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>
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <uapi/linux/netdev.h>
51 #include <linux/hashtable.h>
52 #include <linux/rbtree.h>
53 #include <net/net_trackers.h>
54 #include <net/net_debug.h>
55 #include <net/dropreason.h>
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;
81 void synchronize_net(void);
82 void netdev_set_default_ethtool_ops(struct net_device *dev,
83 const struct ethtool_ops *ops);
84 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
86 /* Backlog congestion levels */
87 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
88 #define NET_RX_DROP 1 /* packet dropped */
90 #define MAX_NEST_DEV 8
93 * Transmit return codes: transmit return codes originate from three different
96 * - qdisc return codes
97 * - driver transmit return codes
100 * Drivers are allowed to return any one of those in their hard_start_xmit()
101 * function. Real network devices commonly used with qdiscs should only return
102 * the driver transmit return codes though - when qdiscs are used, the actual
103 * transmission happens asynchronously, so the value is not propagated to
104 * higher layers. Virtual network devices transmit synchronously; in this case
105 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
106 * others are propagated to higher layers.
109 /* qdisc ->enqueue() return codes. */
110 #define NET_XMIT_SUCCESS 0x00
111 #define NET_XMIT_DROP 0x01 /* skb dropped */
112 #define NET_XMIT_CN 0x02 /* congestion notification */
113 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
115 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
116 * indicates that the device will soon be dropping packets, or already drops
117 * some packets of the same priority; prompting us to send less aggressively. */
118 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
119 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
121 /* Driver transmit return codes */
122 #define NETDEV_TX_MASK 0xf0
125 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
126 NETDEV_TX_OK = 0x00, /* driver took care of packet */
127 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
129 typedef enum netdev_tx netdev_tx_t;
132 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
133 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
135 static inline bool dev_xmit_complete(int rc)
138 * Positive cases with an skb consumed by a driver:
139 * - successful transmission (rc == NETDEV_TX_OK)
140 * - error while transmitting (rc < 0)
141 * - error while queueing to a different device (rc & NET_XMIT_MASK)
143 if (likely(rc < NET_XMIT_MASK))
150 * Compute the worst-case header length according to the protocols
154 #if defined(CONFIG_HYPERV_NET)
155 # define LL_MAX_HEADER 128
156 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
157 # if defined(CONFIG_MAC80211_MESH)
158 # define LL_MAX_HEADER 128
160 # define LL_MAX_HEADER 96
163 # define LL_MAX_HEADER 32
166 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
167 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
168 #define MAX_HEADER LL_MAX_HEADER
170 #define MAX_HEADER (LL_MAX_HEADER + 48)
174 * Old network device statistics. Fields are native words
175 * (unsigned long) so they can be read and written atomically.
178 #define NET_DEV_STAT(FIELD) \
180 unsigned long FIELD; \
181 atomic_long_t __##FIELD; \
184 struct net_device_stats {
185 NET_DEV_STAT(rx_packets);
186 NET_DEV_STAT(tx_packets);
187 NET_DEV_STAT(rx_bytes);
188 NET_DEV_STAT(tx_bytes);
189 NET_DEV_STAT(rx_errors);
190 NET_DEV_STAT(tx_errors);
191 NET_DEV_STAT(rx_dropped);
192 NET_DEV_STAT(tx_dropped);
193 NET_DEV_STAT(multicast);
194 NET_DEV_STAT(collisions);
195 NET_DEV_STAT(rx_length_errors);
196 NET_DEV_STAT(rx_over_errors);
197 NET_DEV_STAT(rx_crc_errors);
198 NET_DEV_STAT(rx_frame_errors);
199 NET_DEV_STAT(rx_fifo_errors);
200 NET_DEV_STAT(rx_missed_errors);
201 NET_DEV_STAT(tx_aborted_errors);
202 NET_DEV_STAT(tx_carrier_errors);
203 NET_DEV_STAT(tx_fifo_errors);
204 NET_DEV_STAT(tx_heartbeat_errors);
205 NET_DEV_STAT(tx_window_errors);
206 NET_DEV_STAT(rx_compressed);
207 NET_DEV_STAT(tx_compressed);
211 /* per-cpu stats, allocated on demand.
212 * Try to fit them in a single cache line, for dev_get_stats() sake.
214 struct net_device_core_stats {
215 unsigned long rx_dropped;
216 unsigned long tx_dropped;
217 unsigned long rx_nohandler;
218 unsigned long rx_otherhost_dropped;
219 } __aligned(4 * sizeof(unsigned long));
221 #include <linux/cache.h>
222 #include <linux/skbuff.h>
225 #include <linux/static_key.h>
226 extern struct static_key_false rps_needed;
227 extern struct static_key_false rfs_needed;
234 struct netdev_hw_addr {
235 struct list_head list;
237 unsigned char addr[MAX_ADDR_LEN];
239 #define NETDEV_HW_ADDR_T_LAN 1
240 #define NETDEV_HW_ADDR_T_SAN 2
241 #define NETDEV_HW_ADDR_T_UNICAST 3
242 #define NETDEV_HW_ADDR_T_MULTICAST 4
247 struct rcu_head rcu_head;
250 struct netdev_hw_addr_list {
251 struct list_head list;
254 /* Auxiliary tree for faster lookup on addition and deletion */
258 #define netdev_hw_addr_list_count(l) ((l)->count)
259 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
260 #define netdev_hw_addr_list_for_each(ha, l) \
261 list_for_each_entry(ha, &(l)->list, list)
263 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
264 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
265 #define netdev_for_each_uc_addr(ha, dev) \
266 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
267 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
268 netdev_for_each_uc_addr((_ha), (_dev)) \
271 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
272 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
273 #define netdev_for_each_mc_addr(ha, dev) \
274 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
275 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
276 netdev_for_each_mc_addr((_ha), (_dev)) \
283 /* cached hardware header; allow for machine alignment needs. */
284 #define HH_DATA_MOD 16
285 #define HH_DATA_OFF(__len) \
286 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
287 #define HH_DATA_ALIGN(__len) \
288 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
289 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
292 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
294 * dev->hard_header_len ? (dev->hard_header_len +
295 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
297 * We could use other alignment values, but we must maintain the
298 * relationship HH alignment <= LL alignment.
300 #define LL_RESERVED_SPACE(dev) \
301 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
302 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
303 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
306 int (*create) (struct sk_buff *skb, struct net_device *dev,
307 unsigned short type, const void *daddr,
308 const void *saddr, unsigned int len);
309 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
310 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
311 void (*cache_update)(struct hh_cache *hh,
312 const struct net_device *dev,
313 const unsigned char *haddr);
314 bool (*validate)(const char *ll_header, unsigned int len);
315 __be16 (*parse_protocol)(const struct sk_buff *skb);
318 /* These flag bits are private to the generic network queueing
319 * layer; they may not be explicitly referenced by any other
323 enum netdev_state_t {
325 __LINK_STATE_PRESENT,
326 __LINK_STATE_NOCARRIER,
327 __LINK_STATE_LINKWATCH_PENDING,
328 __LINK_STATE_DORMANT,
329 __LINK_STATE_TESTING,
333 struct list_head list;
338 * size of gro hash buckets, must less than bit number of
339 * napi_struct::gro_bitmask
341 #define GRO_HASH_BUCKETS 8
344 * Structure for NAPI scheduling similar to tasklet but with weighting
347 /* The poll_list must only be managed by the entity which
348 * changes the state of the NAPI_STATE_SCHED bit. This means
349 * whoever atomically sets that bit can add this napi_struct
350 * to the per-CPU poll_list, and whoever clears that bit
351 * can remove from the list right before clearing the bit.
353 struct list_head poll_list;
357 int defer_hard_irqs_count;
358 unsigned long gro_bitmask;
359 int (*poll)(struct napi_struct *, int);
360 #ifdef CONFIG_NETPOLL
363 struct net_device *dev;
364 struct gro_list gro_hash[GRO_HASH_BUCKETS];
366 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
367 int rx_count; /* length of rx_list */
368 struct hrtimer timer;
369 struct list_head dev_list;
370 struct hlist_node napi_hash_node;
371 unsigned int napi_id;
372 struct task_struct *thread;
376 NAPI_STATE_SCHED, /* Poll is scheduled */
377 NAPI_STATE_MISSED, /* reschedule a napi */
378 NAPI_STATE_DISABLE, /* Disable pending */
379 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
380 NAPI_STATE_LISTED, /* NAPI added to system lists */
381 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
382 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
383 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
384 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
385 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
389 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
390 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
391 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
392 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
393 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
394 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
395 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
396 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
397 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
398 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
408 typedef enum gro_result gro_result_t;
411 * enum rx_handler_result - Possible return values for rx_handlers.
412 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
414 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
415 * case skb->dev was changed by rx_handler.
416 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
417 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
419 * rx_handlers are functions called from inside __netif_receive_skb(), to do
420 * special processing of the skb, prior to delivery to protocol handlers.
422 * Currently, a net_device can only have a single rx_handler registered. Trying
423 * to register a second rx_handler will return -EBUSY.
425 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
426 * To unregister a rx_handler on a net_device, use
427 * netdev_rx_handler_unregister().
429 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
432 * If the rx_handler consumed the skb in some way, it should return
433 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
434 * the skb to be delivered in some other way.
436 * If the rx_handler changed skb->dev, to divert the skb to another
437 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
438 * new device will be called if it exists.
440 * If the rx_handler decides the skb should be ignored, it should return
441 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
442 * are registered on exact device (ptype->dev == skb->dev).
444 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
445 * delivered, it should return RX_HANDLER_PASS.
447 * A device without a registered rx_handler will behave as if rx_handler
448 * returned RX_HANDLER_PASS.
451 enum rx_handler_result {
457 typedef enum rx_handler_result rx_handler_result_t;
458 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
460 void __napi_schedule(struct napi_struct *n);
461 void __napi_schedule_irqoff(struct napi_struct *n);
463 static inline bool napi_disable_pending(struct napi_struct *n)
465 return test_bit(NAPI_STATE_DISABLE, &n->state);
468 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
470 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
473 bool napi_schedule_prep(struct napi_struct *n);
476 * napi_schedule - schedule NAPI poll
479 * Schedule NAPI poll routine to be called if it is not already
482 static inline void napi_schedule(struct napi_struct *n)
484 if (napi_schedule_prep(n))
489 * napi_schedule_irqoff - schedule NAPI poll
492 * Variant of napi_schedule(), assuming hard irqs are masked.
494 static inline void napi_schedule_irqoff(struct napi_struct *n)
496 if (napi_schedule_prep(n))
497 __napi_schedule_irqoff(n);
500 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
501 static inline bool napi_reschedule(struct napi_struct *napi)
503 if (napi_schedule_prep(napi)) {
504 __napi_schedule(napi);
510 bool napi_complete_done(struct napi_struct *n, int work_done);
512 * napi_complete - NAPI processing complete
515 * Mark NAPI processing as complete.
516 * Consider using napi_complete_done() instead.
517 * Return false if device should avoid rearming interrupts.
519 static inline bool napi_complete(struct napi_struct *n)
521 return napi_complete_done(n, 0);
524 int dev_set_threaded(struct net_device *dev, bool threaded);
527 * napi_disable - prevent NAPI from scheduling
530 * Stop NAPI from being scheduled on this context.
531 * Waits till any outstanding processing completes.
533 void napi_disable(struct napi_struct *n);
535 void napi_enable(struct napi_struct *n);
538 * napi_synchronize - wait until NAPI is not running
541 * Wait until NAPI is done being scheduled on this context.
542 * Waits till any outstanding processing completes but
543 * does not disable future activations.
545 static inline void napi_synchronize(const struct napi_struct *n)
547 if (IS_ENABLED(CONFIG_SMP))
548 while (test_bit(NAPI_STATE_SCHED, &n->state))
555 * napi_if_scheduled_mark_missed - if napi is running, set the
559 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
562 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
564 unsigned long val, new;
566 val = READ_ONCE(n->state);
568 if (val & NAPIF_STATE_DISABLE)
571 if (!(val & NAPIF_STATE_SCHED))
574 new = val | NAPIF_STATE_MISSED;
575 } while (!try_cmpxchg(&n->state, &val, new));
580 enum netdev_queue_state_t {
581 __QUEUE_STATE_DRV_XOFF,
582 __QUEUE_STATE_STACK_XOFF,
583 __QUEUE_STATE_FROZEN,
586 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
587 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
588 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
590 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
591 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
593 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
597 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
598 * netif_tx_* functions below are used to manipulate this flag. The
599 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
600 * queue independently. The netif_xmit_*stopped functions below are called
601 * to check if the queue has been stopped by the driver or stack (either
602 * of the XOFF bits are set in the state). Drivers should not need to call
603 * netif_xmit*stopped functions, they should only be using netif_tx_*.
606 struct netdev_queue {
610 struct net_device *dev;
611 netdevice_tracker dev_tracker;
613 struct Qdisc __rcu *qdisc;
614 struct Qdisc *qdisc_sleeping;
618 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
621 unsigned long tx_maxrate;
623 * Number of TX timeouts for this queue
624 * (/sys/class/net/DEV/Q/trans_timeout)
626 atomic_long_t trans_timeout;
628 /* Subordinate device that the queue has been assigned to */
629 struct net_device *sb_dev;
630 #ifdef CONFIG_XDP_SOCKETS
631 struct xsk_buff_pool *pool;
636 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
639 * Time (in jiffies) of last Tx
641 unsigned long trans_start;
648 } ____cacheline_aligned_in_smp;
650 extern int sysctl_fb_tunnels_only_for_init_net;
651 extern int sysctl_devconf_inherit_init_net;
654 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
655 * == 1 : For initns only
658 static inline bool net_has_fallback_tunnels(const struct net *net)
660 #if IS_ENABLED(CONFIG_SYSCTL)
661 int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
663 return !fb_tunnels_only_for_init_net ||
664 (net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
670 static inline int net_inherit_devconf(void)
672 #if IS_ENABLED(CONFIG_SYSCTL)
673 return READ_ONCE(sysctl_devconf_inherit_init_net);
679 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
681 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
688 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
690 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
697 * This structure holds an RPS map which can be of variable length. The
698 * map is an array of CPUs.
705 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
708 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
709 * tail pointer for that CPU's input queue at the time of last enqueue, and
710 * a hardware filter index.
712 struct rps_dev_flow {
715 unsigned int last_qtail;
717 #define RPS_NO_FILTER 0xffff
720 * The rps_dev_flow_table structure contains a table of flow mappings.
722 struct rps_dev_flow_table {
725 struct rps_dev_flow flows[];
727 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
728 ((_num) * sizeof(struct rps_dev_flow)))
731 * The rps_sock_flow_table contains mappings of flows to the last CPU
732 * on which they were processed by the application (set in recvmsg).
733 * Each entry is a 32bit value. Upper part is the high-order bits
734 * of flow hash, lower part is CPU number.
735 * rps_cpu_mask is used to partition the space, depending on number of
736 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
737 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
738 * meaning we use 32-6=26 bits for the hash.
740 struct rps_sock_flow_table {
743 u32 ents[] ____cacheline_aligned_in_smp;
745 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
747 #define RPS_NO_CPU 0xffff
749 extern u32 rps_cpu_mask;
750 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
752 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
756 unsigned int index = hash & table->mask;
757 u32 val = hash & ~rps_cpu_mask;
759 /* We only give a hint, preemption can change CPU under us */
760 val |= raw_smp_processor_id();
762 if (table->ents[index] != val)
763 table->ents[index] = val;
767 #ifdef CONFIG_RFS_ACCEL
768 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
771 #endif /* CONFIG_RPS */
773 /* This structure contains an instance of an RX queue. */
774 struct netdev_rx_queue {
775 struct xdp_rxq_info xdp_rxq;
777 struct rps_map __rcu *rps_map;
778 struct rps_dev_flow_table __rcu *rps_flow_table;
781 struct net_device *dev;
782 netdevice_tracker dev_tracker;
784 #ifdef CONFIG_XDP_SOCKETS
785 struct xsk_buff_pool *pool;
787 } ____cacheline_aligned_in_smp;
790 * RX queue sysfs structures and functions.
792 struct rx_queue_attribute {
793 struct attribute attr;
794 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
795 ssize_t (*store)(struct netdev_rx_queue *queue,
796 const char *buf, size_t len);
799 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
808 * This structure holds an XPS map which can be of variable length. The
809 * map is an array of queues.
813 unsigned int alloc_len;
817 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
818 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
819 - sizeof(struct xps_map)) / sizeof(u16))
822 * This structure holds all XPS maps for device. Maps are indexed by CPU.
824 * We keep track of the number of cpus/rxqs used when the struct is allocated,
825 * in nr_ids. This will help not accessing out-of-bound memory.
827 * We keep track of the number of traffic classes used when the struct is
828 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
829 * not crossing its upper bound, as the original dev->num_tc can be updated in
832 struct xps_dev_maps {
836 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
839 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
840 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
842 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
843 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
845 #endif /* CONFIG_XPS */
847 #define TC_MAX_QUEUE 16
848 #define TC_BITMASK 15
849 /* HW offloaded queuing disciplines txq count and offset maps */
850 struct netdev_tc_txq {
855 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
857 * This structure is to hold information about the device
858 * configured to run FCoE protocol stack.
860 struct netdev_fcoe_hbainfo {
861 char manufacturer[64];
862 char serial_number[64];
863 char hardware_version[64];
864 char driver_version[64];
865 char optionrom_version[64];
866 char firmware_version[64];
868 char model_description[256];
872 #define MAX_PHYS_ITEM_ID_LEN 32
874 /* This structure holds a unique identifier to identify some
875 * physical item (port for example) used by a netdevice.
877 struct netdev_phys_item_id {
878 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
879 unsigned char id_len;
882 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
883 struct netdev_phys_item_id *b)
885 return a->id_len == b->id_len &&
886 memcmp(a->id, b->id, a->id_len) == 0;
889 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
891 struct net_device *sb_dev);
893 enum net_device_path_type {
894 DEV_PATH_ETHERNET = 0,
902 struct net_device_path {
903 enum net_device_path_type type;
904 const struct net_device *dev;
913 DEV_PATH_BR_VLAN_KEEP,
914 DEV_PATH_BR_VLAN_TAG,
915 DEV_PATH_BR_VLAN_UNTAG,
916 DEV_PATH_BR_VLAN_UNTAG_HW,
934 #define NET_DEVICE_PATH_STACK_MAX 5
935 #define NET_DEVICE_PATH_VLAN_MAX 2
937 struct net_device_path_stack {
939 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
942 struct net_device_path_ctx {
943 const struct net_device *dev;
950 } vlan[NET_DEVICE_PATH_VLAN_MAX];
955 TC_SETUP_QDISC_MQPRIO,
958 TC_SETUP_CLSMATCHALL,
968 TC_SETUP_QDISC_TAPRIO,
977 /* These structures hold the attributes of bpf state that are being passed
978 * to the netdevice through the bpf op.
980 enum bpf_netdev_command {
981 /* Set or clear a bpf program used in the earliest stages of packet
982 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
983 * is responsible for calling bpf_prog_put on any old progs that are
984 * stored. In case of error, the callee need not release the new prog
985 * reference, but on success it takes ownership and must bpf_prog_put
986 * when it is no longer used.
990 /* BPF program for offload callbacks, invoked at program load time. */
991 BPF_OFFLOAD_MAP_ALLOC,
992 BPF_OFFLOAD_MAP_FREE,
996 struct bpf_prog_offload_ops;
997 struct netlink_ext_ack;
999 struct xdp_dev_bulk_queue;
1000 struct bpf_xdp_link;
1009 struct bpf_xdp_entity {
1010 struct bpf_prog *prog;
1011 struct bpf_xdp_link *link;
1015 enum bpf_netdev_command command;
1017 /* XDP_SETUP_PROG */
1020 struct bpf_prog *prog;
1021 struct netlink_ext_ack *extack;
1023 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1025 struct bpf_offloaded_map *offmap;
1027 /* XDP_SETUP_XSK_POOL */
1029 struct xsk_buff_pool *pool;
1035 /* Flags for ndo_xsk_wakeup. */
1036 #define XDP_WAKEUP_RX (1 << 0)
1037 #define XDP_WAKEUP_TX (1 << 1)
1039 #ifdef CONFIG_XFRM_OFFLOAD
1040 struct xfrmdev_ops {
1041 int (*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1042 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1043 void (*xdo_dev_state_free) (struct xfrm_state *x);
1044 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1045 struct xfrm_state *x);
1046 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1047 void (*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1048 int (*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1049 void (*xdo_dev_policy_delete) (struct xfrm_policy *x);
1050 void (*xdo_dev_policy_free) (struct xfrm_policy *x);
1054 struct dev_ifalias {
1055 struct rcu_head rcuhead;
1062 struct netdev_net_notifier {
1063 struct list_head list;
1064 struct notifier_block *nb;
1068 * This structure defines the management hooks for network devices.
1069 * The following hooks can be defined; unless noted otherwise, they are
1070 * optional and can be filled with a null pointer.
1072 * int (*ndo_init)(struct net_device *dev);
1073 * This function is called once when a network device is registered.
1074 * The network device can use this for any late stage initialization
1075 * or semantic validation. It can fail with an error code which will
1076 * be propagated back to register_netdev.
1078 * void (*ndo_uninit)(struct net_device *dev);
1079 * This function is called when device is unregistered or when registration
1080 * fails. It is not called if init fails.
1082 * int (*ndo_open)(struct net_device *dev);
1083 * This function is called when a network device transitions to the up
1086 * int (*ndo_stop)(struct net_device *dev);
1087 * This function is called when a network device transitions to the down
1090 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1091 * struct net_device *dev);
1092 * Called when a packet needs to be transmitted.
1093 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1094 * the queue before that can happen; it's for obsolete devices and weird
1095 * corner cases, but the stack really does a non-trivial amount
1096 * of useless work if you return NETDEV_TX_BUSY.
1097 * Required; cannot be NULL.
1099 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1100 * struct net_device *dev
1101 * netdev_features_t features);
1102 * Called by core transmit path to determine if device is capable of
1103 * performing offload operations on a given packet. This is to give
1104 * the device an opportunity to implement any restrictions that cannot
1105 * be otherwise expressed by feature flags. The check is called with
1106 * the set of features that the stack has calculated and it returns
1107 * those the driver believes to be appropriate.
1109 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1110 * struct net_device *sb_dev);
1111 * Called to decide which queue to use when device supports multiple
1114 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1115 * This function is called to allow device receiver to make
1116 * changes to configuration when multicast or promiscuous is enabled.
1118 * void (*ndo_set_rx_mode)(struct net_device *dev);
1119 * This function is called device changes address list filtering.
1120 * If driver handles unicast address filtering, it should set
1121 * IFF_UNICAST_FLT in its priv_flags.
1123 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1124 * This function is called when the Media Access Control address
1125 * needs to be changed. If this interface is not defined, the
1126 * MAC address can not be changed.
1128 * int (*ndo_validate_addr)(struct net_device *dev);
1129 * Test if Media Access Control address is valid for the device.
1131 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1132 * Old-style ioctl entry point. This is used internally by the
1133 * appletalk and ieee802154 subsystems but is no longer called by
1134 * the device ioctl handler.
1136 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1137 * Used by the bonding driver for its device specific ioctls:
1138 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1139 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1141 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1142 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1143 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1145 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1146 * Used to set network devices bus interface parameters. This interface
1147 * is retained for legacy reasons; new devices should use the bus
1148 * interface (PCI) for low level management.
1150 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1151 * Called when a user wants to change the Maximum Transfer Unit
1154 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1155 * Callback used when the transmitter has not made any progress
1156 * for dev->watchdog ticks.
1158 * void (*ndo_get_stats64)(struct net_device *dev,
1159 * struct rtnl_link_stats64 *storage);
1160 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1161 * Called when a user wants to get the network device usage
1162 * statistics. Drivers must do one of the following:
1163 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1164 * rtnl_link_stats64 structure passed by the caller.
1165 * 2. Define @ndo_get_stats to update a net_device_stats structure
1166 * (which should normally be dev->stats) and return a pointer to
1167 * it. The structure may be changed asynchronously only if each
1168 * field is written atomically.
1169 * 3. Update dev->stats asynchronously and atomically, and define
1170 * neither operation.
1172 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1173 * Return true if this device supports offload stats of this attr_id.
1175 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1177 * Get statistics for offload operations by attr_id. Write it into the
1178 * attr_data pointer.
1180 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1181 * If device supports VLAN filtering this function is called when a
1182 * VLAN id is registered.
1184 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1185 * If device supports VLAN filtering this function is called when a
1186 * VLAN id is unregistered.
1188 * void (*ndo_poll_controller)(struct net_device *dev);
1190 * SR-IOV management functions.
1191 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1192 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1193 * u8 qos, __be16 proto);
1194 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1196 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1197 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1198 * int (*ndo_get_vf_config)(struct net_device *dev,
1199 * int vf, struct ifla_vf_info *ivf);
1200 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1201 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1202 * struct nlattr *port[]);
1204 * Enable or disable the VF ability to query its RSS Redirection Table and
1205 * Hash Key. This is needed since on some devices VF share this information
1206 * with PF and querying it may introduce a theoretical security risk.
1207 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1208 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1209 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1211 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1212 * This is always called from the stack with the rtnl lock held and netif
1213 * tx queues stopped. This allows the netdevice to perform queue
1214 * management safely.
1216 * Fiber Channel over Ethernet (FCoE) offload functions.
1217 * int (*ndo_fcoe_enable)(struct net_device *dev);
1218 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1219 * so the underlying device can perform whatever needed configuration or
1220 * initialization to support acceleration of FCoE traffic.
1222 * int (*ndo_fcoe_disable)(struct net_device *dev);
1223 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1224 * so the underlying device can perform whatever needed clean-ups to
1225 * stop supporting acceleration of FCoE traffic.
1227 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1228 * struct scatterlist *sgl, unsigned int sgc);
1229 * Called when the FCoE Initiator wants to initialize an I/O that
1230 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1231 * perform necessary setup and returns 1 to indicate the device is set up
1232 * successfully to perform DDP on this I/O, otherwise this returns 0.
1234 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1235 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1236 * indicated by the FC exchange id 'xid', so the underlying device can
1237 * clean up and reuse resources for later DDP requests.
1239 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1240 * struct scatterlist *sgl, unsigned int sgc);
1241 * Called when the FCoE Target wants to initialize an I/O that
1242 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1243 * perform necessary setup and returns 1 to indicate the device is set up
1244 * successfully to perform DDP on this I/O, otherwise this returns 0.
1246 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1247 * struct netdev_fcoe_hbainfo *hbainfo);
1248 * Called when the FCoE Protocol stack wants information on the underlying
1249 * device. This information is utilized by the FCoE protocol stack to
1250 * register attributes with Fiber Channel management service as per the
1251 * FC-GS Fabric Device Management Information(FDMI) specification.
1253 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1254 * Called when the underlying device wants to override default World Wide
1255 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1256 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1257 * protocol stack to use.
1260 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1261 * u16 rxq_index, u32 flow_id);
1262 * Set hardware filter for RFS. rxq_index is the target queue index;
1263 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1264 * Return the filter ID on success, or a negative error code.
1266 * Slave management functions (for bridge, bonding, etc).
1267 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1268 * Called to make another netdev an underling.
1270 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1271 * Called to release previously enslaved netdev.
1273 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1274 * struct sk_buff *skb,
1276 * Get the xmit slave of master device. If all_slaves is true, function
1277 * assume all the slaves can transmit.
1279 * Feature/offload setting functions.
1280 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1281 * netdev_features_t features);
1282 * Adjusts the requested feature flags according to device-specific
1283 * constraints, and returns the resulting flags. Must not modify
1286 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1287 * Called to update device configuration to new features. Passed
1288 * feature set might be less than what was returned by ndo_fix_features()).
1289 * Must return >0 or -errno if it changed dev->features itself.
1291 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1292 * struct net_device *dev,
1293 * const unsigned char *addr, u16 vid, u16 flags,
1294 * struct netlink_ext_ack *extack);
1295 * Adds an FDB entry to dev for addr.
1296 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1297 * struct net_device *dev,
1298 * const unsigned char *addr, u16 vid)
1299 * Deletes the FDB entry from dev coresponding to addr.
1300 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1301 * struct net_device *dev,
1303 * struct netlink_ext_ack *extack);
1304 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1305 * struct net_device *dev, struct net_device *filter_dev,
1307 * Used to add FDB entries to dump requests. Implementers should add
1308 * entries to skb and update idx with the number of entries.
1310 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1311 * u16 flags, struct netlink_ext_ack *extack)
1312 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1313 * struct net_device *dev, u32 filter_mask,
1315 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1318 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1319 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1320 * which do not represent real hardware may define this to allow their
1321 * userspace components to manage their virtual carrier state. Devices
1322 * that determine carrier state from physical hardware properties (eg
1323 * network cables) or protocol-dependent mechanisms (eg
1324 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1326 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1327 * struct netdev_phys_item_id *ppid);
1328 * Called to get ID of physical port of this device. If driver does
1329 * not implement this, it is assumed that the hw is not able to have
1330 * multiple net devices on single physical port.
1332 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1333 * struct netdev_phys_item_id *ppid)
1334 * Called to get the parent ID of the physical port of this device.
1336 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1337 * struct net_device *dev)
1338 * Called by upper layer devices to accelerate switching or other
1339 * station functionality into hardware. 'pdev is the lowerdev
1340 * to use for the offload and 'dev' is the net device that will
1341 * back the offload. Returns a pointer to the private structure
1342 * the upper layer will maintain.
1343 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1344 * Called by upper layer device to delete the station created
1345 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1346 * the station and priv is the structure returned by the add
1348 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1349 * int queue_index, u32 maxrate);
1350 * Called when a user wants to set a max-rate limitation of specific
1352 * int (*ndo_get_iflink)(const struct net_device *dev);
1353 * Called to get the iflink value of this device.
1354 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1355 * This function is used to get egress tunnel information for given skb.
1356 * This is useful for retrieving outer tunnel header parameters while
1358 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1359 * This function is used to specify the headroom that the skb must
1360 * consider when allocation skb during packet reception. Setting
1361 * appropriate rx headroom value allows avoiding skb head copy on
1362 * forward. Setting a negative value resets the rx headroom to the
1364 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1365 * This function is used to set or query state related to XDP on the
1366 * netdevice and manage BPF offload. See definition of
1367 * enum bpf_netdev_command for details.
1368 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1370 * This function is used to submit @n XDP packets for transmit on a
1371 * netdevice. Returns number of frames successfully transmitted, frames
1372 * that got dropped are freed/returned via xdp_return_frame().
1373 * Returns negative number, means general error invoking ndo, meaning
1374 * no frames were xmit'ed and core-caller will free all frames.
1375 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1376 * struct xdp_buff *xdp);
1377 * Get the xmit slave of master device based on the xdp_buff.
1378 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1379 * This function is used to wake up the softirq, ksoftirqd or kthread
1380 * responsible for sending and/or receiving packets on a specific
1381 * queue id bound to an AF_XDP socket. The flags field specifies if
1382 * only RX, only Tx, or both should be woken up using the flags
1383 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1384 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1386 * Add, change, delete or get information on an IPv4 tunnel.
1387 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1388 * If a device is paired with a peer device, return the peer instance.
1389 * The caller must be under RCU read context.
1390 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1391 * Get the forwarding path to reach the real device from the HW destination address
1392 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1393 * const struct skb_shared_hwtstamps *hwtstamps,
1395 * Get hardware timestamp based on normal/adjustable time or free running
1396 * cycle counter. This function is required if physical clock supports a
1397 * free running cycle counter.
1399 struct net_device_ops {
1400 int (*ndo_init)(struct net_device *dev);
1401 void (*ndo_uninit)(struct net_device *dev);
1402 int (*ndo_open)(struct net_device *dev);
1403 int (*ndo_stop)(struct net_device *dev);
1404 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1405 struct net_device *dev);
1406 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1407 struct net_device *dev,
1408 netdev_features_t features);
1409 u16 (*ndo_select_queue)(struct net_device *dev,
1410 struct sk_buff *skb,
1411 struct net_device *sb_dev);
1412 void (*ndo_change_rx_flags)(struct net_device *dev,
1414 void (*ndo_set_rx_mode)(struct net_device *dev);
1415 int (*ndo_set_mac_address)(struct net_device *dev,
1417 int (*ndo_validate_addr)(struct net_device *dev);
1418 int (*ndo_do_ioctl)(struct net_device *dev,
1419 struct ifreq *ifr, int cmd);
1420 int (*ndo_eth_ioctl)(struct net_device *dev,
1421 struct ifreq *ifr, int cmd);
1422 int (*ndo_siocbond)(struct net_device *dev,
1423 struct ifreq *ifr, int cmd);
1424 int (*ndo_siocwandev)(struct net_device *dev,
1425 struct if_settings *ifs);
1426 int (*ndo_siocdevprivate)(struct net_device *dev,
1428 void __user *data, int cmd);
1429 int (*ndo_set_config)(struct net_device *dev,
1431 int (*ndo_change_mtu)(struct net_device *dev,
1433 int (*ndo_neigh_setup)(struct net_device *dev,
1434 struct neigh_parms *);
1435 void (*ndo_tx_timeout) (struct net_device *dev,
1436 unsigned int txqueue);
1438 void (*ndo_get_stats64)(struct net_device *dev,
1439 struct rtnl_link_stats64 *storage);
1440 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1441 int (*ndo_get_offload_stats)(int attr_id,
1442 const struct net_device *dev,
1444 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1446 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1447 __be16 proto, u16 vid);
1448 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1449 __be16 proto, u16 vid);
1450 #ifdef CONFIG_NET_POLL_CONTROLLER
1451 void (*ndo_poll_controller)(struct net_device *dev);
1452 int (*ndo_netpoll_setup)(struct net_device *dev,
1453 struct netpoll_info *info);
1454 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1456 int (*ndo_set_vf_mac)(struct net_device *dev,
1457 int queue, u8 *mac);
1458 int (*ndo_set_vf_vlan)(struct net_device *dev,
1459 int queue, u16 vlan,
1460 u8 qos, __be16 proto);
1461 int (*ndo_set_vf_rate)(struct net_device *dev,
1462 int vf, int min_tx_rate,
1464 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1465 int vf, bool setting);
1466 int (*ndo_set_vf_trust)(struct net_device *dev,
1467 int vf, bool setting);
1468 int (*ndo_get_vf_config)(struct net_device *dev,
1470 struct ifla_vf_info *ivf);
1471 int (*ndo_set_vf_link_state)(struct net_device *dev,
1472 int vf, int link_state);
1473 int (*ndo_get_vf_stats)(struct net_device *dev,
1475 struct ifla_vf_stats
1477 int (*ndo_set_vf_port)(struct net_device *dev,
1479 struct nlattr *port[]);
1480 int (*ndo_get_vf_port)(struct net_device *dev,
1481 int vf, struct sk_buff *skb);
1482 int (*ndo_get_vf_guid)(struct net_device *dev,
1484 struct ifla_vf_guid *node_guid,
1485 struct ifla_vf_guid *port_guid);
1486 int (*ndo_set_vf_guid)(struct net_device *dev,
1489 int (*ndo_set_vf_rss_query_en)(
1490 struct net_device *dev,
1491 int vf, bool setting);
1492 int (*ndo_setup_tc)(struct net_device *dev,
1493 enum tc_setup_type type,
1495 #if IS_ENABLED(CONFIG_FCOE)
1496 int (*ndo_fcoe_enable)(struct net_device *dev);
1497 int (*ndo_fcoe_disable)(struct net_device *dev);
1498 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1500 struct scatterlist *sgl,
1502 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1504 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1506 struct scatterlist *sgl,
1508 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1509 struct netdev_fcoe_hbainfo *hbainfo);
1512 #if IS_ENABLED(CONFIG_LIBFCOE)
1513 #define NETDEV_FCOE_WWNN 0
1514 #define NETDEV_FCOE_WWPN 1
1515 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1516 u64 *wwn, int type);
1519 #ifdef CONFIG_RFS_ACCEL
1520 int (*ndo_rx_flow_steer)(struct net_device *dev,
1521 const struct sk_buff *skb,
1525 int (*ndo_add_slave)(struct net_device *dev,
1526 struct net_device *slave_dev,
1527 struct netlink_ext_ack *extack);
1528 int (*ndo_del_slave)(struct net_device *dev,
1529 struct net_device *slave_dev);
1530 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1531 struct sk_buff *skb,
1533 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1535 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1536 netdev_features_t features);
1537 int (*ndo_set_features)(struct net_device *dev,
1538 netdev_features_t features);
1539 int (*ndo_neigh_construct)(struct net_device *dev,
1540 struct neighbour *n);
1541 void (*ndo_neigh_destroy)(struct net_device *dev,
1542 struct neighbour *n);
1544 int (*ndo_fdb_add)(struct ndmsg *ndm,
1545 struct nlattr *tb[],
1546 struct net_device *dev,
1547 const unsigned char *addr,
1550 struct netlink_ext_ack *extack);
1551 int (*ndo_fdb_del)(struct ndmsg *ndm,
1552 struct nlattr *tb[],
1553 struct net_device *dev,
1554 const unsigned char *addr,
1555 u16 vid, struct netlink_ext_ack *extack);
1556 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1557 struct nlattr *tb[],
1558 struct net_device *dev,
1560 struct netlink_ext_ack *extack);
1561 int (*ndo_fdb_dump)(struct sk_buff *skb,
1562 struct netlink_callback *cb,
1563 struct net_device *dev,
1564 struct net_device *filter_dev,
1566 int (*ndo_fdb_get)(struct sk_buff *skb,
1567 struct nlattr *tb[],
1568 struct net_device *dev,
1569 const unsigned char *addr,
1570 u16 vid, u32 portid, u32 seq,
1571 struct netlink_ext_ack *extack);
1572 int (*ndo_bridge_setlink)(struct net_device *dev,
1573 struct nlmsghdr *nlh,
1575 struct netlink_ext_ack *extack);
1576 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1578 struct net_device *dev,
1581 int (*ndo_bridge_dellink)(struct net_device *dev,
1582 struct nlmsghdr *nlh,
1584 int (*ndo_change_carrier)(struct net_device *dev,
1586 int (*ndo_get_phys_port_id)(struct net_device *dev,
1587 struct netdev_phys_item_id *ppid);
1588 int (*ndo_get_port_parent_id)(struct net_device *dev,
1589 struct netdev_phys_item_id *ppid);
1590 int (*ndo_get_phys_port_name)(struct net_device *dev,
1591 char *name, size_t len);
1592 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1593 struct net_device *dev);
1594 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1597 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1600 int (*ndo_get_iflink)(const struct net_device *dev);
1601 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1602 struct sk_buff *skb);
1603 void (*ndo_set_rx_headroom)(struct net_device *dev,
1604 int needed_headroom);
1605 int (*ndo_bpf)(struct net_device *dev,
1606 struct netdev_bpf *bpf);
1607 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1608 struct xdp_frame **xdp,
1610 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1611 struct xdp_buff *xdp);
1612 int (*ndo_xsk_wakeup)(struct net_device *dev,
1613 u32 queue_id, u32 flags);
1614 int (*ndo_tunnel_ctl)(struct net_device *dev,
1615 struct ip_tunnel_parm *p, int cmd);
1616 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1617 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1618 struct net_device_path *path);
1619 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1620 const struct skb_shared_hwtstamps *hwtstamps,
1624 struct xdp_metadata_ops {
1625 int (*xmo_rx_timestamp)(const struct xdp_md *ctx, u64 *timestamp);
1626 int (*xmo_rx_hash)(const struct xdp_md *ctx, u32 *hash);
1630 * enum netdev_priv_flags - &struct net_device priv_flags
1632 * These are the &struct net_device, they are only set internally
1633 * by drivers and used in the kernel. These flags are invisible to
1634 * userspace; this means that the order of these flags can change
1635 * during any kernel release.
1637 * You should have a pretty good reason to be extending these flags.
1639 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1640 * @IFF_EBRIDGE: Ethernet bridging device
1641 * @IFF_BONDING: bonding master or slave
1642 * @IFF_ISATAP: ISATAP interface (RFC4214)
1643 * @IFF_WAN_HDLC: WAN HDLC device
1644 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1646 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1647 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1648 * @IFF_MACVLAN_PORT: device used as macvlan port
1649 * @IFF_BRIDGE_PORT: device used as bridge port
1650 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1651 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1652 * @IFF_UNICAST_FLT: Supports unicast filtering
1653 * @IFF_TEAM_PORT: device used as team port
1654 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1655 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1656 * change when it's running
1657 * @IFF_MACVLAN: Macvlan device
1658 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1659 * underlying stacked devices
1660 * @IFF_L3MDEV_MASTER: device is an L3 master device
1661 * @IFF_NO_QUEUE: device can run without qdisc attached
1662 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1663 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1664 * @IFF_TEAM: device is a team device
1665 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1666 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1667 * entity (i.e. the master device for bridged veth)
1668 * @IFF_MACSEC: device is a MACsec device
1669 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1670 * @IFF_FAILOVER: device is a failover master device
1671 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1672 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1673 * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1674 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1675 * skb_headlen(skb) == 0 (data starts from frag0)
1676 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1678 enum netdev_priv_flags {
1679 IFF_802_1Q_VLAN = 1<<0,
1683 IFF_WAN_HDLC = 1<<4,
1684 IFF_XMIT_DST_RELEASE = 1<<5,
1685 IFF_DONT_BRIDGE = 1<<6,
1686 IFF_DISABLE_NETPOLL = 1<<7,
1687 IFF_MACVLAN_PORT = 1<<8,
1688 IFF_BRIDGE_PORT = 1<<9,
1689 IFF_OVS_DATAPATH = 1<<10,
1690 IFF_TX_SKB_SHARING = 1<<11,
1691 IFF_UNICAST_FLT = 1<<12,
1692 IFF_TEAM_PORT = 1<<13,
1693 IFF_SUPP_NOFCS = 1<<14,
1694 IFF_LIVE_ADDR_CHANGE = 1<<15,
1695 IFF_MACVLAN = 1<<16,
1696 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1697 IFF_L3MDEV_MASTER = 1<<18,
1698 IFF_NO_QUEUE = 1<<19,
1699 IFF_OPENVSWITCH = 1<<20,
1700 IFF_L3MDEV_SLAVE = 1<<21,
1702 IFF_RXFH_CONFIGURED = 1<<23,
1703 IFF_PHONY_HEADROOM = 1<<24,
1705 IFF_NO_RX_HANDLER = 1<<26,
1706 IFF_FAILOVER = 1<<27,
1707 IFF_FAILOVER_SLAVE = 1<<28,
1708 IFF_L3MDEV_RX_HANDLER = 1<<29,
1709 IFF_NO_ADDRCONF = BIT_ULL(30),
1710 IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1711 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1714 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1715 #define IFF_EBRIDGE IFF_EBRIDGE
1716 #define IFF_BONDING IFF_BONDING
1717 #define IFF_ISATAP IFF_ISATAP
1718 #define IFF_WAN_HDLC IFF_WAN_HDLC
1719 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1720 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1721 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1722 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1723 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1724 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1725 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1726 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1727 #define IFF_TEAM_PORT IFF_TEAM_PORT
1728 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1729 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1730 #define IFF_MACVLAN IFF_MACVLAN
1731 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1732 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1733 #define IFF_NO_QUEUE IFF_NO_QUEUE
1734 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1735 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1736 #define IFF_TEAM IFF_TEAM
1737 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1738 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1739 #define IFF_MACSEC IFF_MACSEC
1740 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1741 #define IFF_FAILOVER IFF_FAILOVER
1742 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1743 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1744 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1746 /* Specifies the type of the struct net_device::ml_priv pointer */
1747 enum netdev_ml_priv_type {
1753 * struct net_device - The DEVICE structure.
1755 * Actually, this whole structure is a big mistake. It mixes I/O
1756 * data with strictly "high-level" data, and it has to know about
1757 * almost every data structure used in the INET module.
1759 * @name: This is the first field of the "visible" part of this structure
1760 * (i.e. as seen by users in the "Space.c" file). It is the name
1763 * @name_node: Name hashlist node
1764 * @ifalias: SNMP alias
1765 * @mem_end: Shared memory end
1766 * @mem_start: Shared memory start
1767 * @base_addr: Device I/O address
1768 * @irq: Device IRQ number
1770 * @state: Generic network queuing layer state, see netdev_state_t
1771 * @dev_list: The global list of network devices
1772 * @napi_list: List entry used for polling NAPI devices
1773 * @unreg_list: List entry when we are unregistering the
1774 * device; see the function unregister_netdev
1775 * @close_list: List entry used when we are closing the device
1776 * @ptype_all: Device-specific packet handlers for all protocols
1777 * @ptype_specific: Device-specific, protocol-specific packet handlers
1779 * @adj_list: Directly linked devices, like slaves for bonding
1780 * @features: Currently active device features
1781 * @hw_features: User-changeable features
1783 * @wanted_features: User-requested features
1784 * @vlan_features: Mask of features inheritable by VLAN devices
1786 * @hw_enc_features: Mask of features inherited by encapsulating devices
1787 * This field indicates what encapsulation
1788 * offloads the hardware is capable of doing,
1789 * and drivers will need to set them appropriately.
1791 * @mpls_features: Mask of features inheritable by MPLS
1792 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1794 * @ifindex: interface index
1795 * @group: The group the device belongs to
1797 * @stats: Statistics struct, which was left as a legacy, use
1798 * rtnl_link_stats64 instead
1800 * @core_stats: core networking counters,
1801 * do not use this in drivers
1802 * @carrier_up_count: Number of times the carrier has been up
1803 * @carrier_down_count: Number of times the carrier has been down
1805 * @wireless_handlers: List of functions to handle Wireless Extensions,
1807 * see <net/iw_handler.h> for details.
1808 * @wireless_data: Instance data managed by the core of wireless extensions
1810 * @netdev_ops: Includes several pointers to callbacks,
1811 * if one wants to override the ndo_*() functions
1812 * @xdp_metadata_ops: Includes pointers to XDP metadata callbacks.
1813 * @ethtool_ops: Management operations
1814 * @l3mdev_ops: Layer 3 master device operations
1815 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1816 * discovery handling. Necessary for e.g. 6LoWPAN.
1817 * @xfrmdev_ops: Transformation offload operations
1818 * @tlsdev_ops: Transport Layer Security offload operations
1819 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1820 * of Layer 2 headers.
1822 * @flags: Interface flags (a la BSD)
1823 * @xdp_features: XDP capability supported by the device
1824 * @priv_flags: Like 'flags' but invisible to userspace,
1825 * see if.h for the definitions
1826 * @gflags: Global flags ( kept as legacy )
1827 * @padded: How much padding added by alloc_netdev()
1828 * @operstate: RFC2863 operstate
1829 * @link_mode: Mapping policy to operstate
1830 * @if_port: Selectable AUI, TP, ...
1832 * @mtu: Interface MTU value
1833 * @min_mtu: Interface Minimum MTU value
1834 * @max_mtu: Interface Maximum MTU value
1835 * @type: Interface hardware type
1836 * @hard_header_len: Maximum hardware header length.
1837 * @min_header_len: Minimum hardware header length
1839 * @needed_headroom: Extra headroom the hardware may need, but not in all
1840 * cases can this be guaranteed
1841 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1842 * cases can this be guaranteed. Some cases also use
1843 * LL_MAX_HEADER instead to allocate the skb
1845 * interface address info:
1847 * @perm_addr: Permanent hw address
1848 * @addr_assign_type: Hw address assignment type
1849 * @addr_len: Hardware address length
1850 * @upper_level: Maximum depth level of upper devices.
1851 * @lower_level: Maximum depth level of lower devices.
1852 * @neigh_priv_len: Used in neigh_alloc()
1853 * @dev_id: Used to differentiate devices that share
1854 * the same link layer address
1855 * @dev_port: Used to differentiate devices that share
1857 * @addr_list_lock: XXX: need comments on this one
1858 * @name_assign_type: network interface name assignment type
1859 * @uc_promisc: Counter that indicates promiscuous mode
1860 * has been enabled due to the need to listen to
1861 * additional unicast addresses in a device that
1862 * does not implement ndo_set_rx_mode()
1863 * @uc: unicast mac addresses
1864 * @mc: multicast mac addresses
1865 * @dev_addrs: list of device hw addresses
1866 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1867 * @promiscuity: Number of times the NIC is told to work in
1868 * promiscuous mode; if it becomes 0 the NIC will
1869 * exit promiscuous mode
1870 * @allmulti: Counter, enables or disables allmulticast mode
1872 * @vlan_info: VLAN info
1873 * @dsa_ptr: dsa specific data
1874 * @tipc_ptr: TIPC specific data
1875 * @atalk_ptr: AppleTalk link
1876 * @ip_ptr: IPv4 specific data
1877 * @ip6_ptr: IPv6 specific data
1878 * @ax25_ptr: AX.25 specific data
1879 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1880 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1882 * @mpls_ptr: mpls_dev struct pointer
1883 * @mctp_ptr: MCTP specific data
1885 * @dev_addr: Hw address (before bcast,
1886 * because most packets are unicast)
1888 * @_rx: Array of RX queues
1889 * @num_rx_queues: Number of RX queues
1890 * allocated at register_netdev() time
1891 * @real_num_rx_queues: Number of RX queues currently active in device
1892 * @xdp_prog: XDP sockets filter program pointer
1893 * @gro_flush_timeout: timeout for GRO layer in NAPI
1894 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1895 * allow to avoid NIC hard IRQ, on busy queues.
1897 * @rx_handler: handler for received packets
1898 * @rx_handler_data: XXX: need comments on this one
1899 * @miniq_ingress: ingress/clsact qdisc specific data for
1900 * ingress processing
1901 * @ingress_queue: XXX: need comments on this one
1902 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1903 * @broadcast: hw bcast address
1905 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1906 * indexed by RX queue number. Assigned by driver.
1907 * This must only be set if the ndo_rx_flow_steer
1908 * operation is defined
1909 * @index_hlist: Device index hash chain
1911 * @_tx: Array of TX queues
1912 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1913 * @real_num_tx_queues: Number of TX queues currently active in device
1914 * @qdisc: Root qdisc from userspace point of view
1915 * @tx_queue_len: Max frames per queue allowed
1916 * @tx_global_lock: XXX: need comments on this one
1917 * @xdp_bulkq: XDP device bulk queue
1918 * @xps_maps: all CPUs/RXQs maps for XPS device
1920 * @xps_maps: XXX: need comments on this one
1921 * @miniq_egress: clsact qdisc specific data for
1923 * @nf_hooks_egress: netfilter hooks executed for egress packets
1924 * @qdisc_hash: qdisc hash table
1925 * @watchdog_timeo: Represents the timeout that is used by
1926 * the watchdog (see dev_watchdog())
1927 * @watchdog_timer: List of timers
1929 * @proto_down_reason: reason a netdev interface is held down
1930 * @pcpu_refcnt: Number of references to this device
1931 * @dev_refcnt: Number of references to this device
1932 * @refcnt_tracker: Tracker directory for tracked references to this device
1933 * @todo_list: Delayed register/unregister
1934 * @link_watch_list: XXX: need comments on this one
1936 * @reg_state: Register/unregister state machine
1937 * @dismantle: Device is going to be freed
1938 * @rtnl_link_state: This enum represents the phases of creating
1941 * @needs_free_netdev: Should unregister perform free_netdev?
1942 * @priv_destructor: Called from unregister
1943 * @npinfo: XXX: need comments on this one
1944 * @nd_net: Network namespace this network device is inside
1946 * @ml_priv: Mid-layer private
1947 * @ml_priv_type: Mid-layer private type
1948 * @lstats: Loopback statistics
1949 * @tstats: Tunnel statistics
1950 * @dstats: Dummy statistics
1951 * @vstats: Virtual ethernet statistics
1956 * @dm_private: Drop monitor private
1958 * @dev: Class/net/name entry
1959 * @sysfs_groups: Space for optional device, statistics and wireless
1962 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1963 * @rtnl_link_ops: Rtnl_link_ops
1965 * @gso_max_size: Maximum size of generic segmentation offload
1966 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1967 * @gso_max_segs: Maximum number of segments that can be passed to the
1969 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1970 * @gso_ipv4_max_size: Maximum size of generic segmentation offload,
1973 * @dcbnl_ops: Data Center Bridging netlink ops
1974 * @num_tc: Number of traffic classes in the net device
1975 * @tc_to_txq: XXX: need comments on this one
1976 * @prio_tc_map: XXX: need comments on this one
1978 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1980 * @priomap: XXX: need comments on this one
1981 * @phydev: Physical device may attach itself
1982 * for hardware timestamping
1983 * @sfp_bus: attached &struct sfp_bus structure.
1985 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1987 * @proto_down: protocol port state information can be sent to the
1988 * switch driver and used to set the phys state of the
1991 * @wol_enabled: Wake-on-LAN is enabled
1993 * @threaded: napi threaded mode is enabled
1995 * @net_notifier_list: List of per-net netdev notifier block
1996 * that follow this device when it is moved
1997 * to another network namespace.
1999 * @macsec_ops: MACsec offloading ops
2001 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
2002 * offload capabilities of the device
2003 * @udp_tunnel_nic: UDP tunnel offload state
2004 * @xdp_state: stores info on attached XDP BPF programs
2006 * @nested_level: Used as a parameter of spin_lock_nested() of
2007 * dev->addr_list_lock.
2008 * @unlink_list: As netif_addr_lock() can be called recursively,
2009 * keep a list of interfaces to be deleted.
2010 * @gro_max_size: Maximum size of aggregated packet in generic
2011 * receive offload (GRO)
2012 * @gro_ipv4_max_size: Maximum size of aggregated packet in generic
2013 * receive offload (GRO), for IPv4.
2015 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
2016 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
2017 * @watchdog_dev_tracker: refcount tracker used by watchdog.
2018 * @dev_registered_tracker: tracker for reference held while
2020 * @offload_xstats_l3: L3 HW stats for this netdevice.
2022 * @devlink_port: Pointer to related devlink port structure.
2023 * Assigned by a driver before netdev registration using
2024 * SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2025 * during the time netdevice is registered.
2027 * FIXME: cleanup struct net_device such that network protocol info
2032 char name[IFNAMSIZ];
2033 struct netdev_name_node *name_node;
2034 struct dev_ifalias __rcu *ifalias;
2036 * I/O specific fields
2037 * FIXME: Merge these and struct ifmap into one
2039 unsigned long mem_end;
2040 unsigned long mem_start;
2041 unsigned long base_addr;
2044 * Some hardware also needs these fields (state,dev_list,
2045 * napi_list,unreg_list,close_list) but they are not
2046 * part of the usual set specified in Space.c.
2049 unsigned long state;
2051 struct list_head dev_list;
2052 struct list_head napi_list;
2053 struct list_head unreg_list;
2054 struct list_head close_list;
2055 struct list_head ptype_all;
2056 struct list_head ptype_specific;
2059 struct list_head upper;
2060 struct list_head lower;
2063 /* Read-mostly cache-line for fast-path access */
2065 xdp_features_t xdp_features;
2066 unsigned long long priv_flags;
2067 const struct net_device_ops *netdev_ops;
2068 const struct xdp_metadata_ops *xdp_metadata_ops;
2070 unsigned short gflags;
2071 unsigned short hard_header_len;
2073 /* Note : dev->mtu is often read without holding a lock.
2074 * Writers usually hold RTNL.
2075 * It is recommended to use READ_ONCE() to annotate the reads,
2076 * and to use WRITE_ONCE() to annotate the writes.
2079 unsigned short needed_headroom;
2080 unsigned short needed_tailroom;
2082 netdev_features_t features;
2083 netdev_features_t hw_features;
2084 netdev_features_t wanted_features;
2085 netdev_features_t vlan_features;
2086 netdev_features_t hw_enc_features;
2087 netdev_features_t mpls_features;
2088 netdev_features_t gso_partial_features;
2090 unsigned int min_mtu;
2091 unsigned int max_mtu;
2092 unsigned short type;
2093 unsigned char min_header_len;
2094 unsigned char name_assign_type;
2098 struct net_device_stats stats; /* not used by modern drivers */
2100 struct net_device_core_stats __percpu *core_stats;
2102 /* Stats to monitor link on/off, flapping */
2103 atomic_t carrier_up_count;
2104 atomic_t carrier_down_count;
2106 #ifdef CONFIG_WIRELESS_EXT
2107 const struct iw_handler_def *wireless_handlers;
2108 struct iw_public_data *wireless_data;
2110 const struct ethtool_ops *ethtool_ops;
2111 #ifdef CONFIG_NET_L3_MASTER_DEV
2112 const struct l3mdev_ops *l3mdev_ops;
2114 #if IS_ENABLED(CONFIG_IPV6)
2115 const struct ndisc_ops *ndisc_ops;
2118 #ifdef CONFIG_XFRM_OFFLOAD
2119 const struct xfrmdev_ops *xfrmdev_ops;
2122 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2123 const struct tlsdev_ops *tlsdev_ops;
2126 const struct header_ops *header_ops;
2128 unsigned char operstate;
2129 unsigned char link_mode;
2131 unsigned char if_port;
2134 /* Interface address info. */
2135 unsigned char perm_addr[MAX_ADDR_LEN];
2136 unsigned char addr_assign_type;
2137 unsigned char addr_len;
2138 unsigned char upper_level;
2139 unsigned char lower_level;
2141 unsigned short neigh_priv_len;
2142 unsigned short dev_id;
2143 unsigned short dev_port;
2144 unsigned short padded;
2146 spinlock_t addr_list_lock;
2149 struct netdev_hw_addr_list uc;
2150 struct netdev_hw_addr_list mc;
2151 struct netdev_hw_addr_list dev_addrs;
2154 struct kset *queues_kset;
2156 #ifdef CONFIG_LOCKDEP
2157 struct list_head unlink_list;
2159 unsigned int promiscuity;
2160 unsigned int allmulti;
2162 #ifdef CONFIG_LOCKDEP
2163 unsigned char nested_level;
2167 /* Protocol-specific pointers */
2169 struct in_device __rcu *ip_ptr;
2170 struct inet6_dev __rcu *ip6_ptr;
2171 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2172 struct vlan_info __rcu *vlan_info;
2174 #if IS_ENABLED(CONFIG_NET_DSA)
2175 struct dsa_port *dsa_ptr;
2177 #if IS_ENABLED(CONFIG_TIPC)
2178 struct tipc_bearer __rcu *tipc_ptr;
2180 #if IS_ENABLED(CONFIG_ATALK)
2183 #if IS_ENABLED(CONFIG_AX25)
2186 #if IS_ENABLED(CONFIG_CFG80211)
2187 struct wireless_dev *ieee80211_ptr;
2189 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2190 struct wpan_dev *ieee802154_ptr;
2192 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2193 struct mpls_dev __rcu *mpls_ptr;
2195 #if IS_ENABLED(CONFIG_MCTP)
2196 struct mctp_dev __rcu *mctp_ptr;
2200 * Cache lines mostly used on receive path (including eth_type_trans())
2202 /* Interface address info used in eth_type_trans() */
2203 const unsigned char *dev_addr;
2205 struct netdev_rx_queue *_rx;
2206 unsigned int num_rx_queues;
2207 unsigned int real_num_rx_queues;
2209 struct bpf_prog __rcu *xdp_prog;
2210 unsigned long gro_flush_timeout;
2211 int napi_defer_hard_irqs;
2212 #define GRO_LEGACY_MAX_SIZE 65536u
2213 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2214 * and shinfo->gso_segs is a 16bit field.
2216 #define GRO_MAX_SIZE (8 * 65535u)
2217 unsigned int gro_max_size;
2218 unsigned int gro_ipv4_max_size;
2219 rx_handler_func_t __rcu *rx_handler;
2220 void __rcu *rx_handler_data;
2222 #ifdef CONFIG_NET_CLS_ACT
2223 struct mini_Qdisc __rcu *miniq_ingress;
2225 struct netdev_queue __rcu *ingress_queue;
2226 #ifdef CONFIG_NETFILTER_INGRESS
2227 struct nf_hook_entries __rcu *nf_hooks_ingress;
2230 unsigned char broadcast[MAX_ADDR_LEN];
2231 #ifdef CONFIG_RFS_ACCEL
2232 struct cpu_rmap *rx_cpu_rmap;
2234 struct hlist_node index_hlist;
2237 * Cache lines mostly used on transmit path
2239 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2240 unsigned int num_tx_queues;
2241 unsigned int real_num_tx_queues;
2242 struct Qdisc __rcu *qdisc;
2243 unsigned int tx_queue_len;
2244 spinlock_t tx_global_lock;
2246 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2249 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2251 #ifdef CONFIG_NET_CLS_ACT
2252 struct mini_Qdisc __rcu *miniq_egress;
2254 #ifdef CONFIG_NETFILTER_EGRESS
2255 struct nf_hook_entries __rcu *nf_hooks_egress;
2258 #ifdef CONFIG_NET_SCHED
2259 DECLARE_HASHTABLE (qdisc_hash, 4);
2261 /* These may be needed for future network-power-down code. */
2262 struct timer_list watchdog_timer;
2265 u32 proto_down_reason;
2267 struct list_head todo_list;
2269 #ifdef CONFIG_PCPU_DEV_REFCNT
2270 int __percpu *pcpu_refcnt;
2272 refcount_t dev_refcnt;
2274 struct ref_tracker_dir refcnt_tracker;
2276 struct list_head link_watch_list;
2278 enum { NETREG_UNINITIALIZED=0,
2279 NETREG_REGISTERED, /* completed register_netdevice */
2280 NETREG_UNREGISTERING, /* called unregister_netdevice */
2281 NETREG_UNREGISTERED, /* completed unregister todo */
2282 NETREG_RELEASED, /* called free_netdev */
2283 NETREG_DUMMY, /* dummy device for NAPI poll */
2289 RTNL_LINK_INITIALIZED,
2290 RTNL_LINK_INITIALIZING,
2291 } rtnl_link_state:16;
2293 bool needs_free_netdev;
2294 void (*priv_destructor)(struct net_device *dev);
2296 #ifdef CONFIG_NETPOLL
2297 struct netpoll_info __rcu *npinfo;
2300 possible_net_t nd_net;
2302 /* mid-layer private */
2304 enum netdev_ml_priv_type ml_priv_type;
2307 struct pcpu_lstats __percpu *lstats;
2308 struct pcpu_sw_netstats __percpu *tstats;
2309 struct pcpu_dstats __percpu *dstats;
2312 #if IS_ENABLED(CONFIG_GARP)
2313 struct garp_port __rcu *garp_port;
2315 #if IS_ENABLED(CONFIG_MRP)
2316 struct mrp_port __rcu *mrp_port;
2318 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2319 struct dm_hw_stat_delta __rcu *dm_private;
2322 const struct attribute_group *sysfs_groups[4];
2323 const struct attribute_group *sysfs_rx_queue_group;
2325 const struct rtnl_link_ops *rtnl_link_ops;
2327 /* for setting kernel sock attribute on TCP connection setup */
2328 #define GSO_MAX_SEGS 65535u
2329 #define GSO_LEGACY_MAX_SIZE 65536u
2330 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2331 * and shinfo->gso_segs is a 16bit field.
2333 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2335 unsigned int gso_max_size;
2336 #define TSO_LEGACY_MAX_SIZE 65536
2337 #define TSO_MAX_SIZE UINT_MAX
2338 unsigned int tso_max_size;
2340 #define TSO_MAX_SEGS U16_MAX
2342 unsigned int gso_ipv4_max_size;
2345 const struct dcbnl_rtnl_ops *dcbnl_ops;
2348 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2349 u8 prio_tc_map[TC_BITMASK + 1];
2351 #if IS_ENABLED(CONFIG_FCOE)
2352 unsigned int fcoe_ddp_xid;
2354 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2355 struct netprio_map __rcu *priomap;
2357 struct phy_device *phydev;
2358 struct sfp_bus *sfp_bus;
2359 struct lock_class_key *qdisc_tx_busylock;
2361 unsigned wol_enabled:1;
2362 unsigned threaded:1;
2364 struct list_head net_notifier_list;
2366 #if IS_ENABLED(CONFIG_MACSEC)
2367 /* MACsec management functions */
2368 const struct macsec_ops *macsec_ops;
2370 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2371 struct udp_tunnel_nic *udp_tunnel_nic;
2373 /* protected by rtnl_lock */
2374 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2376 u8 dev_addr_shadow[MAX_ADDR_LEN];
2377 netdevice_tracker linkwatch_dev_tracker;
2378 netdevice_tracker watchdog_dev_tracker;
2379 netdevice_tracker dev_registered_tracker;
2380 struct rtnl_hw_stats64 *offload_xstats_l3;
2382 struct devlink_port *devlink_port;
2384 #define to_net_dev(d) container_of(d, struct net_device, dev)
2387 * Driver should use this to assign devlink port instance to a netdevice
2388 * before it registers the netdevice. Therefore devlink_port is static
2389 * during the netdev lifetime after it is registered.
2391 #define SET_NETDEV_DEVLINK_PORT(dev, port) \
2393 WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED); \
2394 ((dev)->devlink_port = (port)); \
2397 static inline bool netif_elide_gro(const struct net_device *dev)
2399 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2404 #define NETDEV_ALIGN 32
2407 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2409 return dev->prio_tc_map[prio & TC_BITMASK];
2413 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2415 if (tc >= dev->num_tc)
2418 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2422 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2423 void netdev_reset_tc(struct net_device *dev);
2424 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2425 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2428 int netdev_get_num_tc(struct net_device *dev)
2433 static inline void net_prefetch(void *p)
2436 #if L1_CACHE_BYTES < 128
2437 prefetch((u8 *)p + L1_CACHE_BYTES);
2441 static inline void net_prefetchw(void *p)
2444 #if L1_CACHE_BYTES < 128
2445 prefetchw((u8 *)p + L1_CACHE_BYTES);
2449 void netdev_unbind_sb_channel(struct net_device *dev,
2450 struct net_device *sb_dev);
2451 int netdev_bind_sb_channel_queue(struct net_device *dev,
2452 struct net_device *sb_dev,
2453 u8 tc, u16 count, u16 offset);
2454 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2455 static inline int netdev_get_sb_channel(struct net_device *dev)
2457 return max_t(int, -dev->num_tc, 0);
2461 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2464 return &dev->_tx[index];
2467 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2468 const struct sk_buff *skb)
2470 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2473 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2474 void (*f)(struct net_device *,
2475 struct netdev_queue *,
2481 for (i = 0; i < dev->num_tx_queues; i++)
2482 f(dev, &dev->_tx[i], arg);
2485 #define netdev_lockdep_set_classes(dev) \
2487 static struct lock_class_key qdisc_tx_busylock_key; \
2488 static struct lock_class_key qdisc_xmit_lock_key; \
2489 static struct lock_class_key dev_addr_list_lock_key; \
2492 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2493 lockdep_set_class(&(dev)->addr_list_lock, \
2494 &dev_addr_list_lock_key); \
2495 for (i = 0; i < (dev)->num_tx_queues; i++) \
2496 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2497 &qdisc_xmit_lock_key); \
2500 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2501 struct net_device *sb_dev);
2502 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2503 struct sk_buff *skb,
2504 struct net_device *sb_dev);
2506 /* returns the headroom that the master device needs to take in account
2507 * when forwarding to this dev
2509 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2511 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2514 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2516 if (dev->netdev_ops->ndo_set_rx_headroom)
2517 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2520 /* set the device rx headroom to the dev's default */
2521 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2523 netdev_set_rx_headroom(dev, -1);
2526 static inline void *netdev_get_ml_priv(struct net_device *dev,
2527 enum netdev_ml_priv_type type)
2529 if (dev->ml_priv_type != type)
2532 return dev->ml_priv;
2535 static inline void netdev_set_ml_priv(struct net_device *dev,
2537 enum netdev_ml_priv_type type)
2539 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2540 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2541 dev->ml_priv_type, type);
2542 WARN(!dev->ml_priv_type && dev->ml_priv,
2543 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2545 dev->ml_priv = ml_priv;
2546 dev->ml_priv_type = type;
2550 * Net namespace inlines
2553 struct net *dev_net(const struct net_device *dev)
2555 return read_pnet(&dev->nd_net);
2559 void dev_net_set(struct net_device *dev, struct net *net)
2561 write_pnet(&dev->nd_net, net);
2565 * netdev_priv - access network device private data
2566 * @dev: network device
2568 * Get network device private data
2570 static inline void *netdev_priv(const struct net_device *dev)
2572 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2575 /* Set the sysfs physical device reference for the network logical device
2576 * if set prior to registration will cause a symlink during initialization.
2578 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2580 /* Set the sysfs device type for the network logical device to allow
2581 * fine-grained identification of different network device types. For
2582 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2584 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2586 /* Default NAPI poll() weight
2587 * Device drivers are strongly advised to not use bigger value
2589 #define NAPI_POLL_WEIGHT 64
2591 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2592 int (*poll)(struct napi_struct *, int), int weight);
2595 * netif_napi_add() - initialize a NAPI context
2596 * @dev: network device
2597 * @napi: NAPI context
2598 * @poll: polling function
2600 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2601 * *any* of the other NAPI-related functions.
2604 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2605 int (*poll)(struct napi_struct *, int))
2607 netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2611 netif_napi_add_tx_weight(struct net_device *dev,
2612 struct napi_struct *napi,
2613 int (*poll)(struct napi_struct *, int),
2616 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2617 netif_napi_add_weight(dev, napi, poll, weight);
2621 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2622 * @dev: network device
2623 * @napi: NAPI context
2624 * @poll: polling function
2626 * This variant of netif_napi_add() should be used from drivers using NAPI
2627 * to exclusively poll a TX queue.
2628 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2630 static inline void netif_napi_add_tx(struct net_device *dev,
2631 struct napi_struct *napi,
2632 int (*poll)(struct napi_struct *, int))
2634 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2638 * __netif_napi_del - remove a NAPI context
2639 * @napi: NAPI context
2641 * Warning: caller must observe RCU grace period before freeing memory
2642 * containing @napi. Drivers might want to call this helper to combine
2643 * all the needed RCU grace periods into a single one.
2645 void __netif_napi_del(struct napi_struct *napi);
2648 * netif_napi_del - remove a NAPI context
2649 * @napi: NAPI context
2651 * netif_napi_del() removes a NAPI context from the network device NAPI list
2653 static inline void netif_napi_del(struct napi_struct *napi)
2655 __netif_napi_del(napi);
2659 struct packet_type {
2660 __be16 type; /* This is really htons(ether_type). */
2661 bool ignore_outgoing;
2662 struct net_device *dev; /* NULL is wildcarded here */
2663 netdevice_tracker dev_tracker;
2664 int (*func) (struct sk_buff *,
2665 struct net_device *,
2666 struct packet_type *,
2667 struct net_device *);
2668 void (*list_func) (struct list_head *,
2669 struct packet_type *,
2670 struct net_device *);
2671 bool (*id_match)(struct packet_type *ptype,
2673 struct net *af_packet_net;
2674 void *af_packet_priv;
2675 struct list_head list;
2678 struct offload_callbacks {
2679 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2680 netdev_features_t features);
2681 struct sk_buff *(*gro_receive)(struct list_head *head,
2682 struct sk_buff *skb);
2683 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2686 struct packet_offload {
2687 __be16 type; /* This is really htons(ether_type). */
2689 struct offload_callbacks callbacks;
2690 struct list_head list;
2693 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2694 struct pcpu_sw_netstats {
2695 u64_stats_t rx_packets;
2696 u64_stats_t rx_bytes;
2697 u64_stats_t tx_packets;
2698 u64_stats_t tx_bytes;
2699 struct u64_stats_sync syncp;
2700 } __aligned(4 * sizeof(u64));
2702 struct pcpu_lstats {
2703 u64_stats_t packets;
2705 struct u64_stats_sync syncp;
2706 } __aligned(2 * sizeof(u64));
2708 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2710 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2712 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2714 u64_stats_update_begin(&tstats->syncp);
2715 u64_stats_add(&tstats->rx_bytes, len);
2716 u64_stats_inc(&tstats->rx_packets);
2717 u64_stats_update_end(&tstats->syncp);
2720 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2721 unsigned int packets,
2724 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2726 u64_stats_update_begin(&tstats->syncp);
2727 u64_stats_add(&tstats->tx_bytes, len);
2728 u64_stats_add(&tstats->tx_packets, packets);
2729 u64_stats_update_end(&tstats->syncp);
2732 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2734 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2736 u64_stats_update_begin(&lstats->syncp);
2737 u64_stats_add(&lstats->bytes, len);
2738 u64_stats_inc(&lstats->packets);
2739 u64_stats_update_end(&lstats->syncp);
2742 #define __netdev_alloc_pcpu_stats(type, gfp) \
2744 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2747 for_each_possible_cpu(__cpu) { \
2748 typeof(type) *stat; \
2749 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2750 u64_stats_init(&stat->syncp); \
2756 #define netdev_alloc_pcpu_stats(type) \
2757 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2759 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2761 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2764 for_each_possible_cpu(__cpu) { \
2765 typeof(type) *stat; \
2766 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2767 u64_stats_init(&stat->syncp); \
2773 enum netdev_lag_tx_type {
2774 NETDEV_LAG_TX_TYPE_UNKNOWN,
2775 NETDEV_LAG_TX_TYPE_RANDOM,
2776 NETDEV_LAG_TX_TYPE_BROADCAST,
2777 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2778 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2779 NETDEV_LAG_TX_TYPE_HASH,
2782 enum netdev_lag_hash {
2783 NETDEV_LAG_HASH_NONE,
2785 NETDEV_LAG_HASH_L34,
2786 NETDEV_LAG_HASH_L23,
2787 NETDEV_LAG_HASH_E23,
2788 NETDEV_LAG_HASH_E34,
2789 NETDEV_LAG_HASH_VLAN_SRCMAC,
2790 NETDEV_LAG_HASH_UNKNOWN,
2793 struct netdev_lag_upper_info {
2794 enum netdev_lag_tx_type tx_type;
2795 enum netdev_lag_hash hash_type;
2798 struct netdev_lag_lower_state_info {
2803 #include <linux/notifier.h>
2805 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2806 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2810 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2812 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2813 detected a hardware crash and restarted
2814 - we can use this eg to kick tcp sessions
2816 NETDEV_CHANGE, /* Notify device state change */
2819 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2820 NETDEV_CHANGEADDR, /* notify after the address change */
2821 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2825 NETDEV_BONDING_FAILOVER,
2827 NETDEV_PRE_TYPE_CHANGE,
2828 NETDEV_POST_TYPE_CHANGE,
2832 NETDEV_NOTIFY_PEERS,
2836 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2837 NETDEV_CHANGEINFODATA,
2838 NETDEV_BONDING_INFO,
2839 NETDEV_PRECHANGEUPPER,
2840 NETDEV_CHANGELOWERSTATE,
2841 NETDEV_UDP_TUNNEL_PUSH_INFO,
2842 NETDEV_UDP_TUNNEL_DROP_INFO,
2843 NETDEV_CHANGE_TX_QUEUE_LEN,
2844 NETDEV_CVLAN_FILTER_PUSH_INFO,
2845 NETDEV_CVLAN_FILTER_DROP_INFO,
2846 NETDEV_SVLAN_FILTER_PUSH_INFO,
2847 NETDEV_SVLAN_FILTER_DROP_INFO,
2848 NETDEV_OFFLOAD_XSTATS_ENABLE,
2849 NETDEV_OFFLOAD_XSTATS_DISABLE,
2850 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2851 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2852 NETDEV_XDP_FEAT_CHANGE,
2854 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2856 int register_netdevice_notifier(struct notifier_block *nb);
2857 int unregister_netdevice_notifier(struct notifier_block *nb);
2858 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2859 int unregister_netdevice_notifier_net(struct net *net,
2860 struct notifier_block *nb);
2861 int register_netdevice_notifier_dev_net(struct net_device *dev,
2862 struct notifier_block *nb,
2863 struct netdev_net_notifier *nn);
2864 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2865 struct notifier_block *nb,
2866 struct netdev_net_notifier *nn);
2868 struct netdev_notifier_info {
2869 struct net_device *dev;
2870 struct netlink_ext_ack *extack;
2873 struct netdev_notifier_info_ext {
2874 struct netdev_notifier_info info; /* must be first */
2880 struct netdev_notifier_change_info {
2881 struct netdev_notifier_info info; /* must be first */
2882 unsigned int flags_changed;
2885 struct netdev_notifier_changeupper_info {
2886 struct netdev_notifier_info info; /* must be first */
2887 struct net_device *upper_dev; /* new upper dev */
2888 bool master; /* is upper dev master */
2889 bool linking; /* is the notification for link or unlink */
2890 void *upper_info; /* upper dev info */
2893 struct netdev_notifier_changelowerstate_info {
2894 struct netdev_notifier_info info; /* must be first */
2895 void *lower_state_info; /* is lower dev state */
2898 struct netdev_notifier_pre_changeaddr_info {
2899 struct netdev_notifier_info info; /* must be first */
2900 const unsigned char *dev_addr;
2903 enum netdev_offload_xstats_type {
2904 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2907 struct netdev_notifier_offload_xstats_info {
2908 struct netdev_notifier_info info; /* must be first */
2909 enum netdev_offload_xstats_type type;
2912 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2913 struct netdev_notifier_offload_xstats_rd *report_delta;
2914 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2915 struct netdev_notifier_offload_xstats_ru *report_used;
2919 int netdev_offload_xstats_enable(struct net_device *dev,
2920 enum netdev_offload_xstats_type type,
2921 struct netlink_ext_ack *extack);
2922 int netdev_offload_xstats_disable(struct net_device *dev,
2923 enum netdev_offload_xstats_type type);
2924 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2925 enum netdev_offload_xstats_type type);
2926 int netdev_offload_xstats_get(struct net_device *dev,
2927 enum netdev_offload_xstats_type type,
2928 struct rtnl_hw_stats64 *stats, bool *used,
2929 struct netlink_ext_ack *extack);
2931 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2932 const struct rtnl_hw_stats64 *stats);
2934 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2935 void netdev_offload_xstats_push_delta(struct net_device *dev,
2936 enum netdev_offload_xstats_type type,
2937 const struct rtnl_hw_stats64 *stats);
2939 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2940 struct net_device *dev)
2943 info->extack = NULL;
2946 static inline struct net_device *
2947 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2952 static inline struct netlink_ext_ack *
2953 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2955 return info->extack;
2958 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2961 extern rwlock_t dev_base_lock; /* Device list lock */
2963 #define for_each_netdev(net, d) \
2964 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2965 #define for_each_netdev_reverse(net, d) \
2966 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2967 #define for_each_netdev_rcu(net, d) \
2968 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2969 #define for_each_netdev_safe(net, d, n) \
2970 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2971 #define for_each_netdev_continue(net, d) \
2972 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2973 #define for_each_netdev_continue_reverse(net, d) \
2974 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2976 #define for_each_netdev_continue_rcu(net, d) \
2977 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2978 #define for_each_netdev_in_bond_rcu(bond, slave) \
2979 for_each_netdev_rcu(&init_net, slave) \
2980 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2981 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2983 static inline struct net_device *next_net_device(struct net_device *dev)
2985 struct list_head *lh;
2989 lh = dev->dev_list.next;
2990 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2993 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2995 struct list_head *lh;
2999 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3000 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3003 static inline struct net_device *first_net_device(struct net *net)
3005 return list_empty(&net->dev_base_head) ? NULL :
3006 net_device_entry(net->dev_base_head.next);
3009 static inline struct net_device *first_net_device_rcu(struct net *net)
3011 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3013 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3016 int netdev_boot_setup_check(struct net_device *dev);
3017 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3018 const char *hwaddr);
3019 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3020 void dev_add_pack(struct packet_type *pt);
3021 void dev_remove_pack(struct packet_type *pt);
3022 void __dev_remove_pack(struct packet_type *pt);
3023 void dev_add_offload(struct packet_offload *po);
3024 void dev_remove_offload(struct packet_offload *po);
3026 int dev_get_iflink(const struct net_device *dev);
3027 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3028 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3029 struct net_device_path_stack *stack);
3030 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3031 unsigned short mask);
3032 struct net_device *dev_get_by_name(struct net *net, const char *name);
3033 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3034 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3035 bool netdev_name_in_use(struct net *net, const char *name);
3036 int dev_alloc_name(struct net_device *dev, const char *name);
3037 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3038 void dev_close(struct net_device *dev);
3039 void dev_close_many(struct list_head *head, bool unlink);
3040 void dev_disable_lro(struct net_device *dev);
3041 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3042 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3043 struct net_device *sb_dev);
3044 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3045 struct net_device *sb_dev);
3047 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3048 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3050 static inline int dev_queue_xmit(struct sk_buff *skb)
3052 return __dev_queue_xmit(skb, NULL);
3055 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3056 struct net_device *sb_dev)
3058 return __dev_queue_xmit(skb, sb_dev);
3061 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3065 ret = __dev_direct_xmit(skb, queue_id);
3066 if (!dev_xmit_complete(ret))
3071 int register_netdevice(struct net_device *dev);
3072 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3073 void unregister_netdevice_many(struct list_head *head);
3074 static inline void unregister_netdevice(struct net_device *dev)
3076 unregister_netdevice_queue(dev, NULL);
3079 int netdev_refcnt_read(const struct net_device *dev);
3080 void free_netdev(struct net_device *dev);
3081 void netdev_freemem(struct net_device *dev);
3082 int init_dummy_netdev(struct net_device *dev);
3084 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3085 struct sk_buff *skb,
3087 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3089 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3090 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3091 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3092 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3093 int dev_restart(struct net_device *dev);
3096 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3097 unsigned short type,
3098 const void *daddr, const void *saddr,
3101 if (!dev->header_ops || !dev->header_ops->create)
3104 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3107 static inline int dev_parse_header(const struct sk_buff *skb,
3108 unsigned char *haddr)
3110 const struct net_device *dev = skb->dev;
3112 if (!dev->header_ops || !dev->header_ops->parse)
3114 return dev->header_ops->parse(skb, haddr);
3117 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3119 const struct net_device *dev = skb->dev;
3121 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3123 return dev->header_ops->parse_protocol(skb);
3126 /* ll_header must have at least hard_header_len allocated */
3127 static inline bool dev_validate_header(const struct net_device *dev,
3128 char *ll_header, int len)
3130 if (likely(len >= dev->hard_header_len))
3132 if (len < dev->min_header_len)
3135 if (capable(CAP_SYS_RAWIO)) {
3136 memset(ll_header + len, 0, dev->hard_header_len - len);
3140 if (dev->header_ops && dev->header_ops->validate)
3141 return dev->header_ops->validate(ll_header, len);
3146 static inline bool dev_has_header(const struct net_device *dev)
3148 return dev->header_ops && dev->header_ops->create;
3152 * Incoming packets are placed on per-CPU queues
3154 struct softnet_data {
3155 struct list_head poll_list;
3156 struct sk_buff_head process_queue;
3159 unsigned int processed;
3160 unsigned int time_squeeze;
3162 struct softnet_data *rps_ipi_list;
3164 #ifdef CONFIG_NET_FLOW_LIMIT
3165 struct sd_flow_limit __rcu *flow_limit;
3167 struct Qdisc *output_queue;
3168 struct Qdisc **output_queue_tailp;
3169 struct sk_buff *completion_queue;
3170 #ifdef CONFIG_XFRM_OFFLOAD
3171 struct sk_buff_head xfrm_backlog;
3173 /* written and read only by owning cpu: */
3177 #ifdef CONFIG_NET_EGRESS
3182 /* input_queue_head should be written by cpu owning this struct,
3183 * and only read by other cpus. Worth using a cache line.
3185 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3187 /* Elements below can be accessed between CPUs for RPS/RFS */
3188 call_single_data_t csd ____cacheline_aligned_in_smp;
3189 struct softnet_data *rps_ipi_next;
3191 unsigned int input_queue_tail;
3193 unsigned int received_rps;
3194 unsigned int dropped;
3195 struct sk_buff_head input_pkt_queue;
3196 struct napi_struct backlog;
3198 /* Another possibly contended cache line */
3199 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3201 int defer_ipi_scheduled;
3202 struct sk_buff *defer_list;
3203 call_single_data_t defer_csd;
3206 static inline void input_queue_head_incr(struct softnet_data *sd)
3209 sd->input_queue_head++;
3213 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3214 unsigned int *qtail)
3217 *qtail = ++sd->input_queue_tail;
3221 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3223 static inline int dev_recursion_level(void)
3225 return this_cpu_read(softnet_data.xmit.recursion);
3228 #define XMIT_RECURSION_LIMIT 8
3229 static inline bool dev_xmit_recursion(void)
3231 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3232 XMIT_RECURSION_LIMIT);
3235 static inline void dev_xmit_recursion_inc(void)
3237 __this_cpu_inc(softnet_data.xmit.recursion);
3240 static inline void dev_xmit_recursion_dec(void)
3242 __this_cpu_dec(softnet_data.xmit.recursion);
3245 void __netif_schedule(struct Qdisc *q);
3246 void netif_schedule_queue(struct netdev_queue *txq);
3248 static inline void netif_tx_schedule_all(struct net_device *dev)
3252 for (i = 0; i < dev->num_tx_queues; i++)
3253 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3256 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3258 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3262 * netif_start_queue - allow transmit
3263 * @dev: network device
3265 * Allow upper layers to call the device hard_start_xmit routine.
3267 static inline void netif_start_queue(struct net_device *dev)
3269 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3272 static inline void netif_tx_start_all_queues(struct net_device *dev)
3276 for (i = 0; i < dev->num_tx_queues; i++) {
3277 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3278 netif_tx_start_queue(txq);
3282 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3285 * netif_wake_queue - restart transmit
3286 * @dev: network device
3288 * Allow upper layers to call the device hard_start_xmit routine.
3289 * Used for flow control when transmit resources are available.
3291 static inline void netif_wake_queue(struct net_device *dev)
3293 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3296 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3300 for (i = 0; i < dev->num_tx_queues; i++) {
3301 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3302 netif_tx_wake_queue(txq);
3306 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3308 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3312 * netif_stop_queue - stop transmitted packets
3313 * @dev: network device
3315 * Stop upper layers calling the device hard_start_xmit routine.
3316 * Used for flow control when transmit resources are unavailable.
3318 static inline void netif_stop_queue(struct net_device *dev)
3320 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3323 void netif_tx_stop_all_queues(struct net_device *dev);
3325 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3327 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3331 * netif_queue_stopped - test if transmit queue is flowblocked
3332 * @dev: network device
3334 * Test if transmit queue on device is currently unable to send.
3336 static inline bool netif_queue_stopped(const struct net_device *dev)
3338 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3341 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3343 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3347 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3349 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3353 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3355 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3359 * netdev_queue_set_dql_min_limit - set dql minimum limit
3360 * @dev_queue: pointer to transmit queue
3361 * @min_limit: dql minimum limit
3363 * Forces xmit_more() to return true until the minimum threshold
3364 * defined by @min_limit is reached (or until the tx queue is
3365 * empty). Warning: to be use with care, misuse will impact the
3368 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3369 unsigned int min_limit)
3372 dev_queue->dql.min_limit = min_limit;
3377 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3378 * @dev_queue: pointer to transmit queue
3380 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3381 * to give appropriate hint to the CPU.
3383 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3386 prefetchw(&dev_queue->dql.num_queued);
3391 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3392 * @dev_queue: pointer to transmit queue
3394 * BQL enabled drivers might use this helper in their TX completion path,
3395 * to give appropriate hint to the CPU.
3397 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3400 prefetchw(&dev_queue->dql.limit);
3405 * netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3406 * @dev_queue: network device queue
3407 * @bytes: number of bytes queued to the device queue
3409 * Report the number of bytes queued for sending/completion to the network
3410 * device hardware queue. @bytes should be a good approximation and should
3411 * exactly match netdev_completed_queue() @bytes.
3412 * This is typically called once per packet, from ndo_start_xmit().
3414 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3418 dql_queued(&dev_queue->dql, bytes);
3420 if (likely(dql_avail(&dev_queue->dql) >= 0))
3423 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3426 * The XOFF flag must be set before checking the dql_avail below,
3427 * because in netdev_tx_completed_queue we update the dql_completed
3428 * before checking the XOFF flag.
3432 /* check again in case another CPU has just made room avail */
3433 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3434 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3438 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3439 * that they should not test BQL status themselves.
3440 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3442 * Returns true if the doorbell must be used to kick the NIC.
3444 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3450 dql_queued(&dev_queue->dql, bytes);
3452 return netif_tx_queue_stopped(dev_queue);
3454 netdev_tx_sent_queue(dev_queue, bytes);
3459 * netdev_sent_queue - report the number of bytes queued to hardware
3460 * @dev: network device
3461 * @bytes: number of bytes queued to the hardware device queue
3463 * Report the number of bytes queued for sending/completion to the network
3464 * device hardware queue#0. @bytes should be a good approximation and should
3465 * exactly match netdev_completed_queue() @bytes.
3466 * This is typically called once per packet, from ndo_start_xmit().
3468 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3470 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3473 static inline bool __netdev_sent_queue(struct net_device *dev,
3477 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3482 * netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3483 * @dev_queue: network device queue
3484 * @pkts: number of packets (currently ignored)
3485 * @bytes: number of bytes dequeued from the device queue
3487 * Must be called at most once per TX completion round (and not per
3488 * individual packet), so that BQL can adjust its limits appropriately.
3490 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3491 unsigned int pkts, unsigned int bytes)
3494 if (unlikely(!bytes))
3497 dql_completed(&dev_queue->dql, bytes);
3500 * Without the memory barrier there is a small possiblity that
3501 * netdev_tx_sent_queue will miss the update and cause the queue to
3502 * be stopped forever
3506 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3509 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3510 netif_schedule_queue(dev_queue);
3515 * netdev_completed_queue - report bytes and packets completed by device
3516 * @dev: network device
3517 * @pkts: actual number of packets sent over the medium
3518 * @bytes: actual number of bytes sent over the medium
3520 * Report the number of bytes and packets transmitted by the network device
3521 * hardware queue over the physical medium, @bytes must exactly match the
3522 * @bytes amount passed to netdev_sent_queue()
3524 static inline void netdev_completed_queue(struct net_device *dev,
3525 unsigned int pkts, unsigned int bytes)
3527 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3530 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3533 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3539 * netdev_reset_queue - reset the packets and bytes count of a network device
3540 * @dev_queue: network device
3542 * Reset the bytes and packet count of a network device and clear the
3543 * software flow control OFF bit for this network device
3545 static inline void netdev_reset_queue(struct net_device *dev_queue)
3547 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3551 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3552 * @dev: network device
3553 * @queue_index: given tx queue index
3555 * Returns 0 if given tx queue index >= number of device tx queues,
3556 * otherwise returns the originally passed tx queue index.
3558 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3560 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3561 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3562 dev->name, queue_index,
3563 dev->real_num_tx_queues);
3571 * netif_running - test if up
3572 * @dev: network device
3574 * Test if the device has been brought up.
3576 static inline bool netif_running(const struct net_device *dev)
3578 return test_bit(__LINK_STATE_START, &dev->state);
3582 * Routines to manage the subqueues on a device. We only need start,
3583 * stop, and a check if it's stopped. All other device management is
3584 * done at the overall netdevice level.
3585 * Also test the device if we're multiqueue.
3589 * netif_start_subqueue - allow sending packets on subqueue
3590 * @dev: network device
3591 * @queue_index: sub queue index
3593 * Start individual transmit queue of a device with multiple transmit queues.
3595 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3597 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3599 netif_tx_start_queue(txq);
3603 * netif_stop_subqueue - stop sending packets on subqueue
3604 * @dev: network device
3605 * @queue_index: sub queue index
3607 * Stop individual transmit queue of a device with multiple transmit queues.
3609 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3611 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3612 netif_tx_stop_queue(txq);
3616 * __netif_subqueue_stopped - test status of subqueue
3617 * @dev: network device
3618 * @queue_index: sub queue index
3620 * Check individual transmit queue of a device with multiple transmit queues.
3622 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3625 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3627 return netif_tx_queue_stopped(txq);
3631 * netif_subqueue_stopped - test status of subqueue
3632 * @dev: network device
3633 * @skb: sub queue buffer pointer
3635 * Check individual transmit queue of a device with multiple transmit queues.
3637 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3638 struct sk_buff *skb)
3640 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3644 * netif_wake_subqueue - allow sending packets on subqueue
3645 * @dev: network device
3646 * @queue_index: sub queue index
3648 * Resume individual transmit queue of a device with multiple transmit queues.
3650 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3652 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3654 netif_tx_wake_queue(txq);
3658 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3660 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3661 u16 index, enum xps_map_type type);
3664 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3665 * @j: CPU/Rx queue index
3666 * @mask: bitmask of all cpus/rx queues
3667 * @nr_bits: number of bits in the bitmask
3669 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3671 static inline bool netif_attr_test_mask(unsigned long j,
3672 const unsigned long *mask,
3673 unsigned int nr_bits)
3675 cpu_max_bits_warn(j, nr_bits);
3676 return test_bit(j, mask);
3680 * netif_attr_test_online - Test for online CPU/Rx queue
3681 * @j: CPU/Rx queue index
3682 * @online_mask: bitmask for CPUs/Rx queues that are online
3683 * @nr_bits: number of bits in the bitmask
3685 * Returns true if a CPU/Rx queue is online.
3687 static inline bool netif_attr_test_online(unsigned long j,
3688 const unsigned long *online_mask,
3689 unsigned int nr_bits)
3691 cpu_max_bits_warn(j, nr_bits);
3694 return test_bit(j, online_mask);
3696 return (j < nr_bits);
3700 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3701 * @n: CPU/Rx queue index
3702 * @srcp: the cpumask/Rx queue mask pointer
3703 * @nr_bits: number of bits in the bitmask
3705 * Returns >= nr_bits if no further CPUs/Rx queues set.
3707 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3708 unsigned int nr_bits)
3710 /* -1 is a legal arg here. */
3712 cpu_max_bits_warn(n, nr_bits);
3715 return find_next_bit(srcp, nr_bits, n + 1);
3721 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3722 * @n: CPU/Rx queue index
3723 * @src1p: the first CPUs/Rx queues mask pointer
3724 * @src2p: the second CPUs/Rx queues mask pointer
3725 * @nr_bits: number of bits in the bitmask
3727 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3729 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3730 const unsigned long *src2p,
3731 unsigned int nr_bits)
3733 /* -1 is a legal arg here. */
3735 cpu_max_bits_warn(n, nr_bits);
3738 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3740 return find_next_bit(src1p, nr_bits, n + 1);
3742 return find_next_bit(src2p, nr_bits, n + 1);
3747 static inline int netif_set_xps_queue(struct net_device *dev,
3748 const struct cpumask *mask,
3754 static inline int __netif_set_xps_queue(struct net_device *dev,
3755 const unsigned long *mask,
3756 u16 index, enum xps_map_type type)
3763 * netif_is_multiqueue - test if device has multiple transmit queues
3764 * @dev: network device
3766 * Check if device has multiple transmit queues
3768 static inline bool netif_is_multiqueue(const struct net_device *dev)
3770 return dev->num_tx_queues > 1;
3773 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3776 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3778 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3781 dev->real_num_rx_queues = rxqs;
3785 int netif_set_real_num_queues(struct net_device *dev,
3786 unsigned int txq, unsigned int rxq);
3788 static inline struct netdev_rx_queue *
3789 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3791 return dev->_rx + rxq;
3795 static inline unsigned int get_netdev_rx_queue_index(
3796 struct netdev_rx_queue *queue)
3798 struct net_device *dev = queue->dev;
3799 int index = queue - dev->_rx;
3801 BUG_ON(index >= dev->num_rx_queues);
3806 int netif_get_num_default_rss_queues(void);
3808 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3809 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3812 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3813 * interrupt context or with hardware interrupts being disabled.
3814 * (in_hardirq() || irqs_disabled())
3816 * We provide four helpers that can be used in following contexts :
3818 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3819 * replacing kfree_skb(skb)
3821 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3822 * Typically used in place of consume_skb(skb) in TX completion path
3824 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3825 * replacing kfree_skb(skb)
3827 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3828 * and consumed a packet. Used in place of consume_skb(skb)
3830 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3832 dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3835 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3837 dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3840 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3842 dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3845 static inline void dev_consume_skb_any(struct sk_buff *skb)
3847 dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3850 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3851 struct bpf_prog *xdp_prog);
3852 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3853 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3854 int netif_rx(struct sk_buff *skb);
3855 int __netif_rx(struct sk_buff *skb);
3857 int netif_receive_skb(struct sk_buff *skb);
3858 int netif_receive_skb_core(struct sk_buff *skb);
3859 void netif_receive_skb_list_internal(struct list_head *head);
3860 void netif_receive_skb_list(struct list_head *head);
3861 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3862 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3863 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3864 void napi_get_frags_check(struct napi_struct *napi);
3865 gro_result_t napi_gro_frags(struct napi_struct *napi);
3866 struct packet_offload *gro_find_receive_by_type(__be16 type);
3867 struct packet_offload *gro_find_complete_by_type(__be16 type);
3869 static inline void napi_free_frags(struct napi_struct *napi)
3871 kfree_skb(napi->skb);
3875 bool netdev_is_rx_handler_busy(struct net_device *dev);
3876 int netdev_rx_handler_register(struct net_device *dev,
3877 rx_handler_func_t *rx_handler,
3878 void *rx_handler_data);
3879 void netdev_rx_handler_unregister(struct net_device *dev);
3881 bool dev_valid_name(const char *name);
3882 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3884 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3886 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3887 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3888 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3889 void __user *data, bool *need_copyout);
3890 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3891 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3892 unsigned int dev_get_flags(const struct net_device *);
3893 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3894 struct netlink_ext_ack *extack);
3895 int dev_change_flags(struct net_device *dev, unsigned int flags,
3896 struct netlink_ext_ack *extack);
3897 int dev_set_alias(struct net_device *, const char *, size_t);
3898 int dev_get_alias(const struct net_device *, char *, size_t);
3899 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3900 const char *pat, int new_ifindex);
3902 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3905 return __dev_change_net_namespace(dev, net, pat, 0);
3907 int __dev_set_mtu(struct net_device *, int);
3908 int dev_set_mtu(struct net_device *, int);
3909 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3910 struct netlink_ext_ack *extack);
3911 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3912 struct netlink_ext_ack *extack);
3913 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3914 struct netlink_ext_ack *extack);
3915 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3916 int dev_get_port_parent_id(struct net_device *dev,
3917 struct netdev_phys_item_id *ppid, bool recurse);
3918 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3919 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3920 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3921 struct netdev_queue *txq, int *ret);
3923 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3924 u8 dev_xdp_prog_count(struct net_device *dev);
3925 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3927 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3928 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3929 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3930 bool is_skb_forwardable(const struct net_device *dev,
3931 const struct sk_buff *skb);
3933 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3934 const struct sk_buff *skb,
3935 const bool check_mtu)
3937 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3940 if (!(dev->flags & IFF_UP))
3946 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3947 if (skb->len <= len)
3950 /* if TSO is enabled, we don't care about the length as the packet
3951 * could be forwarded without being segmented before
3953 if (skb_is_gso(skb))
3959 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3961 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3963 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3964 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3969 return netdev_core_stats_alloc(dev);
3972 #define DEV_CORE_STATS_INC(FIELD) \
3973 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3975 struct net_device_core_stats __percpu *p; \
3977 p = dev_core_stats(dev); \
3979 this_cpu_inc(p->FIELD); \
3981 DEV_CORE_STATS_INC(rx_dropped)
3982 DEV_CORE_STATS_INC(tx_dropped)
3983 DEV_CORE_STATS_INC(rx_nohandler)
3984 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3986 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3987 struct sk_buff *skb,
3988 const bool check_mtu)
3990 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3991 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3992 dev_core_stats_rx_dropped_inc(dev);
3997 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4002 bool dev_nit_active(struct net_device *dev);
4003 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4005 static inline void __dev_put(struct net_device *dev)
4008 #ifdef CONFIG_PCPU_DEV_REFCNT
4009 this_cpu_dec(*dev->pcpu_refcnt);
4011 refcount_dec(&dev->dev_refcnt);
4016 static inline void __dev_hold(struct net_device *dev)
4019 #ifdef CONFIG_PCPU_DEV_REFCNT
4020 this_cpu_inc(*dev->pcpu_refcnt);
4022 refcount_inc(&dev->dev_refcnt);
4027 static inline void __netdev_tracker_alloc(struct net_device *dev,
4028 netdevice_tracker *tracker,
4031 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4032 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4036 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4037 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4039 static inline void netdev_tracker_alloc(struct net_device *dev,
4040 netdevice_tracker *tracker, gfp_t gfp)
4042 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4043 refcount_dec(&dev->refcnt_tracker.no_tracker);
4044 __netdev_tracker_alloc(dev, tracker, gfp);
4048 static inline void netdev_tracker_free(struct net_device *dev,
4049 netdevice_tracker *tracker)
4051 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4052 ref_tracker_free(&dev->refcnt_tracker, tracker);
4056 static inline void netdev_hold(struct net_device *dev,
4057 netdevice_tracker *tracker, gfp_t gfp)
4061 __netdev_tracker_alloc(dev, tracker, gfp);
4065 static inline void netdev_put(struct net_device *dev,
4066 netdevice_tracker *tracker)
4069 netdev_tracker_free(dev, tracker);
4075 * dev_hold - get reference to device
4076 * @dev: network device
4078 * Hold reference to device to keep it from being freed.
4079 * Try using netdev_hold() instead.
4081 static inline void dev_hold(struct net_device *dev)
4083 netdev_hold(dev, NULL, GFP_ATOMIC);
4087 * dev_put - release reference to device
4088 * @dev: network device
4090 * Release reference to device to allow it to be freed.
4091 * Try using netdev_put() instead.
4093 static inline void dev_put(struct net_device *dev)
4095 netdev_put(dev, NULL);
4098 static inline void netdev_ref_replace(struct net_device *odev,
4099 struct net_device *ndev,
4100 netdevice_tracker *tracker,
4104 netdev_tracker_free(odev, tracker);
4110 __netdev_tracker_alloc(ndev, tracker, gfp);
4113 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4114 * and _off may be called from IRQ context, but it is caller
4115 * who is responsible for serialization of these calls.
4117 * The name carrier is inappropriate, these functions should really be
4118 * called netif_lowerlayer_*() because they represent the state of any
4119 * kind of lower layer not just hardware media.
4121 void linkwatch_fire_event(struct net_device *dev);
4124 * netif_carrier_ok - test if carrier present
4125 * @dev: network device
4127 * Check if carrier is present on device
4129 static inline bool netif_carrier_ok(const struct net_device *dev)
4131 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4134 unsigned long dev_trans_start(struct net_device *dev);
4136 void __netdev_watchdog_up(struct net_device *dev);
4138 void netif_carrier_on(struct net_device *dev);
4139 void netif_carrier_off(struct net_device *dev);
4140 void netif_carrier_event(struct net_device *dev);
4143 * netif_dormant_on - mark device as dormant.
4144 * @dev: network device
4146 * Mark device as dormant (as per RFC2863).
4148 * The dormant state indicates that the relevant interface is not
4149 * actually in a condition to pass packets (i.e., it is not 'up') but is
4150 * in a "pending" state, waiting for some external event. For "on-
4151 * demand" interfaces, this new state identifies the situation where the
4152 * interface is waiting for events to place it in the up state.
4154 static inline void netif_dormant_on(struct net_device *dev)
4156 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4157 linkwatch_fire_event(dev);
4161 * netif_dormant_off - set device as not dormant.
4162 * @dev: network device
4164 * Device is not in dormant state.
4166 static inline void netif_dormant_off(struct net_device *dev)
4168 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4169 linkwatch_fire_event(dev);
4173 * netif_dormant - test if device is dormant
4174 * @dev: network device
4176 * Check if device is dormant.
4178 static inline bool netif_dormant(const struct net_device *dev)
4180 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4185 * netif_testing_on - mark device as under test.
4186 * @dev: network device
4188 * Mark device as under test (as per RFC2863).
4190 * The testing state indicates that some test(s) must be performed on
4191 * the interface. After completion, of the test, the interface state
4192 * will change to up, dormant, or down, as appropriate.
4194 static inline void netif_testing_on(struct net_device *dev)
4196 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4197 linkwatch_fire_event(dev);
4201 * netif_testing_off - set device as not under test.
4202 * @dev: network device
4204 * Device is not in testing state.
4206 static inline void netif_testing_off(struct net_device *dev)
4208 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4209 linkwatch_fire_event(dev);
4213 * netif_testing - test if device is under test
4214 * @dev: network device
4216 * Check if device is under test
4218 static inline bool netif_testing(const struct net_device *dev)
4220 return test_bit(__LINK_STATE_TESTING, &dev->state);
4225 * netif_oper_up - test if device is operational
4226 * @dev: network device
4228 * Check if carrier is operational
4230 static inline bool netif_oper_up(const struct net_device *dev)
4232 return (dev->operstate == IF_OPER_UP ||
4233 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4237 * netif_device_present - is device available or removed
4238 * @dev: network device
4240 * Check if device has not been removed from system.
4242 static inline bool netif_device_present(const struct net_device *dev)
4244 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4247 void netif_device_detach(struct net_device *dev);
4249 void netif_device_attach(struct net_device *dev);
4252 * Network interface message level settings
4257 NETIF_MSG_PROBE_BIT,
4259 NETIF_MSG_TIMER_BIT,
4260 NETIF_MSG_IFDOWN_BIT,
4262 NETIF_MSG_RX_ERR_BIT,
4263 NETIF_MSG_TX_ERR_BIT,
4264 NETIF_MSG_TX_QUEUED_BIT,
4266 NETIF_MSG_TX_DONE_BIT,
4267 NETIF_MSG_RX_STATUS_BIT,
4268 NETIF_MSG_PKTDATA_BIT,
4272 /* When you add a new bit above, update netif_msg_class_names array
4273 * in net/ethtool/common.c
4275 NETIF_MSG_CLASS_COUNT,
4277 /* Both ethtool_ops interface and internal driver implementation use u32 */
4278 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4280 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4281 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4283 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4284 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4285 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4286 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4287 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4288 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4289 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4290 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4291 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4292 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4293 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4294 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4295 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4296 #define NETIF_MSG_HW __NETIF_MSG(HW)
4297 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4299 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4300 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4301 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4302 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4303 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4304 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4305 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4306 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4307 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4308 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4309 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4310 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4311 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4312 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4313 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4315 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4318 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4319 return default_msg_enable_bits;
4320 if (debug_value == 0) /* no output */
4322 /* set low N bits */
4323 return (1U << debug_value) - 1;
4326 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4328 spin_lock(&txq->_xmit_lock);
4329 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4330 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4333 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4335 __acquire(&txq->_xmit_lock);
4339 static inline void __netif_tx_release(struct netdev_queue *txq)
4341 __release(&txq->_xmit_lock);
4344 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4346 spin_lock_bh(&txq->_xmit_lock);
4347 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4348 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4351 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4353 bool ok = spin_trylock(&txq->_xmit_lock);
4356 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4357 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4362 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4364 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4365 WRITE_ONCE(txq->xmit_lock_owner, -1);
4366 spin_unlock(&txq->_xmit_lock);
4369 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4371 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4372 WRITE_ONCE(txq->xmit_lock_owner, -1);
4373 spin_unlock_bh(&txq->_xmit_lock);
4377 * txq->trans_start can be read locklessly from dev_watchdog()
4379 static inline void txq_trans_update(struct netdev_queue *txq)
4381 if (txq->xmit_lock_owner != -1)
4382 WRITE_ONCE(txq->trans_start, jiffies);
4385 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4387 unsigned long now = jiffies;
4389 if (READ_ONCE(txq->trans_start) != now)
4390 WRITE_ONCE(txq->trans_start, now);
4393 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4394 static inline void netif_trans_update(struct net_device *dev)
4396 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4398 txq_trans_cond_update(txq);
4402 * netif_tx_lock - grab network device transmit lock
4403 * @dev: network device
4405 * Get network device transmit lock
4407 void netif_tx_lock(struct net_device *dev);
4409 static inline void netif_tx_lock_bh(struct net_device *dev)
4415 void netif_tx_unlock(struct net_device *dev);
4417 static inline void netif_tx_unlock_bh(struct net_device *dev)
4419 netif_tx_unlock(dev);
4423 #define HARD_TX_LOCK(dev, txq, cpu) { \
4424 if ((dev->features & NETIF_F_LLTX) == 0) { \
4425 __netif_tx_lock(txq, cpu); \
4427 __netif_tx_acquire(txq); \
4431 #define HARD_TX_TRYLOCK(dev, txq) \
4432 (((dev->features & NETIF_F_LLTX) == 0) ? \
4433 __netif_tx_trylock(txq) : \
4434 __netif_tx_acquire(txq))
4436 #define HARD_TX_UNLOCK(dev, txq) { \
4437 if ((dev->features & NETIF_F_LLTX) == 0) { \
4438 __netif_tx_unlock(txq); \
4440 __netif_tx_release(txq); \
4444 static inline void netif_tx_disable(struct net_device *dev)
4450 cpu = smp_processor_id();
4451 spin_lock(&dev->tx_global_lock);
4452 for (i = 0; i < dev->num_tx_queues; i++) {
4453 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4455 __netif_tx_lock(txq, cpu);
4456 netif_tx_stop_queue(txq);
4457 __netif_tx_unlock(txq);
4459 spin_unlock(&dev->tx_global_lock);
4463 static inline void netif_addr_lock(struct net_device *dev)
4465 unsigned char nest_level = 0;
4467 #ifdef CONFIG_LOCKDEP
4468 nest_level = dev->nested_level;
4470 spin_lock_nested(&dev->addr_list_lock, nest_level);
4473 static inline void netif_addr_lock_bh(struct net_device *dev)
4475 unsigned char nest_level = 0;
4477 #ifdef CONFIG_LOCKDEP
4478 nest_level = dev->nested_level;
4481 spin_lock_nested(&dev->addr_list_lock, nest_level);
4484 static inline void netif_addr_unlock(struct net_device *dev)
4486 spin_unlock(&dev->addr_list_lock);
4489 static inline void netif_addr_unlock_bh(struct net_device *dev)
4491 spin_unlock_bh(&dev->addr_list_lock);
4495 * dev_addrs walker. Should be used only for read access. Call with
4496 * rcu_read_lock held.
4498 #define for_each_dev_addr(dev, ha) \
4499 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4501 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4503 void ether_setup(struct net_device *dev);
4505 /* Support for loadable net-drivers */
4506 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4507 unsigned char name_assign_type,
4508 void (*setup)(struct net_device *),
4509 unsigned int txqs, unsigned int rxqs);
4510 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4511 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4513 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4514 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4517 int register_netdev(struct net_device *dev);
4518 void unregister_netdev(struct net_device *dev);
4520 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4522 /* General hardware address lists handling functions */
4523 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4524 struct netdev_hw_addr_list *from_list, int addr_len);
4525 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4526 struct netdev_hw_addr_list *from_list, int addr_len);
4527 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4528 struct net_device *dev,
4529 int (*sync)(struct net_device *, const unsigned char *),
4530 int (*unsync)(struct net_device *,
4531 const unsigned char *));
4532 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4533 struct net_device *dev,
4534 int (*sync)(struct net_device *,
4535 const unsigned char *, int),
4536 int (*unsync)(struct net_device *,
4537 const unsigned char *, int));
4538 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4539 struct net_device *dev,
4540 int (*unsync)(struct net_device *,
4541 const unsigned char *, int));
4542 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4543 struct net_device *dev,
4544 int (*unsync)(struct net_device *,
4545 const unsigned char *));
4546 void __hw_addr_init(struct netdev_hw_addr_list *list);
4548 /* Functions used for device addresses handling */
4549 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4550 const void *addr, size_t len);
4553 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4555 dev_addr_mod(dev, 0, addr, len);
4558 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4560 __dev_addr_set(dev, addr, dev->addr_len);
4563 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4564 unsigned char addr_type);
4565 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4566 unsigned char addr_type);
4568 /* Functions used for unicast addresses handling */
4569 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4570 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4571 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4572 int dev_uc_sync(struct net_device *to, struct net_device *from);
4573 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4574 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4575 void dev_uc_flush(struct net_device *dev);
4576 void dev_uc_init(struct net_device *dev);
4579 * __dev_uc_sync - Synchonize device's unicast list
4580 * @dev: device to sync
4581 * @sync: function to call if address should be added
4582 * @unsync: function to call if address should be removed
4584 * Add newly added addresses to the interface, and release
4585 * addresses that have been deleted.
4587 static inline int __dev_uc_sync(struct net_device *dev,
4588 int (*sync)(struct net_device *,
4589 const unsigned char *),
4590 int (*unsync)(struct net_device *,
4591 const unsigned char *))
4593 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4597 * __dev_uc_unsync - Remove synchronized addresses from device
4598 * @dev: device to sync
4599 * @unsync: function to call if address should be removed
4601 * Remove all addresses that were added to the device by dev_uc_sync().
4603 static inline void __dev_uc_unsync(struct net_device *dev,
4604 int (*unsync)(struct net_device *,
4605 const unsigned char *))
4607 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4610 /* Functions used for multicast addresses handling */
4611 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4612 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4613 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4614 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4615 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4616 int dev_mc_sync(struct net_device *to, struct net_device *from);
4617 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4618 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4619 void dev_mc_flush(struct net_device *dev);
4620 void dev_mc_init(struct net_device *dev);
4623 * __dev_mc_sync - Synchonize device's multicast list
4624 * @dev: device to sync
4625 * @sync: function to call if address should be added
4626 * @unsync: function to call if address should be removed
4628 * Add newly added addresses to the interface, and release
4629 * addresses that have been deleted.
4631 static inline int __dev_mc_sync(struct net_device *dev,
4632 int (*sync)(struct net_device *,
4633 const unsigned char *),
4634 int (*unsync)(struct net_device *,
4635 const unsigned char *))
4637 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4641 * __dev_mc_unsync - Remove synchronized addresses from device
4642 * @dev: device to sync
4643 * @unsync: function to call if address should be removed
4645 * Remove all addresses that were added to the device by dev_mc_sync().
4647 static inline void __dev_mc_unsync(struct net_device *dev,
4648 int (*unsync)(struct net_device *,
4649 const unsigned char *))
4651 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4654 /* Functions used for secondary unicast and multicast support */
4655 void dev_set_rx_mode(struct net_device *dev);
4656 int dev_set_promiscuity(struct net_device *dev, int inc);
4657 int dev_set_allmulti(struct net_device *dev, int inc);
4658 void netdev_state_change(struct net_device *dev);
4659 void __netdev_notify_peers(struct net_device *dev);
4660 void netdev_notify_peers(struct net_device *dev);
4661 void netdev_features_change(struct net_device *dev);
4662 /* Load a device via the kmod */
4663 void dev_load(struct net *net, const char *name);
4664 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4665 struct rtnl_link_stats64 *storage);
4666 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4667 const struct net_device_stats *netdev_stats);
4668 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4669 const struct pcpu_sw_netstats __percpu *netstats);
4670 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4672 extern int netdev_max_backlog;
4673 extern int dev_rx_weight;
4674 extern int dev_tx_weight;
4675 extern int gro_normal_batch;
4678 NESTED_SYNC_IMM_BIT,
4679 NESTED_SYNC_TODO_BIT,
4682 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4683 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4685 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4686 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4688 struct netdev_nested_priv {
4689 unsigned char flags;
4693 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4694 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4695 struct list_head **iter);
4697 /* iterate through upper list, must be called under RCU read lock */
4698 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4699 for (iter = &(dev)->adj_list.upper, \
4700 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4702 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4704 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4705 int (*fn)(struct net_device *upper_dev,
4706 struct netdev_nested_priv *priv),
4707 struct netdev_nested_priv *priv);
4709 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4710 struct net_device *upper_dev);
4712 bool netdev_has_any_upper_dev(struct net_device *dev);
4714 void *netdev_lower_get_next_private(struct net_device *dev,
4715 struct list_head **iter);
4716 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4717 struct list_head **iter);
4719 #define netdev_for_each_lower_private(dev, priv, iter) \
4720 for (iter = (dev)->adj_list.lower.next, \
4721 priv = netdev_lower_get_next_private(dev, &(iter)); \
4723 priv = netdev_lower_get_next_private(dev, &(iter)))
4725 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4726 for (iter = &(dev)->adj_list.lower, \
4727 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4729 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4731 void *netdev_lower_get_next(struct net_device *dev,
4732 struct list_head **iter);
4734 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4735 for (iter = (dev)->adj_list.lower.next, \
4736 ldev = netdev_lower_get_next(dev, &(iter)); \
4738 ldev = netdev_lower_get_next(dev, &(iter)))
4740 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4741 struct list_head **iter);
4742 int netdev_walk_all_lower_dev(struct net_device *dev,
4743 int (*fn)(struct net_device *lower_dev,
4744 struct netdev_nested_priv *priv),
4745 struct netdev_nested_priv *priv);
4746 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4747 int (*fn)(struct net_device *lower_dev,
4748 struct netdev_nested_priv *priv),
4749 struct netdev_nested_priv *priv);
4751 void *netdev_adjacent_get_private(struct list_head *adj_list);
4752 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4753 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4754 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4755 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4756 struct netlink_ext_ack *extack);
4757 int netdev_master_upper_dev_link(struct net_device *dev,
4758 struct net_device *upper_dev,
4759 void *upper_priv, void *upper_info,
4760 struct netlink_ext_ack *extack);
4761 void netdev_upper_dev_unlink(struct net_device *dev,
4762 struct net_device *upper_dev);
4763 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4764 struct net_device *new_dev,
4765 struct net_device *dev,
4766 struct netlink_ext_ack *extack);
4767 void netdev_adjacent_change_commit(struct net_device *old_dev,
4768 struct net_device *new_dev,
4769 struct net_device *dev);
4770 void netdev_adjacent_change_abort(struct net_device *old_dev,
4771 struct net_device *new_dev,
4772 struct net_device *dev);
4773 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4774 void *netdev_lower_dev_get_private(struct net_device *dev,
4775 struct net_device *lower_dev);
4776 void netdev_lower_state_changed(struct net_device *lower_dev,
4777 void *lower_state_info);
4779 /* RSS keys are 40 or 52 bytes long */
4780 #define NETDEV_RSS_KEY_LEN 52
4781 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4782 void netdev_rss_key_fill(void *buffer, size_t len);
4784 int skb_checksum_help(struct sk_buff *skb);
4785 int skb_crc32c_csum_help(struct sk_buff *skb);
4786 int skb_csum_hwoffload_help(struct sk_buff *skb,
4787 const netdev_features_t features);
4789 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4790 netdev_features_t features, bool tx_path);
4791 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4792 netdev_features_t features, __be16 type);
4793 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4794 netdev_features_t features);
4796 struct netdev_bonding_info {
4801 struct netdev_notifier_bonding_info {
4802 struct netdev_notifier_info info; /* must be first */
4803 struct netdev_bonding_info bonding_info;
4806 void netdev_bonding_info_change(struct net_device *dev,
4807 struct netdev_bonding_info *bonding_info);
4809 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4810 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4812 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4819 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4821 return __skb_gso_segment(skb, features, true);
4823 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4825 static inline bool can_checksum_protocol(netdev_features_t features,
4828 if (protocol == htons(ETH_P_FCOE))
4829 return !!(features & NETIF_F_FCOE_CRC);
4831 /* Assume this is an IP checksum (not SCTP CRC) */
4833 if (features & NETIF_F_HW_CSUM) {
4834 /* Can checksum everything */
4839 case htons(ETH_P_IP):
4840 return !!(features & NETIF_F_IP_CSUM);
4841 case htons(ETH_P_IPV6):
4842 return !!(features & NETIF_F_IPV6_CSUM);
4849 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4851 static inline void netdev_rx_csum_fault(struct net_device *dev,
4852 struct sk_buff *skb)
4856 /* rx skb timestamps */
4857 void net_enable_timestamp(void);
4858 void net_disable_timestamp(void);
4860 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4861 const struct skb_shared_hwtstamps *hwtstamps,
4864 const struct net_device_ops *ops = dev->netdev_ops;
4866 if (ops->ndo_get_tstamp)
4867 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4869 return hwtstamps->hwtstamp;
4872 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4873 struct sk_buff *skb, struct net_device *dev,
4876 __this_cpu_write(softnet_data.xmit.more, more);
4877 return ops->ndo_start_xmit(skb, dev);
4880 static inline bool netdev_xmit_more(void)
4882 return __this_cpu_read(softnet_data.xmit.more);
4885 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4886 struct netdev_queue *txq, bool more)
4888 const struct net_device_ops *ops = dev->netdev_ops;
4891 rc = __netdev_start_xmit(ops, skb, dev, more);
4892 if (rc == NETDEV_TX_OK)
4893 txq_trans_update(txq);
4898 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4900 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4903 extern const struct kobj_ns_type_operations net_ns_type_operations;
4905 const char *netdev_drivername(const struct net_device *dev);
4907 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4908 netdev_features_t f2)
4910 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4911 if (f1 & NETIF_F_HW_CSUM)
4912 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4914 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4920 static inline netdev_features_t netdev_get_wanted_features(
4921 struct net_device *dev)
4923 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4925 netdev_features_t netdev_increment_features(netdev_features_t all,
4926 netdev_features_t one, netdev_features_t mask);
4928 /* Allow TSO being used on stacked device :
4929 * Performing the GSO segmentation before last device
4930 * is a performance improvement.
4932 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4933 netdev_features_t mask)
4935 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4938 int __netdev_update_features(struct net_device *dev);
4939 void netdev_update_features(struct net_device *dev);
4940 void netdev_change_features(struct net_device *dev);
4942 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4943 struct net_device *dev);
4945 netdev_features_t passthru_features_check(struct sk_buff *skb,
4946 struct net_device *dev,
4947 netdev_features_t features);
4948 netdev_features_t netif_skb_features(struct sk_buff *skb);
4950 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4952 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4954 /* check flags correspondence */
4955 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4956 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4957 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4958 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4959 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4960 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4961 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4962 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4963 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4964 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4965 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4966 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4967 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4968 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4969 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4970 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4971 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4972 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4973 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4975 return (features & feature) == feature;
4978 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4980 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4981 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4984 static inline bool netif_needs_gso(struct sk_buff *skb,
4985 netdev_features_t features)
4987 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4988 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4989 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4992 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4993 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
4994 void netif_inherit_tso_max(struct net_device *to,
4995 const struct net_device *from);
4997 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4998 int pulled_hlen, u16 mac_offset,
5001 skb->protocol = protocol;
5002 skb->encapsulation = 1;
5003 skb_push(skb, pulled_hlen);
5004 skb_reset_transport_header(skb);
5005 skb->mac_header = mac_offset;
5006 skb->network_header = skb->mac_header + mac_len;
5007 skb->mac_len = mac_len;
5010 static inline bool netif_is_macsec(const struct net_device *dev)
5012 return dev->priv_flags & IFF_MACSEC;
5015 static inline bool netif_is_macvlan(const struct net_device *dev)
5017 return dev->priv_flags & IFF_MACVLAN;
5020 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5022 return dev->priv_flags & IFF_MACVLAN_PORT;
5025 static inline bool netif_is_bond_master(const struct net_device *dev)
5027 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5030 static inline bool netif_is_bond_slave(const struct net_device *dev)
5032 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5035 static inline bool netif_supports_nofcs(struct net_device *dev)
5037 return dev->priv_flags & IFF_SUPP_NOFCS;
5040 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5042 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5045 static inline bool netif_is_l3_master(const struct net_device *dev)
5047 return dev->priv_flags & IFF_L3MDEV_MASTER;
5050 static inline bool netif_is_l3_slave(const struct net_device *dev)
5052 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5055 static inline bool netif_is_bridge_master(const struct net_device *dev)
5057 return dev->priv_flags & IFF_EBRIDGE;
5060 static inline bool netif_is_bridge_port(const struct net_device *dev)
5062 return dev->priv_flags & IFF_BRIDGE_PORT;
5065 static inline bool netif_is_ovs_master(const struct net_device *dev)
5067 return dev->priv_flags & IFF_OPENVSWITCH;
5070 static inline bool netif_is_ovs_port(const struct net_device *dev)
5072 return dev->priv_flags & IFF_OVS_DATAPATH;
5075 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5077 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5080 static inline bool netif_is_team_master(const struct net_device *dev)
5082 return dev->priv_flags & IFF_TEAM;
5085 static inline bool netif_is_team_port(const struct net_device *dev)
5087 return dev->priv_flags & IFF_TEAM_PORT;
5090 static inline bool netif_is_lag_master(const struct net_device *dev)
5092 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5095 static inline bool netif_is_lag_port(const struct net_device *dev)
5097 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5100 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5102 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5105 static inline bool netif_is_failover(const struct net_device *dev)
5107 return dev->priv_flags & IFF_FAILOVER;
5110 static inline bool netif_is_failover_slave(const struct net_device *dev)
5112 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5115 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5116 static inline void netif_keep_dst(struct net_device *dev)
5118 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5121 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5122 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5124 /* TODO: reserve and use an additional IFF bit, if we get more users */
5125 return netif_is_macsec(dev);
5128 extern struct pernet_operations __net_initdata loopback_net_ops;
5130 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5132 /* netdev_printk helpers, similar to dev_printk */
5134 static inline const char *netdev_name(const struct net_device *dev)
5136 if (!dev->name[0] || strchr(dev->name, '%'))
5137 return "(unnamed net_device)";
5141 static inline const char *netdev_reg_state(const struct net_device *dev)
5143 switch (dev->reg_state) {
5144 case NETREG_UNINITIALIZED: return " (uninitialized)";
5145 case NETREG_REGISTERED: return "";
5146 case NETREG_UNREGISTERING: return " (unregistering)";
5147 case NETREG_UNREGISTERED: return " (unregistered)";
5148 case NETREG_RELEASED: return " (released)";
5149 case NETREG_DUMMY: return " (dummy)";
5152 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5153 return " (unknown)";
5156 #define MODULE_ALIAS_NETDEV(device) \
5157 MODULE_ALIAS("netdev-" device)
5160 * netdev_WARN() acts like dev_printk(), but with the key difference
5161 * of using a WARN/WARN_ON to get the message out, including the
5162 * file/line information and a backtrace.
5164 #define netdev_WARN(dev, format, args...) \
5165 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5166 netdev_reg_state(dev), ##args)
5168 #define netdev_WARN_ONCE(dev, format, args...) \
5169 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5170 netdev_reg_state(dev), ##args)
5173 * The list of packet types we will receive (as opposed to discard)
5174 * and the routines to invoke.
5176 * Why 16. Because with 16 the only overlap we get on a hash of the
5177 * low nibble of the protocol value is RARP/SNAP/X.25.
5191 #define PTYPE_HASH_SIZE (16)
5192 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5194 extern struct list_head ptype_all __read_mostly;
5195 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5197 extern struct net_device *blackhole_netdev;
5199 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5200 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5201 #define DEV_STATS_ADD(DEV, FIELD, VAL) \
5202 atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5204 #endif /* _LINUX_NETDEVICE_H */