct net_device *dev,
1553 const struct sk_buff *skb,
1557 int (*ndo_add_slave)(struct net_device *dev,
1558 struct net_device *slave_dev,
1559 struct netlink_ext_ack *extack);
1560 int (*ndo_del_slave)(struct net_device *dev,
1561 struct net_device *slave_dev);
1562 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1563 struct sk_buff *skb,
1565 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1567 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1568 netdev_features_t features);
1569 int (*ndo_set_features)(struct net_device *dev,
1570 netdev_features_t features);
1571 int (*ndo_neigh_construct)(struct net_device *dev,
1572 struct neighbour *n);
1573 void (*ndo_neigh_destroy)(struct net_device *dev,
1574 struct neighbour *n);
1576 int (*ndo_fdb_add)(struct ndmsg *ndm,
1577 struct nlattr *tb[],
1578 struct net_device *dev,
1579 const unsigned char *addr,
1582 struct netlink_ext_ack *extack);
1583 int (*ndo_fdb_del)(struct ndmsg *ndm,
1584 struct nlattr *tb[],
1585 struct net_device *dev,
1586 const unsigned char *addr,
1587 u16 vid, struct netlink_ext_ack *extack);
1588 int (*ndo_fdb_del_bulk)(struct nlmsghdr *nlh,
1589 struct net_device *dev,
1590 struct netlink_ext_ack *extack);
1591 int (*ndo_fdb_dump)(struct sk_buff *skb,
1592 struct netlink_callback *cb,
1593 struct net_device *dev,
1594 struct net_device *filter_dev,
1596 int (*ndo_fdb_get)(struct sk_buff *skb,
1597 struct nlattr *tb[],
1598 struct net_device *dev,
1599 const unsigned char *addr,
1600 u16 vid, u32 portid, u32 seq,
1601 struct netlink_ext_ack *extack);
1602 int (*ndo_mdb_add)(struct net_device *dev,
1603 struct nlattr *tb[],
1605 struct netlink_ext_ack *extack);
1606 int (*ndo_mdb_del)(struct net_device *dev,
1607 struct nlattr *tb[],
1608 struct netlink_ext_ack *extack);
1609 int (*ndo_mdb_dump)(struct net_device *dev,
1610 struct sk_buff *skb,
1611 struct netlink_callback *cb);
1612 int (*ndo_mdb_get)(struct net_device *dev,
1613 struct nlattr *tb[], u32 portid,
1615 struct netlink_ext_ack *extack);
1616 int (*ndo_bridge_setlink)(struct net_device *dev,
1617 struct nlmsghdr *nlh,
1619 struct netlink_ext_ack *extack);
1620 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1622 struct net_device *dev,
1625 int (*ndo_bridge_dellink)(struct net_device *dev,
1626 struct nlmsghdr *nlh,
1628 int (*ndo_change_carrier)(struct net_device *dev,
1630 int (*ndo_get_phys_port_id)(struct net_device *dev,
1631 struct netdev_phys_item_id *ppid);
1632 int (*ndo_get_port_parent_id)(struct net_device *dev,
1633 struct netdev_phys_item_id *ppid);
1634 int (*ndo_get_phys_port_name)(struct net_device *dev,
1635 char *name, size_t len);
1636 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1637 struct net_device *dev);
1638 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1641 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1644 int (*ndo_get_iflink)(const struct net_device *dev);
1645 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1646 struct sk_buff *skb);
1647 void (*ndo_set_rx_headroom)(struct net_device *dev,
1648 int needed_headroom);
1649 int (*ndo_bpf)(struct net_device *dev,
1650 struct netdev_bpf *bpf);
1651 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1652 struct xdp_frame **xdp,
1654 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1655 struct xdp_buff *xdp);
1656 int (*ndo_xsk_wakeup)(struct net_device *dev,
1657 u32 queue_id, u32 flags);
1658 int (*ndo_tunnel_ctl)(struct net_device *dev,
1659 struct ip_tunnel_parm *p, int cmd);
1660 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1661 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1662 struct net_device_path *path);
1663 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1664 const struct skb_shared_hwtstamps *hwtstamps,
1666 int (*ndo_hwtstamp_get)(struct net_device *dev,
1667 struct kernel_hwtstamp_config *kernel_config);
1668 int (*ndo_hwtstamp_set)(struct net_device *dev,
1669 struct kernel_hwtstamp_config *kernel_config,
1670 struct netlink_ext_ack *extack);
1674 * enum netdev_priv_flags - &struct net_device priv_flags
1676 * These are the &struct net_device, they are only set internally
1677 * by drivers and used in the kernel. These flags are invisible to
1678 * userspace; this means that the order of these flags can change
1679 * during any kernel release.
1681 * You should have a pretty good reason to be extending these flags.
1683 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1684 * @IFF_EBRIDGE: Ethernet bridging device
1685 * @IFF_BONDING: bonding master or slave
1686 * @IFF_ISATAP: ISATAP interface (RFC4214)
1687 * @IFF_WAN_HDLC: WAN HDLC device
1688 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1690 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1691 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1692 * @IFF_MACVLAN_PORT: device used as macvlan port
1693 * @IFF_BRIDGE_PORT: device used as bridge port
1694 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1695 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1696 * @IFF_UNICAST_FLT: Supports unicast filtering
1697 * @IFF_TEAM_PORT: device used as team port
1698 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1699 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1700 * change when it's running
1701 * @IFF_MACVLAN: Macvlan device
1702 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1703 * underlying stacked devices
1704 * @IFF_L3MDEV_MASTER: device is an L3 master device
1705 * @IFF_NO_QUEUE: device can run without qdisc attached
1706 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1707 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1708 * @IFF_TEAM: device is a team device
1709 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1710 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1711 * entity (i.e. the master device for bridged veth)
1712 * @IFF_MACSEC: device is a MACsec device
1713 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1714 * @IFF_FAILOVER: device is a failover master device
1715 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1716 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1717 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1718 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1719 * skb_headlen(skb) == 0 (data starts from frag0)
1720 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1721 * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1722 * ndo_hwtstamp_set() for all timestamp requests regardless of source,
1723 * even if those aren't HWTSTAMP_SOURCE_NETDEV.
1725 enum netdev_priv_flags {
1726 IFF_802_1Q_VLAN = 1<<0,
1730 IFF_WAN_HDLC = 1<<4,
1731 IFF_XMIT_DST_RELEASE = 1<<5,
1732 IFF_DONT_BRIDGE = 1<<6,
1733 IFF_DISABLE_NETPOLL = 1<<7,
1734 IFF_MACVLAN_PORT = 1<<8,
1735 IFF_BRIDGE_PORT = 1<<9,
1736 IFF_OVS_DATAPATH = 1<<10,
1737 IFF_TX_SKB_SHARING = 1<<11,
1738 IFF_UNICAST_FLT = 1<<12,
1739 IFF_TEAM_PORT = 1<<13,
1740 IFF_SUPP_NOFCS = 1<<14,
1741 IFF_LIVE_ADDR_CHANGE = 1<<15,
1742 IFF_MACVLAN = 1<<16,
1743 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1744 IFF_L3MDEV_MASTER = 1<<18,
1745 IFF_NO_QUEUE = 1<<19,
1746 IFF_OPENVSWITCH = 1<<20,
1747 IFF_L3MDEV_SLAVE = 1<<21,
1749 IFF_RXFH_CONFIGURED = 1<<23,
1750 IFF_PHONY_HEADROOM = 1<<24,
1752 IFF_NO_RX_HANDLER = 1<<26,
1753 IFF_FAILOVER = 1<<27,
1754 IFF_FAILOVER_SLAVE = 1<<28,
1755 IFF_L3MDEV_RX_HANDLER = 1<<29,
1756 IFF_NO_ADDRCONF = BIT_ULL(30),
1757 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1758 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1759 IFF_SEE_ALL_HWTSTAMP_REQUESTS = BIT_ULL(33),
1762 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1763 #define IFF_EBRIDGE IFF_EBRIDGE
1764 #define IFF_BONDING IFF_BONDING
1765 #define IFF_ISATAP IFF_ISATAP
1766 #define IFF_WAN_HDLC IFF_WAN_HDLC
1767 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1768 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1769 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1770 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1771 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1772 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1773 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1774 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1775 #define IFF_TEAM_PORT IFF_TEAM_PORT
1776 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1777 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1778 #define IFF_MACVLAN IFF_MACVLAN
1779 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1780 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1781 #define IFF_NO_QUEUE IFF_NO_QUEUE
1782 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1783 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1784 #define IFF_TEAM IFF_TEAM
1785 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1786 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1787 #define IFF_MACSEC IFF_MACSEC
1788 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1789 #define IFF_FAILOVER IFF_FAILOVER
1790 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1791 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1792 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1794 /* Specifies the type of the struct net_device::ml_priv pointer */
1795 enum netdev_ml_priv_type {
1801 * struct net_device - The DEVICE structure.
1803 * Actually, this whole structure is a big mistake. It mixes I/O
1804 * data with strictly "high-level" data, and it has to know about
1805 * almost every data structure used in the INET module.
1807 * @name: This is the first field of the "visible" part of this structure
1808 * (i.e. as seen by users in the "Space.c" file). It is the name
1811 * @name_node: Name hashlist node
1812 * @ifalias: SNMP alias
1813 * @mem_end: Shared memory end
1814 * @mem_start: Shared memory start
1815 * @base_addr: Device I/O address
1816 * @irq: Device IRQ number
1818 * @state: Generic network queuing layer state, see netdev_state_t
1819 * @dev_list: The global list of network devices
1820 * @napi_list: List entry used for polling NAPI devices
1821 * @unreg_list: List entry when we are unregistering the
1822 * device; see the function unregister_netdev
1823 * @close_list: List entry used when we are closing the device
1824 * @ptype_all: Device-specific packet handlers for all protocols
1825 * @ptype_specific: Device-specific, protocol-specific packet handlers
1827 * @adj_list: Directly linked devices, like slaves for bonding
1828 * @features: Currently active device features
1829 * @hw_features: User-changeable features
1831 * @wanted_features: User-requested features
1832 * @vlan_features: Mask of features inheritable by VLAN devices
1834 * @hw_enc_features: Mask of features inherited by encapsulating devices
1835 * This field indicates what encapsulation
1836 * offloads the hardware is capable of doing,
1837 * and drivers will need to set them appropriately.
1839 * @mpls_features: Mask of features inheritable by MPLS
1840 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1842 * @ifindex: interface index
1843 * @group: The group the device belongs to
1845 * @stats: Statistics struct, which was left as a legacy, use
1846 * rtnl_link_stats64 instead
1848 * @core_stats: core networking counters,
1849 * do not use this in drivers
1850 * @carrier_up_count: Number of times the carrier has been up
1851 * @carrier_down_count: Number of times the carrier has been down
1853 * @wireless_handlers: List of functions to handle Wireless Extensions,
1855 * see <net/iw_handler.h> for details.
1856 * @wireless_data: Instance data managed by the core of wireless extensions
1858 * @netdev_ops: Includes several pointers to callbacks,
1859 * if one wants to override the ndo_*() functions
1860 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1861 * @ethtool_ops: Management operations
1862 * @l3mdev_ops: Layer 3 master device operations
1863 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1864 * discovery handling. Necessary for e.g. 6LoWPAN.
1865 * @xfrmdev_ops: Transformation offload operations
1866 * @tlsdev_ops: Transport Layer Security offload operations
1867 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1868 * of Layer 2 headers.
1870 * @flags: Interface flags (a la BSD)
1871 * @xdp_features: XDP capability supported by the device
1872 * @priv_flags: Like 'flags' but invisible to userspace,
1873 * see if.h for the definitions
1874 * @gflags: Global flags ( kept as legacy )
1875 * @padded: How much padding added by alloc_netdev()
1876 * @operstate: RFC2863 operstate
1877 * @link_mode: Mapping policy to operstate
1878 * @if_port: Selectable AUI, TP, ...
1880 * @mtu: Interface MTU value
1881 * @min_mtu: Interface Minimum MTU value
1882 * @max_mtu: Interface Maximum MTU value
1883 * @type: Interface hardware type
1884 * @hard_header_len: Maximum hardware header length.
1885 * @min_header_len: Minimum hardware header length
1887 * @needed_headroom: Extra headroom the hardware may need, but not in all
1888 * cases can this be guaranteed
1889 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1890 * cases can this be guaranteed. Some cases also use
1891 * LL_MAX_HEADER instead to allocate the skb
1893 * interface address info:
1895 * @perm_addr: Permanent hw address
1896 * @addr_assign_type: Hw address assignment type
1897 * @addr_len: Hardware address length
1898 * @upper_level: Maximum depth level of upper devices.
1899 * @lower_level: Maximum depth level of lower devices.
1900 * @neigh_priv_len: Used in neigh_alloc()
1901 * @dev_id: Used to differentiate devices that share
1902 * the same link layer address
1903 * @dev_port: Used to differentiate devices that share
1905 * @addr_list_lock: XXX: need comments on this one
1906 * @name_assign_type: network interface name assignment type
1907 * @uc_promisc: Counter that indicates promiscuous mode
1908 * has been enabled due to the need to listen to
1909 * additional unicast addresses in a device that
1910 * does not implement ndo_set_rx_mode()
1911 * @uc: unicast mac addresses
1912 * @mc: multicast mac addresses
1913 * @dev_addrs: list of device hw addresses
1914 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1915 * @promiscuity: Number of times the NIC is told to work in
1916 * promiscuous mode; if it becomes 0 the NIC will
1917 * exit promiscuous mode
1918 * @allmulti: Counter, enables or disables allmulticast mode
1920 * @vlan_info: VLAN info
1921 * @dsa_ptr: dsa specific data
1922 * @tipc_ptr: TIPC specific data
1923 * @atalk_ptr: AppleTalk link
1924 * @ip_ptr: IPv4 specific data
1925 * @ip6_ptr: IPv6 specific data
1926 * @ax25_ptr: AX.25 specific data
1927 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1928 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1930 * @mpls_ptr: mpls_dev struct pointer
1931 * @mctp_ptr: MCTP specific data
1933 * @dev_addr: Hw address (before bcast,
1934 * because most packets are unicast)
1936 * @_rx: Array of RX queues
1937 * @num_rx_queues: Number of RX queues
1938 * allocated at register_netdev() time
1939 * @real_num_rx_queues: Number of RX queues currently active in device
1940 * @xdp_prog: XDP sockets filter program pointer
1941 * @gro_flush_timeout: timeout for GRO layer in NAPI
1942 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1943 * allow to avoid NIC hard IRQ, on busy queues.
1945 * @rx_handler: handler for received packets
1946 * @rx_handler_data: XXX: need comments on this one
1947 * @tcx_ingress: BPF & clsact qdisc specific data for ingress processing
1948 * @ingress_queue: XXX: need comments on this one
1949 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1950 * @broadcast: hw bcast address
1952 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1953 * indexed by RX queue number. Assigned by driver.
1954 * This must only be set if the ndo_rx_flow_steer
1955 * operation is defined
1956 * @index_hlist: Device index hash chain
1958 * @_tx: Array of TX queues
1959 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1960 * @real_num_tx_queues: Number of TX queues currently active in device
1961 * @qdisc: Root qdisc from userspace point of view
1962 * @tx_queue_len: Max frames per queue allowed
1963 * @tx_global_lock: XXX: need comments on this one
1964 * @xdp_bulkq: XDP device bulk queue
1965 * @xps_maps: all CPUs/RXQs maps for XPS device
1967 * @xps_maps: XXX: need comments on this one
1968 * @tcx_egress: BPF & clsact qdisc specific data for egress processing
1969 * @nf_hooks_egress: netfilter hooks executed for egress packets
1970 * @qdisc_hash: qdisc hash table
1971 * @watchdog_timeo: Represents the timeout that is used by
1972 * the watchdog (see dev_watchdog())
1973 * @watchdog_timer: List of timers
1975 * @proto_down_reason: reason a netdev interface is held down
1976 * @pcpu_refcnt: Number of references to this device
1977 * @dev_refcnt: Number of references to this device
1978 * @refcnt_tracker: Tracker directory for tracked references to this device
1979 * @todo_list: Delayed register/unregister
1980 * @link_watch_list: XXX: need comments on this one
1982 * @reg_state: Register/unregister state machine
1983 * @dismantle: Device is going to be freed
1984 * @rtnl_link_state: This enum represents the phases of creating
1987 * @needs_free_netdev: Should unregister perform free_netdev?
1988 * @priv_destructor: Called from unregister
1989 * @npinfo: XXX: need comments on this one
1990 * @nd_net: Network namespace this network device is inside
1992 * @ml_priv: Mid-layer private
1993 * @ml_priv_type: Mid-layer private type
1994 * @lstats: Loopback statistics
1995 * @tstats: Tunnel statistics
1996 * @dstats: Dummy statistics
1997 * @vstats: Virtual ethernet statistics
2002 * @dm_private: Drop monitor private
2004 * @dev: Class/net/name entry
2005 * @sysfs_groups: Space for optional device, statistics and wireless
2008 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
2009 * @rtnl_link_ops: Rtnl_link_ops
2011 * @gso_max_size: Maximum size of generic segmentation offload
2012 * @tso_max_size: Device (as in HW) limit on the max TSO request size
2013 * @gso_max_segs: Maximum number of segments that can be passed to the
2015 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
2016 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
2019 * @dcbnl_ops: Data Center Bridging netlink ops
2020 * @num_tc: Number of traffic classes in the net device
2021 * @tc_to_txq: XXX: need comments on this one
2022 * @prio_tc_map: XXX: need comments on this one
2024 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
2026 * @priomap: XXX: need comments on this one
2027 * @phydev: Physical device may attach itself
2028 * for hardware timestamping
2029 * @sfp_bus: attached &struct sfp_bus structure.
2031 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2033 * @proto_down: protocol port state information can be sent to the
2034 * switch driver and used to set the phys state of the
2037 * @wol_enabled: Wake-on-LAN is enabled
2039 * @threaded: napi threaded mode is enabled
2041 * @net_notifier_list: List of per-net netdev notifier block
2042 * that follow this device when it is moved
2043 * to another network namespace.
2045 * @macsec_ops: MACsec offloading ops
2047 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2048 * offload capabilities of the device
2049 * @udp_tunnel_nic: UDP tunnel offload state
2050 * @xdp_state: stores info on attached XDP BPF programs
2052 * @nested_level: Used as a parameter of spin_lock_nested() of
2053 * dev->addr_list_lock.
2054 * @unlink_list: As netif_addr_lock() can be called recursively,
2055 * keep a list of interfaces to be deleted.
2056 * @gro_max_size: Maximum size of aggregated packet in generic
2057 * receive offload (GRO)
2058 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2059 * receive offload (GRO), for IPv4.
2060 * @xdp_zc_max_segs: Maximum number of segments supported by AF_XDP
2063 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2064 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2065 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2066 * @dev_registered_tracker: tracker for reference held while
2068 * @offload_xstats_l3: L3 HW stats for this netdevice.
2070 * @devlink_port: Pointer to related devlink port structure.
2071 * Assigned by a driver before netdev registration using
2072 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2073 * during the time netdevice is registered.
2075 * @dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2076 * where the clock is recovered.
2078 * FIXME: cleanup struct net_device such that network protocol info
2083 char name[IFNAMSIZ];
2084 struct netdev_name_node *name_node;
2085 struct dev_ifalias __rcu *ifalias;
2087 * I/O specific fields
2088 * FIXME: Merge these and struct ifmap into one
2090 unsigned long mem_end;
2091 unsigned long mem_start;
2092 unsigned long base_addr;
2095 * Some hardware also needs these fields (state,dev_list,
2096 * napi_list,unreg_list,close_list) but they are not
2097 * part of the usual set specified in Space.c.
2100 unsigned long state;
2102 struct list_head dev_list;
2103 struct list_head napi_list;
2104 struct list_head unreg_list;
2105 struct list_head close_list;
2106 struct list_head ptype_all;
2107 struct list_head ptype_specific;
2110 struct list_head upper;
2111 struct list_head lower;
2114 /* Read-mostly cache-line for fast-path access */
2116 xdp_features_t xdp_features;
2117 unsigned long long priv_flags;
2118 const struct net_device_ops *netdev_ops;
2119 const struct xdp_metadata_ops *xdp_metadata_ops;
2121 unsigned short gflags;
2122 unsigned short hard_header_len;
2124 /* Note : dev->mtu is often read without holding a lock.
2125 * Writers usually hold RTNL.
2126 * It is recommended to use READ_ONCE() to annotate the reads,
2127 * and to use WRITE_ONCE() to annotate the writes.
2130 unsigned short needed_headroom;
2131 unsigned short needed_tailroom;
2133 netdev_features_t features;
2134 netdev_features_t hw_features;
2135 netdev_features_t wanted_features;
2136 netdev_features_t vlan_features;
2137 netdev_features_t hw_enc_features;
2138 netdev_features_t mpls_features;
2139 netdev_features_t gso_partial_features;
2141 unsigned int min_mtu;
2142 unsigned int max_mtu;
2143 unsigned short type;
2144 unsigned char min_header_len;
2145 unsigned char name_assign_type;
2149 struct net_device_stats stats; /* not used by modern drivers */
2151 struct net_device_core_stats __percpu *core_stats;
2153 /* Stats to monitor link on/off, flapping */
2154 atomic_t carrier_up_count;
2155 atomic_t carrier_down_count;
2157 #ifdef CONFIG_WIRELESS_EXT
2158 const struct iw_handler_def *wireless_handlers;
2159 struct iw_public_data *wireless_data;
2161 const struct ethtool_ops *ethtool_ops;
2162 #ifdef CONFIG_NET_L3_MASTER_DEV
2163 const struct l3mdev_ops *l3mdev_ops;
2165 #if IS_ENABLED(CONFIG_IPV6)
2166 const struct ndisc_ops *ndisc_ops;
2169 #ifdef CONFIG_XFRM_OFFLOAD
2170 const struct xfrmdev_ops *xfrmdev_ops;
2173 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2174 const struct tlsdev_ops *tlsdev_ops;
2177 const struct header_ops *header_ops;
2179 unsigned char operstate;
2180 unsigned char link_mode;
2182 unsigned char if_port;
2185 /* Interface address info. */
2186 unsigned char perm_addr[MAX_ADDR_LEN];
2187 unsigned char addr_assign_type;
2188 unsigned char addr_len;
2189 unsigned char upper_level;
2190 unsigned char lower_level;
2192 unsigned short neigh_priv_len;
2193 unsigned short dev_id;
2194 unsigned short dev_port;
2195 unsigned short padded;
2197 spinlock_t addr_list_lock;
2200 struct netdev_hw_addr_list uc;
2201 struct netdev_hw_addr_list mc;
2202 struct netdev_hw_addr_list dev_addrs;
2205 struct kset *queues_kset;
2207 #ifdef CONFIG_LOCKDEP
2208 struct list_head unlink_list;
2210 unsigned int promiscuity;
2211 unsigned int allmulti;
2213 #ifdef CONFIG_LOCKDEP
2214 unsigned char nested_level;
2218 /* Protocol-specific pointers */
2220 struct in_device __rcu *ip_ptr;
2221 struct inet6_dev __rcu *ip6_ptr;
2222 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2223 struct vlan_info __rcu *vlan_info;
2225 #if IS_ENABLED(CONFIG_NET_DSA)
2226 struct dsa_port *dsa_ptr;
2228 #if IS_ENABLED(CONFIG_TIPC)
2229 struct tipc_bearer __rcu *tipc_ptr;
2231 #if IS_ENABLED(CONFIG_ATALK)
2234 #if IS_ENABLED(CONFIG_AX25)
2237 #if IS_ENABLED(CONFIG_CFG80211)
2238 struct wireless_dev *ieee80211_ptr;
2240 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2241 struct wpan_dev *ieee802154_ptr;
2243 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2244 struct mpls_dev __rcu *mpls_ptr;
2246 #if IS_ENABLED(CONFIG_MCTP)
2247 struct mctp_dev __rcu *mctp_ptr;
2251 * Cache lines mostly used on receive path (including eth_type_trans())
2253 /* Interface address info used in eth_type_trans() */
2254 const unsigned char *dev_addr;
2256 struct netdev_rx_queue *_rx;
2257 unsigned int num_rx_queues;
2258 unsigned int real_num_rx_queues;
2260 struct bpf_prog __rcu *xdp_prog;
2261 unsigned long gro_flush_timeout;
2262 int napi_defer_hard_irqs;
2263 #define GRO_LEGACY_MAX_SIZE 65536u
2264 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2265 * and shinfo->gso_segs is a 16bit field.
2267 #define GRO_MAX_SIZE (8 * 65535u)
2268 unsigned int gro_max_size;
2269 unsigned int gro_ipv4_max_size;
2270 unsigned int xdp_zc_max_segs;
2271 rx_handler_func_t __rcu *rx_handler;
2272 void __rcu *rx_handler_data;
2273 #ifdef CONFIG_NET_XGRESS
2274 struct bpf_mprog_entry __rcu *tcx_ingress;
2276 struct netdev_queue __rcu *ingress_queue;
2277 #ifdef CONFIG_NETFILTER_INGRESS
2278 struct nf_hook_entries __rcu *nf_hooks_ingress;
2281 unsigned char broadcast[MAX_ADDR_LEN];
2282 #ifdef CONFIG_RFS_ACCEL
2283 struct cpu_rmap *rx_cpu_rmap;
2285 struct hlist_node index_hlist;
2288 * Cache lines mostly used on transmit path
2290 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2291 unsigned int num_tx_queues;
2292 unsigned int real_num_tx_queues;
2293 struct Qdisc __rcu *qdisc;
2294 unsigned int tx_queue_len;
2295 spinlock_t tx_global_lock;
2297 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2300 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2302 #ifdef CONFIG_NET_XGRESS
2303 struct bpf_mprog_entry __rcu *tcx_egress;
2305 #ifdef CONFIG_NETFILTER_EGRESS
2306 struct nf_hook_entries __rcu *nf_hooks_egress;
2309 #ifdef CONFIG_NET_SCHED
2310 DECLARE_HASHTABLE (qdisc_hash, 4);
2312 /* These may be needed for future network-power-down code. */
2313 struct timer_list watchdog_timer;
2316 u32 proto_down_reason;
2318 struct list_head todo_list;
2320 #ifdef CONFIG_PCPU_DEV_REFCNT
2321 int __percpu *pcpu_refcnt;
2323 refcount_t dev_refcnt;
2325 struct ref_tracker_dir refcnt_tracker;
2327 struct list_head link_watch_list;
2329 enum { NETREG_UNINITIALIZED=0,
2330 NETREG_REGISTERED, /* completed register_netdevice */
2331 NETREG_UNREGISTERING, /* called unregister_netdevice */
2332 NETREG_UNREGISTERED, /* completed unregister todo */
2333 NETREG_RELEASED, /* called free_netdev */
2334 NETREG_DUMMY, /* dummy device for NAPI poll */
2340 RTNL_LINK_INITIALIZED,
2341 RTNL_LINK_INITIALIZING,
2342 } rtnl_link_state:16;
2344 bool needs_free_netdev;
2345 void (*priv_destructor)(struct net_device *dev);
2347 #ifdef CONFIG_NETPOLL
2348 struct netpoll_info __rcu *npinfo;
2351 possible_net_t nd_net;
2353 /* mid-layer private */
2355 enum netdev_ml_priv_type ml_priv_type;
2358 struct pcpu_lstats __percpu *lstats;
2359 struct pcpu_sw_netstats __percpu *tstats;
2360 struct pcpu_dstats __percpu *dstats;
2363 #if IS_ENABLED(CONFIG_GARP)
2364 struct garp_port __rcu *garp_port;
2366 #if IS_ENABLED(CONFIG_MRP)
2367 struct mrp_port __rcu *mrp_port;
2369 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2370 struct dm_hw_stat_delta __rcu *dm_private;
2373 const struct attribute_group *sysfs_groups[4];
2374 const struct attribute_group *sysfs_rx_queue_group;
2376 const struct rtnl_link_ops *rtnl_link_ops;
2378 /* for setting kernel sock attribute on TCP connection setup */
2379 #define GSO_MAX_SEGS 65535u
2380 #define GSO_LEGACY_MAX_SIZE 65536u
2381 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2382 * and shinfo->gso_segs is a 16bit field.
2384 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2386 unsigned int gso_max_size;
2387 #define TSO_LEGACY_MAX_SIZE 65536
2388 #define TSO_MAX_SIZE UINT_MAX
2389 unsigned int tso_max_size;
2391 #define TSO_MAX_SEGS U16_MAX
2393 unsigned int gso_ipv4_max_size;
2396 const struct dcbnl_rtnl_ops *dcbnl_ops;
2399 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2400 u8 prio_tc_map[TC_BITMASK + 1];
2402 #if IS_ENABLED(CONFIG_FCOE)
2403 unsigned int fcoe_ddp_xid;
2405 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2406 struct netprio_map __rcu *priomap;
2408 struct phy_device *phydev;
2409 struct sfp_bus *sfp_bus;
2410 struct lock_class_key *qdisc_tx_busylock;
2412 unsigned wol_enabled:1;
2413 unsigned threaded:1;
2415 struct list_head net_notifier_list;
2417 #if IS_ENABLED(CONFIG_MACSEC)
2418 /* MACsec management functions */
2419 const struct macsec_ops *macsec_ops;
2421 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2422 struct udp_tunnel_nic *udp_tunnel_nic;
2424 /* protected by rtnl_lock */
2425 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2427 u8 dev_addr_shadow[MAX_ADDR_LEN];
2428 netdevice_tracker linkwatch_dev_tracker;
2429 netdevice_tracker watchdog_dev_tracker;
2430 netdevice_tracker dev_registered_tracker;
2431 struct rtnl_hw_stats64 *offload_xstats_l3;
2433 struct devlink_port *devlink_port;
2435 #if IS_ENABLED(CONFIG_DPLL)
2436 struct dpll_pin *dpll_pin;
2439 #define to_net_dev(d) container_of(d, struct net_device, dev)
2442 * Driver should use this to assign devlink port instance to a netdevice
2443 * before it registers the netdevice. Therefore devlink_port is static
2444 * during the netdev lifetime after it is registered.
2446 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2448 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2449 ((dev)->devlink_port = (port)); \
2452 static inline bool netif_elide_gro(const struct net_device *dev)
2454 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2459 #define NETDEV_ALIGN 32
2462 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2464 return dev->prio_tc_map[prio & TC_BITMASK];
2468 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2470 if (tc >= dev->num_tc)
2473 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2477 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2478 void netdev_reset_tc(struct net_device *dev);
2479 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2480 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2483 int netdev_get_num_tc(struct net_device *dev)
2488 static inline void net_prefetch(void *p)
2491 #if L1_CACHE_BYTES < 128
2492 prefetch((u8 *)p + L1_CACHE_BYTES);
2496 static inline void net_prefetchw(void *p)
2499 #if L1_CACHE_BYTES < 128
2500 prefetchw((u8 *)p + L1_CACHE_BYTES);
2504 void netdev_unbind_sb_channel(struct net_device *dev,
2505 struct net_device *sb_dev);
2506 int netdev_bind_sb_channel_queue(struct net_device *dev,
2507 struct net_device *sb_dev,
2508 u8 tc, u16 count, u16 offset);
2509 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2510 static inline int netdev_get_sb_channel(struct net_device *dev)
2512 return max_t(int, -dev->num_tc, 0);
2516 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2519 DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2520 return &dev->_tx[index];
2523 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2524 const struct sk_buff *skb)
2526 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2529 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2530 void (*f)(struct net_device *,
2531 struct netdev_queue *,
2537 for (i = 0; i < dev->num_tx_queues; i++)
2538 f(dev, &dev->_tx[i], arg);
2541 #define netdev_lockdep_set_classes(dev) \
2543 static struct lock_class_key qdisc_tx_busylock_key; \
2544 static struct lock_class_key qdisc_xmit_lock_key; \
2545 static struct lock_class_key dev_addr_list_lock_key; \
2548 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2549 lockdep_set_class(&(dev)->addr_list_lock, \
2550 &dev_addr_list_lock_key); \
2551 for (i = 0; i < (dev)->num_tx_queues; i++) \
2552 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2553 &qdisc_xmit_lock_key); \
2556 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2557 struct net_device *sb_dev);
2558 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2559 struct sk_buff *skb,
2560 struct net_device *sb_dev);
2562 /* returns the headroom that the master device needs to take in account
2563 * when forwarding to this dev
2565 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2567 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2570 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2572 if (dev->netdev_ops->ndo_set_rx_headroom)
2573 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2576 /* set the device rx headroom to the dev's default */
2577 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2579 netdev_set_rx_headroom(dev, -1);
2582 static inline void *netdev_get_ml_priv(struct net_device *dev,
2583 enum netdev_ml_priv_type type)
2585 if (dev->ml_priv_type != type)
2588 return dev->ml_priv;
2591 static inline void netdev_set_ml_priv(struct net_device *dev,
2593 enum netdev_ml_priv_type type)
2595 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2596 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2597 dev->ml_priv_type, type);
2598 WARN(!dev->ml_priv_type && dev->ml_priv,
2599 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2601 dev->ml_priv = ml_priv;
2602 dev->ml_priv_type = type;
2606 * Net namespace inlines
2609 struct net *dev_net(const struct net_device *dev)
2611 return read_pnet(&dev->nd_net);
2615 void dev_net_set(struct net_device *dev, struct net *net)
2617 write_pnet(&dev->nd_net, net);
2621 * netdev_priv - access network device private data
2622 * @dev: network device
2624 * Get network device private data
2626 static inline void *netdev_priv(const struct net_device *dev)
2628 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2631 /* Set the sysfs physical device reference for the network logical device
2632 * if set prior to registration will cause a symlink during initialization.
2634 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2636 /* Set the sysfs device type for the network logical device to allow
2637 * fine-grained identification of different network device types. For
2638 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2640 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2642 /* Default NAPI poll() weight
2643 * Device drivers are strongly advised to not use bigger value
2645 #define NAPI_POLL_WEIGHT 64
2647 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2648 int (*poll)(struct napi_struct *, int), int weight);
2651 * netif_napi_add() - initialize a NAPI context
2652 * @dev: network device
2653 * @napi: NAPI context
2654 * @poll: polling function
2656 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2657 * *any* of the other NAPI-related functions.
2660 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2661 int (*poll)(struct napi_struct *, int))
2663 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2667 netif_napi_add_tx_weight(struct net_device *dev,
2668 struct napi_struct *napi,
2669 int (*poll)(struct napi_struct *, int),
2672 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2673 netif_napi_add_weight(dev, napi, poll, weight);
2677 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2678 * @dev: network device
2679 * @napi: NAPI context
2680 * @poll: polling function
2682 * This variant of netif_napi_add() should be used from drivers using NAPI
2683 * to exclusively poll a TX queue.
2684 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2686 static inline void netif_napi_add_tx(struct net_device *dev,
2687 struct napi_struct *napi,
2688 int (*poll)(struct napi_struct *, int))
2690 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2694 * __netif_napi_del - remove a NAPI context
2695 * @napi: NAPI context
2697 * Warning: caller must observe RCU grace period before freeing memory
2698 * containing @napi. Drivers might want to call this helper to combine
2699 * all the needed RCU grace periods into a single one.
2701 void __netif_napi_del(struct napi_struct *napi);
2704 * netif_napi_del - remove a NAPI context
2705 * @napi: NAPI context
2707 * netif_napi_del() removes a NAPI context from the network device NAPI list
2709 static inline void netif_napi_del(struct napi_struct *napi)
2711 __netif_napi_del(napi);
2715 struct packet_type {
2716 __be16 type; /* This is really htons(ether_type). */
2717 bool ignore_outgoing;
2718 struct net_device *dev; /* NULL is wildcarded here */
2719 netdevice_tracker dev_tracker;
2720 int (*func) (struct sk_buff *,
2721 struct net_device *,
2722 struct packet_type *,
2723 struct net_device *);
2724 void (*list_func) (struct list_head *,
2725 struct packet_type *,
2726 struct net_device *);
2727 bool (*id_match)(struct packet_type *ptype,
2729 struct net *af_packet_net;
2730 void *af_packet_priv;
2731 struct list_head list;
2734 struct offload_callbacks {
2735 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2736 netdev_features_t features);
2737 struct sk_buff *(*gro_receive)(struct list_head *head,
2738 struct sk_buff *skb);
2739 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2742 struct packet_offload {
2743 __be16 type; /* This is really htons(ether_type). */
2745 struct offload_callbacks callbacks;
2746 struct list_head list;
2749 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2750 struct pcpu_sw_netstats {
2751 u64_stats_t rx_packets;
2752 u64_stats_t rx_bytes;
2753 u64_stats_t tx_packets;
2754 u64_stats_t tx_bytes;
2755 struct u64_stats_sync syncp;
2756 } __aligned(4 * sizeof(u64));
2758 struct pcpu_lstats {
2759 u64_stats_t packets;
2761 struct u64_stats_sync syncp;
2762 } __aligned(2 * sizeof(u64));
2764 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2766 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2768 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2770 u64_stats_update_begin(&tstats->syncp);
2771 u64_stats_add(&tstats->rx_bytes, len);
2772 u64_stats_inc(&tstats->rx_packets);
2773 u64_stats_update_end(&tstats->syncp);
2776 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2777 unsigned int packets,
2780 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2782 u64_stats_update_begin(&tstats->syncp);
2783 u64_stats_add(&tstats->tx_bytes, len);
2784 u64_stats_add(&tstats->tx_packets, packets);
2785 u64_stats_update_end(&tstats->syncp);
2788 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2790 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2792 u64_stats_update_begin(&lstats->syncp);
2793 u64_stats_add(&lstats->bytes, len);
2794 u64_stats_inc(&lstats->packets);
2795 u64_stats_update_end(&lstats->syncp);
2798 #define __netdev_alloc_pcpu_stats(type, gfp) \
2800 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2803 for_each_possible_cpu(__cpu) { \
2804 typeof(type) *stat; \
2805 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2806 u64_stats_init(&stat->syncp); \
2812 #define netdev_alloc_pcpu_stats(type) \
2813 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2815 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2817 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2820 for_each_possible_cpu(__cpu) { \
2821 typeof(type) *stat; \
2822 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2823 u64_stats_init(&stat->syncp); \
2829 enum netdev_lag_tx_type {
2830 NETDEV_LAG_TX_TYPE_UNKNOWN,
2831 NETDEV_LAG_TX_TYPE_RANDOM,
2832 NETDEV_LAG_TX_TYPE_BROADCAST,
2833 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2834 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2835 NETDEV_LAG_TX_TYPE_HASH,
2838 enum netdev_lag_hash {
2839 NETDEV_LAG_HASH_NONE,
2841 NETDEV_LAG_HASH_L34,
2842 NETDEV_LAG_HASH_L23,
2843 NETDEV_LAG_HASH_E23,
2844 NETDEV_LAG_HASH_E34,
2845 NETDEV_LAG_HASH_VLAN_SRCMAC,
2846 NETDEV_LAG_HASH_UNKNOWN,
2849 struct netdev_lag_upper_info {
2850 enum netdev_lag_tx_type tx_type;
2851 enum netdev_lag_hash hash_type;
2854 struct netdev_lag_lower_state_info {
2859 #include <linux/notifier.h>
2861 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2862 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2866 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2868 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2869 detected a hardware crash and restarted
2870 - we can use this eg to kick tcp sessions
2872 NETDEV_CHANGE, /* Notify device state change */
2875 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2876 NETDEV_CHANGEADDR, /* notify after the address change */
2877 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2881 NETDEV_BONDING_FAILOVER,
2883 NETDEV_PRE_TYPE_CHANGE,
2884 NETDEV_POST_TYPE_CHANGE,
2888 NETDEV_NOTIFY_PEERS,
2892 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2893 NETDEV_CHANGEINFODATA,
2894 NETDEV_BONDING_INFO,
2895 NETDEV_PRECHANGEUPPER,
2896 NETDEV_CHANGELOWERSTATE,
2897 NETDEV_UDP_TUNNEL_PUSH_INFO,
2898 NETDEV_UDP_TUNNEL_DROP_INFO,
2899 NETDEV_CHANGE_TX_QUEUE_LEN,
2900 NETDEV_CVLAN_FILTER_PUSH_INFO,
2901 NETDEV_CVLAN_FILTER_DROP_INFO,
2902 NETDEV_SVLAN_FILTER_PUSH_INFO,
2903 NETDEV_SVLAN_FILTER_DROP_INFO,
2904 NETDEV_OFFLOAD_XSTATS_ENABLE,
2905 NETDEV_OFFLOAD_XSTATS_DISABLE,
2906 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2907 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2908 NETDEV_XDP_FEAT_CHANGE,
2910 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2912 int register_netdevice_notifier(struct notifier_block *nb);
2913 int unregister_netdevice_notifier(struct notifier_block *nb);
2914 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2915 int unregister_netdevice_notifier_net(struct net *net,
2916 struct notifier_block *nb);
2917 int register_netdevice_notifier_dev_net(struct net_device *dev,
2918 struct notifier_block *nb,
2919 struct netdev_net_notifier *nn);
2920 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2921 struct notifier_block *nb,
2922 struct netdev_net_notifier *nn);
2924 struct netdev_notifier_info {
2925 struct net_device *dev;
2926 struct netlink_ext_ack *extack;
2929 struct netdev_notifier_info_ext {
2930 struct netdev_notifier_info info; /* must be first */
2936 struct netdev_notifier_change_info {
2937 struct netdev_notifier_info info; /* must be first */
2938 unsigned int flags_changed;
2941 struct netdev_notifier_changeupper_info {
2942 struct netdev_notifier_info info; /* must be first */
2943 struct net_device *upper_dev; /* new upper dev */
2944 bool master; /* is upper dev master */
2945 bool linking; /* is the notification for link or unlink */
2946 void *upper_info; /* upper dev info */
2949 struct netdev_notifier_changelowerstate_info {
2950 struct netdev_notifier_info info; /* must be first */
2951 void *lower_state_info; /* is lower dev state */
2954 struct netdev_notifier_pre_changeaddr_info {
2955 struct netdev_notifier_info info; /* must be first */
2956 const unsigned char *dev_addr;
2959 enum netdev_offload_xstats_type {
2960 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2963 struct netdev_notifier_offload_xstats_info {
2964 struct netdev_notifier_info info; /* must be first */
2965 enum netdev_offload_xstats_type type;
2968 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2969 struct netdev_notifier_offload_xstats_rd *report_delta;
2970 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2971 struct netdev_notifier_offload_xstats_ru *report_used;
2975 int netdev_offload_xstats_enable(struct net_device *dev,
2976 enum netdev_offload_xstats_type type,
2977 struct netlink_ext_ack *extack);
2978 int netdev_offload_xstats_disable(struct net_device *dev,
2979 enum netdev_offload_xstats_type type);
2980 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2981 enum netdev_offload_xstats_type type);
2982 int netdev_offload_xstats_get(struct net_device *dev,
2983 enum netdev_offload_xstats_type type,
2984 struct rtnl_hw_stats64 *stats, bool *used,
2985 struct netlink_ext_ack *extack);
2987 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2988 const struct rtnl_hw_stats64 *stats);
2990 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2991 void netdev_offload_xstats_push_delta(struct net_device *dev,
2992 enum netdev_offload_xstats_type type,
2993 const struct rtnl_hw_stats64 *stats);
2995 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2996 struct net_device *dev)
2999 info->extack = NULL;
3002 static inline struct net_device *
3003 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
3008 static inline struct netlink_ext_ack *
3009 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
3011 return info->extack;
3014 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
3015 int call_netdevice_notifiers_info(unsigned long val,
3016 struct netdev_notifier_info *info);
3018 extern rwlock_t dev_base_lock; /* Device list lock */
3020 #define for_each_netdev(net, d) \
3021 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3022 #define for_each_netdev_reverse(net, d) \
3023 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3024 #define for_each_netdev_rcu(net, d) \
3025 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3026 #define for_each_netdev_safe(net, d, n) \
3027 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3028 #define for_each_netdev_continue(net, d) \
3029 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3030 #define for_each_netdev_continue_reverse(net, d) \
3031 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3033 #define for_each_netdev_continue_rcu(net, d) \
3034 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3035 #define for_each_netdev_in_bond_rcu(bond, slave) \
3036 for_each_netdev_rcu(&init_net, slave) \
3037 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3038 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
3040 #define for_each_netdev_dump(net, d, ifindex) \
3041 xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3043 static inline struct net_device *next_net_device(struct net_device *dev)
3045 struct list_head *lh;
3049 lh = dev->dev_list.next;
3050 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3053 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3055 struct list_head *lh;
3059 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3060 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3063 static inline struct net_device *first_net_device(struct net *net)
3065 return list_empty(&net->dev_base_head) ? NULL :
3066 net_device_entry(net->dev_base_head.next);
3069 static inline struct net_device *first_net_device_rcu(struct net *net)
3071 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3073 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3076 int netdev_boot_setup_check(struct net_device *dev);
3077 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3078 const char *hwaddr);
3079 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3080 void dev_add_pack(struct packet_type *pt);
3081 void dev_remove_pack(struct packet_type *pt);
3082 void __dev_remove_pack(struct packet_type *pt);
3083 void dev_add_offload(struct packet_offload *po);
3084 void dev_remove_offload(struct packet_offload *po);
3086 int dev_get_iflink(const struct net_device *dev);
3087 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3088 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3089 struct net_device_path_stack *stack);
3090 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3091 unsigned short mask);
3092 struct net_device *dev_get_by_name(struct net *net, const char *name);
3093 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3094 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3095 bool netdev_name_in_use(struct net *net, const char *name);
3096 int dev_alloc_name(struct net_device *dev, const char *name);
3097 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3098 void dev_close(struct net_device *dev);
3099 void dev_close_many(struct list_head *head, bool unlink);
3100 void dev_disable_lro(struct net_device *dev);
3101 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3102 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3103 struct net_device *sb_dev);
3104 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3105 struct net_device *sb_dev);
3107 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3108 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3110 static inline int dev_queue_xmit(struct sk_buff *skb)
3112 return __dev_queue_xmit(skb, NULL);
3115 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3116 struct net_device *sb_dev)
3118 return __dev_queue_xmit(skb, sb_dev);
3121 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3125 ret = __dev_direct_xmit(skb, queue_id);
3126 if (!dev_xmit_complete(ret))
3131 int register_netdevice(struct net_device *dev);
3132 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3133 void unregister_netdevice_many(struct list_head *head);
3134 static inline void unregister_netdevice(struct net_device *dev)
3136 unregister_netdevice_queue(dev, NULL);
3139 int netdev_refcnt_read(const struct net_device *dev);
3140 void free_netdev(struct net_device *dev);
3141 void netdev_freemem(struct net_device *dev);
3142 int init_dummy_netdev(struct net_device *dev);
3144 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3145 struct sk_buff *skb,
3147 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3149 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3150 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3151 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3152 netdevice_tracker *tracker, gfp_t gfp);
3153 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3154 netdevice_tracker *tracker, gfp_t gfp);
3155 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3156 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3158 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3159 unsigned short type,
3160 const void *daddr, const void *saddr,
3163 if (!dev->header_ops || !dev->header_ops->create)
3166 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3169 static inline int dev_parse_header(const struct sk_buff *skb,
3170 unsigned char *haddr)
3172 const struct net_device *dev = skb->dev;
3174 if (!dev->header_ops || !dev->header_ops->parse)
3176 return dev->header_ops->parse(skb, haddr);
3179 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3181 const struct net_device *dev = skb->dev;
3183 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3185 return dev->header_ops->parse_protocol(skb);
3188 /* ll_header must have at least hard_header_len allocated */
3189 static inline bool dev_validate_header(const struct net_device *dev,
3190 char *ll_header, int len)
3192 if (likely(len >= dev->hard_header_len))
3194 if (len < dev->min_header_len)
3197 if (capable(CAP_SYS_RAWIO)) {
3198 memset(ll_header + len, 0, dev->hard_header_len - len);
3202 if (dev->header_ops && dev->header_ops->validate)
3203 return dev->header_ops->validate(ll_header, len);
3208 static inline bool dev_has_header(const struct net_device *dev)
3210 return dev->header_ops && dev->header_ops->create;
3214 * Incoming packets are placed on per-CPU queues
3216 struct softnet_data {
3217 struct list_head poll_list;
3218 struct sk_buff_head process_queue;
3221 unsigned int processed;
3222 unsigned int time_squeeze;
3224 struct softnet_data *rps_ipi_list;
3227 bool in_net_rx_action;
3228 bool in_napi_threaded_poll;
3230 #ifdef CONFIG_NET_FLOW_LIMIT
3231 struct sd_flow_limit __rcu *flow_limit;
3233 struct Qdisc *output_queue;
3234 struct Qdisc **output_queue_tailp;
3235 struct sk_buff *completion_queue;
3236 #ifdef CONFIG_XFRM_OFFLOAD
3237 struct sk_buff_head xfrm_backlog;
3239 /* written and read only by owning cpu: */
3243 #ifdef CONFIG_NET_EGRESS
3248 /* input_queue_head should be written by cpu owning this struct,
3249 * and only read by other cpus. Worth using a cache line.
3251 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3253 /* Elements below can be accessed between CPUs for RPS/RFS */
3254 call_single_data_t csd ____cacheline_aligned_in_smp;
3255 struct softnet_data *rps_ipi_next;
3257 unsigned int input_queue_tail;
3259 unsigned int received_rps;
3260 unsigned int dropped;
3261 struct sk_buff_head input_pkt_queue;
3262 struct napi_struct backlog;
3264 /* Another possibly contended cache line */
3265 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3267 int defer_ipi_scheduled;
3268 struct sk_buff *defer_list;
3269 call_single_data_t defer_csd;
3272 static inline void input_queue_head_incr(struct softnet_data *sd)
3275 sd->input_queue_head++;
3279 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3280 unsigned int *qtail)
3283 *qtail = ++sd->input_queue_tail;
3287 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3289 static inline int dev_recursion_level(void)
3291 return this_cpu_read(softnet_data.xmit.recursion);
3294 #define XMIT_RECURSION_LIMIT 8
3295 static inline bool dev_xmit_recursion(void)
3297 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3298 XMIT_RECURSION_LIMIT);
3301 static inline void dev_xmit_recursion_inc(void)
3303 __this_cpu_inc(softnet_data.xmit.recursion);
3306 static inline void dev_xmit_recursion_dec(void)
3308 __this_cpu_dec(softnet_data.xmit.recursion);
3311 void __netif_schedule(struct Qdisc *q);
3312 void netif_schedule_queue(struct netdev_queue *txq);
3314 static inline void netif_tx_schedule_all(struct net_device *dev)
3318 for (i = 0; i < dev->num_tx_queues; i++)
3319 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3322 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3324 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3328 * netif_start_queue - allow transmit
3329 * @dev: network device
3331 * Allow upper layers to call the device hard_start_xmit routine.
3333 static inline void netif_start_queue(struct net_device *dev)
3335 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3338 static inline void netif_tx_start_all_queues(struct net_device *dev)
3342 for (i = 0; i < dev->num_tx_queues; i++) {
3343 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3344 netif_tx_start_queue(txq);
3348 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3351 * netif_wake_queue - restart transmit
3352 * @dev: network device
3354 * Allow upper layers to call the device hard_start_xmit routine.
3355 * Used for flow control when transmit resources are available.
3357 static inline void netif_wake_queue(struct net_device *dev)
3359 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3362 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3366 for (i = 0; i < dev->num_tx_queues; i++) {
3367 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3368 netif_tx_wake_queue(txq);
3372 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3374 /* Must be an atomic op see netif_txq_try_stop() */
3375 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3379 * netif_stop_queue - stop transmitted packets
3380 * @dev: network device
3382 * Stop upper layers calling the device hard_start_xmit routine.
3383 * Used for flow control when transmit resources are unavailable.
3385 static inline void netif_stop_queue(struct net_device *dev)
3387 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3390 void netif_tx_stop_all_queues(struct net_device *dev);
3392 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3394 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3398 * netif_queue_stopped - test if transmit queue is flowblocked
3399 * @dev: network device
3401 * Test if transmit queue on device is currently unable to send.
3403 static inline bool netif_queue_stopped(const struct net_device *dev)
3405 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3408 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3410 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3414 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3416 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3420 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3422 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3426 * netdev_queue_set_dql_min_limit - set dql minimum limit
3427 * @dev_queue: pointer to transmit queue
3428 * @min_limit: dql minimum limit
3430 * Forces xmit_more() to return true until the minimum threshold
3431 * defined by @min_limit is reached (or until the tx queue is
3432 * empty). Warning: to be use with care, misuse will impact the
3435 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3436 unsigned int min_limit)
3439 dev_queue->dql.min_limit = min_limit;
3444 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3445 * @dev_queue: pointer to transmit queue
3447 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3448 * to give appropriate hint to the CPU.
3450 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3453 prefetchw(&dev_queue->dql.num_queued);
3458 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3459 * @dev_queue: pointer to transmit queue
3461 * BQL enabled drivers might use this helper in their TX completion path,
3462 * to give appropriate hint to the CPU.
3464 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3467 prefetchw(&dev_queue->dql.limit);
3472 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3473 * @dev_queue: network device queue
3474 * @bytes: number of bytes queued to the device queue
3476 * Report the number of bytes queued for sending/completion to the network
3477 * device hardware queue. @bytes should be a good approximation and should
3478 * exactly match netdev_completed_queue() @bytes.
3479 * This is typically called once per packet, from ndo_start_xmit().
3481 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3485 dql_queued(&dev_queue->dql, bytes);
3487 if (likely(dql_avail(&dev_queue->dql) >= 0))
3490 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3493 * The XOFF flag must be set before checking the dql_avail below,
3494 * because in netdev_tx_completed_queue we update the dql_completed
3495 * before checking the XOFF flag.
3499 /* check again in case another CPU has just made room avail */
3500 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3501 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3505 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3506 * that they should not test BQL status themselves.
3507 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3509 * Returns true if the doorbell must be used to kick the NIC.
3511 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3517 dql_queued(&dev_queue->dql, bytes);
3519 return netif_tx_queue_stopped(dev_queue);
3521 netdev_tx_sent_queue(dev_queue, bytes);
3526 * netdev_sent_queue - report the number of bytes queued to hardware
3527 * @dev: network device
3528 * @bytes: number of bytes queued to the hardware device queue
3530 * Report the number of bytes queued for sending/completion to the network
3531 * device hardware queue#0. @bytes should be a good approximation and should
3532 * exactly match netdev_completed_queue() @bytes.
3533 * This is typically called once per packet, from ndo_start_xmit().
3535 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3537 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3540 static inline bool __netdev_sent_queue(struct net_device *dev,
3544 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3549 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3550 * @dev_queue: network device queue
3551 * @pkts: number of packets (currently ignored)
3552 * @bytes: number of bytes dequeued from the device queue
3554 * Must be called at most once per TX completion round (and not per
3555 * individual packet), so that BQL can adjust its limits appropriately.
3557 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3558 unsigned int pkts, unsigned int bytes)
3561 if (unlikely(!bytes))
3564 dql_completed(&dev_queue->dql, bytes);
3567 * Without the memory barrier there is a small possiblity that
3568 * netdev_tx_sent_queue will miss the update and cause the queue to
3569 * be stopped forever
3571 smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3573 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3576 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3577 netif_schedule_queue(dev_queue);
3582 * netdev_completed_queue - report bytes and packets completed by device
3583 * @dev: network device
3584 * @pkts: actual number of packets sent over the medium
3585 * @bytes: actual number of bytes sent over the medium
3587 * Report the number of bytes and packets transmitted by the network device
3588 * hardware queue over the physical medium, @bytes must exactly match the
3589 * @bytes amount passed to netdev_sent_queue()
3591 static inline void netdev_completed_queue(struct net_device *dev,
3592 unsigned int pkts, unsigned int bytes)
3594 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3597 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3600 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3606 * netdev_reset_queue - reset the packets and bytes count of a network device
3607 * @dev_queue: network device
3609 * Reset the bytes and packet count of a network device and clear the
3610 * software flow control OFF bit for this network device
3612 static inline void netdev_reset_queue(struct net_device *dev_queue)
3614 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3618 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3619 * @dev: network device
3620 * @queue_index: given tx queue index
3622 * Returns 0 if given tx queue index >= number of device tx queues,
3623 * otherwise returns the originally passed tx queue index.
3625 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3627 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3628 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3629 dev->name, queue_index,
3630 dev->real_num_tx_queues);
3638 * netif_running - test if up
3639 * @dev: network device
3641 * Test if the device has been brought up.
3643 static inline bool netif_running(const struct net_device *dev)
3645 return test_bit(__LINK_STATE_START, &dev->state);
3649 * Routines to manage the subqueues on a device. We only need start,
3650 * stop, and a check if it's stopped. All other device management is
3651 * done at the overall netdevice level.
3652 * Also test the device if we're multiqueue.
3656 * netif_start_subqueue - allow sending packets on subqueue
3657 * @dev: network device
3658 * @queue_index: sub queue index
3660 * Start individual transmit queue of a device with multiple transmit queues.
3662 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3664 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3666 netif_tx_start_queue(txq);
3670 * netif_stop_subqueue - stop sending packets on subqueue
3671 * @dev: network device
3672 * @queue_index: sub queue index
3674 * Stop individual transmit queue of a device with multiple transmit queues.
3676 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3678 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3679 netif_tx_stop_queue(txq);
3683 * __netif_subqueue_stopped - test status of subqueue
3684 * @dev: network device
3685 * @queue_index: sub queue index
3687 * Check individual transmit queue of a device with multiple transmit queues.
3689 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3692 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3694 return netif_tx_queue_stopped(txq);
3698 * netif_subqueue_stopped - test status of subqueue
3699 * @dev: network device
3700 * @skb: sub queue buffer pointer
3702 * Check individual transmit queue of a device with multiple transmit queues.
3704 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3705 struct sk_buff *skb)
3707 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3711 * netif_wake_subqueue - allow sending packets on subqueue
3712 * @dev: network device
3713 * @queue_index: sub queue index
3715 * Resume individual transmit queue of a device with multiple transmit queues.
3717 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3719 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3721 netif_tx_wake_queue(txq);
3725 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3727 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3728 u16 index, enum xps_map_type type);
3731 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3732 * @j: CPU/Rx queue index
3733 * @mask: bitmask of all cpus/rx queues
3734 * @nr_bits: number of bits in the bitmask
3736 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3738 static inline bool netif_attr_test_mask(unsigned long j,
3739 const unsigned long *mask,
3740 unsigned int nr_bits)
3742 cpu_max_bits_warn(j, nr_bits);
3743 return test_bit(j, mask);
3747 * netif_attr_test_online - Test for online CPU/Rx queue
3748 * @j: CPU/Rx queue index
3749 * @online_mask: bitmask for CPUs/Rx queues that are online
3750 * @nr_bits: number of bits in the bitmask
3752 * Returns true if a CPU/Rx queue is online.
3754 static inline bool netif_attr_test_online(unsigned long j,
3755 const unsigned long *online_mask,
3756 unsigned int nr_bits)
3758 cpu_max_bits_warn(j, nr_bits);
3761 return test_bit(j, online_mask);
3763 return (j < nr_bits);
3767 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3768 * @n: CPU/Rx queue index
3769 * @srcp: the cpumask/Rx queue mask pointer
3770 * @nr_bits: number of bits in the bitmask
3772 * Returns >= nr_bits if no further CPUs/Rx queues set.
3774 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3775 unsigned int nr_bits)
3777 /* -1 is a legal arg here. */
3779 cpu_max_bits_warn(n, nr_bits);
3782 return find_next_bit(srcp, nr_bits, n + 1);
3788 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3789 * @n: CPU/Rx queue index
3790 * @src1p: the first CPUs/Rx queues mask pointer
3791 * @src2p: the second CPUs/Rx queues mask pointer
3792 * @nr_bits: number of bits in the bitmask
3794 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3796 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3797 const unsigned long *src2p,
3798 unsigned int nr_bits)
3800 /* -1 is a legal arg here. */
3802 cpu_max_bits_warn(n, nr_bits);
3805 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3807 return find_next_bit(src1p, nr_bits, n + 1);
3809 return find_next_bit(src2p, nr_bits, n + 1);
3814 static inline int netif_set_xps_queue(struct net_device *dev,
3815 const struct cpumask *mask,
3821 static inline int __netif_set_xps_queue(struct net_device *dev,
3822 const unsigned long *mask,
3823 u16 index, enum xps_map_type type)
3830 * netif_is_multiqueue - test if device has multiple transmit queues
3831 * @dev: network device
3833 * Check if device has multiple transmit queues
3835 static inline bool netif_is_multiqueue(const struct net_device *dev)
3837 return dev->num_tx_queues > 1;
3840 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3843 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3845 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3848 dev->real_num_rx_queues = rxqs;
3852 int netif_set_real_num_queues(struct net_device *dev,
3853 unsigned int txq, unsigned int rxq);
3855 int netif_get_num_default_rss_queues(void);
3857 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3858 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3861 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3862 * interrupt context or with hardware interrupts being disabled.
3863 * (in_hardirq() || irqs_disabled())
3865 * We provide four helpers that can be used in following contexts :
3867 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3868 * replacing kfree_skb(skb)
3870 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3871 * Typically used in place of consume_skb(skb) in TX completion path
3873 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3874 * replacing kfree_skb(skb)
3876 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3877 * and consumed a packet. Used in place of consume_skb(skb)
3879 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3881 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3884 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3886 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3889 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3891 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3894 static inline void dev_consume_skb_any(struct sk_buff *skb)
3896 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3899 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3900 struct bpf_prog *xdp_prog);
3901 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3902 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3903 int netif_rx(struct sk_buff *skb);
3904 int __netif_rx(struct sk_buff *skb);
3906 int netif_receive_skb(struct sk_buff *skb);
3907 int netif_receive_skb_core(struct sk_buff *skb);
3908 void netif_receive_skb_list_internal(struct list_head *head);
3909 void netif_receive_skb_list(struct list_head *head);
3910 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3911 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3912 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3913 void napi_get_frags_check(struct napi_struct *napi);
3914 gro_result_t napi_gro_frags(struct napi_struct *napi);
3915 struct packet_offload *gro_find_receive_by_type(__be16 type);
3916 struct packet_offload *gro_find_complete_by_type(__be16 type);
3918 static inline void napi_free_frags(struct napi_struct *napi)
3920 kfree_skb(napi->skb);
3924 bool netdev_is_rx_handler_busy(struct net_device *dev);
3925 int netdev_rx_handler_register(struct net_device *dev,
3926 rx_handler_func_t *rx_handler,
3927 void *rx_handler_data);
3928 void netdev_rx_handler_unregister(struct net_device *dev);
3930 bool dev_valid_name(const char *name);
3931 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3933 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3935 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3936 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3937 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3938 void __user *data, bool *need_copyout);
3939 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3940 int generic_hwtstamp_get_lower(struct net_device *dev,
3941 struct kernel_hwtstamp_config *kernel_cfg);
3942 int generic_hwtstamp_set_lower(struct net_device *dev,
3943 struct kernel_hwtstamp_config *kernel_cfg,
3944 struct netlink_ext_ack *extack);
3945 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3946 unsigned int dev_get_flags(const struct net_device *);
3947 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3948 struct netlink_ext_ack *extack);
3949 int dev_change_flags(struct net_device *dev, unsigned int flags,
3950 struct netlink_ext_ack *extack);
3951 int dev_set_alias(struct net_device *, const char *, size_t);
3952 int dev_get_alias(const struct net_device *, char *, size_t);
3953 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3954 const char *pat, int new_ifindex);
3956 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3959 return __dev_change_net_namespace(dev, net, pat, 0);
3961 int __dev_set_mtu(struct net_device *, int);
3962 int dev_set_mtu(struct net_device *, int);
3963 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3964 struct netlink_ext_ack *extack);
3965 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3966 struct netlink_ext_ack *extack);
3967 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3968 struct netlink_ext_ack *extack);
3969 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3970 int dev_get_port_parent_id(struct net_device *dev,
3971 struct netdev_phys_item_id *ppid, bool recurse);
3972 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3973 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
3974 void netdev_dpll_pin_clear(struct net_device *dev);
3976 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
3978 #if IS_ENABLED(CONFIG_DPLL)
3979 return dev->dpll_pin;
3985 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3986 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3987 struct netdev_queue *txq, int *ret);
3989 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3990 u8 dev_xdp_prog_count(struct net_device *dev);
3991 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3993 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3994 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3995 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3996 bool is_skb_forwardable(const struct net_device *dev,
3997 const struct sk_buff *skb);
3999 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4000 const struct sk_buff *skb,
4001 const bool check_mtu)
4003 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4006 if (!(dev->flags & IFF_UP))
4012 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4013 if (skb->len <= len)
4016 /* if TSO is enabled, we don't care about the length as the packet
4017 * could be forwarded without being segmented before
4019 if (skb_is_gso(skb))
4025 void netdev_core_stats_inc(struct net_device *dev, u32 offset);
4027 #define DEV_CORE_STATS_INC(FIELD) \
4028 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
4030 netdev_core_stats_inc(dev, \
4031 offsetof(struct net_device_core_stats, FIELD)); \
4033 DEV_CORE_STATS_INC(rx_dropped)
4034 DEV_CORE_STATS_INC(tx_dropped)
4035 DEV_CORE_STATS_INC(rx_nohandler)
4036 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4037 #undef DEV_CORE_STATS_INC
4039 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4040 struct sk_buff *skb,
4041 const bool check_mtu)
4043 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4044 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4045 dev_core_stats_rx_dropped_inc(dev);
4050 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4055 bool dev_nit_active(struct net_device *dev);
4056 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4058 static inline void __dev_put(struct net_device *dev)
4061 #ifdef CONFIG_PCPU_DEV_REFCNT
4062 this_cpu_dec(*dev->pcpu_refcnt);
4064 refcount_dec(&dev->dev_refcnt);
4069 static inline void __dev_hold(struct net_device *dev)
4072 #ifdef CONFIG_PCPU_DEV_REFCNT
4073 this_cpu_inc(*dev->pcpu_refcnt);
4075 refcount_inc(&dev->dev_refcnt);
4080 static inline void __netdev_tracker_alloc(struct net_device *dev,
4081 netdevice_tracker *tracker,
4084 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4085 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4089 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4090 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4092 static inline void netdev_tracker_alloc(struct net_device *dev,
4093 netdevice_tracker *tracker, gfp_t gfp)
4095 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4096 refcount_dec(&dev->refcnt_tracker.no_tracker);
4097 __netdev_tracker_alloc(dev, tracker, gfp);
4101 static inline void netdev_tracker_free(struct net_device *dev,
4102 netdevice_tracker *tracker)
4104 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4105 ref_tracker_free(&dev->refcnt_tracker, tracker);
4109 static inline void netdev_hold(struct net_device *dev,
4110 netdevice_tracker *tracker, gfp_t gfp)
4114 __netdev_tracker_alloc(dev, tracker, gfp);
4118 static inline void netdev_put(struct net_device *dev,
4119 netdevice_tracker *tracker)
4122 netdev_tracker_free(dev, tracker);
4128 * dev_hold - get reference to device
4129 * @dev: network device
4131 * Hold reference to device to keep it from being freed.
4132 * Try using netdev_hold() instead.
4134 static inline void dev_hold(struct net_device *dev)
4136 netdev_hold(dev, NULL, GFP_ATOMIC);
4140 * dev_put - release reference to device
4141 * @dev: network device
4143 * Release reference to device to allow it to be freed.
4144 * Try using netdev_put() instead.
4146 static inline void dev_put(struct net_device *dev)
4148 netdev_put(dev, NULL);
4151 static inline void netdev_ref_replace(struct net_device *odev,
4152 struct net_device *ndev,
4153 netdevice_tracker *tracker,
4157 netdev_tracker_free(odev, tracker);
4163 __netdev_tracker_alloc(ndev, tracker, gfp);
4166 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4167 * and _off may be called from IRQ context, but it is caller
4168 * who is responsible for serialization of these calls.
4170 * The name carrier is inappropriate, these functions should really be
4171 * called netif_lowerlayer_*() because they represent the state of any
4172 * kind of lower layer not just hardware media.
4174 void linkwatch_fire_event(struct net_device *dev);
4177 * netif_carrier_ok - test if carrier present
4178 * @dev: network device
4180 * Check if carrier is present on device
4182 static inline bool netif_carrier_ok(const struct net_device *dev)
4184 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4187 unsigned long dev_trans_start(struct net_device *dev);
4189 void __netdev_watchdog_up(struct net_device *dev);
4191 void netif_carrier_on(struct net_device *dev);
4192 void netif_carrier_off(struct net_device *dev);
4193 void netif_carrier_event(struct net_device *dev);
4196 * netif_dormant_on - mark device as dormant.
4197 * @dev: network device
4199 * Mark device as dormant (as per RFC2863).
4201 * The dormant state indicates that the relevant interface is not
4202 * actually in a condition to pass packets (i.e., it is not 'up') but is
4203 * in a "pending" state, waiting for some external event. For "on-
4204 * demand" interfaces, this new state identifies the situation where the
4205 * interface is waiting for events to place it in the up state.
4207 static inline void netif_dormant_on(struct net_device *dev)
4209 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4210 linkwatch_fire_event(dev);
4214 * netif_dormant_off - set device as not dormant.
4215 * @dev: network device
4217 * Device is not in dormant state.
4219 static inline void netif_dormant_off(struct net_device *dev)
4221 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4222 linkwatch_fire_event(dev);
4226 * netif_dormant - test if device is dormant
4227 * @dev: network device
4229 * Check if device is dormant.
4231 static inline bool netif_dormant(const struct net_device *dev)
4233 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4238 * netif_testing_on - mark device as under test.
4239 * @dev: network device
4241 * Mark device as under test (as per RFC2863).
4243 * The testing state indicates that some test(s) must be performed on
4244 * the interface. After completion, of the test, the interface state
4245 * will change to up, dormant, or down, as appropriate.
4247 static inline void netif_testing_on(struct net_device *dev)
4249 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4250 linkwatch_fire_event(dev);
4254 * netif_testing_off - set device as not under test.
4255 * @dev: network device
4257 * Device is not in testing state.
4259 static inline void netif_testing_off(struct net_device *dev)
4261 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4262 linkwatch_fire_event(dev);
4266 * netif_testing - test if device is under test
4267 * @dev: network device
4269 * Check if device is under test
4271 static inline bool netif_testing(const struct net_device *dev)
4273 return test_bit(__LINK_STATE_TESTING, &dev->state);
4278 * netif_oper_up - test if device is operational
4279 * @dev: network device
4281 * Check if carrier is operational
4283 static inline bool netif_oper_up(const struct net_device *dev)
4285 return (dev->operstate == IF_OPER_UP ||
4286 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4290 * netif_device_present - is device available or removed
4291 * @dev: network device
4293 * Check if device has not been removed from system.
4295 static inline bool netif_device_present(const struct net_device *dev)
4297 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4300 void netif_device_detach(struct net_device *dev);
4302 void netif_device_attach(struct net_device *dev);
4305 * Network interface message level settings
4310 NETIF_MSG_PROBE_BIT,
4312 NETIF_MSG_TIMER_BIT,
4313 NETIF_MSG_IFDOWN_BIT,
4315 NETIF_MSG_RX_ERR_BIT,
4316 NETIF_MSG_TX_ERR_BIT,
4317 NETIF_MSG_TX_QUEUED_BIT,
4319 NETIF_MSG_TX_DONE_BIT,
4320 NETIF_MSG_RX_STATUS_BIT,
4321 NETIF_MSG_PKTDATA_BIT,
4325 /* When you add a new bit above, update netif_msg_class_names array
4326 * in net/ethtool/common.c
4328 NETIF_MSG_CLASS_COUNT,
4330 /* Both ethtool_ops interface and internal driver implementation use u32 */
4331 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4333 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4334 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4336 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4337 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4338 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4339 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4340 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4341 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4342 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4343 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4344 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4345 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4346 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4347 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4348 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4349 #define NETIF_MSG_HW __NETIF_MSG(HW)
4350 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4352 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4353 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4354 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4355 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4356 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4357 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4358 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4359 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4360 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4361 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4362 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4363 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4364 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4365 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4366 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4368 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4371 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4372 return default_msg_enable_bits;
4373 if (debug_value == 0) /* no output */
4375 /* set low N bits */
4376 return (1U << debug_value) - 1;
4379 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4381 spin_lock(&txq->_xmit_lock);
4382 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4383 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4386 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4388 __acquire(&txq->_xmit_lock);
4392 static inline void __netif_tx_release(struct netdev_queue *txq)
4394 __release(&txq->_xmit_lock);
4397 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4399 spin_lock_bh(&txq->_xmit_lock);
4400 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4401 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4404 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4406 bool ok = spin_trylock(&txq->_xmit_lock);
4409 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4410 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4415 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4417 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4418 WRITE_ONCE(txq->xmit_lock_owner, -1);
4419 spin_unlock(&txq->_xmit_lock);
4422 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4424 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4425 WRITE_ONCE(txq->xmit_lock_owner, -1);
4426 spin_unlock_bh(&txq->_xmit_lock);
4430 * txq->trans_start can be read locklessly from dev_watchdog()
4432 static inline void txq_trans_update(struct netdev_queue *txq)
4434 if (txq->xmit_lock_owner != -1)
4435 WRITE_ONCE(txq->trans_start, jiffies);
4438 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4440 unsigned long now = jiffies;
4442 if (READ_ONCE(txq->trans_start) != now)
4443 WRITE_ONCE(txq->trans_start, now);
4446 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4447 static inline void netif_trans_update(struct net_device *dev)
4449 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4451 txq_trans_cond_update(txq);
4455 * netif_tx_lock - grab network device transmit lock
4456 * @dev: network device
4458 * Get network device transmit lock
4460 void netif_tx_lock(struct net_device *dev);
4462 static inline void netif_tx_lock_bh(struct net_device *dev)
4468 void netif_tx_unlock(struct net_device *dev);
4470 static inline void netif_tx_unlock_bh(struct net_device *dev)
4472 netif_tx_unlock(dev);
4476 #define HARD_TX_LOCK(dev, txq, cpu) { \
4477 if ((dev->features & NETIF_F_LLTX) == 0) { \
4478 __netif_tx_lock(txq, cpu); \
4480 __netif_tx_acquire(txq); \
4484 #define HARD_TX_TRYLOCK(dev, txq) \
4485 (((dev->features & NETIF_F_LLTX) == 0) ? \
4486 __netif_tx_trylock(txq) : \
4487 __netif_tx_acquire(txq))
4489 #define HARD_TX_UNLOCK(dev, txq) { \
4490 if ((dev->features & NETIF_F_LLTX) == 0) { \
4491 __netif_tx_unlock(txq); \
4493 __netif_tx_release(txq); \
4497 static inline void netif_tx_disable(struct net_device *dev)
4503 cpu = smp_processor_id();
4504 spin_lock(&dev->tx_global_lock);
4505 for (i = 0; i < dev->num_tx_queues; i++) {
4506 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4508 __netif_tx_lock(txq, cpu);
4509 netif_tx_stop_queue(txq);
4510 __netif_tx_unlock(txq);
4512 spin_unlock(&dev->tx_global_lock);
4516 static inline void netif_addr_lock(struct net_device *dev)
4518 unsigned char nest_level = 0;
4520 #ifdef CONFIG_LOCKDEP
4521 nest_level = dev->nested_level;
4523 spin_lock_nested(&dev->addr_list_lock, nest_level);
4526 static inline void netif_addr_lock_bh(struct net_device *dev)
4528 unsigned char nest_level = 0;
4530 #ifdef CONFIG_LOCKDEP
4531 nest_level = dev->nested_level;
4534 spin_lock_nested(&dev->addr_list_lock, nest_level);
4537 static inline void netif_addr_unlock(struct net_device *dev)
4539 spin_unlock(&dev->addr_list_lock);
4542 static inline void netif_addr_unlock_bh(struct net_device *dev)
4544 spin_unlock_bh(&dev->addr_list_lock);
4548 * dev_addrs walker. Should be used only for read access. Call with
4549 * rcu_read_lock held.
4551 #define for_each_dev_addr(dev, ha) \
4552 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4554 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4556 void ether_setup(struct net_device *dev);
4558 /* Support for loadable net-drivers */
4559 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4560 unsigned char name_assign_type,
4561 void (*setup)(struct net_device *),
4562 unsigned int txqs, unsigned int rxqs);
4563 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4564 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4566 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4567 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4570 int register_netdev(struct net_device *dev);
4571 void unregister_netdev(struct net_device *dev);
4573 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4575 /* General hardware address lists handling functions */
4576 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4577 struct netdev_hw_addr_list *from_list, int addr_len);
4578 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4579 struct netdev_hw_addr_list *from_list, int addr_len);
4580 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4581 struct net_device *dev,
4582 int (*sync)(struct net_device *, const unsigned char *),
4583 int (*unsync)(struct net_device *,
4584 const unsigned char *));
4585 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4586 struct net_device *dev,
4587 int (*sync)(struct net_device *,
4588 const unsigned char *, int),
4589 int (*unsync)(struct net_device *,
4590 const unsigned char *, int));
4591 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4592 struct net_device *dev,
4593 int (*unsync)(struct net_device *,
4594 const unsigned char *, int));
4595 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4596 struct net_device *dev,
4597 int (*unsync)(struct net_device *,
4598 const unsigned char *));
4599 void __hw_addr_init(struct netdev_hw_addr_list *list);
4601 /* Functions used for device addresses handling */
4602 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4603 const void *addr, size_t len);
4606 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4608 dev_addr_mod(dev, 0, addr, len);
4611 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4613 __dev_addr_set(dev, addr, dev->addr_len);
4616 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4617 unsigned char addr_type);
4618 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4619 unsigned char addr_type);
4621 /* Functions used for unicast addresses handling */
4622 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4623 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4624 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4625 int dev_uc_sync(struct net_device *to, struct net_device *from);
4626 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4627 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4628 void dev_uc_flush(struct net_device *dev);
4629 void dev_uc_init(struct net_device *dev);
4632 * __dev_uc_sync - Synchonize device's unicast list
4633 * @dev: device to sync
4634 * @sync: function to call if address should be added
4635 * @unsync: function to call if address should be removed
4637 * Add newly added addresses to the interface, and release
4638 * addresses that have been deleted.
4640 static inline int __dev_uc_sync(struct net_device *dev,
4641 int (*sync)(struct net_device *,
4642 const unsigned char *),
4643 int (*unsync)(struct net_device *,
4644 const unsigned char *))
4646 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4650 * __dev_uc_unsync - Remove synchronized addresses from device
4651 * @dev: device to sync
4652 * @unsync: function to call if address should be removed
4654 * Remove all addresses that were added to the device by dev_uc_sync().
4656 static inline void __dev_uc_unsync(struct net_device *dev,
4657 int (*unsync)(struct net_device *,
4658 const unsigned char *))
4660 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4663 /* Functions used for multicast addresses handling */
4664 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4665 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4666 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4667 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4668 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4669 int dev_mc_sync(struct net_device *to, struct net_device *from);
4670 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4671 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4672 void dev_mc_flush(struct net_device *dev);
4673 void dev_mc_init(struct net_device *dev);
4676 * __dev_mc_sync - Synchonize device's multicast list
4677 * @dev: device to sync
4678 * @sync: function to call if address should be added
4679 * @unsync: function to call if address should be removed
4681 * Add newly added addresses to the interface, and release
4682 * addresses that have been deleted.
4684 static inline int __dev_mc_sync(struct net_device *dev,
4685 int (*sync)(struct net_device *,
4686 const unsigned char *),
4687 int (*unsync)(struct net_device *,
4688 const unsigned char *))
4690 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4694 * __dev_mc_unsync - Remove synchronized addresses from device
4695 * @dev: device to sync
4696 * @unsync: function to call if address should be removed
4698 * Remove all addresses that were added to the device by dev_mc_sync().
4700 static inline void __dev_mc_unsync(struct net_device *dev,
4701 int (*unsync)(struct net_device *,
4702 const unsigned char *))
4704 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4707 /* Functions used for secondary unicast and multicast support */
4708 void dev_set_rx_mode(struct net_device *dev);
4709 int dev_set_promiscuity(struct net_device *dev, int inc);
4710 int dev_set_allmulti(struct net_device *dev, int inc);
4711 void netdev_state_change(struct net_device *dev);
4712 void __netdev_notify_peers(struct net_device *dev);
4713 void netdev_notify_peers(struct net_device *dev);
4714 void netdev_features_change(struct net_device *dev);
4715 /* Load a device via the kmod */
4716 void dev_load(struct net *net, const char *name);
4717 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4718 struct rtnl_link_stats64 *storage);
4719 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4720 const struct net_device_stats *netdev_stats);
4721 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4722 const struct pcpu_sw_netstats __percpu *netstats);
4723 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4725 extern int netdev_max_backlog;
4726 extern int dev_rx_weight;
4727 extern int dev_tx_weight;
4728 extern int gro_normal_batch;
4731 NESTED_SYNC_IMM_BIT,
4732 NESTED_SYNC_TODO_BIT,
4735 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4736 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4738 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4739 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4741 struct netdev_nested_priv {
4742 unsigned char flags;
4746 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4747 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4748 struct list_head **iter);
4750 /* iterate through upper list, must be called under RCU read lock */
4751 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4752 for (iter = &(dev)->adj_list.upper, \
4753 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4755 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4757 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4758 int (*fn)(struct net_device *upper_dev,
4759 struct netdev_nested_priv *priv),
4760 struct netdev_nested_priv *priv);
4762 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4763 struct net_device *upper_dev);
4765 bool netdev_has_any_upper_dev(struct net_device *dev);
4767 void *netdev_lower_get_next_private(struct net_device *dev,
4768 struct list_head **iter);
4769 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4770 struct list_head **iter);
4772 #define netdev_for_each_lower_private(dev, priv, iter) \
4773 for (iter = (dev)->adj_list.lower.next, \
4774 priv = netdev_lower_get_next_private(dev, &(iter)); \
4776 priv = netdev_lower_get_next_private(dev, &(iter)))
4778 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4779 for (iter = &(dev)->adj_list.lower, \
4780 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4782 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4784 void *netdev_lower_get_next(struct net_device *dev,
4785 struct list_head **iter);
4787 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4788 for (iter = (dev)->adj_list.lower.next, \
4789 ldev = netdev_lower_get_next(dev, &(iter)); \
4791 ldev = netdev_lower_get_next(dev, &(iter)))
4793 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4794 struct list_head **iter);
4795 int netdev_walk_all_lower_dev(struct net_device *dev,
4796 int (*fn)(struct net_device *lower_dev,
4797 struct netdev_nested_priv *priv),
4798 struct netdev_nested_priv *priv);
4799 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4800 int (*fn)(struct net_device *lower_dev,
4801 struct netdev_nested_priv *priv),
4802 struct netdev_nested_priv *priv);
4804 void *netdev_adjacent_get_private(struct list_head *adj_list);
4805 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4806 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4807 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4808 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4809 struct netlink_ext_ack *extack);
4810 int netdev_master_upper_dev_link(struct net_device *dev,
4811 struct net_device *upper_dev,
4812 void *upper_priv, void *upper_info,
4813 struct netlink_ext_ack *extack);
4814 void netdev_upper_dev_unlink(struct net_device *dev,
4815 struct net_device *upper_dev);
4816 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4817 struct net_device *new_dev,
4818 struct net_device *dev,
4819 struct netlink_ext_ack *extack);
4820 void netdev_adjacent_change_commit(struct net_device *old_dev,
4821 struct net_device *new_dev,
4822 struct net_device *dev);
4823 void netdev_adjacent_change_abort(struct net_device *old_dev,
4824 struct net_device *new_dev,
4825 struct net_device *dev);
4826 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4827 void *netdev_lower_dev_get_private(struct net_device *dev,
4828 struct net_device *lower_dev);
4829 void netdev_lower_state_changed(struct net_device *lower_dev,
4830 void *lower_state_info);
4832 /* RSS keys are 40 or 52 bytes long */
4833 #define NETDEV_RSS_KEY_LEN 52
4834 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4835 void netdev_rss_key_fill(void *buffer, size_t len);
4837 int skb_checksum_help(struct sk_buff *skb);
4838 int skb_crc32c_csum_help(struct sk_buff *skb);
4839 int skb_csum_hwoffload_help(struct sk_buff *skb,
4840 const netdev_features_t features);
4842 struct netdev_bonding_info {
4847 struct netdev_notifier_bonding_info {
4848 struct netdev_notifier_info info; /* must be first */
4849 struct netdev_bonding_info bonding_info;
4852 void netdev_bonding_info_change(struct net_device *dev,
4853 struct netdev_bonding_info *bonding_info);
4855 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4856 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4858 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4864 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4866 static inline bool can_checksum_protocol(netdev_features_t features,
4869 if (protocol == htons(ETH_P_FCOE))
4870 return !!(features & NETIF_F_FCOE_CRC);
4872 /* Assume this is an IP checksum (not SCTP CRC) */
4874 if (features & NETIF_F_HW_CSUM) {
4875 /* Can checksum everything */
4880 case htons(ETH_P_IP):
4881 return !!(features & NETIF_F_IP_CSUM);
4882 case htons(ETH_P_IPV6):
4883 return !!(features & NETIF_F_IPV6_CSUM);
4890 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4892 static inline void netdev_rx_csum_fault(struct net_device *dev,
4893 struct sk_buff *skb)
4897 /* rx skb timestamps */
4898 void net_enable_timestamp(void);
4899 void net_disable_timestamp(void);
4901 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4902 const struct skb_shared_hwtstamps *hwtstamps,
4905 const struct net_device_ops *ops = dev->netdev_ops;
4907 if (ops->ndo_get_tstamp)
4908 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4910 return hwtstamps->hwtstamp;
4913 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4914 struct sk_buff *skb, struct net_device *dev,
4917 __this_cpu_write(softnet_data.xmit.more, more);
4918 return ops->ndo_start_xmit(skb, dev);
4921 static inline bool netdev_xmit_more(void)
4923 return __this_cpu_read(softnet_data.xmit.more);
4926 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4927 struct netdev_queue *txq, bool more)
4929 const struct net_device_ops *ops = dev->netdev_ops;
4932 rc = __netdev_start_xmit(ops, skb, dev, more);
4933 if (rc == NETDEV_TX_OK)
4934 txq_trans_update(txq);
4939 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4941 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4944 extern const struct kobj_ns_type_operations net_ns_type_operations;
4946 const char *netdev_drivername(const struct net_device *dev);
4948 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4949 netdev_features_t f2)
4951 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4952 if (f1 & NETIF_F_HW_CSUM)
4953 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4955 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4961 static inline netdev_features_t netdev_get_wanted_features(
4962 struct net_device *dev)
4964 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4966 netdev_features_t netdev_increment_features(netdev_features_t all,
4967 netdev_features_t one, netdev_features_t mask);
4969 /* Allow TSO being used on stacked device :
4970 * Performing the GSO segmentation before last device
4971 * is a performance improvement.
4973 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4974 netdev_features_t mask)
4976 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4979 int __netdev_update_features(struct net_device *dev);
4980 void netdev_update_features(struct net_device *dev);
4981 void netdev_change_features(struct net_device *dev);
4983 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4984 struct net_device *dev);
4986 netdev_features_t passthru_features_check(struct sk_buff *skb,
4987 struct net_device *dev,
4988 netdev_features_t features);
4989 netdev_features_t netif_skb_features(struct sk_buff *skb);
4990 void skb_warn_bad_offload(const struct sk_buff *skb);
4992 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4994 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4996 /* check flags correspondence */
4997 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4998 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4999 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5000 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5001 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5002 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5003 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5004 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5005 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5006 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5007 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5008 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5009 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5010 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5011 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5012 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5013 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5014 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5015 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5017 return (features & feature) == feature;
5020 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5022 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5023 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5026 static inline bool netif_needs_gso(struct sk_buff *skb,
5027 netdev_features_t features)
5029 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5030 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5031 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5034 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5035 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5036 void netif_inherit_tso_max(struct net_device *to,
5037 const struct net_device *from);
5039 static inline bool netif_is_macsec(const struct net_device *dev)
5041 return dev->priv_flags & IFF_MACSEC;
5044 static inline bool netif_is_macvlan(const struct net_device *dev)
5046 return dev->priv_flags & IFF_MACVLAN;
5049 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5051 return dev->priv_flags & IFF_MACVLAN_PORT;
5054 static inline bool netif_is_bond_master(const struct net_device *dev)
5056 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5059 static inline bool netif_is_bond_slave(const struct net_device *dev)
5061 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5064 static inline bool netif_supports_nofcs(struct net_device *dev)
5066 return dev->priv_flags & IFF_SUPP_NOFCS;
5069 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5071 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5074 static inline bool netif_is_l3_master(const struct net_device *dev)
5076 return dev->priv_flags & IFF_L3MDEV_MASTER;
5079 static inline bool netif_is_l3_slave(const struct net_device *dev)
5081 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5084 static inline int dev_sdif(const struct net_device *dev)
5086 #ifdef CONFIG_NET_L3_MASTER_DEV
5087 if (netif_is_l3_slave(dev))
5088 return dev->ifindex;
5093 static inline bool netif_is_bridge_master(const struct net_device *dev)
5095 return dev->priv_flags & IFF_EBRIDGE;
5098 static inline bool netif_is_bridge_port(const struct net_device *dev)
5100 return dev->priv_flags & IFF_BRIDGE_PORT;
5103 static inline bool netif_is_ovs_master(const struct net_device *dev)
5105 return dev->priv_flags & IFF_OPENVSWITCH;
5108 static inline bool netif_is_ovs_port(const struct net_device *dev)
5110 return dev->priv_flags & IFF_OVS_DATAPATH;
5113 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5115 return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5118 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5120 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5123 static inline bool netif_is_team_master(const struct net_device *dev)
5125 return dev->priv_flags & IFF_TEAM;
5128 static inline bool netif_is_team_port(const struct net_device *dev)
5130 return dev->priv_flags & IFF_TEAM_PORT;
5133 static inline bool netif_is_lag_master(const struct net_device *dev)
5135 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5138 static inline bool netif_is_lag_port(const struct net_device *dev)
5140 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5143 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5145 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5148 static inline bool netif_is_failover(const struct net_device *dev)
5150 return dev->priv_flags & IFF_FAILOVER;
5153 static inline bool netif_is_failover_slave(const struct net_device *dev)
5155 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5158 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5159 static inline void netif_keep_dst(struct net_device *dev)
5161 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5164 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5165 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5167 /* TODO: reserve and use an additional IFF bit, if we get more users */
5168 return netif_is_macsec(dev);
5171 extern struct pernet_operations __net_initdata loopback_net_ops;
5173 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5175 /* netdev_printk helpers, similar to dev_printk */
5177 static inline const char *netdev_name(const struct net_device *dev)
5179 if (!dev->name[0] || strchr(dev->name, '%'))
5180 return "(unnamed net_device)";
5184 static inline const char *netdev_reg_state(const struct net_device *dev)
5186 switch (dev->reg_state) {
5187 case NETREG_UNINITIALIZED: return " (uninitialized)";
5188 case NETREG_REGISTERED: return "";
5189 case NETREG_UNREGISTERING: return " (unregistering)";
5190 case NETREG_UNREGISTERED: return " (unregistered)";
5191 case NETREG_RELEASED: return " (released)";
5192 case NETREG_DUMMY: return " (dummy)";
5195 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5196 return " (unknown)";
5199 #define MODULE_ALIAS_NETDEV(device) \
5200 MODULE_ALIAS("netdev-" device)
5203 * netdev_WARN() acts like dev_printk(), but with the key difference
5204 * of using a WARN/WARN_ON to get the message out, including the
5205 * file/line information and a backtrace.
5207 #define netdev_WARN(dev, format, args...) \
5208 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5209 netdev_reg_state(dev), ##args)
5211 #define netdev_WARN_ONCE(dev, format, args...) \
5212 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5213 netdev_reg_state(dev), ##args)
5216 * The list of packet types we will receive (as opposed to discard)
5217 * and the routines to invoke.
5219 * Why 16. Because with 16 the only overlap we get on a hash of the
5220 * low nibble of the protocol value is RARP/SNAP/X.25.
5234 #define PTYPE_HASH_SIZE (16)
5235 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5237 extern struct list_head ptype_all __read_mostly;
5238 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5240 extern struct net_device *blackhole_netdev;
5242 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5243 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5244 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5245 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5246 #define DEV_STATS_READ(DEV, FIELD) atomic_long_read(&(DEV)->stats.__##FIELD)
5248 #endif /* _LINUX_NETDEVICE_H */