1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 #include <asm/local.h>
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
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 <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
60 struct ip_tunnel_parm;
61 struct macsec_context;
63 struct netdev_name_node;
68 /* 802.15.4 specific */
71 /* UDP Tunnel offloads */
72 struct udp_tunnel_info;
73 struct udp_tunnel_nic_info;
74 struct udp_tunnel_nic;
78 void synchronize_net(void);
79 void netdev_set_default_ethtool_ops(struct net_device *dev,
80 const struct ethtool_ops *ops);
82 /* Backlog congestion levels */
83 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
84 #define NET_RX_DROP 1 /* packet dropped */
86 #define MAX_NEST_DEV 8
89 * Transmit return codes: transmit return codes originate from three different
92 * - qdisc return codes
93 * - driver transmit return codes
96 * Drivers are allowed to return any one of those in their hard_start_xmit()
97 * function. Real network devices commonly used with qdiscs should only return
98 * the driver transmit return codes though - when qdiscs are used, the actual
99 * transmission happens asynchronously, so the value is not propagated to
100 * higher layers. Virtual network devices transmit synchronously; in this case
101 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
102 * others are propagated to higher layers.
105 /* qdisc ->enqueue() return codes. */
106 #define NET_XMIT_SUCCESS 0x00
107 #define NET_XMIT_DROP 0x01 /* skb dropped */
108 #define NET_XMIT_CN 0x02 /* congestion notification */
109 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
111 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
112 * indicates that the device will soon be dropping packets, or already drops
113 * some packets of the same priority; prompting us to send less aggressively. */
114 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
115 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
117 /* Driver transmit return codes */
118 #define NETDEV_TX_MASK 0xf0
121 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
122 NETDEV_TX_OK = 0x00, /* driver took care of packet */
123 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
125 typedef enum netdev_tx netdev_tx_t;
128 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
129 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
131 static inline bool dev_xmit_complete(int rc)
134 * Positive cases with an skb consumed by a driver:
135 * - successful transmission (rc == NETDEV_TX_OK)
136 * - error while transmitting (rc < 0)
137 * - error while queueing to a different device (rc & NET_XMIT_MASK)
139 if (likely(rc < NET_XMIT_MASK))
146 * Compute the worst-case header length according to the protocols
150 #if defined(CONFIG_HYPERV_NET)
151 # define LL_MAX_HEADER 128
152 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
153 # if defined(CONFIG_MAC80211_MESH)
154 # define LL_MAX_HEADER 128
156 # define LL_MAX_HEADER 96
159 # define LL_MAX_HEADER 32
162 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
163 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
164 #define MAX_HEADER LL_MAX_HEADER
166 #define MAX_HEADER (LL_MAX_HEADER + 48)
170 * Old network device statistics. Fields are native words
171 * (unsigned long) so they can be read and written atomically.
174 struct net_device_stats {
175 unsigned long rx_packets;
176 unsigned long tx_packets;
177 unsigned long rx_bytes;
178 unsigned long tx_bytes;
179 unsigned long rx_errors;
180 unsigned long tx_errors;
181 unsigned long rx_dropped;
182 unsigned long tx_dropped;
183 unsigned long multicast;
184 unsigned long collisions;
185 unsigned long rx_length_errors;
186 unsigned long rx_over_errors;
187 unsigned long rx_crc_errors;
188 unsigned long rx_frame_errors;
189 unsigned long rx_fifo_errors;
190 unsigned long rx_missed_errors;
191 unsigned long tx_aborted_errors;
192 unsigned long tx_carrier_errors;
193 unsigned long tx_fifo_errors;
194 unsigned long tx_heartbeat_errors;
195 unsigned long tx_window_errors;
196 unsigned long rx_compressed;
197 unsigned long tx_compressed;
200 /* per-cpu stats, allocated on demand.
201 * Try to fit them in a single cache line, for dev_get_stats() sake.
203 struct net_device_core_stats {
204 unsigned long rx_dropped;
205 unsigned long tx_dropped;
206 unsigned long rx_nohandler;
207 unsigned long rx_otherhost_dropped;
208 } __aligned(4 * sizeof(unsigned long));
210 #include <linux/cache.h>
211 #include <linux/skbuff.h>
214 #include <linux/static_key.h>
215 extern struct static_key_false rps_needed;
216 extern struct static_key_false rfs_needed;
223 struct netdev_hw_addr {
224 struct list_head list;
226 unsigned char addr[MAX_ADDR_LEN];
228 #define NETDEV_HW_ADDR_T_LAN 1
229 #define NETDEV_HW_ADDR_T_SAN 2
230 #define NETDEV_HW_ADDR_T_UNICAST 3
231 #define NETDEV_HW_ADDR_T_MULTICAST 4
236 struct rcu_head rcu_head;
239 struct netdev_hw_addr_list {
240 struct list_head list;
243 /* Auxiliary tree for faster lookup on addition and deletion */
247 #define netdev_hw_addr_list_count(l) ((l)->count)
248 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
249 #define netdev_hw_addr_list_for_each(ha, l) \
250 list_for_each_entry(ha, &(l)->list, list)
252 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
253 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
254 #define netdev_for_each_uc_addr(ha, dev) \
255 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
257 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
258 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
259 #define netdev_for_each_mc_addr(ha, dev) \
260 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
266 /* cached hardware header; allow for machine alignment needs. */
267 #define HH_DATA_MOD 16
268 #define HH_DATA_OFF(__len) \
269 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
270 #define HH_DATA_ALIGN(__len) \
271 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
272 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
275 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
277 * dev->hard_header_len ? (dev->hard_header_len +
278 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
280 * We could use other alignment values, but we must maintain the
281 * relationship HH alignment <= LL alignment.
283 #define LL_RESERVED_SPACE(dev) \
284 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
285 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
286 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
289 int (*create) (struct sk_buff *skb, struct net_device *dev,
290 unsigned short type, const void *daddr,
291 const void *saddr, unsigned int len);
292 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
293 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
294 void (*cache_update)(struct hh_cache *hh,
295 const struct net_device *dev,
296 const unsigned char *haddr);
297 bool (*validate)(const char *ll_header, unsigned int len);
298 __be16 (*parse_protocol)(const struct sk_buff *skb);
301 /* These flag bits are private to the generic network queueing
302 * layer; they may not be explicitly referenced by any other
306 enum netdev_state_t {
308 __LINK_STATE_PRESENT,
309 __LINK_STATE_NOCARRIER,
310 __LINK_STATE_LINKWATCH_PENDING,
311 __LINK_STATE_DORMANT,
312 __LINK_STATE_TESTING,
316 struct list_head list;
321 * size of gro hash buckets, must less than bit number of
322 * napi_struct::gro_bitmask
324 #define GRO_HASH_BUCKETS 8
327 * Structure for NAPI scheduling similar to tasklet but with weighting
330 /* The poll_list must only be managed by the entity which
331 * changes the state of the NAPI_STATE_SCHED bit. This means
332 * whoever atomically sets that bit can add this napi_struct
333 * to the per-CPU poll_list, and whoever clears that bit
334 * can remove from the list right before clearing the bit.
336 struct list_head poll_list;
340 int defer_hard_irqs_count;
341 unsigned long gro_bitmask;
342 int (*poll)(struct napi_struct *, int);
343 #ifdef CONFIG_NETPOLL
346 struct net_device *dev;
347 struct gro_list gro_hash[GRO_HASH_BUCKETS];
349 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
350 int rx_count; /* length of rx_list */
351 struct hrtimer timer;
352 struct list_head dev_list;
353 struct hlist_node napi_hash_node;
354 unsigned int napi_id;
355 struct task_struct *thread;
359 NAPI_STATE_SCHED, /* Poll is scheduled */
360 NAPI_STATE_MISSED, /* reschedule a napi */
361 NAPI_STATE_DISABLE, /* Disable pending */
362 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
363 NAPI_STATE_LISTED, /* NAPI added to system lists */
364 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
365 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
366 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
367 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
368 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
372 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
373 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
374 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
375 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
376 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
377 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
378 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
379 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
380 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
381 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
391 typedef enum gro_result gro_result_t;
394 * enum rx_handler_result - Possible return values for rx_handlers.
395 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
397 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
398 * case skb->dev was changed by rx_handler.
399 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
400 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
402 * rx_handlers are functions called from inside __netif_receive_skb(), to do
403 * special processing of the skb, prior to delivery to protocol handlers.
405 * Currently, a net_device can only have a single rx_handler registered. Trying
406 * to register a second rx_handler will return -EBUSY.
408 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
409 * To unregister a rx_handler on a net_device, use
410 * netdev_rx_handler_unregister().
412 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
415 * If the rx_handler consumed the skb in some way, it should return
416 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
417 * the skb to be delivered in some other way.
419 * If the rx_handler changed skb->dev, to divert the skb to another
420 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
421 * new device will be called if it exists.
423 * If the rx_handler decides the skb should be ignored, it should return
424 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
425 * are registered on exact device (ptype->dev == skb->dev).
427 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
428 * delivered, it should return RX_HANDLER_PASS.
430 * A device without a registered rx_handler will behave as if rx_handler
431 * returned RX_HANDLER_PASS.
434 enum rx_handler_result {
440 typedef enum rx_handler_result rx_handler_result_t;
441 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
443 void __napi_schedule(struct napi_struct *n);
444 void __napi_schedule_irqoff(struct napi_struct *n);
446 static inline bool napi_disable_pending(struct napi_struct *n)
448 return test_bit(NAPI_STATE_DISABLE, &n->state);
451 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
453 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
456 bool napi_schedule_prep(struct napi_struct *n);
459 * napi_schedule - schedule NAPI poll
462 * Schedule NAPI poll routine to be called if it is not already
465 static inline void napi_schedule(struct napi_struct *n)
467 if (napi_schedule_prep(n))
472 * napi_schedule_irqoff - schedule NAPI poll
475 * Variant of napi_schedule(), assuming hard irqs are masked.
477 static inline void napi_schedule_irqoff(struct napi_struct *n)
479 if (napi_schedule_prep(n))
480 __napi_schedule_irqoff(n);
483 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
484 static inline bool napi_reschedule(struct napi_struct *napi)
486 if (napi_schedule_prep(napi)) {
487 __napi_schedule(napi);
493 bool napi_complete_done(struct napi_struct *n, int work_done);
495 * napi_complete - NAPI processing complete
498 * Mark NAPI processing as complete.
499 * Consider using napi_complete_done() instead.
500 * Return false if device should avoid rearming interrupts.
502 static inline bool napi_complete(struct napi_struct *n)
504 return napi_complete_done(n, 0);
507 int dev_set_threaded(struct net_device *dev, bool threaded);
510 * napi_disable - prevent NAPI from scheduling
513 * Stop NAPI from being scheduled on this context.
514 * Waits till any outstanding processing completes.
516 void napi_disable(struct napi_struct *n);
518 void napi_enable(struct napi_struct *n);
521 * napi_synchronize - wait until NAPI is not running
524 * Wait until NAPI is done being scheduled on this context.
525 * Waits till any outstanding processing completes but
526 * does not disable future activations.
528 static inline void napi_synchronize(const struct napi_struct *n)
530 if (IS_ENABLED(CONFIG_SMP))
531 while (test_bit(NAPI_STATE_SCHED, &n->state))
538 * napi_if_scheduled_mark_missed - if napi is running, set the
542 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
547 unsigned long val, new;
550 val = READ_ONCE(n->state);
551 if (val & NAPIF_STATE_DISABLE)
554 if (!(val & NAPIF_STATE_SCHED))
557 new = val | NAPIF_STATE_MISSED;
558 } while (cmpxchg(&n->state, val, new) != val);
563 enum netdev_queue_state_t {
564 __QUEUE_STATE_DRV_XOFF,
565 __QUEUE_STATE_STACK_XOFF,
566 __QUEUE_STATE_FROZEN,
569 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
570 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
571 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
573 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
580 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
581 * netif_tx_* functions below are used to manipulate this flag. The
582 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
583 * queue independently. The netif_xmit_*stopped functions below are called
584 * to check if the queue has been stopped by the driver or stack (either
585 * of the XOFF bits are set in the state). Drivers should not need to call
586 * netif_xmit*stopped functions, they should only be using netif_tx_*.
589 struct netdev_queue {
593 struct net_device *dev;
594 netdevice_tracker dev_tracker;
596 struct Qdisc __rcu *qdisc;
597 struct Qdisc *qdisc_sleeping;
601 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
604 unsigned long tx_maxrate;
606 * Number of TX timeouts for this queue
607 * (/sys/class/net/DEV/Q/trans_timeout)
609 atomic_long_t trans_timeout;
611 /* Subordinate device that the queue has been assigned to */
612 struct net_device *sb_dev;
613 #ifdef CONFIG_XDP_SOCKETS
614 struct xsk_buff_pool *pool;
619 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
622 * Time (in jiffies) of last Tx
624 unsigned long trans_start;
631 } ____cacheline_aligned_in_smp;
633 extern int sysctl_fb_tunnels_only_for_init_net;
634 extern int sysctl_devconf_inherit_init_net;
637 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
638 * == 1 : For initns only
641 static inline bool net_has_fallback_tunnels(const struct net *net)
643 return !IS_ENABLED(CONFIG_SYSCTL) ||
644 !sysctl_fb_tunnels_only_for_init_net ||
645 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
648 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
659 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
666 * This structure holds an RPS map which can be of variable length. The
667 * map is an array of CPUs.
674 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
677 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
678 * tail pointer for that CPU's input queue at the time of last enqueue, and
679 * a hardware filter index.
681 struct rps_dev_flow {
684 unsigned int last_qtail;
686 #define RPS_NO_FILTER 0xffff
689 * The rps_dev_flow_table structure contains a table of flow mappings.
691 struct rps_dev_flow_table {
694 struct rps_dev_flow flows[];
696 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
697 ((_num) * sizeof(struct rps_dev_flow)))
700 * The rps_sock_flow_table contains mappings of flows to the last CPU
701 * on which they were processed by the application (set in recvmsg).
702 * Each entry is a 32bit value. Upper part is the high-order bits
703 * of flow hash, lower part is CPU number.
704 * rps_cpu_mask is used to partition the space, depending on number of
705 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
706 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
707 * meaning we use 32-6=26 bits for the hash.
709 struct rps_sock_flow_table {
712 u32 ents[] ____cacheline_aligned_in_smp;
714 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
716 #define RPS_NO_CPU 0xffff
718 extern u32 rps_cpu_mask;
719 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
721 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
725 unsigned int index = hash & table->mask;
726 u32 val = hash & ~rps_cpu_mask;
728 /* We only give a hint, preemption can change CPU under us */
729 val |= raw_smp_processor_id();
731 if (table->ents[index] != val)
732 table->ents[index] = val;
736 #ifdef CONFIG_RFS_ACCEL
737 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
740 #endif /* CONFIG_RPS */
742 /* This structure contains an instance of an RX queue. */
743 struct netdev_rx_queue {
744 struct xdp_rxq_info xdp_rxq;
746 struct rps_map __rcu *rps_map;
747 struct rps_dev_flow_table __rcu *rps_flow_table;
750 struct net_device *dev;
751 netdevice_tracker dev_tracker;
753 #ifdef CONFIG_XDP_SOCKETS
754 struct xsk_buff_pool *pool;
756 } ____cacheline_aligned_in_smp;
759 * RX queue sysfs structures and functions.
761 struct rx_queue_attribute {
762 struct attribute attr;
763 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
764 ssize_t (*store)(struct netdev_rx_queue *queue,
765 const char *buf, size_t len);
768 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
777 * This structure holds an XPS map which can be of variable length. The
778 * map is an array of queues.
782 unsigned int alloc_len;
786 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
787 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
788 - sizeof(struct xps_map)) / sizeof(u16))
791 * This structure holds all XPS maps for device. Maps are indexed by CPU.
793 * We keep track of the number of cpus/rxqs used when the struct is allocated,
794 * in nr_ids. This will help not accessing out-of-bound memory.
796 * We keep track of the number of traffic classes used when the struct is
797 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
798 * not crossing its upper bound, as the original dev->num_tc can be updated in
801 struct xps_dev_maps {
805 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
808 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
809 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
811 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
812 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
814 #endif /* CONFIG_XPS */
816 #define TC_MAX_QUEUE 16
817 #define TC_BITMASK 15
818 /* HW offloaded queuing disciplines txq count and offset maps */
819 struct netdev_tc_txq {
824 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
826 * This structure is to hold information about the device
827 * configured to run FCoE protocol stack.
829 struct netdev_fcoe_hbainfo {
830 char manufacturer[64];
831 char serial_number[64];
832 char hardware_version[64];
833 char driver_version[64];
834 char optionrom_version[64];
835 char firmware_version[64];
837 char model_description[256];
841 #define MAX_PHYS_ITEM_ID_LEN 32
843 /* This structure holds a unique identifier to identify some
844 * physical item (port for example) used by a netdevice.
846 struct netdev_phys_item_id {
847 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
848 unsigned char id_len;
851 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
852 struct netdev_phys_item_id *b)
854 return a->id_len == b->id_len &&
855 memcmp(a->id, b->id, a->id_len) == 0;
858 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
860 struct net_device *sb_dev);
862 enum net_device_path_type {
863 DEV_PATH_ETHERNET = 0,
871 struct net_device_path {
872 enum net_device_path_type type;
873 const struct net_device *dev;
882 DEV_PATH_BR_VLAN_KEEP,
883 DEV_PATH_BR_VLAN_TAG,
884 DEV_PATH_BR_VLAN_UNTAG,
885 DEV_PATH_BR_VLAN_UNTAG_HW,
903 #define NET_DEVICE_PATH_STACK_MAX 5
904 #define NET_DEVICE_PATH_VLAN_MAX 2
906 struct net_device_path_stack {
908 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
911 struct net_device_path_ctx {
912 const struct net_device *dev;
919 } vlan[NET_DEVICE_PATH_VLAN_MAX];
923 TC_SETUP_QDISC_MQPRIO,
926 TC_SETUP_CLSMATCHALL,
936 TC_SETUP_QDISC_TAPRIO,
945 /* These structures hold the attributes of bpf state that are being passed
946 * to the netdevice through the bpf op.
948 enum bpf_netdev_command {
949 /* Set or clear a bpf program used in the earliest stages of packet
950 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
951 * is responsible for calling bpf_prog_put on any old progs that are
952 * stored. In case of error, the callee need not release the new prog
953 * reference, but on success it takes ownership and must bpf_prog_put
954 * when it is no longer used.
958 /* BPF program for offload callbacks, invoked at program load time. */
959 BPF_OFFLOAD_MAP_ALLOC,
960 BPF_OFFLOAD_MAP_FREE,
964 struct bpf_prog_offload_ops;
965 struct netlink_ext_ack;
967 struct xdp_dev_bulk_queue;
977 struct bpf_xdp_entity {
978 struct bpf_prog *prog;
979 struct bpf_xdp_link *link;
983 enum bpf_netdev_command command;
988 struct bpf_prog *prog;
989 struct netlink_ext_ack *extack;
991 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
993 struct bpf_offloaded_map *offmap;
995 /* XDP_SETUP_XSK_POOL */
997 struct xsk_buff_pool *pool;
1003 /* Flags for ndo_xsk_wakeup. */
1004 #define XDP_WAKEUP_RX (1 << 0)
1005 #define XDP_WAKEUP_TX (1 << 1)
1007 #ifdef CONFIG_XFRM_OFFLOAD
1008 struct xfrmdev_ops {
1009 int (*xdo_dev_state_add) (struct xfrm_state *x);
1010 void (*xdo_dev_state_delete) (struct xfrm_state *x);
1011 void (*xdo_dev_state_free) (struct xfrm_state *x);
1012 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
1013 struct xfrm_state *x);
1014 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1018 struct dev_ifalias {
1019 struct rcu_head rcuhead;
1026 struct netdev_net_notifier {
1027 struct list_head list;
1028 struct notifier_block *nb;
1032 * This structure defines the management hooks for network devices.
1033 * The following hooks can be defined; unless noted otherwise, they are
1034 * optional and can be filled with a null pointer.
1036 * int (*ndo_init)(struct net_device *dev);
1037 * This function is called once when a network device is registered.
1038 * The network device can use this for any late stage initialization
1039 * or semantic validation. It can fail with an error code which will
1040 * be propagated back to register_netdev.
1042 * void (*ndo_uninit)(struct net_device *dev);
1043 * This function is called when device is unregistered or when registration
1044 * fails. It is not called if init fails.
1046 * int (*ndo_open)(struct net_device *dev);
1047 * This function is called when a network device transitions to the up
1050 * int (*ndo_stop)(struct net_device *dev);
1051 * This function is called when a network device transitions to the down
1054 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1055 * struct net_device *dev);
1056 * Called when a packet needs to be transmitted.
1057 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1058 * the queue before that can happen; it's for obsolete devices and weird
1059 * corner cases, but the stack really does a non-trivial amount
1060 * of useless work if you return NETDEV_TX_BUSY.
1061 * Required; cannot be NULL.
1063 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1064 * struct net_device *dev
1065 * netdev_features_t features);
1066 * Called by core transmit path to determine if device is capable of
1067 * performing offload operations on a given packet. This is to give
1068 * the device an opportunity to implement any restrictions that cannot
1069 * be otherwise expressed by feature flags. The check is called with
1070 * the set of features that the stack has calculated and it returns
1071 * those the driver believes to be appropriate.
1073 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1074 * struct net_device *sb_dev);
1075 * Called to decide which queue to use when device supports multiple
1078 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1079 * This function is called to allow device receiver to make
1080 * changes to configuration when multicast or promiscuous is enabled.
1082 * void (*ndo_set_rx_mode)(struct net_device *dev);
1083 * This function is called device changes address list filtering.
1084 * If driver handles unicast address filtering, it should set
1085 * IFF_UNICAST_FLT in its priv_flags.
1087 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1088 * This function is called when the Media Access Control address
1089 * needs to be changed. If this interface is not defined, the
1090 * MAC address can not be changed.
1092 * int (*ndo_validate_addr)(struct net_device *dev);
1093 * Test if Media Access Control address is valid for the device.
1095 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1096 * Old-style ioctl entry point. This is used internally by the
1097 * appletalk and ieee802154 subsystems but is no longer called by
1098 * the device ioctl handler.
1100 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1101 * Used by the bonding driver for its device specific ioctls:
1102 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1103 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1105 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1106 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1107 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1109 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1110 * Used to set network devices bus interface parameters. This interface
1111 * is retained for legacy reasons; new devices should use the bus
1112 * interface (PCI) for low level management.
1114 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1115 * Called when a user wants to change the Maximum Transfer Unit
1118 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1119 * Callback used when the transmitter has not made any progress
1120 * for dev->watchdog ticks.
1122 * void (*ndo_get_stats64)(struct net_device *dev,
1123 * struct rtnl_link_stats64 *storage);
1124 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1125 * Called when a user wants to get the network device usage
1126 * statistics. Drivers must do one of the following:
1127 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1128 * rtnl_link_stats64 structure passed by the caller.
1129 * 2. Define @ndo_get_stats to update a net_device_stats structure
1130 * (which should normally be dev->stats) and return a pointer to
1131 * it. The structure may be changed asynchronously only if each
1132 * field is written atomically.
1133 * 3. Update dev->stats asynchronously and atomically, and define
1134 * neither operation.
1136 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1137 * Return true if this device supports offload stats of this attr_id.
1139 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1141 * Get statistics for offload operations by attr_id. Write it into the
1142 * attr_data pointer.
1144 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1145 * If device supports VLAN filtering this function is called when a
1146 * VLAN id is registered.
1148 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1149 * If device supports VLAN filtering this function is called when a
1150 * VLAN id is unregistered.
1152 * void (*ndo_poll_controller)(struct net_device *dev);
1154 * SR-IOV management functions.
1155 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1156 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1157 * u8 qos, __be16 proto);
1158 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1160 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1161 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1162 * int (*ndo_get_vf_config)(struct net_device *dev,
1163 * int vf, struct ifla_vf_info *ivf);
1164 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1165 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1166 * struct nlattr *port[]);
1168 * Enable or disable the VF ability to query its RSS Redirection Table and
1169 * Hash Key. This is needed since on some devices VF share this information
1170 * with PF and querying it may introduce a theoretical security risk.
1171 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1172 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1173 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1175 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1176 * This is always called from the stack with the rtnl lock held and netif
1177 * tx queues stopped. This allows the netdevice to perform queue
1178 * management safely.
1180 * Fiber Channel over Ethernet (FCoE) offload functions.
1181 * int (*ndo_fcoe_enable)(struct net_device *dev);
1182 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1183 * so the underlying device can perform whatever needed configuration or
1184 * initialization to support acceleration of FCoE traffic.
1186 * int (*ndo_fcoe_disable)(struct net_device *dev);
1187 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1188 * so the underlying device can perform whatever needed clean-ups to
1189 * stop supporting acceleration of FCoE traffic.
1191 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1192 * struct scatterlist *sgl, unsigned int sgc);
1193 * Called when the FCoE Initiator wants to initialize an I/O that
1194 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1195 * perform necessary setup and returns 1 to indicate the device is set up
1196 * successfully to perform DDP on this I/O, otherwise this returns 0.
1198 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1199 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1200 * indicated by the FC exchange id 'xid', so the underlying device can
1201 * clean up and reuse resources for later DDP requests.
1203 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1204 * struct scatterlist *sgl, unsigned int sgc);
1205 * Called when the FCoE Target wants to initialize an I/O that
1206 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1207 * perform necessary setup and returns 1 to indicate the device is set up
1208 * successfully to perform DDP on this I/O, otherwise this returns 0.
1210 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1211 * struct netdev_fcoe_hbainfo *hbainfo);
1212 * Called when the FCoE Protocol stack wants information on the underlying
1213 * device. This information is utilized by the FCoE protocol stack to
1214 * register attributes with Fiber Channel management service as per the
1215 * FC-GS Fabric Device Management Information(FDMI) specification.
1217 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1218 * Called when the underlying device wants to override default World Wide
1219 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1220 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1221 * protocol stack to use.
1224 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1225 * u16 rxq_index, u32 flow_id);
1226 * Set hardware filter for RFS. rxq_index is the target queue index;
1227 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1228 * Return the filter ID on success, or a negative error code.
1230 * Slave management functions (for bridge, bonding, etc).
1231 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1232 * Called to make another netdev an underling.
1234 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1235 * Called to release previously enslaved netdev.
1237 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1238 * struct sk_buff *skb,
1240 * Get the xmit slave of master device. If all_slaves is true, function
1241 * assume all the slaves can transmit.
1243 * Feature/offload setting functions.
1244 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1245 * netdev_features_t features);
1246 * Adjusts the requested feature flags according to device-specific
1247 * constraints, and returns the resulting flags. Must not modify
1250 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1251 * Called to update device configuration to new features. Passed
1252 * feature set might be less than what was returned by ndo_fix_features()).
1253 * Must return >0 or -errno if it changed dev->features itself.
1255 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1256 * struct net_device *dev,
1257 * const unsigned char *addr, u16 vid, u16 flags,
1258 * struct netlink_ext_ack *extack);
1259 * Adds an FDB entry to dev for addr.
1260 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1261 * struct net_device *dev,
1262 * const unsigned char *addr, u16 vid)
1263 * Deletes the FDB entry from dev coresponding to addr.
1264 * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1265 * struct net_device *dev,
1267 * struct netlink_ext_ack *extack);
1268 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1269 * struct net_device *dev, struct net_device *filter_dev,
1271 * Used to add FDB entries to dump requests. Implementers should add
1272 * entries to skb and update idx with the number of entries.
1274 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1275 * u16 flags, struct netlink_ext_ack *extack)
1276 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1277 * struct net_device *dev, u32 filter_mask,
1279 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1282 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1283 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1284 * which do not represent real hardware may define this to allow their
1285 * userspace components to manage their virtual carrier state. Devices
1286 * that determine carrier state from physical hardware properties (eg
1287 * network cables) or protocol-dependent mechanisms (eg
1288 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1290 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1291 * struct netdev_phys_item_id *ppid);
1292 * Called to get ID of physical port of this device. If driver does
1293 * not implement this, it is assumed that the hw is not able to have
1294 * multiple net devices on single physical port.
1296 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1297 * struct netdev_phys_item_id *ppid)
1298 * Called to get the parent ID of the physical port of this device.
1300 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1301 * struct net_device *dev)
1302 * Called by upper layer devices to accelerate switching or other
1303 * station functionality into hardware. 'pdev is the lowerdev
1304 * to use for the offload and 'dev' is the net device that will
1305 * back the offload. Returns a pointer to the private structure
1306 * the upper layer will maintain.
1307 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1308 * Called by upper layer device to delete the station created
1309 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1310 * the station and priv is the structure returned by the add
1312 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1313 * int queue_index, u32 maxrate);
1314 * Called when a user wants to set a max-rate limitation of specific
1316 * int (*ndo_get_iflink)(const struct net_device *dev);
1317 * Called to get the iflink value of this device.
1318 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1319 * This function is used to get egress tunnel information for given skb.
1320 * This is useful for retrieving outer tunnel header parameters while
1322 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1323 * This function is used to specify the headroom that the skb must
1324 * consider when allocation skb during packet reception. Setting
1325 * appropriate rx headroom value allows avoiding skb head copy on
1326 * forward. Setting a negative value resets the rx headroom to the
1328 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1329 * This function is used to set or query state related to XDP on the
1330 * netdevice and manage BPF offload. See definition of
1331 * enum bpf_netdev_command for details.
1332 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1334 * This function is used to submit @n XDP packets for transmit on a
1335 * netdevice. Returns number of frames successfully transmitted, frames
1336 * that got dropped are freed/returned via xdp_return_frame().
1337 * Returns negative number, means general error invoking ndo, meaning
1338 * no frames were xmit'ed and core-caller will free all frames.
1339 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1340 * struct xdp_buff *xdp);
1341 * Get the xmit slave of master device based on the xdp_buff.
1342 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1343 * This function is used to wake up the softirq, ksoftirqd or kthread
1344 * responsible for sending and/or receiving packets on a specific
1345 * queue id bound to an AF_XDP socket. The flags field specifies if
1346 * only RX, only Tx, or both should be woken up using the flags
1347 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1348 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1349 * Get devlink port instance associated with a given netdev.
1350 * Called with a reference on the netdevice and devlink locks only,
1351 * rtnl_lock is not held.
1352 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1354 * Add, change, delete or get information on an IPv4 tunnel.
1355 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1356 * If a device is paired with a peer device, return the peer instance.
1357 * The caller must be under RCU read context.
1358 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1359 * Get the forwarding path to reach the real device from the HW destination address
1360 * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1361 * const struct skb_shared_hwtstamps *hwtstamps,
1363 * Get hardware timestamp based on normal/adjustable time or free running
1364 * cycle counter. This function is required if physical clock supports a
1365 * free running cycle counter.
1367 struct net_device_ops {
1368 int (*ndo_init)(struct net_device *dev);
1369 void (*ndo_uninit)(struct net_device *dev);
1370 int (*ndo_open)(struct net_device *dev);
1371 int (*ndo_stop)(struct net_device *dev);
1372 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1373 struct net_device *dev);
1374 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1375 struct net_device *dev,
1376 netdev_features_t features);
1377 u16 (*ndo_select_queue)(struct net_device *dev,
1378 struct sk_buff *skb,
1379 struct net_device *sb_dev);
1380 void (*ndo_change_rx_flags)(struct net_device *dev,
1382 void (*ndo_set_rx_mode)(struct net_device *dev);
1383 int (*ndo_set_mac_address)(struct net_device *dev,
1385 int (*ndo_validate_addr)(struct net_device *dev);
1386 int (*ndo_do_ioctl)(struct net_device *dev,
1387 struct ifreq *ifr, int cmd);
1388 int (*ndo_eth_ioctl)(struct net_device *dev,
1389 struct ifreq *ifr, int cmd);
1390 int (*ndo_siocbond)(struct net_device *dev,
1391 struct ifreq *ifr, int cmd);
1392 int (*ndo_siocwandev)(struct net_device *dev,
1393 struct if_settings *ifs);
1394 int (*ndo_siocdevprivate)(struct net_device *dev,
1396 void __user *data, int cmd);
1397 int (*ndo_set_config)(struct net_device *dev,
1399 int (*ndo_change_mtu)(struct net_device *dev,
1401 int (*ndo_neigh_setup)(struct net_device *dev,
1402 struct neigh_parms *);
1403 void (*ndo_tx_timeout) (struct net_device *dev,
1404 unsigned int txqueue);
1406 void (*ndo_get_stats64)(struct net_device *dev,
1407 struct rtnl_link_stats64 *storage);
1408 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1409 int (*ndo_get_offload_stats)(int attr_id,
1410 const struct net_device *dev,
1412 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1414 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1415 __be16 proto, u16 vid);
1416 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1417 __be16 proto, u16 vid);
1418 #ifdef CONFIG_NET_POLL_CONTROLLER
1419 void (*ndo_poll_controller)(struct net_device *dev);
1420 int (*ndo_netpoll_setup)(struct net_device *dev,
1421 struct netpoll_info *info);
1422 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1424 int (*ndo_set_vf_mac)(struct net_device *dev,
1425 int queue, u8 *mac);
1426 int (*ndo_set_vf_vlan)(struct net_device *dev,
1427 int queue, u16 vlan,
1428 u8 qos, __be16 proto);
1429 int (*ndo_set_vf_rate)(struct net_device *dev,
1430 int vf, int min_tx_rate,
1432 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1433 int vf, bool setting);
1434 int (*ndo_set_vf_trust)(struct net_device *dev,
1435 int vf, bool setting);
1436 int (*ndo_get_vf_config)(struct net_device *dev,
1438 struct ifla_vf_info *ivf);
1439 int (*ndo_set_vf_link_state)(struct net_device *dev,
1440 int vf, int link_state);
1441 int (*ndo_get_vf_stats)(struct net_device *dev,
1443 struct ifla_vf_stats
1445 int (*ndo_set_vf_port)(struct net_device *dev,
1447 struct nlattr *port[]);
1448 int (*ndo_get_vf_port)(struct net_device *dev,
1449 int vf, struct sk_buff *skb);
1450 int (*ndo_get_vf_guid)(struct net_device *dev,
1452 struct ifla_vf_guid *node_guid,
1453 struct ifla_vf_guid *port_guid);
1454 int (*ndo_set_vf_guid)(struct net_device *dev,
1457 int (*ndo_set_vf_rss_query_en)(
1458 struct net_device *dev,
1459 int vf, bool setting);
1460 int (*ndo_setup_tc)(struct net_device *dev,
1461 enum tc_setup_type type,
1463 #if IS_ENABLED(CONFIG_FCOE)
1464 int (*ndo_fcoe_enable)(struct net_device *dev);
1465 int (*ndo_fcoe_disable)(struct net_device *dev);
1466 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1468 struct scatterlist *sgl,
1470 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1472 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1474 struct scatterlist *sgl,
1476 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1477 struct netdev_fcoe_hbainfo *hbainfo);
1480 #if IS_ENABLED(CONFIG_LIBFCOE)
1481 #define NETDEV_FCOE_WWNN 0
1482 #define NETDEV_FCOE_WWPN 1
1483 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1484 u64 *wwn, int type);
1487 #ifdef CONFIG_RFS_ACCEL
1488 int (*ndo_rx_flow_steer)(struct net_device *dev,
1489 const struct sk_buff *skb,
1493 int (*ndo_add_slave)(struct net_device *dev,
1494 struct net_device *slave_dev,
1495 struct netlink_ext_ack *extack);
1496 int (*ndo_del_slave)(struct net_device *dev,
1497 struct net_device *slave_dev);
1498 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1499 struct sk_buff *skb,
1501 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1503 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1504 netdev_features_t features);
1505 int (*ndo_set_features)(struct net_device *dev,
1506 netdev_features_t features);
1507 int (*ndo_neigh_construct)(struct net_device *dev,
1508 struct neighbour *n);
1509 void (*ndo_neigh_destroy)(struct net_device *dev,
1510 struct neighbour *n);
1512 int (*ndo_fdb_add)(struct ndmsg *ndm,
1513 struct nlattr *tb[],
1514 struct net_device *dev,
1515 const unsigned char *addr,
1518 struct netlink_ext_ack *extack);
1519 int (*ndo_fdb_del)(struct ndmsg *ndm,
1520 struct nlattr *tb[],
1521 struct net_device *dev,
1522 const unsigned char *addr,
1523 u16 vid, struct netlink_ext_ack *extack);
1524 int (*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1525 struct nlattr *tb[],
1526 struct net_device *dev,
1528 struct netlink_ext_ack *extack);
1529 int (*ndo_fdb_dump)(struct sk_buff *skb,
1530 struct netlink_callback *cb,
1531 struct net_device *dev,
1532 struct net_device *filter_dev,
1534 int (*ndo_fdb_get)(struct sk_buff *skb,
1535 struct nlattr *tb[],
1536 struct net_device *dev,
1537 const unsigned char *addr,
1538 u16 vid, u32 portid, u32 seq,
1539 struct netlink_ext_ack *extack);
1540 int (*ndo_bridge_setlink)(struct net_device *dev,
1541 struct nlmsghdr *nlh,
1543 struct netlink_ext_ack *extack);
1544 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1546 struct net_device *dev,
1549 int (*ndo_bridge_dellink)(struct net_device *dev,
1550 struct nlmsghdr *nlh,
1552 int (*ndo_change_carrier)(struct net_device *dev,
1554 int (*ndo_get_phys_port_id)(struct net_device *dev,
1555 struct netdev_phys_item_id *ppid);
1556 int (*ndo_get_port_parent_id)(struct net_device *dev,
1557 struct netdev_phys_item_id *ppid);
1558 int (*ndo_get_phys_port_name)(struct net_device *dev,
1559 char *name, size_t len);
1560 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1561 struct net_device *dev);
1562 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1565 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1568 int (*ndo_get_iflink)(const struct net_device *dev);
1569 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1570 struct sk_buff *skb);
1571 void (*ndo_set_rx_headroom)(struct net_device *dev,
1572 int needed_headroom);
1573 int (*ndo_bpf)(struct net_device *dev,
1574 struct netdev_bpf *bpf);
1575 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1576 struct xdp_frame **xdp,
1578 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1579 struct xdp_buff *xdp);
1580 int (*ndo_xsk_wakeup)(struct net_device *dev,
1581 u32 queue_id, u32 flags);
1582 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1583 int (*ndo_tunnel_ctl)(struct net_device *dev,
1584 struct ip_tunnel_parm *p, int cmd);
1585 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1586 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1587 struct net_device_path *path);
1588 ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1589 const struct skb_shared_hwtstamps *hwtstamps,
1594 * enum netdev_priv_flags - &struct net_device priv_flags
1596 * These are the &struct net_device, they are only set internally
1597 * by drivers and used in the kernel. These flags are invisible to
1598 * userspace; this means that the order of these flags can change
1599 * during any kernel release.
1601 * You should have a pretty good reason to be extending these flags.
1603 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1604 * @IFF_EBRIDGE: Ethernet bridging device
1605 * @IFF_BONDING: bonding master or slave
1606 * @IFF_ISATAP: ISATAP interface (RFC4214)
1607 * @IFF_WAN_HDLC: WAN HDLC device
1608 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1610 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1611 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1612 * @IFF_MACVLAN_PORT: device used as macvlan port
1613 * @IFF_BRIDGE_PORT: device used as bridge port
1614 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1615 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1616 * @IFF_UNICAST_FLT: Supports unicast filtering
1617 * @IFF_TEAM_PORT: device used as team port
1618 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1619 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1620 * change when it's running
1621 * @IFF_MACVLAN: Macvlan device
1622 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1623 * underlying stacked devices
1624 * @IFF_L3MDEV_MASTER: device is an L3 master device
1625 * @IFF_NO_QUEUE: device can run without qdisc attached
1626 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1627 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1628 * @IFF_TEAM: device is a team device
1629 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1630 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1631 * entity (i.e. the master device for bridged veth)
1632 * @IFF_MACSEC: device is a MACsec device
1633 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1634 * @IFF_FAILOVER: device is a failover master device
1635 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1636 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1637 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1638 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1639 * skb_headlen(skb) == 0 (data starts from frag0)
1640 * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1642 enum netdev_priv_flags {
1643 IFF_802_1Q_VLAN = 1<<0,
1647 IFF_WAN_HDLC = 1<<4,
1648 IFF_XMIT_DST_RELEASE = 1<<5,
1649 IFF_DONT_BRIDGE = 1<<6,
1650 IFF_DISABLE_NETPOLL = 1<<7,
1651 IFF_MACVLAN_PORT = 1<<8,
1652 IFF_BRIDGE_PORT = 1<<9,
1653 IFF_OVS_DATAPATH = 1<<10,
1654 IFF_TX_SKB_SHARING = 1<<11,
1655 IFF_UNICAST_FLT = 1<<12,
1656 IFF_TEAM_PORT = 1<<13,
1657 IFF_SUPP_NOFCS = 1<<14,
1658 IFF_LIVE_ADDR_CHANGE = 1<<15,
1659 IFF_MACVLAN = 1<<16,
1660 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1661 IFF_L3MDEV_MASTER = 1<<18,
1662 IFF_NO_QUEUE = 1<<19,
1663 IFF_OPENVSWITCH = 1<<20,
1664 IFF_L3MDEV_SLAVE = 1<<21,
1666 IFF_RXFH_CONFIGURED = 1<<23,
1667 IFF_PHONY_HEADROOM = 1<<24,
1669 IFF_NO_RX_HANDLER = 1<<26,
1670 IFF_FAILOVER = 1<<27,
1671 IFF_FAILOVER_SLAVE = 1<<28,
1672 IFF_L3MDEV_RX_HANDLER = 1<<29,
1673 IFF_LIVE_RENAME_OK = 1<<30,
1674 IFF_TX_SKB_NO_LINEAR = 1<<31,
1675 IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1678 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1679 #define IFF_EBRIDGE IFF_EBRIDGE
1680 #define IFF_BONDING IFF_BONDING
1681 #define IFF_ISATAP IFF_ISATAP
1682 #define IFF_WAN_HDLC IFF_WAN_HDLC
1683 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1684 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1685 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1686 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1687 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1688 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1689 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1690 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1691 #define IFF_TEAM_PORT IFF_TEAM_PORT
1692 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1693 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1694 #define IFF_MACVLAN IFF_MACVLAN
1695 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1696 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1697 #define IFF_NO_QUEUE IFF_NO_QUEUE
1698 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1699 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1700 #define IFF_TEAM IFF_TEAM
1701 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1702 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1703 #define IFF_MACSEC IFF_MACSEC
1704 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1705 #define IFF_FAILOVER IFF_FAILOVER
1706 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1707 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1708 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1709 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1711 /* Specifies the type of the struct net_device::ml_priv pointer */
1712 enum netdev_ml_priv_type {
1718 * struct net_device - The DEVICE structure.
1720 * Actually, this whole structure is a big mistake. It mixes I/O
1721 * data with strictly "high-level" data, and it has to know about
1722 * almost every data structure used in the INET module.
1724 * @name: This is the first field of the "visible" part of this structure
1725 * (i.e. as seen by users in the "Space.c" file). It is the name
1728 * @name_node: Name hashlist node
1729 * @ifalias: SNMP alias
1730 * @mem_end: Shared memory end
1731 * @mem_start: Shared memory start
1732 * @base_addr: Device I/O address
1733 * @irq: Device IRQ number
1735 * @state: Generic network queuing layer state, see netdev_state_t
1736 * @dev_list: The global list of network devices
1737 * @napi_list: List entry used for polling NAPI devices
1738 * @unreg_list: List entry when we are unregistering the
1739 * device; see the function unregister_netdev
1740 * @close_list: List entry used when we are closing the device
1741 * @ptype_all: Device-specific packet handlers for all protocols
1742 * @ptype_specific: Device-specific, protocol-specific packet handlers
1744 * @adj_list: Directly linked devices, like slaves for bonding
1745 * @features: Currently active device features
1746 * @hw_features: User-changeable features
1748 * @wanted_features: User-requested features
1749 * @vlan_features: Mask of features inheritable by VLAN devices
1751 * @hw_enc_features: Mask of features inherited by encapsulating devices
1752 * This field indicates what encapsulation
1753 * offloads the hardware is capable of doing,
1754 * and drivers will need to set them appropriately.
1756 * @mpls_features: Mask of features inheritable by MPLS
1757 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1759 * @ifindex: interface index
1760 * @group: The group the device belongs to
1762 * @stats: Statistics struct, which was left as a legacy, use
1763 * rtnl_link_stats64 instead
1765 * @core_stats: core networking counters,
1766 * do not use this in drivers
1767 * @carrier_up_count: Number of times the carrier has been up
1768 * @carrier_down_count: Number of times the carrier has been down
1770 * @wireless_handlers: List of functions to handle Wireless Extensions,
1772 * see <net/iw_handler.h> for details.
1773 * @wireless_data: Instance data managed by the core of wireless extensions
1775 * @netdev_ops: Includes several pointers to callbacks,
1776 * if one wants to override the ndo_*() functions
1777 * @ethtool_ops: Management operations
1778 * @l3mdev_ops: Layer 3 master device operations
1779 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1780 * discovery handling. Necessary for e.g. 6LoWPAN.
1781 * @xfrmdev_ops: Transformation offload operations
1782 * @tlsdev_ops: Transport Layer Security offload operations
1783 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1784 * of Layer 2 headers.
1786 * @flags: Interface flags (a la BSD)
1787 * @priv_flags: Like 'flags' but invisible to userspace,
1788 * see if.h for the definitions
1789 * @gflags: Global flags ( kept as legacy )
1790 * @padded: How much padding added by alloc_netdev()
1791 * @operstate: RFC2863 operstate
1792 * @link_mode: Mapping policy to operstate
1793 * @if_port: Selectable AUI, TP, ...
1795 * @mtu: Interface MTU value
1796 * @min_mtu: Interface Minimum MTU value
1797 * @max_mtu: Interface Maximum MTU value
1798 * @type: Interface hardware type
1799 * @hard_header_len: Maximum hardware header length.
1800 * @min_header_len: Minimum hardware header length
1802 * @needed_headroom: Extra headroom the hardware may need, but not in all
1803 * cases can this be guaranteed
1804 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1805 * cases can this be guaranteed. Some cases also use
1806 * LL_MAX_HEADER instead to allocate the skb
1808 * interface address info:
1810 * @perm_addr: Permanent hw address
1811 * @addr_assign_type: Hw address assignment type
1812 * @addr_len: Hardware address length
1813 * @upper_level: Maximum depth level of upper devices.
1814 * @lower_level: Maximum depth level of lower devices.
1815 * @neigh_priv_len: Used in neigh_alloc()
1816 * @dev_id: Used to differentiate devices that share
1817 * the same link layer address
1818 * @dev_port: Used to differentiate devices that share
1820 * @addr_list_lock: XXX: need comments on this one
1821 * @name_assign_type: network interface name assignment type
1822 * @uc_promisc: Counter that indicates promiscuous mode
1823 * has been enabled due to the need to listen to
1824 * additional unicast addresses in a device that
1825 * does not implement ndo_set_rx_mode()
1826 * @uc: unicast mac addresses
1827 * @mc: multicast mac addresses
1828 * @dev_addrs: list of device hw addresses
1829 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1830 * @promiscuity: Number of times the NIC is told to work in
1831 * promiscuous mode; if it becomes 0 the NIC will
1832 * exit promiscuous mode
1833 * @allmulti: Counter, enables or disables allmulticast mode
1835 * @vlan_info: VLAN info
1836 * @dsa_ptr: dsa specific data
1837 * @tipc_ptr: TIPC specific data
1838 * @atalk_ptr: AppleTalk link
1839 * @ip_ptr: IPv4 specific data
1840 * @dn_ptr: DECnet specific data
1841 * @ip6_ptr: IPv6 specific data
1842 * @ax25_ptr: AX.25 specific data
1843 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1844 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1846 * @mpls_ptr: mpls_dev struct pointer
1847 * @mctp_ptr: MCTP specific data
1849 * @dev_addr: Hw address (before bcast,
1850 * because most packets are unicast)
1852 * @_rx: Array of RX queues
1853 * @num_rx_queues: Number of RX queues
1854 * allocated at register_netdev() time
1855 * @real_num_rx_queues: Number of RX queues currently active in device
1856 * @xdp_prog: XDP sockets filter program pointer
1857 * @gro_flush_timeout: timeout for GRO layer in NAPI
1858 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1859 * allow to avoid NIC hard IRQ, on busy queues.
1861 * @rx_handler: handler for received packets
1862 * @rx_handler_data: XXX: need comments on this one
1863 * @miniq_ingress: ingress/clsact qdisc specific data for
1864 * ingress processing
1865 * @ingress_queue: XXX: need comments on this one
1866 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1867 * @broadcast: hw bcast address
1869 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1870 * indexed by RX queue number. Assigned by driver.
1871 * This must only be set if the ndo_rx_flow_steer
1872 * operation is defined
1873 * @index_hlist: Device index hash chain
1875 * @_tx: Array of TX queues
1876 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1877 * @real_num_tx_queues: Number of TX queues currently active in device
1878 * @qdisc: Root qdisc from userspace point of view
1879 * @tx_queue_len: Max frames per queue allowed
1880 * @tx_global_lock: XXX: need comments on this one
1881 * @xdp_bulkq: XDP device bulk queue
1882 * @xps_maps: all CPUs/RXQs maps for XPS device
1884 * @xps_maps: XXX: need comments on this one
1885 * @miniq_egress: clsact qdisc specific data for
1887 * @nf_hooks_egress: netfilter hooks executed for egress packets
1888 * @qdisc_hash: qdisc hash table
1889 * @watchdog_timeo: Represents the timeout that is used by
1890 * the watchdog (see dev_watchdog())
1891 * @watchdog_timer: List of timers
1893 * @proto_down_reason: reason a netdev interface is held down
1894 * @pcpu_refcnt: Number of references to this device
1895 * @dev_refcnt: Number of references to this device
1896 * @refcnt_tracker: Tracker directory for tracked references to this device
1897 * @todo_list: Delayed register/unregister
1898 * @link_watch_list: XXX: need comments on this one
1900 * @reg_state: Register/unregister state machine
1901 * @dismantle: Device is going to be freed
1902 * @rtnl_link_state: This enum represents the phases of creating
1905 * @needs_free_netdev: Should unregister perform free_netdev?
1906 * @priv_destructor: Called from unregister
1907 * @npinfo: XXX: need comments on this one
1908 * @nd_net: Network namespace this network device is inside
1910 * @ml_priv: Mid-layer private
1911 * @ml_priv_type: Mid-layer private type
1912 * @lstats: Loopback statistics
1913 * @tstats: Tunnel statistics
1914 * @dstats: Dummy statistics
1915 * @vstats: Virtual ethernet statistics
1920 * @dm_private: Drop monitor private
1922 * @dev: Class/net/name entry
1923 * @sysfs_groups: Space for optional device, statistics and wireless
1926 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1927 * @rtnl_link_ops: Rtnl_link_ops
1929 * @gso_max_size: Maximum size of generic segmentation offload
1930 * @tso_max_size: Device (as in HW) limit on the max TSO request size
1931 * @gso_max_segs: Maximum number of segments that can be passed to the
1933 * @tso_max_segs: Device (as in HW) limit on the max TSO segment count
1935 * @dcbnl_ops: Data Center Bridging netlink ops
1936 * @num_tc: Number of traffic classes in the net device
1937 * @tc_to_txq: XXX: need comments on this one
1938 * @prio_tc_map: XXX: need comments on this one
1940 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1942 * @priomap: XXX: need comments on this one
1943 * @phydev: Physical device may attach itself
1944 * for hardware timestamping
1945 * @sfp_bus: attached &struct sfp_bus structure.
1947 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1949 * @proto_down: protocol port state information can be sent to the
1950 * switch driver and used to set the phys state of the
1953 * @wol_enabled: Wake-on-LAN is enabled
1955 * @threaded: napi threaded mode is enabled
1957 * @net_notifier_list: List of per-net netdev notifier block
1958 * that follow this device when it is moved
1959 * to another network namespace.
1961 * @macsec_ops: MACsec offloading ops
1963 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1964 * offload capabilities of the device
1965 * @udp_tunnel_nic: UDP tunnel offload state
1966 * @xdp_state: stores info on attached XDP BPF programs
1968 * @nested_level: Used as a parameter of spin_lock_nested() of
1969 * dev->addr_list_lock.
1970 * @unlink_list: As netif_addr_lock() can be called recursively,
1971 * keep a list of interfaces to be deleted.
1972 * @gro_max_size: Maximum size of aggregated packet in generic
1973 * receive offload (GRO)
1975 * @dev_addr_shadow: Copy of @dev_addr to catch direct writes.
1976 * @linkwatch_dev_tracker: refcount tracker used by linkwatch.
1977 * @watchdog_dev_tracker: refcount tracker used by watchdog.
1978 * @dev_registered_tracker: tracker for reference held while
1980 * @offload_xstats_l3: L3 HW stats for this netdevice.
1982 * FIXME: cleanup struct net_device such that network protocol info
1987 char name[IFNAMSIZ];
1988 struct netdev_name_node *name_node;
1989 struct dev_ifalias __rcu *ifalias;
1991 * I/O specific fields
1992 * FIXME: Merge these and struct ifmap into one
1994 unsigned long mem_end;
1995 unsigned long mem_start;
1996 unsigned long base_addr;
1999 * Some hardware also needs these fields (state,dev_list,
2000 * napi_list,unreg_list,close_list) but they are not
2001 * part of the usual set specified in Space.c.
2004 unsigned long state;
2006 struct list_head dev_list;
2007 struct list_head napi_list;
2008 struct list_head unreg_list;
2009 struct list_head close_list;
2010 struct list_head ptype_all;
2011 struct list_head ptype_specific;
2014 struct list_head upper;
2015 struct list_head lower;
2018 /* Read-mostly cache-line for fast-path access */
2020 unsigned long long priv_flags;
2021 const struct net_device_ops *netdev_ops;
2023 unsigned short gflags;
2024 unsigned short hard_header_len;
2026 /* Note : dev->mtu is often read without holding a lock.
2027 * Writers usually hold RTNL.
2028 * It is recommended to use READ_ONCE() to annotate the reads,
2029 * and to use WRITE_ONCE() to annotate the writes.
2032 unsigned short needed_headroom;
2033 unsigned short needed_tailroom;
2035 netdev_features_t features;
2036 netdev_features_t hw_features;
2037 netdev_features_t wanted_features;
2038 netdev_features_t vlan_features;
2039 netdev_features_t hw_enc_features;
2040 netdev_features_t mpls_features;
2041 netdev_features_t gso_partial_features;
2043 unsigned int min_mtu;
2044 unsigned int max_mtu;
2045 unsigned short type;
2046 unsigned char min_header_len;
2047 unsigned char name_assign_type;
2051 struct net_device_stats stats; /* not used by modern drivers */
2053 struct net_device_core_stats __percpu *core_stats;
2055 /* Stats to monitor link on/off, flapping */
2056 atomic_t carrier_up_count;
2057 atomic_t carrier_down_count;
2059 #ifdef CONFIG_WIRELESS_EXT
2060 const struct iw_handler_def *wireless_handlers;
2061 struct iw_public_data *wireless_data;
2063 const struct ethtool_ops *ethtool_ops;
2064 #ifdef CONFIG_NET_L3_MASTER_DEV
2065 const struct l3mdev_ops *l3mdev_ops;
2067 #if IS_ENABLED(CONFIG_IPV6)
2068 const struct ndisc_ops *ndisc_ops;
2071 #ifdef CONFIG_XFRM_OFFLOAD
2072 const struct xfrmdev_ops *xfrmdev_ops;
2075 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2076 const struct tlsdev_ops *tlsdev_ops;
2079 const struct header_ops *header_ops;
2081 unsigned char operstate;
2082 unsigned char link_mode;
2084 unsigned char if_port;
2087 /* Interface address info. */
2088 unsigned char perm_addr[MAX_ADDR_LEN];
2089 unsigned char addr_assign_type;
2090 unsigned char addr_len;
2091 unsigned char upper_level;
2092 unsigned char lower_level;
2094 unsigned short neigh_priv_len;
2095 unsigned short dev_id;
2096 unsigned short dev_port;
2097 unsigned short padded;
2099 spinlock_t addr_list_lock;
2102 struct netdev_hw_addr_list uc;
2103 struct netdev_hw_addr_list mc;
2104 struct netdev_hw_addr_list dev_addrs;
2107 struct kset *queues_kset;
2109 #ifdef CONFIG_LOCKDEP
2110 struct list_head unlink_list;
2112 unsigned int promiscuity;
2113 unsigned int allmulti;
2115 #ifdef CONFIG_LOCKDEP
2116 unsigned char nested_level;
2120 /* Protocol-specific pointers */
2122 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2123 struct vlan_info __rcu *vlan_info;
2125 #if IS_ENABLED(CONFIG_NET_DSA)
2126 struct dsa_port *dsa_ptr;
2128 #if IS_ENABLED(CONFIG_TIPC)
2129 struct tipc_bearer __rcu *tipc_ptr;
2131 #if IS_ENABLED(CONFIG_ATALK)
2134 struct in_device __rcu *ip_ptr;
2135 #if IS_ENABLED(CONFIG_DECNET)
2136 struct dn_dev __rcu *dn_ptr;
2138 struct inet6_dev __rcu *ip6_ptr;
2139 #if IS_ENABLED(CONFIG_AX25)
2142 struct wireless_dev *ieee80211_ptr;
2143 struct wpan_dev *ieee802154_ptr;
2144 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2145 struct mpls_dev __rcu *mpls_ptr;
2147 #if IS_ENABLED(CONFIG_MCTP)
2148 struct mctp_dev __rcu *mctp_ptr;
2152 * Cache lines mostly used on receive path (including eth_type_trans())
2154 /* Interface address info used in eth_type_trans() */
2155 const unsigned char *dev_addr;
2157 struct netdev_rx_queue *_rx;
2158 unsigned int num_rx_queues;
2159 unsigned int real_num_rx_queues;
2161 struct bpf_prog __rcu *xdp_prog;
2162 unsigned long gro_flush_timeout;
2163 int napi_defer_hard_irqs;
2164 #define GRO_MAX_SIZE 65536
2165 unsigned int gro_max_size;
2166 rx_handler_func_t __rcu *rx_handler;
2167 void __rcu *rx_handler_data;
2169 #ifdef CONFIG_NET_CLS_ACT
2170 struct mini_Qdisc __rcu *miniq_ingress;
2172 struct netdev_queue __rcu *ingress_queue;
2173 #ifdef CONFIG_NETFILTER_INGRESS
2174 struct nf_hook_entries __rcu *nf_hooks_ingress;
2177 unsigned char broadcast[MAX_ADDR_LEN];
2178 #ifdef CONFIG_RFS_ACCEL
2179 struct cpu_rmap *rx_cpu_rmap;
2181 struct hlist_node index_hlist;
2184 * Cache lines mostly used on transmit path
2186 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2187 unsigned int num_tx_queues;
2188 unsigned int real_num_tx_queues;
2189 struct Qdisc __rcu *qdisc;
2190 unsigned int tx_queue_len;
2191 spinlock_t tx_global_lock;
2193 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2196 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2198 #ifdef CONFIG_NET_CLS_ACT
2199 struct mini_Qdisc __rcu *miniq_egress;
2201 #ifdef CONFIG_NETFILTER_EGRESS
2202 struct nf_hook_entries __rcu *nf_hooks_egress;
2205 #ifdef CONFIG_NET_SCHED
2206 DECLARE_HASHTABLE (qdisc_hash, 4);
2208 /* These may be needed for future network-power-down code. */
2209 struct timer_list watchdog_timer;
2212 u32 proto_down_reason;
2214 struct list_head todo_list;
2216 #ifdef CONFIG_PCPU_DEV_REFCNT
2217 int __percpu *pcpu_refcnt;
2219 refcount_t dev_refcnt;
2221 struct ref_tracker_dir refcnt_tracker;
2223 struct list_head link_watch_list;
2225 enum { NETREG_UNINITIALIZED=0,
2226 NETREG_REGISTERED, /* completed register_netdevice */
2227 NETREG_UNREGISTERING, /* called unregister_netdevice */
2228 NETREG_UNREGISTERED, /* completed unregister todo */
2229 NETREG_RELEASED, /* called free_netdev */
2230 NETREG_DUMMY, /* dummy device for NAPI poll */
2236 RTNL_LINK_INITIALIZED,
2237 RTNL_LINK_INITIALIZING,
2238 } rtnl_link_state:16;
2240 bool needs_free_netdev;
2241 void (*priv_destructor)(struct net_device *dev);
2243 #ifdef CONFIG_NETPOLL
2244 struct netpoll_info __rcu *npinfo;
2247 possible_net_t nd_net;
2249 /* mid-layer private */
2251 enum netdev_ml_priv_type ml_priv_type;
2254 struct pcpu_lstats __percpu *lstats;
2255 struct pcpu_sw_netstats __percpu *tstats;
2256 struct pcpu_dstats __percpu *dstats;
2259 #if IS_ENABLED(CONFIG_GARP)
2260 struct garp_port __rcu *garp_port;
2262 #if IS_ENABLED(CONFIG_MRP)
2263 struct mrp_port __rcu *mrp_port;
2265 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2266 struct dm_hw_stat_delta __rcu *dm_private;
2269 const struct attribute_group *sysfs_groups[4];
2270 const struct attribute_group *sysfs_rx_queue_group;
2272 const struct rtnl_link_ops *rtnl_link_ops;
2274 /* for setting kernel sock attribute on TCP connection setup */
2275 #define GSO_MAX_SEGS 65535u
2276 #define GSO_LEGACY_MAX_SIZE 65536u
2277 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2278 * and shinfo->gso_segs is a 16bit field.
2280 #define GSO_MAX_SIZE (8 * GSO_MAX_SEGS)
2282 unsigned int gso_max_size;
2283 #define TSO_LEGACY_MAX_SIZE 65536
2284 #define TSO_MAX_SIZE UINT_MAX
2285 unsigned int tso_max_size;
2287 #define TSO_MAX_SEGS U16_MAX
2291 const struct dcbnl_rtnl_ops *dcbnl_ops;
2294 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2295 u8 prio_tc_map[TC_BITMASK + 1];
2297 #if IS_ENABLED(CONFIG_FCOE)
2298 unsigned int fcoe_ddp_xid;
2300 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2301 struct netprio_map __rcu *priomap;
2303 struct phy_device *phydev;
2304 struct sfp_bus *sfp_bus;
2305 struct lock_class_key *qdisc_tx_busylock;
2307 unsigned wol_enabled:1;
2308 unsigned threaded:1;
2310 struct list_head net_notifier_list;
2312 #if IS_ENABLED(CONFIG_MACSEC)
2313 /* MACsec management functions */
2314 const struct macsec_ops *macsec_ops;
2316 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2317 struct udp_tunnel_nic *udp_tunnel_nic;
2319 /* protected by rtnl_lock */
2320 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2322 u8 dev_addr_shadow[MAX_ADDR_LEN];
2323 netdevice_tracker linkwatch_dev_tracker;
2324 netdevice_tracker watchdog_dev_tracker;
2325 netdevice_tracker dev_registered_tracker;
2326 struct rtnl_hw_stats64 *offload_xstats_l3;
2328 #define to_net_dev(d) container_of(d, struct net_device, dev)
2330 static inline bool netif_elide_gro(const struct net_device *dev)
2332 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2337 #define NETDEV_ALIGN 32
2340 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2342 return dev->prio_tc_map[prio & TC_BITMASK];
2346 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2348 if (tc >= dev->num_tc)
2351 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2355 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2356 void netdev_reset_tc(struct net_device *dev);
2357 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2358 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2361 int netdev_get_num_tc(struct net_device *dev)
2366 static inline void net_prefetch(void *p)
2369 #if L1_CACHE_BYTES < 128
2370 prefetch((u8 *)p + L1_CACHE_BYTES);
2374 static inline void net_prefetchw(void *p)
2377 #if L1_CACHE_BYTES < 128
2378 prefetchw((u8 *)p + L1_CACHE_BYTES);
2382 void netdev_unbind_sb_channel(struct net_device *dev,
2383 struct net_device *sb_dev);
2384 int netdev_bind_sb_channel_queue(struct net_device *dev,
2385 struct net_device *sb_dev,
2386 u8 tc, u16 count, u16 offset);
2387 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2388 static inline int netdev_get_sb_channel(struct net_device *dev)
2390 return max_t(int, -dev->num_tc, 0);
2394 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2397 return &dev->_tx[index];
2400 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2401 const struct sk_buff *skb)
2403 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2406 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2407 void (*f)(struct net_device *,
2408 struct netdev_queue *,
2414 for (i = 0; i < dev->num_tx_queues; i++)
2415 f(dev, &dev->_tx[i], arg);
2418 #define netdev_lockdep_set_classes(dev) \
2420 static struct lock_class_key qdisc_tx_busylock_key; \
2421 static struct lock_class_key qdisc_xmit_lock_key; \
2422 static struct lock_class_key dev_addr_list_lock_key; \
2425 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2426 lockdep_set_class(&(dev)->addr_list_lock, \
2427 &dev_addr_list_lock_key); \
2428 for (i = 0; i < (dev)->num_tx_queues; i++) \
2429 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2430 &qdisc_xmit_lock_key); \
2433 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2434 struct net_device *sb_dev);
2435 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2436 struct sk_buff *skb,
2437 struct net_device *sb_dev);
2439 /* returns the headroom that the master device needs to take in account
2440 * when forwarding to this dev
2442 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2444 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2447 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2449 if (dev->netdev_ops->ndo_set_rx_headroom)
2450 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2453 /* set the device rx headroom to the dev's default */
2454 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2456 netdev_set_rx_headroom(dev, -1);
2459 static inline void *netdev_get_ml_priv(struct net_device *dev,
2460 enum netdev_ml_priv_type type)
2462 if (dev->ml_priv_type != type)
2465 return dev->ml_priv;
2468 static inline void netdev_set_ml_priv(struct net_device *dev,
2470 enum netdev_ml_priv_type type)
2472 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2473 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2474 dev->ml_priv_type, type);
2475 WARN(!dev->ml_priv_type && dev->ml_priv,
2476 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2478 dev->ml_priv = ml_priv;
2479 dev->ml_priv_type = type;
2483 * Net namespace inlines
2486 struct net *dev_net(const struct net_device *dev)
2488 return read_pnet(&dev->nd_net);
2492 void dev_net_set(struct net_device *dev, struct net *net)
2494 write_pnet(&dev->nd_net, net);
2498 * netdev_priv - access network device private data
2499 * @dev: network device
2501 * Get network device private data
2503 static inline void *netdev_priv(const struct net_device *dev)
2505 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2508 /* Set the sysfs physical device reference for the network logical device
2509 * if set prior to registration will cause a symlink during initialization.
2511 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2513 /* Set the sysfs device type for the network logical device to allow
2514 * fine-grained identification of different network device types. For
2515 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2517 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2519 /* Default NAPI poll() weight
2520 * Device drivers are strongly advised to not use bigger value
2522 #define NAPI_POLL_WEIGHT 64
2524 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2525 int (*poll)(struct napi_struct *, int), int weight);
2528 * netif_napi_add() - initialize a NAPI context
2529 * @dev: network device
2530 * @napi: NAPI context
2531 * @poll: polling function
2532 * @weight: default weight
2534 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2535 * *any* of the other NAPI-related functions.
2538 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2539 int (*poll)(struct napi_struct *, int), int weight)
2541 netif_napi_add_weight(dev, napi, poll, weight);
2545 netif_napi_add_tx_weight(struct net_device *dev,
2546 struct napi_struct *napi,
2547 int (*poll)(struct napi_struct *, int),
2550 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2551 netif_napi_add_weight(dev, napi, poll, weight);
2554 #define netif_tx_napi_add netif_napi_add_tx_weight
2557 * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2558 * @dev: network device
2559 * @napi: NAPI context
2560 * @poll: polling function
2562 * This variant of netif_napi_add() should be used from drivers using NAPI
2563 * to exclusively poll a TX queue.
2564 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2566 static inline void netif_napi_add_tx(struct net_device *dev,
2567 struct napi_struct *napi,
2568 int (*poll)(struct napi_struct *, int))
2570 netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2574 * __netif_napi_del - remove a NAPI context
2575 * @napi: NAPI context
2577 * Warning: caller must observe RCU grace period before freeing memory
2578 * containing @napi. Drivers might want to call this helper to combine
2579 * all the needed RCU grace periods into a single one.
2581 void __netif_napi_del(struct napi_struct *napi);
2584 * netif_napi_del - remove a NAPI context
2585 * @napi: NAPI context
2587 * netif_napi_del() removes a NAPI context from the network device NAPI list
2589 static inline void netif_napi_del(struct napi_struct *napi)
2591 __netif_napi_del(napi);
2595 struct packet_type {
2596 __be16 type; /* This is really htons(ether_type). */
2597 bool ignore_outgoing;
2598 struct net_device *dev; /* NULL is wildcarded here */
2599 netdevice_tracker dev_tracker;
2600 int (*func) (struct sk_buff *,
2601 struct net_device *,
2602 struct packet_type *,
2603 struct net_device *);
2604 void (*list_func) (struct list_head *,
2605 struct packet_type *,
2606 struct net_device *);
2607 bool (*id_match)(struct packet_type *ptype,
2609 struct net *af_packet_net;
2610 void *af_packet_priv;
2611 struct list_head list;
2614 struct offload_callbacks {
2615 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2616 netdev_features_t features);
2617 struct sk_buff *(*gro_receive)(struct list_head *head,
2618 struct sk_buff *skb);
2619 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2622 struct packet_offload {
2623 __be16 type; /* This is really htons(ether_type). */
2625 struct offload_callbacks callbacks;
2626 struct list_head list;
2629 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2630 struct pcpu_sw_netstats {
2635 struct u64_stats_sync syncp;
2636 } __aligned(4 * sizeof(u64));
2638 struct pcpu_lstats {
2639 u64_stats_t packets;
2641 struct u64_stats_sync syncp;
2642 } __aligned(2 * sizeof(u64));
2644 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2646 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2648 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2650 u64_stats_update_begin(&tstats->syncp);
2651 tstats->rx_bytes += len;
2652 tstats->rx_packets++;
2653 u64_stats_update_end(&tstats->syncp);
2656 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2657 unsigned int packets,
2660 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2662 u64_stats_update_begin(&tstats->syncp);
2663 tstats->tx_bytes += len;
2664 tstats->tx_packets += packets;
2665 u64_stats_update_end(&tstats->syncp);
2668 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2670 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2672 u64_stats_update_begin(&lstats->syncp);
2673 u64_stats_add(&lstats->bytes, len);
2674 u64_stats_inc(&lstats->packets);
2675 u64_stats_update_end(&lstats->syncp);
2678 #define __netdev_alloc_pcpu_stats(type, gfp) \
2680 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2683 for_each_possible_cpu(__cpu) { \
2684 typeof(type) *stat; \
2685 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2686 u64_stats_init(&stat->syncp); \
2692 #define netdev_alloc_pcpu_stats(type) \
2693 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2695 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2697 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2700 for_each_possible_cpu(__cpu) { \
2701 typeof(type) *stat; \
2702 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2703 u64_stats_init(&stat->syncp); \
2709 enum netdev_lag_tx_type {
2710 NETDEV_LAG_TX_TYPE_UNKNOWN,
2711 NETDEV_LAG_TX_TYPE_RANDOM,
2712 NETDEV_LAG_TX_TYPE_BROADCAST,
2713 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2714 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2715 NETDEV_LAG_TX_TYPE_HASH,
2718 enum netdev_lag_hash {
2719 NETDEV_LAG_HASH_NONE,
2721 NETDEV_LAG_HASH_L34,
2722 NETDEV_LAG_HASH_L23,
2723 NETDEV_LAG_HASH_E23,
2724 NETDEV_LAG_HASH_E34,
2725 NETDEV_LAG_HASH_VLAN_SRCMAC,
2726 NETDEV_LAG_HASH_UNKNOWN,
2729 struct netdev_lag_upper_info {
2730 enum netdev_lag_tx_type tx_type;
2731 enum netdev_lag_hash hash_type;
2734 struct netdev_lag_lower_state_info {
2739 #include <linux/notifier.h>
2741 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2742 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2746 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2748 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2749 detected a hardware crash and restarted
2750 - we can use this eg to kick tcp sessions
2752 NETDEV_CHANGE, /* Notify device state change */
2755 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2756 NETDEV_CHANGEADDR, /* notify after the address change */
2757 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2761 NETDEV_BONDING_FAILOVER,
2763 NETDEV_PRE_TYPE_CHANGE,
2764 NETDEV_POST_TYPE_CHANGE,
2767 NETDEV_NOTIFY_PEERS,
2771 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2772 NETDEV_CHANGEINFODATA,
2773 NETDEV_BONDING_INFO,
2774 NETDEV_PRECHANGEUPPER,
2775 NETDEV_CHANGELOWERSTATE,
2776 NETDEV_UDP_TUNNEL_PUSH_INFO,
2777 NETDEV_UDP_TUNNEL_DROP_INFO,
2778 NETDEV_CHANGE_TX_QUEUE_LEN,
2779 NETDEV_CVLAN_FILTER_PUSH_INFO,
2780 NETDEV_CVLAN_FILTER_DROP_INFO,
2781 NETDEV_SVLAN_FILTER_PUSH_INFO,
2782 NETDEV_SVLAN_FILTER_DROP_INFO,
2783 NETDEV_OFFLOAD_XSTATS_ENABLE,
2784 NETDEV_OFFLOAD_XSTATS_DISABLE,
2785 NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2786 NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2788 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2790 int register_netdevice_notifier(struct notifier_block *nb);
2791 int unregister_netdevice_notifier(struct notifier_block *nb);
2792 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2793 int unregister_netdevice_notifier_net(struct net *net,
2794 struct notifier_block *nb);
2795 int register_netdevice_notifier_dev_net(struct net_device *dev,
2796 struct notifier_block *nb,
2797 struct netdev_net_notifier *nn);
2798 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2799 struct notifier_block *nb,
2800 struct netdev_net_notifier *nn);
2802 struct netdev_notifier_info {
2803 struct net_device *dev;
2804 struct netlink_ext_ack *extack;
2807 struct netdev_notifier_info_ext {
2808 struct netdev_notifier_info info; /* must be first */
2814 struct netdev_notifier_change_info {
2815 struct netdev_notifier_info info; /* must be first */
2816 unsigned int flags_changed;
2819 struct netdev_notifier_changeupper_info {
2820 struct netdev_notifier_info info; /* must be first */
2821 struct net_device *upper_dev; /* new upper dev */
2822 bool master; /* is upper dev master */
2823 bool linking; /* is the notification for link or unlink */
2824 void *upper_info; /* upper dev info */
2827 struct netdev_notifier_changelowerstate_info {
2828 struct netdev_notifier_info info; /* must be first */
2829 void *lower_state_info; /* is lower dev state */
2832 struct netdev_notifier_pre_changeaddr_info {
2833 struct netdev_notifier_info info; /* must be first */
2834 const unsigned char *dev_addr;
2837 enum netdev_offload_xstats_type {
2838 NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2841 struct netdev_notifier_offload_xstats_info {
2842 struct netdev_notifier_info info; /* must be first */
2843 enum netdev_offload_xstats_type type;
2846 /* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2847 struct netdev_notifier_offload_xstats_rd *report_delta;
2848 /* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2849 struct netdev_notifier_offload_xstats_ru *report_used;
2853 int netdev_offload_xstats_enable(struct net_device *dev,
2854 enum netdev_offload_xstats_type type,
2855 struct netlink_ext_ack *extack);
2856 int netdev_offload_xstats_disable(struct net_device *dev,
2857 enum netdev_offload_xstats_type type);
2858 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2859 enum netdev_offload_xstats_type type);
2860 int netdev_offload_xstats_get(struct net_device *dev,
2861 enum netdev_offload_xstats_type type,
2862 struct rtnl_hw_stats64 *stats, bool *used,
2863 struct netlink_ext_ack *extack);
2865 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2866 const struct rtnl_hw_stats64 *stats);
2868 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2869 void netdev_offload_xstats_push_delta(struct net_device *dev,
2870 enum netdev_offload_xstats_type type,
2871 const struct rtnl_hw_stats64 *stats);
2873 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2874 struct net_device *dev)
2877 info->extack = NULL;
2880 static inline struct net_device *
2881 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2886 static inline struct netlink_ext_ack *
2887 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2889 return info->extack;
2892 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2895 extern rwlock_t dev_base_lock; /* Device list lock */
2897 #define for_each_netdev(net, d) \
2898 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2899 #define for_each_netdev_reverse(net, d) \
2900 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2901 #define for_each_netdev_rcu(net, d) \
2902 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2903 #define for_each_netdev_safe(net, d, n) \
2904 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2905 #define for_each_netdev_continue(net, d) \
2906 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2907 #define for_each_netdev_continue_reverse(net, d) \
2908 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2910 #define for_each_netdev_continue_rcu(net, d) \
2911 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2912 #define for_each_netdev_in_bond_rcu(bond, slave) \
2913 for_each_netdev_rcu(&init_net, slave) \
2914 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2915 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2917 static inline struct net_device *next_net_device(struct net_device *dev)
2919 struct list_head *lh;
2923 lh = dev->dev_list.next;
2924 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2927 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2929 struct list_head *lh;
2933 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2934 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2937 static inline struct net_device *first_net_device(struct net *net)
2939 return list_empty(&net->dev_base_head) ? NULL :
2940 net_device_entry(net->dev_base_head.next);
2943 static inline struct net_device *first_net_device_rcu(struct net *net)
2945 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2947 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2950 int netdev_boot_setup_check(struct net_device *dev);
2951 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2952 const char *hwaddr);
2953 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2954 void dev_add_pack(struct packet_type *pt);
2955 void dev_remove_pack(struct packet_type *pt);
2956 void __dev_remove_pack(struct packet_type *pt);
2957 void dev_add_offload(struct packet_offload *po);
2958 void dev_remove_offload(struct packet_offload *po);
2960 int dev_get_iflink(const struct net_device *dev);
2961 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2962 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2963 struct net_device_path_stack *stack);
2964 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2965 unsigned short mask);
2966 struct net_device *dev_get_by_name(struct net *net, const char *name);
2967 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2968 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2969 bool netdev_name_in_use(struct net *net, const char *name);
2970 int dev_alloc_name(struct net_device *dev, const char *name);
2971 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2972 void dev_close(struct net_device *dev);
2973 void dev_close_many(struct list_head *head, bool unlink);
2974 void dev_disable_lro(struct net_device *dev);
2975 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2976 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2977 struct net_device *sb_dev);
2978 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2979 struct net_device *sb_dev);
2981 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
2982 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2984 static inline int dev_queue_xmit(struct sk_buff *skb)
2986 return __dev_queue_xmit(skb, NULL);
2989 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
2990 struct net_device *sb_dev)
2992 return __dev_queue_xmit(skb, sb_dev);
2995 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2999 ret = __dev_direct_xmit(skb, queue_id);
3000 if (!dev_xmit_complete(ret))
3005 int register_netdevice(struct net_device *dev);
3006 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3007 void unregister_netdevice_many(struct list_head *head);
3008 static inline void unregister_netdevice(struct net_device *dev)
3010 unregister_netdevice_queue(dev, NULL);
3013 int netdev_refcnt_read(const struct net_device *dev);
3014 void free_netdev(struct net_device *dev);
3015 void netdev_freemem(struct net_device *dev);
3016 int init_dummy_netdev(struct net_device *dev);
3018 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3019 struct sk_buff *skb,
3021 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3023 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3024 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3025 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3026 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3027 int dev_restart(struct net_device *dev);
3030 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3031 unsigned short type,
3032 const void *daddr, const void *saddr,
3035 if (!dev->header_ops || !dev->header_ops->create)
3038 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3041 static inline int dev_parse_header(const struct sk_buff *skb,
3042 unsigned char *haddr)
3044 const struct net_device *dev = skb->dev;
3046 if (!dev->header_ops || !dev->header_ops->parse)
3048 return dev->header_ops->parse(skb, haddr);
3051 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3053 const struct net_device *dev = skb->dev;
3055 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3057 return dev->header_ops->parse_protocol(skb);
3060 /* ll_header must have at least hard_header_len allocated */
3061 static inline bool dev_validate_header(const struct net_device *dev,
3062 char *ll_header, int len)
3064 if (likely(len >= dev->hard_header_len))
3066 if (len < dev->min_header_len)
3069 if (capable(CAP_SYS_RAWIO)) {
3070 memset(ll_header + len, 0, dev->hard_header_len - len);
3074 if (dev->header_ops && dev->header_ops->validate)
3075 return dev->header_ops->validate(ll_header, len);
3080 static inline bool dev_has_header(const struct net_device *dev)
3082 return dev->header_ops && dev->header_ops->create;
3086 * Incoming packets are placed on per-CPU queues
3088 struct softnet_data {
3089 struct list_head poll_list;
3090 struct sk_buff_head process_queue;
3093 unsigned int processed;
3094 unsigned int time_squeeze;
3095 unsigned int received_rps;
3097 struct softnet_data *rps_ipi_list;
3099 #ifdef CONFIG_NET_FLOW_LIMIT
3100 struct sd_flow_limit __rcu *flow_limit;
3102 struct Qdisc *output_queue;
3103 struct Qdisc **output_queue_tailp;
3104 struct sk_buff *completion_queue;
3105 #ifdef CONFIG_XFRM_OFFLOAD
3106 struct sk_buff_head xfrm_backlog;
3108 /* written and read only by owning cpu: */
3112 #ifdef CONFIG_NET_EGRESS
3117 /* input_queue_head should be written by cpu owning this struct,
3118 * and only read by other cpus. Worth using a cache line.
3120 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3122 /* Elements below can be accessed between CPUs for RPS/RFS */
3123 call_single_data_t csd ____cacheline_aligned_in_smp;
3124 struct softnet_data *rps_ipi_next;
3126 unsigned int input_queue_tail;
3128 unsigned int dropped;
3129 struct sk_buff_head input_pkt_queue;
3130 struct napi_struct backlog;
3132 /* Another possibly contended cache line */
3133 spinlock_t defer_lock ____cacheline_aligned_in_smp;
3135 struct sk_buff *defer_list;
3136 call_single_data_t defer_csd;
3139 static inline void input_queue_head_incr(struct softnet_data *sd)
3142 sd->input_queue_head++;
3146 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3147 unsigned int *qtail)
3150 *qtail = ++sd->input_queue_tail;
3154 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3156 static inline int dev_recursion_level(void)
3158 return this_cpu_read(softnet_data.xmit.recursion);
3161 #define XMIT_RECURSION_LIMIT 8
3162 static inline bool dev_xmit_recursion(void)
3164 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3165 XMIT_RECURSION_LIMIT);
3168 static inline void dev_xmit_recursion_inc(void)
3170 __this_cpu_inc(softnet_data.xmit.recursion);
3173 static inline void dev_xmit_recursion_dec(void)
3175 __this_cpu_dec(softnet_data.xmit.recursion);
3178 void __netif_schedule(struct Qdisc *q);
3179 void netif_schedule_queue(struct netdev_queue *txq);
3181 static inline void netif_tx_schedule_all(struct net_device *dev)
3185 for (i = 0; i < dev->num_tx_queues; i++)
3186 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3189 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3191 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3195 * netif_start_queue - allow transmit
3196 * @dev: network device
3198 * Allow upper layers to call the device hard_start_xmit routine.
3200 static inline void netif_start_queue(struct net_device *dev)
3202 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3205 static inline void netif_tx_start_all_queues(struct net_device *dev)
3209 for (i = 0; i < dev->num_tx_queues; i++) {
3210 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3211 netif_tx_start_queue(txq);
3215 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3218 * netif_wake_queue - restart transmit
3219 * @dev: network device
3221 * Allow upper layers to call the device hard_start_xmit routine.
3222 * Used for flow control when transmit resources are available.
3224 static inline void netif_wake_queue(struct net_device *dev)
3226 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3229 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3233 for (i = 0; i < dev->num_tx_queues; i++) {
3234 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3235 netif_tx_wake_queue(txq);
3239 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3241 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3245 * netif_stop_queue - stop transmitted packets
3246 * @dev: network device
3248 * Stop upper layers calling the device hard_start_xmit routine.
3249 * Used for flow control when transmit resources are unavailable.
3251 static inline void netif_stop_queue(struct net_device *dev)
3253 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3256 void netif_tx_stop_all_queues(struct net_device *dev);
3258 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3260 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3264 * netif_queue_stopped - test if transmit queue is flowblocked
3265 * @dev: network device
3267 * Test if transmit queue on device is currently unable to send.
3269 static inline bool netif_queue_stopped(const struct net_device *dev)
3271 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3274 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3276 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3280 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3282 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3286 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3288 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3292 * netdev_queue_set_dql_min_limit - set dql minimum limit
3293 * @dev_queue: pointer to transmit queue
3294 * @min_limit: dql minimum limit
3296 * Forces xmit_more() to return true until the minimum threshold
3297 * defined by @min_limit is reached (or until the tx queue is
3298 * empty). Warning: to be use with care, misuse will impact the
3301 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3302 unsigned int min_limit)
3305 dev_queue->dql.min_limit = min_limit;
3310 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3311 * @dev_queue: pointer to transmit queue
3313 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3314 * to give appropriate hint to the CPU.
3316 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3319 prefetchw(&dev_queue->dql.num_queued);
3324 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3325 * @dev_queue: pointer to transmit queue
3327 * BQL enabled drivers might use this helper in their TX completion path,
3328 * to give appropriate hint to the CPU.
3330 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3333 prefetchw(&dev_queue->dql.limit);
3337 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3341 dql_queued(&dev_queue->dql, bytes);
3343 if (likely(dql_avail(&dev_queue->dql) >= 0))
3346 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3349 * The XOFF flag must be set before checking the dql_avail below,
3350 * because in netdev_tx_completed_queue we update the dql_completed
3351 * before checking the XOFF flag.
3355 /* check again in case another CPU has just made room avail */
3356 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3357 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3361 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3362 * that they should not test BQL status themselves.
3363 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3365 * Returns true if the doorbell must be used to kick the NIC.
3367 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3373 dql_queued(&dev_queue->dql, bytes);
3375 return netif_tx_queue_stopped(dev_queue);
3377 netdev_tx_sent_queue(dev_queue, bytes);
3382 * netdev_sent_queue - report the number of bytes queued to hardware
3383 * @dev: network device
3384 * @bytes: number of bytes queued to the hardware device queue
3386 * Report the number of bytes queued for sending/completion to the network
3387 * device hardware queue. @bytes should be a good approximation and should
3388 * exactly match netdev_completed_queue() @bytes
3390 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3392 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3395 static inline bool __netdev_sent_queue(struct net_device *dev,
3399 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3403 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3404 unsigned int pkts, unsigned int bytes)
3407 if (unlikely(!bytes))
3410 dql_completed(&dev_queue->dql, bytes);
3413 * Without the memory barrier there is a small possiblity that
3414 * netdev_tx_sent_queue will miss the update and cause the queue to
3415 * be stopped forever
3419 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3422 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3423 netif_schedule_queue(dev_queue);
3428 * netdev_completed_queue - report bytes and packets completed by device
3429 * @dev: network device
3430 * @pkts: actual number of packets sent over the medium
3431 * @bytes: actual number of bytes sent over the medium
3433 * Report the number of bytes and packets transmitted by the network device
3434 * hardware queue over the physical medium, @bytes must exactly match the
3435 * @bytes amount passed to netdev_sent_queue()
3437 static inline void netdev_completed_queue(struct net_device *dev,
3438 unsigned int pkts, unsigned int bytes)
3440 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3443 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3446 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3452 * netdev_reset_queue - reset the packets and bytes count of a network device
3453 * @dev_queue: network device
3455 * Reset the bytes and packet count of a network device and clear the
3456 * software flow control OFF bit for this network device
3458 static inline void netdev_reset_queue(struct net_device *dev_queue)
3460 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3464 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3465 * @dev: network device
3466 * @queue_index: given tx queue index
3468 * Returns 0 if given tx queue index >= number of device tx queues,
3469 * otherwise returns the originally passed tx queue index.
3471 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3473 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3474 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3475 dev->name, queue_index,
3476 dev->real_num_tx_queues);
3484 * netif_running - test if up
3485 * @dev: network device
3487 * Test if the device has been brought up.
3489 static inline bool netif_running(const struct net_device *dev)
3491 return test_bit(__LINK_STATE_START, &dev->state);
3495 * Routines to manage the subqueues on a device. We only need start,
3496 * stop, and a check if it's stopped. All other device management is
3497 * done at the overall netdevice level.
3498 * Also test the device if we're multiqueue.
3502 * netif_start_subqueue - allow sending packets on subqueue
3503 * @dev: network device
3504 * @queue_index: sub queue index
3506 * Start individual transmit queue of a device with multiple transmit queues.
3508 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3510 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3512 netif_tx_start_queue(txq);
3516 * netif_stop_subqueue - stop sending packets on subqueue
3517 * @dev: network device
3518 * @queue_index: sub queue index
3520 * Stop individual transmit queue of a device with multiple transmit queues.
3522 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3524 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3525 netif_tx_stop_queue(txq);
3529 * __netif_subqueue_stopped - test status of subqueue
3530 * @dev: network device
3531 * @queue_index: sub queue index
3533 * Check individual transmit queue of a device with multiple transmit queues.
3535 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3538 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3540 return netif_tx_queue_stopped(txq);
3544 * netif_subqueue_stopped - test status of subqueue
3545 * @dev: network device
3546 * @skb: sub queue buffer pointer
3548 * Check individual transmit queue of a device with multiple transmit queues.
3550 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3551 struct sk_buff *skb)
3553 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3557 * netif_wake_subqueue - allow sending packets on subqueue
3558 * @dev: network device
3559 * @queue_index: sub queue index
3561 * Resume individual transmit queue of a device with multiple transmit queues.
3563 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3565 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3567 netif_tx_wake_queue(txq);
3571 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3573 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3574 u16 index, enum xps_map_type type);
3577 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3578 * @j: CPU/Rx queue index
3579 * @mask: bitmask of all cpus/rx queues
3580 * @nr_bits: number of bits in the bitmask
3582 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3584 static inline bool netif_attr_test_mask(unsigned long j,
3585 const unsigned long *mask,
3586 unsigned int nr_bits)
3588 cpu_max_bits_warn(j, nr_bits);
3589 return test_bit(j, mask);
3593 * netif_attr_test_online - Test for online CPU/Rx queue
3594 * @j: CPU/Rx queue index
3595 * @online_mask: bitmask for CPUs/Rx queues that are online
3596 * @nr_bits: number of bits in the bitmask
3598 * Returns true if a CPU/Rx queue is online.
3600 static inline bool netif_attr_test_online(unsigned long j,
3601 const unsigned long *online_mask,
3602 unsigned int nr_bits)
3604 cpu_max_bits_warn(j, nr_bits);
3607 return test_bit(j, online_mask);
3609 return (j < nr_bits);
3613 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3614 * @n: CPU/Rx queue index
3615 * @srcp: the cpumask/Rx queue mask pointer
3616 * @nr_bits: number of bits in the bitmask
3618 * Returns >= nr_bits if no further CPUs/Rx queues set.
3620 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3621 unsigned int nr_bits)
3623 /* -1 is a legal arg here. */
3625 cpu_max_bits_warn(n, nr_bits);
3628 return find_next_bit(srcp, nr_bits, n + 1);
3634 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3635 * @n: CPU/Rx queue index
3636 * @src1p: the first CPUs/Rx queues mask pointer
3637 * @src2p: the second CPUs/Rx queues mask pointer
3638 * @nr_bits: number of bits in the bitmask
3640 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3642 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3643 const unsigned long *src2p,
3644 unsigned int nr_bits)
3646 /* -1 is a legal arg here. */
3648 cpu_max_bits_warn(n, nr_bits);
3651 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3653 return find_next_bit(src1p, nr_bits, n + 1);
3655 return find_next_bit(src2p, nr_bits, n + 1);
3660 static inline int netif_set_xps_queue(struct net_device *dev,
3661 const struct cpumask *mask,
3667 static inline int __netif_set_xps_queue(struct net_device *dev,
3668 const unsigned long *mask,
3669 u16 index, enum xps_map_type type)
3676 * netif_is_multiqueue - test if device has multiple transmit queues
3677 * @dev: network device
3679 * Check if device has multiple transmit queues
3681 static inline bool netif_is_multiqueue(const struct net_device *dev)
3683 return dev->num_tx_queues > 1;
3686 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3689 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3691 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3694 dev->real_num_rx_queues = rxqs;
3698 int netif_set_real_num_queues(struct net_device *dev,
3699 unsigned int txq, unsigned int rxq);
3701 static inline struct netdev_rx_queue *
3702 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3704 return dev->_rx + rxq;
3708 static inline unsigned int get_netdev_rx_queue_index(
3709 struct netdev_rx_queue *queue)
3711 struct net_device *dev = queue->dev;
3712 int index = queue - dev->_rx;
3714 BUG_ON(index >= dev->num_rx_queues);
3719 int netif_get_num_default_rss_queues(void);
3721 enum skb_free_reason {
3722 SKB_REASON_CONSUMED,
3726 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3727 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3730 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3731 * interrupt context or with hardware interrupts being disabled.
3732 * (in_hardirq() || irqs_disabled())
3734 * We provide four helpers that can be used in following contexts :
3736 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3737 * replacing kfree_skb(skb)
3739 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3740 * Typically used in place of consume_skb(skb) in TX completion path
3742 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3743 * replacing kfree_skb(skb)
3745 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3746 * and consumed a packet. Used in place of consume_skb(skb)
3748 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3750 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3753 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3755 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3758 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3760 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3763 static inline void dev_consume_skb_any(struct sk_buff *skb)
3765 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3768 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3769 struct bpf_prog *xdp_prog);
3770 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3771 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3772 int netif_rx(struct sk_buff *skb);
3773 int __netif_rx(struct sk_buff *skb);
3775 int netif_receive_skb(struct sk_buff *skb);
3776 int netif_receive_skb_core(struct sk_buff *skb);
3777 void netif_receive_skb_list_internal(struct list_head *head);
3778 void netif_receive_skb_list(struct list_head *head);
3779 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3780 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3781 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3782 gro_result_t napi_gro_frags(struct napi_struct *napi);
3783 struct packet_offload *gro_find_receive_by_type(__be16 type);
3784 struct packet_offload *gro_find_complete_by_type(__be16 type);
3786 static inline void napi_free_frags(struct napi_struct *napi)
3788 kfree_skb(napi->skb);
3792 bool netdev_is_rx_handler_busy(struct net_device *dev);
3793 int netdev_rx_handler_register(struct net_device *dev,
3794 rx_handler_func_t *rx_handler,
3795 void *rx_handler_data);
3796 void netdev_rx_handler_unregister(struct net_device *dev);
3798 bool dev_valid_name(const char *name);
3799 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3801 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3803 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3804 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3805 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3806 void __user *data, bool *need_copyout);
3807 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3808 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3809 unsigned int dev_get_flags(const struct net_device *);
3810 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3811 struct netlink_ext_ack *extack);
3812 int dev_change_flags(struct net_device *dev, unsigned int flags,
3813 struct netlink_ext_ack *extack);
3814 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3815 unsigned int gchanges);
3816 int dev_set_alias(struct net_device *, const char *, size_t);
3817 int dev_get_alias(const struct net_device *, char *, size_t);
3818 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3819 const char *pat, int new_ifindex);
3821 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3824 return __dev_change_net_namespace(dev, net, pat, 0);
3826 int __dev_set_mtu(struct net_device *, int);
3827 int dev_set_mtu(struct net_device *, int);
3828 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3829 struct netlink_ext_ack *extack);
3830 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3831 struct netlink_ext_ack *extack);
3832 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3833 struct netlink_ext_ack *extack);
3834 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3835 int dev_get_port_parent_id(struct net_device *dev,
3836 struct netdev_phys_item_id *ppid, bool recurse);
3837 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3838 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3839 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3840 struct netdev_queue *txq, int *ret);
3842 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3843 u8 dev_xdp_prog_count(struct net_device *dev);
3844 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3846 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3847 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3848 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3849 bool is_skb_forwardable(const struct net_device *dev,
3850 const struct sk_buff *skb);
3852 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3853 const struct sk_buff *skb,
3854 const bool check_mtu)
3856 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3859 if (!(dev->flags & IFF_UP))
3865 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3866 if (skb->len <= len)
3869 /* if TSO is enabled, we don't care about the length as the packet
3870 * could be forwarded without being segmented before
3872 if (skb_is_gso(skb))
3878 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
3880 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
3882 /* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
3883 struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
3888 return netdev_core_stats_alloc(dev);
3891 #define DEV_CORE_STATS_INC(FIELD) \
3892 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev) \
3894 struct net_device_core_stats __percpu *p; \
3896 p = dev_core_stats(dev); \
3898 this_cpu_inc(p->FIELD); \
3900 DEV_CORE_STATS_INC(rx_dropped)
3901 DEV_CORE_STATS_INC(tx_dropped)
3902 DEV_CORE_STATS_INC(rx_nohandler)
3903 DEV_CORE_STATS_INC(rx_otherhost_dropped)
3905 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3906 struct sk_buff *skb,
3907 const bool check_mtu)
3909 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3910 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
3911 dev_core_stats_rx_dropped_inc(dev);
3916 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
3921 bool dev_nit_active(struct net_device *dev);
3922 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3924 static inline void __dev_put(struct net_device *dev)
3927 #ifdef CONFIG_PCPU_DEV_REFCNT
3928 this_cpu_dec(*dev->pcpu_refcnt);
3930 refcount_dec(&dev->dev_refcnt);
3935 static inline void __dev_hold(struct net_device *dev)
3938 #ifdef CONFIG_PCPU_DEV_REFCNT
3939 this_cpu_inc(*dev->pcpu_refcnt);
3941 refcount_inc(&dev->dev_refcnt);
3946 static inline void __netdev_tracker_alloc(struct net_device *dev,
3947 netdevice_tracker *tracker,
3950 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3951 ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
3955 /* netdev_tracker_alloc() can upgrade a prior untracked reference
3956 * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
3958 static inline void netdev_tracker_alloc(struct net_device *dev,
3959 netdevice_tracker *tracker, gfp_t gfp)
3961 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3962 refcount_dec(&dev->refcnt_tracker.no_tracker);
3963 __netdev_tracker_alloc(dev, tracker, gfp);
3967 static inline void netdev_tracker_free(struct net_device *dev,
3968 netdevice_tracker *tracker)
3970 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
3971 ref_tracker_free(&dev->refcnt_tracker, tracker);
3975 static inline void dev_hold_track(struct net_device *dev,
3976 netdevice_tracker *tracker, gfp_t gfp)
3980 __netdev_tracker_alloc(dev, tracker, gfp);
3984 static inline void dev_put_track(struct net_device *dev,
3985 netdevice_tracker *tracker)
3988 netdev_tracker_free(dev, tracker);
3994 * dev_hold - get reference to device
3995 * @dev: network device
3997 * Hold reference to device to keep it from being freed.
3998 * Try using dev_hold_track() instead.
4000 static inline void dev_hold(struct net_device *dev)
4002 dev_hold_track(dev, NULL, GFP_ATOMIC);
4006 * dev_put - release reference to device
4007 * @dev: network device
4009 * Release reference to device to allow it to be freed.
4010 * Try using dev_put_track() instead.
4012 static inline void dev_put(struct net_device *dev)
4014 dev_put_track(dev, NULL);
4017 static inline void dev_replace_track(struct net_device *odev,
4018 struct net_device *ndev,
4019 netdevice_tracker *tracker,
4023 netdev_tracker_free(odev, tracker);
4029 __netdev_tracker_alloc(ndev, tracker, gfp);
4032 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4033 * and _off may be called from IRQ context, but it is caller
4034 * who is responsible for serialization of these calls.
4036 * The name carrier is inappropriate, these functions should really be
4037 * called netif_lowerlayer_*() because they represent the state of any
4038 * kind of lower layer not just hardware media.
4040 void linkwatch_fire_event(struct net_device *dev);
4043 * netif_carrier_ok - test if carrier present
4044 * @dev: network device
4046 * Check if carrier is present on device
4048 static inline bool netif_carrier_ok(const struct net_device *dev)
4050 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4053 unsigned long dev_trans_start(struct net_device *dev);
4055 void __netdev_watchdog_up(struct net_device *dev);
4057 void netif_carrier_on(struct net_device *dev);
4058 void netif_carrier_off(struct net_device *dev);
4059 void netif_carrier_event(struct net_device *dev);
4062 * netif_dormant_on - mark device as dormant.
4063 * @dev: network device
4065 * Mark device as dormant (as per RFC2863).
4067 * The dormant state indicates that the relevant interface is not
4068 * actually in a condition to pass packets (i.e., it is not 'up') but is
4069 * in a "pending" state, waiting for some external event. For "on-
4070 * demand" interfaces, this new state identifies the situation where the
4071 * interface is waiting for events to place it in the up state.
4073 static inline void netif_dormant_on(struct net_device *dev)
4075 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4076 linkwatch_fire_event(dev);
4080 * netif_dormant_off - set device as not dormant.
4081 * @dev: network device
4083 * Device is not in dormant state.
4085 static inline void netif_dormant_off(struct net_device *dev)
4087 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4088 linkwatch_fire_event(dev);
4092 * netif_dormant - test if device is dormant
4093 * @dev: network device
4095 * Check if device is dormant.
4097 static inline bool netif_dormant(const struct net_device *dev)
4099 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4104 * netif_testing_on - mark device as under test.
4105 * @dev: network device
4107 * Mark device as under test (as per RFC2863).
4109 * The testing state indicates that some test(s) must be performed on
4110 * the interface. After completion, of the test, the interface state
4111 * will change to up, dormant, or down, as appropriate.
4113 static inline void netif_testing_on(struct net_device *dev)
4115 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4116 linkwatch_fire_event(dev);
4120 * netif_testing_off - set device as not under test.
4121 * @dev: network device
4123 * Device is not in testing state.
4125 static inline void netif_testing_off(struct net_device *dev)
4127 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4128 linkwatch_fire_event(dev);
4132 * netif_testing - test if device is under test
4133 * @dev: network device
4135 * Check if device is under test
4137 static inline bool netif_testing(const struct net_device *dev)
4139 return test_bit(__LINK_STATE_TESTING, &dev->state);
4144 * netif_oper_up - test if device is operational
4145 * @dev: network device
4147 * Check if carrier is operational
4149 static inline bool netif_oper_up(const struct net_device *dev)
4151 return (dev->operstate == IF_OPER_UP ||
4152 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4156 * netif_device_present - is device available or removed
4157 * @dev: network device
4159 * Check if device has not been removed from system.
4161 static inline bool netif_device_present(const struct net_device *dev)
4163 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4166 void netif_device_detach(struct net_device *dev);
4168 void netif_device_attach(struct net_device *dev);
4171 * Network interface message level settings
4176 NETIF_MSG_PROBE_BIT,
4178 NETIF_MSG_TIMER_BIT,
4179 NETIF_MSG_IFDOWN_BIT,
4181 NETIF_MSG_RX_ERR_BIT,
4182 NETIF_MSG_TX_ERR_BIT,
4183 NETIF_MSG_TX_QUEUED_BIT,
4185 NETIF_MSG_TX_DONE_BIT,
4186 NETIF_MSG_RX_STATUS_BIT,
4187 NETIF_MSG_PKTDATA_BIT,
4191 /* When you add a new bit above, update netif_msg_class_names array
4192 * in net/ethtool/common.c
4194 NETIF_MSG_CLASS_COUNT,
4196 /* Both ethtool_ops interface and internal driver implementation use u32 */
4197 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4199 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4200 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4202 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4203 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4204 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4205 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4206 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4207 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4208 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4209 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4210 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4211 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4212 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4213 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4214 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4215 #define NETIF_MSG_HW __NETIF_MSG(HW)
4216 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4218 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4219 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4220 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4221 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4222 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4223 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4224 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4225 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4226 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4227 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4228 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4229 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4230 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4231 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4232 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4234 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4237 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4238 return default_msg_enable_bits;
4239 if (debug_value == 0) /* no output */
4241 /* set low N bits */
4242 return (1U << debug_value) - 1;
4245 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4247 spin_lock(&txq->_xmit_lock);
4248 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4249 WRITE_ONCE(txq->xmit_lock_owner, cpu);
4252 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4254 __acquire(&txq->_xmit_lock);
4258 static inline void __netif_tx_release(struct netdev_queue *txq)
4260 __release(&txq->_xmit_lock);
4263 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4265 spin_lock_bh(&txq->_xmit_lock);
4266 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4267 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4270 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4272 bool ok = spin_trylock(&txq->_xmit_lock);
4275 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4276 WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4281 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4283 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4284 WRITE_ONCE(txq->xmit_lock_owner, -1);
4285 spin_unlock(&txq->_xmit_lock);
4288 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4290 /* Pairs with READ_ONCE() in __dev_queue_xmit() */
4291 WRITE_ONCE(txq->xmit_lock_owner, -1);
4292 spin_unlock_bh(&txq->_xmit_lock);
4296 * txq->trans_start can be read locklessly from dev_watchdog()
4298 static inline void txq_trans_update(struct netdev_queue *txq)
4300 if (txq->xmit_lock_owner != -1)
4301 WRITE_ONCE(txq->trans_start, jiffies);
4304 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4306 unsigned long now = jiffies;
4308 if (READ_ONCE(txq->trans_start) != now)
4309 WRITE_ONCE(txq->trans_start, now);
4312 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4313 static inline void netif_trans_update(struct net_device *dev)
4315 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4317 txq_trans_cond_update(txq);
4321 * netif_tx_lock - grab network device transmit lock
4322 * @dev: network device
4324 * Get network device transmit lock
4326 void netif_tx_lock(struct net_device *dev);
4328 static inline void netif_tx_lock_bh(struct net_device *dev)
4334 void netif_tx_unlock(struct net_device *dev);
4336 static inline void netif_tx_unlock_bh(struct net_device *dev)
4338 netif_tx_unlock(dev);
4342 #define HARD_TX_LOCK(dev, txq, cpu) { \
4343 if ((dev->features & NETIF_F_LLTX) == 0) { \
4344 __netif_tx_lock(txq, cpu); \
4346 __netif_tx_acquire(txq); \
4350 #define HARD_TX_TRYLOCK(dev, txq) \
4351 (((dev->features & NETIF_F_LLTX) == 0) ? \
4352 __netif_tx_trylock(txq) : \
4353 __netif_tx_acquire(txq))
4355 #define HARD_TX_UNLOCK(dev, txq) { \
4356 if ((dev->features & NETIF_F_LLTX) == 0) { \
4357 __netif_tx_unlock(txq); \
4359 __netif_tx_release(txq); \
4363 static inline void netif_tx_disable(struct net_device *dev)
4369 cpu = smp_processor_id();
4370 spin_lock(&dev->tx_global_lock);
4371 for (i = 0; i < dev->num_tx_queues; i++) {
4372 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4374 __netif_tx_lock(txq, cpu);
4375 netif_tx_stop_queue(txq);
4376 __netif_tx_unlock(txq);
4378 spin_unlock(&dev->tx_global_lock);
4382 static inline void netif_addr_lock(struct net_device *dev)
4384 unsigned char nest_level = 0;
4386 #ifdef CONFIG_LOCKDEP
4387 nest_level = dev->nested_level;
4389 spin_lock_nested(&dev->addr_list_lock, nest_level);
4392 static inline void netif_addr_lock_bh(struct net_device *dev)
4394 unsigned char nest_level = 0;
4396 #ifdef CONFIG_LOCKDEP
4397 nest_level = dev->nested_level;
4400 spin_lock_nested(&dev->addr_list_lock, nest_level);
4403 static inline void netif_addr_unlock(struct net_device *dev)
4405 spin_unlock(&dev->addr_list_lock);
4408 static inline void netif_addr_unlock_bh(struct net_device *dev)
4410 spin_unlock_bh(&dev->addr_list_lock);
4414 * dev_addrs walker. Should be used only for read access. Call with
4415 * rcu_read_lock held.
4417 #define for_each_dev_addr(dev, ha) \
4418 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4420 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4422 void ether_setup(struct net_device *dev);
4424 /* Support for loadable net-drivers */
4425 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4426 unsigned char name_assign_type,
4427 void (*setup)(struct net_device *),
4428 unsigned int txqs, unsigned int rxqs);
4429 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4430 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4432 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4433 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4436 int register_netdev(struct net_device *dev);
4437 void unregister_netdev(struct net_device *dev);
4439 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4441 /* General hardware address lists handling functions */
4442 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4443 struct netdev_hw_addr_list *from_list, int addr_len);
4444 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4445 struct netdev_hw_addr_list *from_list, int addr_len);
4446 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4447 struct net_device *dev,
4448 int (*sync)(struct net_device *, const unsigned char *),
4449 int (*unsync)(struct net_device *,
4450 const unsigned char *));
4451 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4452 struct net_device *dev,
4453 int (*sync)(struct net_device *,
4454 const unsigned char *, int),
4455 int (*unsync)(struct net_device *,
4456 const unsigned char *, int));
4457 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4458 struct net_device *dev,
4459 int (*unsync)(struct net_device *,
4460 const unsigned char *, int));
4461 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4462 struct net_device *dev,
4463 int (*unsync)(struct net_device *,
4464 const unsigned char *));
4465 void __hw_addr_init(struct netdev_hw_addr_list *list);
4467 /* Functions used for device addresses handling */
4468 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4469 const void *addr, size_t len);
4472 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4474 dev_addr_mod(dev, 0, addr, len);
4477 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4479 __dev_addr_set(dev, addr, dev->addr_len);
4482 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4483 unsigned char addr_type);
4484 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4485 unsigned char addr_type);
4487 /* Functions used for unicast addresses handling */
4488 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4489 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4490 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4491 int dev_uc_sync(struct net_device *to, struct net_device *from);
4492 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4493 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4494 void dev_uc_flush(struct net_device *dev);
4495 void dev_uc_init(struct net_device *dev);
4498 * __dev_uc_sync - Synchonize device's unicast list
4499 * @dev: device to sync
4500 * @sync: function to call if address should be added
4501 * @unsync: function to call if address should be removed
4503 * Add newly added addresses to the interface, and release
4504 * addresses that have been deleted.
4506 static inline int __dev_uc_sync(struct net_device *dev,
4507 int (*sync)(struct net_device *,
4508 const unsigned char *),
4509 int (*unsync)(struct net_device *,
4510 const unsigned char *))
4512 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4516 * __dev_uc_unsync - Remove synchronized addresses from device
4517 * @dev: device to sync
4518 * @unsync: function to call if address should be removed
4520 * Remove all addresses that were added to the device by dev_uc_sync().
4522 static inline void __dev_uc_unsync(struct net_device *dev,
4523 int (*unsync)(struct net_device *,
4524 const unsigned char *))
4526 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4529 /* Functions used for multicast addresses handling */
4530 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4531 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4532 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4533 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4534 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4535 int dev_mc_sync(struct net_device *to, struct net_device *from);
4536 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4537 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4538 void dev_mc_flush(struct net_device *dev);
4539 void dev_mc_init(struct net_device *dev);
4542 * __dev_mc_sync - Synchonize device's multicast list
4543 * @dev: device to sync
4544 * @sync: function to call if address should be added
4545 * @unsync: function to call if address should be removed
4547 * Add newly added addresses to the interface, and release
4548 * addresses that have been deleted.
4550 static inline int __dev_mc_sync(struct net_device *dev,
4551 int (*sync)(struct net_device *,
4552 const unsigned char *),
4553 int (*unsync)(struct net_device *,
4554 const unsigned char *))
4556 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4560 * __dev_mc_unsync - Remove synchronized addresses from device
4561 * @dev: device to sync
4562 * @unsync: function to call if address should be removed
4564 * Remove all addresses that were added to the device by dev_mc_sync().
4566 static inline void __dev_mc_unsync(struct net_device *dev,
4567 int (*unsync)(struct net_device *,
4568 const unsigned char *))
4570 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4573 /* Functions used for secondary unicast and multicast support */
4574 void dev_set_rx_mode(struct net_device *dev);
4575 int dev_set_promiscuity(struct net_device *dev, int inc);
4576 int dev_set_allmulti(struct net_device *dev, int inc);
4577 void netdev_state_change(struct net_device *dev);
4578 void __netdev_notify_peers(struct net_device *dev);
4579 void netdev_notify_peers(struct net_device *dev);
4580 void netdev_features_change(struct net_device *dev);
4581 /* Load a device via the kmod */
4582 void dev_load(struct net *net, const char *name);
4583 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4584 struct rtnl_link_stats64 *storage);
4585 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4586 const struct net_device_stats *netdev_stats);
4587 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4588 const struct pcpu_sw_netstats __percpu *netstats);
4589 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4591 extern int netdev_max_backlog;
4592 extern int dev_rx_weight;
4593 extern int dev_tx_weight;
4594 extern int gro_normal_batch;
4597 NESTED_SYNC_IMM_BIT,
4598 NESTED_SYNC_TODO_BIT,
4601 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4602 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4604 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4605 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4607 struct netdev_nested_priv {
4608 unsigned char flags;
4612 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4613 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4614 struct list_head **iter);
4616 /* iterate through upper list, must be called under RCU read lock */
4617 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4618 for (iter = &(dev)->adj_list.upper, \
4619 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4621 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4623 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4624 int (*fn)(struct net_device *upper_dev,
4625 struct netdev_nested_priv *priv),
4626 struct netdev_nested_priv *priv);
4628 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4629 struct net_device *upper_dev);
4631 bool netdev_has_any_upper_dev(struct net_device *dev);
4633 void *netdev_lower_get_next_private(struct net_device *dev,
4634 struct list_head **iter);
4635 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4636 struct list_head **iter);
4638 #define netdev_for_each_lower_private(dev, priv, iter) \
4639 for (iter = (dev)->adj_list.lower.next, \
4640 priv = netdev_lower_get_next_private(dev, &(iter)); \
4642 priv = netdev_lower_get_next_private(dev, &(iter)))
4644 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4645 for (iter = &(dev)->adj_list.lower, \
4646 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4648 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4650 void *netdev_lower_get_next(struct net_device *dev,
4651 struct list_head **iter);
4653 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4654 for (iter = (dev)->adj_list.lower.next, \
4655 ldev = netdev_lower_get_next(dev, &(iter)); \
4657 ldev = netdev_lower_get_next(dev, &(iter)))
4659 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4660 struct list_head **iter);
4661 int netdev_walk_all_lower_dev(struct net_device *dev,
4662 int (*fn)(struct net_device *lower_dev,
4663 struct netdev_nested_priv *priv),
4664 struct netdev_nested_priv *priv);
4665 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4666 int (*fn)(struct net_device *lower_dev,
4667 struct netdev_nested_priv *priv),
4668 struct netdev_nested_priv *priv);
4670 void *netdev_adjacent_get_private(struct list_head *adj_list);
4671 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4672 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4673 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4674 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4675 struct netlink_ext_ack *extack);
4676 int netdev_master_upper_dev_link(struct net_device *dev,
4677 struct net_device *upper_dev,
4678 void *upper_priv, void *upper_info,
4679 struct netlink_ext_ack *extack);
4680 void netdev_upper_dev_unlink(struct net_device *dev,
4681 struct net_device *upper_dev);
4682 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4683 struct net_device *new_dev,
4684 struct net_device *dev,
4685 struct netlink_ext_ack *extack);
4686 void netdev_adjacent_change_commit(struct net_device *old_dev,
4687 struct net_device *new_dev,
4688 struct net_device *dev);
4689 void netdev_adjacent_change_abort(struct net_device *old_dev,
4690 struct net_device *new_dev,
4691 struct net_device *dev);
4692 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4693 void *netdev_lower_dev_get_private(struct net_device *dev,
4694 struct net_device *lower_dev);
4695 void netdev_lower_state_changed(struct net_device *lower_dev,
4696 void *lower_state_info);
4698 /* RSS keys are 40 or 52 bytes long */
4699 #define NETDEV_RSS_KEY_LEN 52
4700 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4701 void netdev_rss_key_fill(void *buffer, size_t len);
4703 int skb_checksum_help(struct sk_buff *skb);
4704 int skb_crc32c_csum_help(struct sk_buff *skb);
4705 int skb_csum_hwoffload_help(struct sk_buff *skb,
4706 const netdev_features_t features);
4708 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4709 netdev_features_t features, bool tx_path);
4710 struct sk_buff *skb_eth_gso_segment(struct sk_buff *skb,
4711 netdev_features_t features, __be16 type);
4712 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4713 netdev_features_t features);
4715 struct netdev_bonding_info {
4720 struct netdev_notifier_bonding_info {
4721 struct netdev_notifier_info info; /* must be first */
4722 struct netdev_bonding_info bonding_info;
4725 void netdev_bonding_info_change(struct net_device *dev,
4726 struct netdev_bonding_info *bonding_info);
4728 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4729 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4731 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4738 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4740 return __skb_gso_segment(skb, features, true);
4742 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4744 static inline bool can_checksum_protocol(netdev_features_t features,
4747 if (protocol == htons(ETH_P_FCOE))
4748 return !!(features & NETIF_F_FCOE_CRC);
4750 /* Assume this is an IP checksum (not SCTP CRC) */
4752 if (features & NETIF_F_HW_CSUM) {
4753 /* Can checksum everything */
4758 case htons(ETH_P_IP):
4759 return !!(features & NETIF_F_IP_CSUM);
4760 case htons(ETH_P_IPV6):
4761 return !!(features & NETIF_F_IPV6_CSUM);
4768 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4770 static inline void netdev_rx_csum_fault(struct net_device *dev,
4771 struct sk_buff *skb)
4775 /* rx skb timestamps */
4776 void net_enable_timestamp(void);
4777 void net_disable_timestamp(void);
4779 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4780 const struct skb_shared_hwtstamps *hwtstamps,
4783 const struct net_device_ops *ops = dev->netdev_ops;
4785 if (ops->ndo_get_tstamp)
4786 return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4788 return hwtstamps->hwtstamp;
4791 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4792 struct sk_buff *skb, struct net_device *dev,
4795 __this_cpu_write(softnet_data.xmit.more, more);
4796 return ops->ndo_start_xmit(skb, dev);
4799 static inline bool netdev_xmit_more(void)
4801 return __this_cpu_read(softnet_data.xmit.more);
4804 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4805 struct netdev_queue *txq, bool more)
4807 const struct net_device_ops *ops = dev->netdev_ops;
4810 rc = __netdev_start_xmit(ops, skb, dev, more);
4811 if (rc == NETDEV_TX_OK)
4812 txq_trans_update(txq);
4817 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4819 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4822 extern const struct kobj_ns_type_operations net_ns_type_operations;
4824 const char *netdev_drivername(const struct net_device *dev);
4826 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4827 netdev_features_t f2)
4829 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4830 if (f1 & NETIF_F_HW_CSUM)
4831 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4833 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4839 static inline netdev_features_t netdev_get_wanted_features(
4840 struct net_device *dev)
4842 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4844 netdev_features_t netdev_increment_features(netdev_features_t all,
4845 netdev_features_t one, netdev_features_t mask);
4847 /* Allow TSO being used on stacked device :
4848 * Performing the GSO segmentation before last device
4849 * is a performance improvement.
4851 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4852 netdev_features_t mask)
4854 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4857 int __netdev_update_features(struct net_device *dev);
4858 void netdev_update_features(struct net_device *dev);
4859 void netdev_change_features(struct net_device *dev);
4861 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4862 struct net_device *dev);
4864 netdev_features_t passthru_features_check(struct sk_buff *skb,
4865 struct net_device *dev,
4866 netdev_features_t features);
4867 netdev_features_t netif_skb_features(struct sk_buff *skb);
4869 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4871 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4873 /* check flags correspondence */
4874 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4875 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4876 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4877 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4878 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4879 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4880 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4881 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4882 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4883 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4884 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4885 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4886 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4887 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4888 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4889 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4890 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4891 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4892 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4894 return (features & feature) == feature;
4897 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4899 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4900 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4903 static inline bool netif_needs_gso(struct sk_buff *skb,
4904 netdev_features_t features)
4906 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4907 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4908 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4911 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
4912 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
4913 void netif_inherit_tso_max(struct net_device *to,
4914 const struct net_device *from);
4916 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4917 int pulled_hlen, u16 mac_offset,
4920 skb->protocol = protocol;
4921 skb->encapsulation = 1;
4922 skb_push(skb, pulled_hlen);
4923 skb_reset_transport_header(skb);
4924 skb->mac_header = mac_offset;
4925 skb->network_header = skb->mac_header + mac_len;
4926 skb->mac_len = mac_len;
4929 static inline bool netif_is_macsec(const struct net_device *dev)
4931 return dev->priv_flags & IFF_MACSEC;
4934 static inline bool netif_is_macvlan(const struct net_device *dev)
4936 return dev->priv_flags & IFF_MACVLAN;
4939 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4941 return dev->priv_flags & IFF_MACVLAN_PORT;
4944 static inline bool netif_is_bond_master(const struct net_device *dev)
4946 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4949 static inline bool netif_is_bond_slave(const struct net_device *dev)
4951 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4954 static inline bool netif_supports_nofcs(struct net_device *dev)
4956 return dev->priv_flags & IFF_SUPP_NOFCS;
4959 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4961 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4964 static inline bool netif_is_l3_master(const struct net_device *dev)
4966 return dev->priv_flags & IFF_L3MDEV_MASTER;
4969 static inline bool netif_is_l3_slave(const struct net_device *dev)
4971 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4974 static inline bool netif_is_bridge_master(const struct net_device *dev)
4976 return dev->priv_flags & IFF_EBRIDGE;
4979 static inline bool netif_is_bridge_port(const struct net_device *dev)
4981 return dev->priv_flags & IFF_BRIDGE_PORT;
4984 static inline bool netif_is_ovs_master(const struct net_device *dev)
4986 return dev->priv_flags & IFF_OPENVSWITCH;
4989 static inline bool netif_is_ovs_port(const struct net_device *dev)
4991 return dev->priv_flags & IFF_OVS_DATAPATH;
4994 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4996 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4999 static inline bool netif_is_team_master(const struct net_device *dev)
5001 return dev->priv_flags & IFF_TEAM;
5004 static inline bool netif_is_team_port(const struct net_device *dev)
5006 return dev->priv_flags & IFF_TEAM_PORT;
5009 static inline bool netif_is_lag_master(const struct net_device *dev)
5011 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5014 static inline bool netif_is_lag_port(const struct net_device *dev)
5016 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5019 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5021 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5024 static inline bool netif_is_failover(const struct net_device *dev)
5026 return dev->priv_flags & IFF_FAILOVER;
5029 static inline bool netif_is_failover_slave(const struct net_device *dev)
5031 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5034 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5035 static inline void netif_keep_dst(struct net_device *dev)
5037 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5040 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5041 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5043 /* TODO: reserve and use an additional IFF bit, if we get more users */
5044 return netif_is_macsec(dev);
5047 extern struct pernet_operations __net_initdata loopback_net_ops;
5049 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5051 /* netdev_printk helpers, similar to dev_printk */
5053 static inline const char *netdev_name(const struct net_device *dev)
5055 if (!dev->name[0] || strchr(dev->name, '%'))
5056 return "(unnamed net_device)";
5060 static inline bool netdev_unregistering(const struct net_device *dev)
5062 return dev->reg_state == NETREG_UNREGISTERING;
5065 static inline const char *netdev_reg_state(const struct net_device *dev)
5067 switch (dev->reg_state) {
5068 case NETREG_UNINITIALIZED: return " (uninitialized)";
5069 case NETREG_REGISTERED: return "";
5070 case NETREG_UNREGISTERING: return " (unregistering)";
5071 case NETREG_UNREGISTERED: return " (unregistered)";
5072 case NETREG_RELEASED: return " (released)";
5073 case NETREG_DUMMY: return " (dummy)";
5076 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5077 return " (unknown)";
5080 #define MODULE_ALIAS_NETDEV(device) \
5081 MODULE_ALIAS("netdev-" device)
5084 * netdev_WARN() acts like dev_printk(), but with the key difference
5085 * of using a WARN/WARN_ON to get the message out, including the
5086 * file/line information and a backtrace.
5088 #define netdev_WARN(dev, format, args...) \
5089 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5090 netdev_reg_state(dev), ##args)
5092 #define netdev_WARN_ONCE(dev, format, args...) \
5093 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5094 netdev_reg_state(dev), ##args)
5097 * The list of packet types we will receive (as opposed to discard)
5098 * and the routines to invoke.
5100 * Why 16. Because with 16 the only overlap we get on a hash of the
5101 * low nibble of the protocol value is RARP/SNAP/X.25.
5115 #define PTYPE_HASH_SIZE (16)
5116 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5118 extern struct list_head ptype_all __read_mostly;
5119 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5121 extern struct net_device *blackhole_netdev;
5123 #endif /* _LINUX_NETDEVICE_H */