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>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <net/net_namespace.h>
39 #include <net/dcbnl.h>
41 #include <net/netprio_cgroup.h>
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <linux/hashtable.h>
56 struct ip_tunnel_parm;
57 struct macsec_context;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct udp_tunnel_nic_info;
69 struct udp_tunnel_nic;
73 void synchronize_net(void);
74 void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
77 /* Backlog congestion levels */
78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79 #define NET_RX_DROP 1 /* packet dropped */
81 #define MAX_NEST_DEV 8
84 * Transmit return codes: transmit return codes originate from three different
87 * - qdisc return codes
88 * - driver transmit return codes
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
100 /* qdisc ->enqueue() return codes. */
101 #define NET_XMIT_SUCCESS 0x00
102 #define NET_XMIT_DROP 0x01 /* skb dropped */
103 #define NET_XMIT_CN 0x02 /* congestion notification */
104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
112 /* Driver transmit return codes */
113 #define NETDEV_TX_MASK 0xf0
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
120 typedef enum netdev_tx netdev_tx_t;
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
126 static inline bool dev_xmit_complete(int rc)
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
134 if (likely(rc < NET_XMIT_MASK))
141 * Compute the worst-case header length according to the protocols
145 #if defined(CONFIG_HYPERV_NET)
146 # define LL_MAX_HEADER 128
147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
151 # define LL_MAX_HEADER 96
154 # define LL_MAX_HEADER 32
157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159 #define MAX_HEADER LL_MAX_HEADER
161 #define MAX_HEADER (LL_MAX_HEADER + 48)
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
169 struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
196 #include <linux/cache.h>
197 #include <linux/skbuff.h>
200 #include <linux/static_key.h>
201 extern struct static_key_false rps_needed;
202 extern struct static_key_false rfs_needed;
209 struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
213 #define NETDEV_HW_ADDR_T_LAN 1
214 #define NETDEV_HW_ADDR_T_SAN 2
215 #define NETDEV_HW_ADDR_T_UNICAST 3
216 #define NETDEV_HW_ADDR_T_MULTICAST 4
221 struct rcu_head rcu_head;
224 struct netdev_hw_addr_list {
225 struct list_head list;
229 #define netdev_hw_addr_list_count(l) ((l)->count)
230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231 #define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236 #define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241 #define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
248 /* cached hardware header; allow for machine alignment needs. */
249 #define HH_DATA_MOD 16
250 #define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252 #define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
265 #define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
283 /* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
288 enum netdev_state_t {
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
302 struct netdev_boot_setup {
306 #define NETDEV_BOOT_SETUP_MAX 8
308 int __init netdev_boot_setup(char *str);
311 struct list_head list;
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
319 #define GRO_HASH_BUCKETS 8
322 * Structure for NAPI scheduling similar to tasklet but with weighting
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
331 struct list_head poll_list;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338 #ifdef CONFIG_NETPOLL
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
350 struct task_struct *thread;
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_MISSED, /* reschedule a napi */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
358 NAPI_STATE_LISTED, /* NAPI added to system lists */
359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
363 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
367 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
368 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
369 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
370 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
371 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
372 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
373 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
374 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
375 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
376 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
386 typedef enum gro_result gro_result_t;
389 * enum rx_handler_result - Possible return values for rx_handlers.
390 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
392 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
393 * case skb->dev was changed by rx_handler.
394 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
395 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
397 * rx_handlers are functions called from inside __netif_receive_skb(), to do
398 * special processing of the skb, prior to delivery to protocol handlers.
400 * Currently, a net_device can only have a single rx_handler registered. Trying
401 * to register a second rx_handler will return -EBUSY.
403 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
404 * To unregister a rx_handler on a net_device, use
405 * netdev_rx_handler_unregister().
407 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
410 * If the rx_handler consumed the skb in some way, it should return
411 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
412 * the skb to be delivered in some other way.
414 * If the rx_handler changed skb->dev, to divert the skb to another
415 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
416 * new device will be called if it exists.
418 * If the rx_handler decides the skb should be ignored, it should return
419 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
420 * are registered on exact device (ptype->dev == skb->dev).
422 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
423 * delivered, it should return RX_HANDLER_PASS.
425 * A device without a registered rx_handler will behave as if rx_handler
426 * returned RX_HANDLER_PASS.
429 enum rx_handler_result {
435 typedef enum rx_handler_result rx_handler_result_t;
436 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
438 void __napi_schedule(struct napi_struct *n);
439 void __napi_schedule_irqoff(struct napi_struct *n);
441 static inline bool napi_disable_pending(struct napi_struct *n)
443 return test_bit(NAPI_STATE_DISABLE, &n->state);
446 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
448 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
451 bool napi_schedule_prep(struct napi_struct *n);
454 * napi_schedule - schedule NAPI poll
457 * Schedule NAPI poll routine to be called if it is not already
460 static inline void napi_schedule(struct napi_struct *n)
462 if (napi_schedule_prep(n))
467 * napi_schedule_irqoff - schedule NAPI poll
470 * Variant of napi_schedule(), assuming hard irqs are masked.
472 static inline void napi_schedule_irqoff(struct napi_struct *n)
474 if (napi_schedule_prep(n))
475 __napi_schedule_irqoff(n);
478 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
479 static inline bool napi_reschedule(struct napi_struct *napi)
481 if (napi_schedule_prep(napi)) {
482 __napi_schedule(napi);
488 bool napi_complete_done(struct napi_struct *n, int work_done);
490 * napi_complete - NAPI processing complete
493 * Mark NAPI processing as complete.
494 * Consider using napi_complete_done() instead.
495 * Return false if device should avoid rearming interrupts.
497 static inline bool napi_complete(struct napi_struct *n)
499 return napi_complete_done(n, 0);
502 int dev_set_threaded(struct net_device *dev, bool threaded);
505 * napi_disable - prevent NAPI from scheduling
508 * Stop NAPI from being scheduled on this context.
509 * Waits till any outstanding processing completes.
511 void napi_disable(struct napi_struct *n);
513 void napi_enable(struct napi_struct *n);
516 * napi_synchronize - wait until NAPI is not running
519 * Wait until NAPI is done being scheduled on this context.
520 * Waits till any outstanding processing completes but
521 * does not disable future activations.
523 static inline void napi_synchronize(const struct napi_struct *n)
525 if (IS_ENABLED(CONFIG_SMP))
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
533 * napi_if_scheduled_mark_missed - if napi is running, set the
537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
540 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
542 unsigned long val, new;
545 val = READ_ONCE(n->state);
546 if (val & NAPIF_STATE_DISABLE)
549 if (!(val & NAPIF_STATE_SCHED))
552 new = val | NAPIF_STATE_MISSED;
553 } while (cmpxchg(&n->state, val, new) != val);
558 enum netdev_queue_state_t {
559 __QUEUE_STATE_DRV_XOFF,
560 __QUEUE_STATE_STACK_XOFF,
561 __QUEUE_STATE_FROZEN,
564 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
565 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
566 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
568 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
569 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
571 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
576 * netif_tx_* functions below are used to manipulate this flag. The
577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
578 * queue independently. The netif_xmit_*stopped functions below are called
579 * to check if the queue has been stopped by the driver or stack (either
580 * of the XOFF bits are set in the state). Drivers should not need to call
581 * netif_xmit*stopped functions, they should only be using netif_tx_*.
584 struct netdev_queue {
588 struct net_device *dev;
589 struct Qdisc __rcu *qdisc;
590 struct Qdisc *qdisc_sleeping;
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 unsigned long tx_maxrate;
599 * Number of TX timeouts for this queue
600 * (/sys/class/net/DEV/Q/trans_timeout)
602 unsigned long trans_timeout;
604 /* Subordinate device that the queue has been assigned to */
605 struct net_device *sb_dev;
606 #ifdef CONFIG_XDP_SOCKETS
607 struct xsk_buff_pool *pool;
612 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
615 * Time (in jiffies) of last Tx
617 unsigned long trans_start;
624 } ____cacheline_aligned_in_smp;
626 extern int sysctl_fb_tunnels_only_for_init_net;
627 extern int sysctl_devconf_inherit_init_net;
630 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
631 * == 1 : For initns only
634 static inline bool net_has_fallback_tunnels(const struct net *net)
636 return !IS_ENABLED(CONFIG_SYSCTL) ||
637 !sysctl_fb_tunnels_only_for_init_net ||
638 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
674 struct rps_dev_flow {
677 unsigned int last_qtail;
679 #define RPS_NO_FILTER 0xffff
682 * The rps_dev_flow_table structure contains a table of flow mappings.
684 struct rps_dev_flow_table {
687 struct rps_dev_flow flows[];
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
702 struct rps_sock_flow_table {
705 u32 ents[] ____cacheline_aligned_in_smp;
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
709 #define RPS_NO_CPU 0xffff
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
724 if (table->ents[index] != val)
725 table->ents[index] = val;
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 #endif /* CONFIG_RPS */
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xsk_buff_pool *pool;
747 } ____cacheline_aligned_in_smp;
750 * RX queue sysfs structures and functions.
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
759 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
768 * This structure holds an XPS map which can be of variable length. The
769 * map is an array of queues.
773 unsigned int alloc_len;
777 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
778 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
779 - sizeof(struct xps_map)) / sizeof(u16))
782 * This structure holds all XPS maps for device. Maps are indexed by CPU.
784 * We keep track of the number of cpus/rxqs used when the struct is allocated,
785 * in nr_ids. This will help not accessing out-of-bound memory.
787 * We keep track of the number of traffic classes used when the struct is
788 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
789 * not crossing its upper bound, as the original dev->num_tc can be updated in
792 struct xps_dev_maps {
796 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
799 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
800 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
802 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
803 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
805 #endif /* CONFIG_XPS */
807 #define TC_MAX_QUEUE 16
808 #define TC_BITMASK 15
809 /* HW offloaded queuing disciplines txq count and offset maps */
810 struct netdev_tc_txq {
815 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
817 * This structure is to hold information about the device
818 * configured to run FCoE protocol stack.
820 struct netdev_fcoe_hbainfo {
821 char manufacturer[64];
822 char serial_number[64];
823 char hardware_version[64];
824 char driver_version[64];
825 char optionrom_version[64];
826 char firmware_version[64];
828 char model_description[256];
832 #define MAX_PHYS_ITEM_ID_LEN 32
834 /* This structure holds a unique identifier to identify some
835 * physical item (port for example) used by a netdevice.
837 struct netdev_phys_item_id {
838 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
839 unsigned char id_len;
842 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
843 struct netdev_phys_item_id *b)
845 return a->id_len == b->id_len &&
846 memcmp(a->id, b->id, a->id_len) == 0;
849 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
851 struct net_device *sb_dev);
853 enum net_device_path_type {
854 DEV_PATH_ETHERNET = 0,
861 struct net_device_path {
862 enum net_device_path_type type;
863 const struct net_device *dev;
872 DEV_PATH_BR_VLAN_KEEP,
873 DEV_PATH_BR_VLAN_TAG,
874 DEV_PATH_BR_VLAN_UNTAG,
875 DEV_PATH_BR_VLAN_UNTAG_HW,
887 #define NET_DEVICE_PATH_STACK_MAX 5
888 #define NET_DEVICE_PATH_VLAN_MAX 2
890 struct net_device_path_stack {
892 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
895 struct net_device_path_ctx {
896 const struct net_device *dev;
903 } vlan[NET_DEVICE_PATH_VLAN_MAX];
907 TC_SETUP_QDISC_MQPRIO,
910 TC_SETUP_CLSMATCHALL,
920 TC_SETUP_QDISC_TAPRIO,
928 /* These structures hold the attributes of bpf state that are being passed
929 * to the netdevice through the bpf op.
931 enum bpf_netdev_command {
932 /* Set or clear a bpf program used in the earliest stages of packet
933 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
934 * is responsible for calling bpf_prog_put on any old progs that are
935 * stored. In case of error, the callee need not release the new prog
936 * reference, but on success it takes ownership and must bpf_prog_put
937 * when it is no longer used.
941 /* BPF program for offload callbacks, invoked at program load time. */
942 BPF_OFFLOAD_MAP_ALLOC,
943 BPF_OFFLOAD_MAP_FREE,
947 struct bpf_prog_offload_ops;
948 struct netlink_ext_ack;
950 struct xdp_dev_bulk_queue;
960 struct bpf_xdp_entity {
961 struct bpf_prog *prog;
962 struct bpf_xdp_link *link;
966 enum bpf_netdev_command command;
971 struct bpf_prog *prog;
972 struct netlink_ext_ack *extack;
974 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
976 struct bpf_offloaded_map *offmap;
978 /* XDP_SETUP_XSK_POOL */
980 struct xsk_buff_pool *pool;
986 /* Flags for ndo_xsk_wakeup. */
987 #define XDP_WAKEUP_RX (1 << 0)
988 #define XDP_WAKEUP_TX (1 << 1)
990 #ifdef CONFIG_XFRM_OFFLOAD
992 int (*xdo_dev_state_add) (struct xfrm_state *x);
993 void (*xdo_dev_state_delete) (struct xfrm_state *x);
994 void (*xdo_dev_state_free) (struct xfrm_state *x);
995 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
996 struct xfrm_state *x);
997 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1001 struct dev_ifalias {
1002 struct rcu_head rcuhead;
1009 struct netdev_name_node {
1010 struct hlist_node hlist;
1011 struct list_head list;
1012 struct net_device *dev;
1016 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1017 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1019 struct netdev_net_notifier {
1020 struct list_head list;
1021 struct notifier_block *nb;
1025 * This structure defines the management hooks for network devices.
1026 * The following hooks can be defined; unless noted otherwise, they are
1027 * optional and can be filled with a null pointer.
1029 * int (*ndo_init)(struct net_device *dev);
1030 * This function is called once when a network device is registered.
1031 * The network device can use this for any late stage initialization
1032 * or semantic validation. It can fail with an error code which will
1033 * be propagated back to register_netdev.
1035 * void (*ndo_uninit)(struct net_device *dev);
1036 * This function is called when device is unregistered or when registration
1037 * fails. It is not called if init fails.
1039 * int (*ndo_open)(struct net_device *dev);
1040 * This function is called when a network device transitions to the up
1043 * int (*ndo_stop)(struct net_device *dev);
1044 * This function is called when a network device transitions to the down
1047 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1048 * struct net_device *dev);
1049 * Called when a packet needs to be transmitted.
1050 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1051 * the queue before that can happen; it's for obsolete devices and weird
1052 * corner cases, but the stack really does a non-trivial amount
1053 * of useless work if you return NETDEV_TX_BUSY.
1054 * Required; cannot be NULL.
1056 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1057 * struct net_device *dev
1058 * netdev_features_t features);
1059 * Called by core transmit path to determine if device is capable of
1060 * performing offload operations on a given packet. This is to give
1061 * the device an opportunity to implement any restrictions that cannot
1062 * be otherwise expressed by feature flags. The check is called with
1063 * the set of features that the stack has calculated and it returns
1064 * those the driver believes to be appropriate.
1066 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1067 * struct net_device *sb_dev);
1068 * Called to decide which queue to use when device supports multiple
1071 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1072 * This function is called to allow device receiver to make
1073 * changes to configuration when multicast or promiscuous is enabled.
1075 * void (*ndo_set_rx_mode)(struct net_device *dev);
1076 * This function is called device changes address list filtering.
1077 * If driver handles unicast address filtering, it should set
1078 * IFF_UNICAST_FLT in its priv_flags.
1080 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1081 * This function is called when the Media Access Control address
1082 * needs to be changed. If this interface is not defined, the
1083 * MAC address can not be changed.
1085 * int (*ndo_validate_addr)(struct net_device *dev);
1086 * Test if Media Access Control address is valid for the device.
1088 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1089 * Old-style ioctl entry point. This is used internally by the
1090 * appletalk and ieee802154 subsystems but is no longer called by
1091 * the device ioctl handler.
1093 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1094 * Used by the bonding driver for its device specific ioctls:
1095 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1096 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1098 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1099 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1100 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1102 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1103 * Used to set network devices bus interface parameters. This interface
1104 * is retained for legacy reasons; new devices should use the bus
1105 * interface (PCI) for low level management.
1107 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1108 * Called when a user wants to change the Maximum Transfer Unit
1111 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1112 * Callback used when the transmitter has not made any progress
1113 * for dev->watchdog ticks.
1115 * void (*ndo_get_stats64)(struct net_device *dev,
1116 * struct rtnl_link_stats64 *storage);
1117 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1118 * Called when a user wants to get the network device usage
1119 * statistics. Drivers must do one of the following:
1120 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1121 * rtnl_link_stats64 structure passed by the caller.
1122 * 2. Define @ndo_get_stats to update a net_device_stats structure
1123 * (which should normally be dev->stats) and return a pointer to
1124 * it. The structure may be changed asynchronously only if each
1125 * field is written atomically.
1126 * 3. Update dev->stats asynchronously and atomically, and define
1127 * neither operation.
1129 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1130 * Return true if this device supports offload stats of this attr_id.
1132 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1134 * Get statistics for offload operations by attr_id. Write it into the
1135 * attr_data pointer.
1137 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1138 * If device supports VLAN filtering this function is called when a
1139 * VLAN id is registered.
1141 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1142 * If device supports VLAN filtering this function is called when a
1143 * VLAN id is unregistered.
1145 * void (*ndo_poll_controller)(struct net_device *dev);
1147 * SR-IOV management functions.
1148 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1149 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1150 * u8 qos, __be16 proto);
1151 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1153 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1154 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1155 * int (*ndo_get_vf_config)(struct net_device *dev,
1156 * int vf, struct ifla_vf_info *ivf);
1157 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1158 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1159 * struct nlattr *port[]);
1161 * Enable or disable the VF ability to query its RSS Redirection Table and
1162 * Hash Key. This is needed since on some devices VF share this information
1163 * with PF and querying it may introduce a theoretical security risk.
1164 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1165 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1166 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1168 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1169 * This is always called from the stack with the rtnl lock held and netif
1170 * tx queues stopped. This allows the netdevice to perform queue
1171 * management safely.
1173 * Fiber Channel over Ethernet (FCoE) offload functions.
1174 * int (*ndo_fcoe_enable)(struct net_device *dev);
1175 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1176 * so the underlying device can perform whatever needed configuration or
1177 * initialization to support acceleration of FCoE traffic.
1179 * int (*ndo_fcoe_disable)(struct net_device *dev);
1180 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1181 * so the underlying device can perform whatever needed clean-ups to
1182 * stop supporting acceleration of FCoE traffic.
1184 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1185 * struct scatterlist *sgl, unsigned int sgc);
1186 * Called when the FCoE Initiator wants to initialize an I/O that
1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1188 * perform necessary setup and returns 1 to indicate the device is set up
1189 * successfully to perform DDP on this I/O, otherwise this returns 0.
1191 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1192 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1193 * indicated by the FC exchange id 'xid', so the underlying device can
1194 * clean up and reuse resources for later DDP requests.
1196 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1197 * struct scatterlist *sgl, unsigned int sgc);
1198 * Called when the FCoE Target wants to initialize an I/O that
1199 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1200 * perform necessary setup and returns 1 to indicate the device is set up
1201 * successfully to perform DDP on this I/O, otherwise this returns 0.
1203 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1204 * struct netdev_fcoe_hbainfo *hbainfo);
1205 * Called when the FCoE Protocol stack wants information on the underlying
1206 * device. This information is utilized by the FCoE protocol stack to
1207 * register attributes with Fiber Channel management service as per the
1208 * FC-GS Fabric Device Management Information(FDMI) specification.
1210 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1211 * Called when the underlying device wants to override default World Wide
1212 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1213 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1214 * protocol stack to use.
1217 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1218 * u16 rxq_index, u32 flow_id);
1219 * Set hardware filter for RFS. rxq_index is the target queue index;
1220 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1221 * Return the filter ID on success, or a negative error code.
1223 * Slave management functions (for bridge, bonding, etc).
1224 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1225 * Called to make another netdev an underling.
1227 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1228 * Called to release previously enslaved netdev.
1230 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1231 * struct sk_buff *skb,
1233 * Get the xmit slave of master device. If all_slaves is true, function
1234 * assume all the slaves can transmit.
1236 * Feature/offload setting functions.
1237 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1238 * netdev_features_t features);
1239 * Adjusts the requested feature flags according to device-specific
1240 * constraints, and returns the resulting flags. Must not modify
1243 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1244 * Called to update device configuration to new features. Passed
1245 * feature set might be less than what was returned by ndo_fix_features()).
1246 * Must return >0 or -errno if it changed dev->features itself.
1248 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1249 * struct net_device *dev,
1250 * const unsigned char *addr, u16 vid, u16 flags,
1251 * struct netlink_ext_ack *extack);
1252 * Adds an FDB entry to dev for addr.
1253 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1254 * struct net_device *dev,
1255 * const unsigned char *addr, u16 vid)
1256 * Deletes the FDB entry from dev coresponding to addr.
1257 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1258 * struct net_device *dev, struct net_device *filter_dev,
1260 * Used to add FDB entries to dump requests. Implementers should add
1261 * entries to skb and update idx with the number of entries.
1263 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1264 * u16 flags, struct netlink_ext_ack *extack)
1265 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1266 * struct net_device *dev, u32 filter_mask,
1268 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1271 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1272 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1273 * which do not represent real hardware may define this to allow their
1274 * userspace components to manage their virtual carrier state. Devices
1275 * that determine carrier state from physical hardware properties (eg
1276 * network cables) or protocol-dependent mechanisms (eg
1277 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1279 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1280 * struct netdev_phys_item_id *ppid);
1281 * Called to get ID of physical port of this device. If driver does
1282 * not implement this, it is assumed that the hw is not able to have
1283 * multiple net devices on single physical port.
1285 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1286 * struct netdev_phys_item_id *ppid)
1287 * Called to get the parent ID of the physical port of this device.
1289 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1290 * struct net_device *dev)
1291 * Called by upper layer devices to accelerate switching or other
1292 * station functionality into hardware. 'pdev is the lowerdev
1293 * to use for the offload and 'dev' is the net device that will
1294 * back the offload. Returns a pointer to the private structure
1295 * the upper layer will maintain.
1296 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1297 * Called by upper layer device to delete the station created
1298 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1299 * the station and priv is the structure returned by the add
1301 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1302 * int queue_index, u32 maxrate);
1303 * Called when a user wants to set a max-rate limitation of specific
1305 * int (*ndo_get_iflink)(const struct net_device *dev);
1306 * Called to get the iflink value of this device.
1307 * void (*ndo_change_proto_down)(struct net_device *dev,
1309 * This function is used to pass protocol port error state information
1310 * to the switch driver. The switch driver can react to the proto_down
1311 * by doing a phys down on the associated switch port.
1312 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1313 * This function is used to get egress tunnel information for given skb.
1314 * This is useful for retrieving outer tunnel header parameters while
1316 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1317 * This function is used to specify the headroom that the skb must
1318 * consider when allocation skb during packet reception. Setting
1319 * appropriate rx headroom value allows avoiding skb head copy on
1320 * forward. Setting a negative value resets the rx headroom to the
1322 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1323 * This function is used to set or query state related to XDP on the
1324 * netdevice and manage BPF offload. See definition of
1325 * enum bpf_netdev_command for details.
1326 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1328 * This function is used to submit @n XDP packets for transmit on a
1329 * netdevice. Returns number of frames successfully transmitted, frames
1330 * that got dropped are freed/returned via xdp_return_frame().
1331 * Returns negative number, means general error invoking ndo, meaning
1332 * no frames were xmit'ed and core-caller will free all frames.
1333 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1334 * This function is used to wake up the softirq, ksoftirqd or kthread
1335 * responsible for sending and/or receiving packets on a specific
1336 * queue id bound to an AF_XDP socket. The flags field specifies if
1337 * only RX, only Tx, or both should be woken up using the flags
1338 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1339 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1340 * Get devlink port instance associated with a given netdev.
1341 * Called with a reference on the netdevice and devlink locks only,
1342 * rtnl_lock is not held.
1343 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1345 * Add, change, delete or get information on an IPv4 tunnel.
1346 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1347 * If a device is paired with a peer device, return the peer instance.
1348 * The caller must be under RCU read context.
1349 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1350 * Get the forwarding path to reach the real device from the HW destination address
1352 struct net_device_ops {
1353 int (*ndo_init)(struct net_device *dev);
1354 void (*ndo_uninit)(struct net_device *dev);
1355 int (*ndo_open)(struct net_device *dev);
1356 int (*ndo_stop)(struct net_device *dev);
1357 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1358 struct net_device *dev);
1359 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1360 struct net_device *dev,
1361 netdev_features_t features);
1362 u16 (*ndo_select_queue)(struct net_device *dev,
1363 struct sk_buff *skb,
1364 struct net_device *sb_dev);
1365 void (*ndo_change_rx_flags)(struct net_device *dev,
1367 void (*ndo_set_rx_mode)(struct net_device *dev);
1368 int (*ndo_set_mac_address)(struct net_device *dev,
1370 int (*ndo_validate_addr)(struct net_device *dev);
1371 int (*ndo_do_ioctl)(struct net_device *dev,
1372 struct ifreq *ifr, int cmd);
1373 int (*ndo_eth_ioctl)(struct net_device *dev,
1374 struct ifreq *ifr, int cmd);
1375 int (*ndo_siocbond)(struct net_device *dev,
1376 struct ifreq *ifr, int cmd);
1377 int (*ndo_siocwandev)(struct net_device *dev,
1378 struct if_settings *ifs);
1379 int (*ndo_siocdevprivate)(struct net_device *dev,
1381 void __user *data, int cmd);
1382 int (*ndo_set_config)(struct net_device *dev,
1384 int (*ndo_change_mtu)(struct net_device *dev,
1386 int (*ndo_neigh_setup)(struct net_device *dev,
1387 struct neigh_parms *);
1388 void (*ndo_tx_timeout) (struct net_device *dev,
1389 unsigned int txqueue);
1391 void (*ndo_get_stats64)(struct net_device *dev,
1392 struct rtnl_link_stats64 *storage);
1393 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1394 int (*ndo_get_offload_stats)(int attr_id,
1395 const struct net_device *dev,
1397 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1399 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1400 __be16 proto, u16 vid);
1401 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1402 __be16 proto, u16 vid);
1403 #ifdef CONFIG_NET_POLL_CONTROLLER
1404 void (*ndo_poll_controller)(struct net_device *dev);
1405 int (*ndo_netpoll_setup)(struct net_device *dev,
1406 struct netpoll_info *info);
1407 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1409 int (*ndo_set_vf_mac)(struct net_device *dev,
1410 int queue, u8 *mac);
1411 int (*ndo_set_vf_vlan)(struct net_device *dev,
1412 int queue, u16 vlan,
1413 u8 qos, __be16 proto);
1414 int (*ndo_set_vf_rate)(struct net_device *dev,
1415 int vf, int min_tx_rate,
1417 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1418 int vf, bool setting);
1419 int (*ndo_set_vf_trust)(struct net_device *dev,
1420 int vf, bool setting);
1421 int (*ndo_get_vf_config)(struct net_device *dev,
1423 struct ifla_vf_info *ivf);
1424 int (*ndo_set_vf_link_state)(struct net_device *dev,
1425 int vf, int link_state);
1426 int (*ndo_get_vf_stats)(struct net_device *dev,
1428 struct ifla_vf_stats
1430 int (*ndo_set_vf_port)(struct net_device *dev,
1432 struct nlattr *port[]);
1433 int (*ndo_get_vf_port)(struct net_device *dev,
1434 int vf, struct sk_buff *skb);
1435 int (*ndo_get_vf_guid)(struct net_device *dev,
1437 struct ifla_vf_guid *node_guid,
1438 struct ifla_vf_guid *port_guid);
1439 int (*ndo_set_vf_guid)(struct net_device *dev,
1442 int (*ndo_set_vf_rss_query_en)(
1443 struct net_device *dev,
1444 int vf, bool setting);
1445 int (*ndo_setup_tc)(struct net_device *dev,
1446 enum tc_setup_type type,
1448 #if IS_ENABLED(CONFIG_FCOE)
1449 int (*ndo_fcoe_enable)(struct net_device *dev);
1450 int (*ndo_fcoe_disable)(struct net_device *dev);
1451 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1453 struct scatterlist *sgl,
1455 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1457 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1459 struct scatterlist *sgl,
1461 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1462 struct netdev_fcoe_hbainfo *hbainfo);
1465 #if IS_ENABLED(CONFIG_LIBFCOE)
1466 #define NETDEV_FCOE_WWNN 0
1467 #define NETDEV_FCOE_WWPN 1
1468 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1469 u64 *wwn, int type);
1472 #ifdef CONFIG_RFS_ACCEL
1473 int (*ndo_rx_flow_steer)(struct net_device *dev,
1474 const struct sk_buff *skb,
1478 int (*ndo_add_slave)(struct net_device *dev,
1479 struct net_device *slave_dev,
1480 struct netlink_ext_ack *extack);
1481 int (*ndo_del_slave)(struct net_device *dev,
1482 struct net_device *slave_dev);
1483 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1484 struct sk_buff *skb,
1486 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1488 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1489 netdev_features_t features);
1490 int (*ndo_set_features)(struct net_device *dev,
1491 netdev_features_t features);
1492 int (*ndo_neigh_construct)(struct net_device *dev,
1493 struct neighbour *n);
1494 void (*ndo_neigh_destroy)(struct net_device *dev,
1495 struct neighbour *n);
1497 int (*ndo_fdb_add)(struct ndmsg *ndm,
1498 struct nlattr *tb[],
1499 struct net_device *dev,
1500 const unsigned char *addr,
1503 struct netlink_ext_ack *extack);
1504 int (*ndo_fdb_del)(struct ndmsg *ndm,
1505 struct nlattr *tb[],
1506 struct net_device *dev,
1507 const unsigned char *addr,
1509 int (*ndo_fdb_dump)(struct sk_buff *skb,
1510 struct netlink_callback *cb,
1511 struct net_device *dev,
1512 struct net_device *filter_dev,
1514 int (*ndo_fdb_get)(struct sk_buff *skb,
1515 struct nlattr *tb[],
1516 struct net_device *dev,
1517 const unsigned char *addr,
1518 u16 vid, u32 portid, u32 seq,
1519 struct netlink_ext_ack *extack);
1520 int (*ndo_bridge_setlink)(struct net_device *dev,
1521 struct nlmsghdr *nlh,
1523 struct netlink_ext_ack *extack);
1524 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1526 struct net_device *dev,
1529 int (*ndo_bridge_dellink)(struct net_device *dev,
1530 struct nlmsghdr *nlh,
1532 int (*ndo_change_carrier)(struct net_device *dev,
1534 int (*ndo_get_phys_port_id)(struct net_device *dev,
1535 struct netdev_phys_item_id *ppid);
1536 int (*ndo_get_port_parent_id)(struct net_device *dev,
1537 struct netdev_phys_item_id *ppid);
1538 int (*ndo_get_phys_port_name)(struct net_device *dev,
1539 char *name, size_t len);
1540 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1541 struct net_device *dev);
1542 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1545 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1548 int (*ndo_get_iflink)(const struct net_device *dev);
1549 int (*ndo_change_proto_down)(struct net_device *dev,
1551 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1552 struct sk_buff *skb);
1553 void (*ndo_set_rx_headroom)(struct net_device *dev,
1554 int needed_headroom);
1555 int (*ndo_bpf)(struct net_device *dev,
1556 struct netdev_bpf *bpf);
1557 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1558 struct xdp_frame **xdp,
1560 int (*ndo_xsk_wakeup)(struct net_device *dev,
1561 u32 queue_id, u32 flags);
1562 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1563 int (*ndo_tunnel_ctl)(struct net_device *dev,
1564 struct ip_tunnel_parm *p, int cmd);
1565 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1566 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1567 struct net_device_path *path);
1571 * enum netdev_priv_flags - &struct net_device priv_flags
1573 * These are the &struct net_device, they are only set internally
1574 * by drivers and used in the kernel. These flags are invisible to
1575 * userspace; this means that the order of these flags can change
1576 * during any kernel release.
1578 * You should have a pretty good reason to be extending these flags.
1580 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1581 * @IFF_EBRIDGE: Ethernet bridging device
1582 * @IFF_BONDING: bonding master or slave
1583 * @IFF_ISATAP: ISATAP interface (RFC4214)
1584 * @IFF_WAN_HDLC: WAN HDLC device
1585 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1587 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1588 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1589 * @IFF_MACVLAN_PORT: device used as macvlan port
1590 * @IFF_BRIDGE_PORT: device used as bridge port
1591 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1592 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1593 * @IFF_UNICAST_FLT: Supports unicast filtering
1594 * @IFF_TEAM_PORT: device used as team port
1595 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1596 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1597 * change when it's running
1598 * @IFF_MACVLAN: Macvlan device
1599 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1600 * underlying stacked devices
1601 * @IFF_L3MDEV_MASTER: device is an L3 master device
1602 * @IFF_NO_QUEUE: device can run without qdisc attached
1603 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1604 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1605 * @IFF_TEAM: device is a team device
1606 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1607 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1608 * entity (i.e. the master device for bridged veth)
1609 * @IFF_MACSEC: device is a MACsec device
1610 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1611 * @IFF_FAILOVER: device is a failover master device
1612 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1613 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1614 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1615 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1616 * skb_headlen(skb) == 0 (data starts from frag0)
1618 enum netdev_priv_flags {
1619 IFF_802_1Q_VLAN = 1<<0,
1623 IFF_WAN_HDLC = 1<<4,
1624 IFF_XMIT_DST_RELEASE = 1<<5,
1625 IFF_DONT_BRIDGE = 1<<6,
1626 IFF_DISABLE_NETPOLL = 1<<7,
1627 IFF_MACVLAN_PORT = 1<<8,
1628 IFF_BRIDGE_PORT = 1<<9,
1629 IFF_OVS_DATAPATH = 1<<10,
1630 IFF_TX_SKB_SHARING = 1<<11,
1631 IFF_UNICAST_FLT = 1<<12,
1632 IFF_TEAM_PORT = 1<<13,
1633 IFF_SUPP_NOFCS = 1<<14,
1634 IFF_LIVE_ADDR_CHANGE = 1<<15,
1635 IFF_MACVLAN = 1<<16,
1636 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1637 IFF_L3MDEV_MASTER = 1<<18,
1638 IFF_NO_QUEUE = 1<<19,
1639 IFF_OPENVSWITCH = 1<<20,
1640 IFF_L3MDEV_SLAVE = 1<<21,
1642 IFF_RXFH_CONFIGURED = 1<<23,
1643 IFF_PHONY_HEADROOM = 1<<24,
1645 IFF_NO_RX_HANDLER = 1<<26,
1646 IFF_FAILOVER = 1<<27,
1647 IFF_FAILOVER_SLAVE = 1<<28,
1648 IFF_L3MDEV_RX_HANDLER = 1<<29,
1649 IFF_LIVE_RENAME_OK = 1<<30,
1650 IFF_TX_SKB_NO_LINEAR = 1<<31,
1653 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1654 #define IFF_EBRIDGE IFF_EBRIDGE
1655 #define IFF_BONDING IFF_BONDING
1656 #define IFF_ISATAP IFF_ISATAP
1657 #define IFF_WAN_HDLC IFF_WAN_HDLC
1658 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1659 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1660 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1661 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1662 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1663 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1664 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1665 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1666 #define IFF_TEAM_PORT IFF_TEAM_PORT
1667 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1668 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1669 #define IFF_MACVLAN IFF_MACVLAN
1670 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1671 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1672 #define IFF_NO_QUEUE IFF_NO_QUEUE
1673 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1674 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1675 #define IFF_TEAM IFF_TEAM
1676 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1677 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1678 #define IFF_MACSEC IFF_MACSEC
1679 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1680 #define IFF_FAILOVER IFF_FAILOVER
1681 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1682 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1683 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1684 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1686 /* Specifies the type of the struct net_device::ml_priv pointer */
1687 enum netdev_ml_priv_type {
1693 * struct net_device - The DEVICE structure.
1695 * Actually, this whole structure is a big mistake. It mixes I/O
1696 * data with strictly "high-level" data, and it has to know about
1697 * almost every data structure used in the INET module.
1699 * @name: This is the first field of the "visible" part of this structure
1700 * (i.e. as seen by users in the "Space.c" file). It is the name
1703 * @name_node: Name hashlist node
1704 * @ifalias: SNMP alias
1705 * @mem_end: Shared memory end
1706 * @mem_start: Shared memory start
1707 * @base_addr: Device I/O address
1708 * @irq: Device IRQ number
1710 * @state: Generic network queuing layer state, see netdev_state_t
1711 * @dev_list: The global list of network devices
1712 * @napi_list: List entry used for polling NAPI devices
1713 * @unreg_list: List entry when we are unregistering the
1714 * device; see the function unregister_netdev
1715 * @close_list: List entry used when we are closing the device
1716 * @ptype_all: Device-specific packet handlers for all protocols
1717 * @ptype_specific: Device-specific, protocol-specific packet handlers
1719 * @adj_list: Directly linked devices, like slaves for bonding
1720 * @features: Currently active device features
1721 * @hw_features: User-changeable features
1723 * @wanted_features: User-requested features
1724 * @vlan_features: Mask of features inheritable by VLAN devices
1726 * @hw_enc_features: Mask of features inherited by encapsulating devices
1727 * This field indicates what encapsulation
1728 * offloads the hardware is capable of doing,
1729 * and drivers will need to set them appropriately.
1731 * @mpls_features: Mask of features inheritable by MPLS
1732 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1734 * @ifindex: interface index
1735 * @group: The group the device belongs to
1737 * @stats: Statistics struct, which was left as a legacy, use
1738 * rtnl_link_stats64 instead
1740 * @rx_dropped: Dropped packets by core network,
1741 * do not use this in drivers
1742 * @tx_dropped: Dropped packets by core network,
1743 * do not use this in drivers
1744 * @rx_nohandler: nohandler dropped packets by core network on
1745 * inactive devices, do not use this in drivers
1746 * @carrier_up_count: Number of times the carrier has been up
1747 * @carrier_down_count: Number of times the carrier has been down
1749 * @wireless_handlers: List of functions to handle Wireless Extensions,
1751 * see <net/iw_handler.h> for details.
1752 * @wireless_data: Instance data managed by the core of wireless extensions
1754 * @netdev_ops: Includes several pointers to callbacks,
1755 * if one wants to override the ndo_*() functions
1756 * @ethtool_ops: Management operations
1757 * @l3mdev_ops: Layer 3 master device operations
1758 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1759 * discovery handling. Necessary for e.g. 6LoWPAN.
1760 * @xfrmdev_ops: Transformation offload operations
1761 * @tlsdev_ops: Transport Layer Security offload operations
1762 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1763 * of Layer 2 headers.
1765 * @flags: Interface flags (a la BSD)
1766 * @priv_flags: Like 'flags' but invisible to userspace,
1767 * see if.h for the definitions
1768 * @gflags: Global flags ( kept as legacy )
1769 * @padded: How much padding added by alloc_netdev()
1770 * @operstate: RFC2863 operstate
1771 * @link_mode: Mapping policy to operstate
1772 * @if_port: Selectable AUI, TP, ...
1774 * @mtu: Interface MTU value
1775 * @min_mtu: Interface Minimum MTU value
1776 * @max_mtu: Interface Maximum MTU value
1777 * @type: Interface hardware type
1778 * @hard_header_len: Maximum hardware header length.
1779 * @min_header_len: Minimum hardware header length
1781 * @needed_headroom: Extra headroom the hardware may need, but not in all
1782 * cases can this be guaranteed
1783 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1784 * cases can this be guaranteed. Some cases also use
1785 * LL_MAX_HEADER instead to allocate the skb
1787 * interface address info:
1789 * @perm_addr: Permanent hw address
1790 * @addr_assign_type: Hw address assignment type
1791 * @addr_len: Hardware address length
1792 * @upper_level: Maximum depth level of upper devices.
1793 * @lower_level: Maximum depth level of lower devices.
1794 * @neigh_priv_len: Used in neigh_alloc()
1795 * @dev_id: Used to differentiate devices that share
1796 * the same link layer address
1797 * @dev_port: Used to differentiate devices that share
1799 * @addr_list_lock: XXX: need comments on this one
1800 * @name_assign_type: network interface name assignment type
1801 * @uc_promisc: Counter that indicates promiscuous mode
1802 * has been enabled due to the need to listen to
1803 * additional unicast addresses in a device that
1804 * does not implement ndo_set_rx_mode()
1805 * @uc: unicast mac addresses
1806 * @mc: multicast mac addresses
1807 * @dev_addrs: list of device hw addresses
1808 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1809 * @promiscuity: Number of times the NIC is told to work in
1810 * promiscuous mode; if it becomes 0 the NIC will
1811 * exit promiscuous mode
1812 * @allmulti: Counter, enables or disables allmulticast mode
1814 * @vlan_info: VLAN info
1815 * @dsa_ptr: dsa specific data
1816 * @tipc_ptr: TIPC specific data
1817 * @atalk_ptr: AppleTalk link
1818 * @ip_ptr: IPv4 specific data
1819 * @dn_ptr: DECnet specific data
1820 * @ip6_ptr: IPv6 specific data
1821 * @ax25_ptr: AX.25 specific data
1822 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1823 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1825 * @mpls_ptr: mpls_dev struct pointer
1826 * @mctp_ptr: MCTP specific data
1828 * @dev_addr: Hw address (before bcast,
1829 * because most packets are unicast)
1831 * @_rx: Array of RX queues
1832 * @num_rx_queues: Number of RX queues
1833 * allocated at register_netdev() time
1834 * @real_num_rx_queues: Number of RX queues currently active in device
1835 * @xdp_prog: XDP sockets filter program pointer
1836 * @gro_flush_timeout: timeout for GRO layer in NAPI
1837 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1838 * allow to avoid NIC hard IRQ, on busy queues.
1840 * @rx_handler: handler for received packets
1841 * @rx_handler_data: XXX: need comments on this one
1842 * @miniq_ingress: ingress/clsact qdisc specific data for
1843 * ingress processing
1844 * @ingress_queue: XXX: need comments on this one
1845 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1846 * @broadcast: hw bcast address
1848 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1849 * indexed by RX queue number. Assigned by driver.
1850 * This must only be set if the ndo_rx_flow_steer
1851 * operation is defined
1852 * @index_hlist: Device index hash chain
1854 * @_tx: Array of TX queues
1855 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1856 * @real_num_tx_queues: Number of TX queues currently active in device
1857 * @qdisc: Root qdisc from userspace point of view
1858 * @tx_queue_len: Max frames per queue allowed
1859 * @tx_global_lock: XXX: need comments on this one
1860 * @xdp_bulkq: XDP device bulk queue
1861 * @xps_maps: all CPUs/RXQs maps for XPS device
1863 * @xps_maps: XXX: need comments on this one
1864 * @miniq_egress: clsact qdisc specific data for
1866 * @qdisc_hash: qdisc hash table
1867 * @watchdog_timeo: Represents the timeout that is used by
1868 * the watchdog (see dev_watchdog())
1869 * @watchdog_timer: List of timers
1871 * @proto_down_reason: reason a netdev interface is held down
1872 * @pcpu_refcnt: Number of references to this device
1873 * @dev_refcnt: Number of references to this device
1874 * @todo_list: Delayed register/unregister
1875 * @link_watch_list: XXX: need comments on this one
1877 * @reg_state: Register/unregister state machine
1878 * @dismantle: Device is going to be freed
1879 * @rtnl_link_state: This enum represents the phases of creating
1882 * @needs_free_netdev: Should unregister perform free_netdev?
1883 * @priv_destructor: Called from unregister
1884 * @npinfo: XXX: need comments on this one
1885 * @nd_net: Network namespace this network device is inside
1887 * @ml_priv: Mid-layer private
1888 * @ml_priv_type: Mid-layer private type
1889 * @lstats: Loopback statistics
1890 * @tstats: Tunnel statistics
1891 * @dstats: Dummy statistics
1892 * @vstats: Virtual ethernet statistics
1897 * @dev: Class/net/name entry
1898 * @sysfs_groups: Space for optional device, statistics and wireless
1901 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1902 * @rtnl_link_ops: Rtnl_link_ops
1904 * @gso_max_size: Maximum size of generic segmentation offload
1905 * @gso_max_segs: Maximum number of segments that can be passed to the
1908 * @dcbnl_ops: Data Center Bridging netlink ops
1909 * @num_tc: Number of traffic classes in the net device
1910 * @tc_to_txq: XXX: need comments on this one
1911 * @prio_tc_map: XXX: need comments on this one
1913 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1915 * @priomap: XXX: need comments on this one
1916 * @phydev: Physical device may attach itself
1917 * for hardware timestamping
1918 * @sfp_bus: attached &struct sfp_bus structure.
1920 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1921 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1923 * @proto_down: protocol port state information can be sent to the
1924 * switch driver and used to set the phys state of the
1927 * @wol_enabled: Wake-on-LAN is enabled
1929 * @threaded: napi threaded mode is enabled
1931 * @net_notifier_list: List of per-net netdev notifier block
1932 * that follow this device when it is moved
1933 * to another network namespace.
1935 * @macsec_ops: MACsec offloading ops
1937 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1938 * offload capabilities of the device
1939 * @udp_tunnel_nic: UDP tunnel offload state
1940 * @xdp_state: stores info on attached XDP BPF programs
1942 * @nested_level: Used as as a parameter of spin_lock_nested() of
1943 * dev->addr_list_lock.
1944 * @unlink_list: As netif_addr_lock() can be called recursively,
1945 * keep a list of interfaces to be deleted.
1947 * FIXME: cleanup struct net_device such that network protocol info
1952 char name[IFNAMSIZ];
1953 struct netdev_name_node *name_node;
1954 struct dev_ifalias __rcu *ifalias;
1956 * I/O specific fields
1957 * FIXME: Merge these and struct ifmap into one
1959 unsigned long mem_end;
1960 unsigned long mem_start;
1961 unsigned long base_addr;
1964 * Some hardware also needs these fields (state,dev_list,
1965 * napi_list,unreg_list,close_list) but they are not
1966 * part of the usual set specified in Space.c.
1969 unsigned long state;
1971 struct list_head dev_list;
1972 struct list_head napi_list;
1973 struct list_head unreg_list;
1974 struct list_head close_list;
1975 struct list_head ptype_all;
1976 struct list_head ptype_specific;
1979 struct list_head upper;
1980 struct list_head lower;
1983 /* Read-mostly cache-line for fast-path access */
1985 unsigned int priv_flags;
1986 const struct net_device_ops *netdev_ops;
1988 unsigned short gflags;
1989 unsigned short hard_header_len;
1991 /* Note : dev->mtu is often read without holding a lock.
1992 * Writers usually hold RTNL.
1993 * It is recommended to use READ_ONCE() to annotate the reads,
1994 * and to use WRITE_ONCE() to annotate the writes.
1997 unsigned short needed_headroom;
1998 unsigned short needed_tailroom;
2000 netdev_features_t features;
2001 netdev_features_t hw_features;
2002 netdev_features_t wanted_features;
2003 netdev_features_t vlan_features;
2004 netdev_features_t hw_enc_features;
2005 netdev_features_t mpls_features;
2006 netdev_features_t gso_partial_features;
2008 unsigned int min_mtu;
2009 unsigned int max_mtu;
2010 unsigned short type;
2011 unsigned char min_header_len;
2012 unsigned char name_assign_type;
2016 struct net_device_stats stats; /* not used by modern drivers */
2018 atomic_long_t rx_dropped;
2019 atomic_long_t tx_dropped;
2020 atomic_long_t rx_nohandler;
2022 /* Stats to monitor link on/off, flapping */
2023 atomic_t carrier_up_count;
2024 atomic_t carrier_down_count;
2026 #ifdef CONFIG_WIRELESS_EXT
2027 const struct iw_handler_def *wireless_handlers;
2028 struct iw_public_data *wireless_data;
2030 const struct ethtool_ops *ethtool_ops;
2031 #ifdef CONFIG_NET_L3_MASTER_DEV
2032 const struct l3mdev_ops *l3mdev_ops;
2034 #if IS_ENABLED(CONFIG_IPV6)
2035 const struct ndisc_ops *ndisc_ops;
2038 #ifdef CONFIG_XFRM_OFFLOAD
2039 const struct xfrmdev_ops *xfrmdev_ops;
2042 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2043 const struct tlsdev_ops *tlsdev_ops;
2046 const struct header_ops *header_ops;
2048 unsigned char operstate;
2049 unsigned char link_mode;
2051 unsigned char if_port;
2054 /* Interface address info. */
2055 unsigned char perm_addr[MAX_ADDR_LEN];
2056 unsigned char addr_assign_type;
2057 unsigned char addr_len;
2058 unsigned char upper_level;
2059 unsigned char lower_level;
2061 unsigned short neigh_priv_len;
2062 unsigned short dev_id;
2063 unsigned short dev_port;
2064 unsigned short padded;
2066 spinlock_t addr_list_lock;
2069 struct netdev_hw_addr_list uc;
2070 struct netdev_hw_addr_list mc;
2071 struct netdev_hw_addr_list dev_addrs;
2074 struct kset *queues_kset;
2076 #ifdef CONFIG_LOCKDEP
2077 struct list_head unlink_list;
2079 unsigned int promiscuity;
2080 unsigned int allmulti;
2082 #ifdef CONFIG_LOCKDEP
2083 unsigned char nested_level;
2087 /* Protocol-specific pointers */
2089 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2090 struct vlan_info __rcu *vlan_info;
2092 #if IS_ENABLED(CONFIG_NET_DSA)
2093 struct dsa_port *dsa_ptr;
2095 #if IS_ENABLED(CONFIG_TIPC)
2096 struct tipc_bearer __rcu *tipc_ptr;
2098 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2101 struct in_device __rcu *ip_ptr;
2102 #if IS_ENABLED(CONFIG_DECNET)
2103 struct dn_dev __rcu *dn_ptr;
2105 struct inet6_dev __rcu *ip6_ptr;
2106 #if IS_ENABLED(CONFIG_AX25)
2109 struct wireless_dev *ieee80211_ptr;
2110 struct wpan_dev *ieee802154_ptr;
2111 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2112 struct mpls_dev __rcu *mpls_ptr;
2114 #if IS_ENABLED(CONFIG_MCTP)
2115 struct mctp_dev __rcu *mctp_ptr;
2119 * Cache lines mostly used on receive path (including eth_type_trans())
2121 /* Interface address info used in eth_type_trans() */
2122 unsigned char *dev_addr;
2124 struct netdev_rx_queue *_rx;
2125 unsigned int num_rx_queues;
2126 unsigned int real_num_rx_queues;
2128 struct bpf_prog __rcu *xdp_prog;
2129 unsigned long gro_flush_timeout;
2130 int napi_defer_hard_irqs;
2131 rx_handler_func_t __rcu *rx_handler;
2132 void __rcu *rx_handler_data;
2134 #ifdef CONFIG_NET_CLS_ACT
2135 struct mini_Qdisc __rcu *miniq_ingress;
2137 struct netdev_queue __rcu *ingress_queue;
2138 #ifdef CONFIG_NETFILTER_INGRESS
2139 struct nf_hook_entries __rcu *nf_hooks_ingress;
2142 unsigned char broadcast[MAX_ADDR_LEN];
2143 #ifdef CONFIG_RFS_ACCEL
2144 struct cpu_rmap *rx_cpu_rmap;
2146 struct hlist_node index_hlist;
2149 * Cache lines mostly used on transmit path
2151 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2152 unsigned int num_tx_queues;
2153 unsigned int real_num_tx_queues;
2154 struct Qdisc *qdisc;
2155 unsigned int tx_queue_len;
2156 spinlock_t tx_global_lock;
2158 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2161 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2163 #ifdef CONFIG_NET_CLS_ACT
2164 struct mini_Qdisc __rcu *miniq_egress;
2167 #ifdef CONFIG_NET_SCHED
2168 DECLARE_HASHTABLE (qdisc_hash, 4);
2170 /* These may be needed for future network-power-down code. */
2171 struct timer_list watchdog_timer;
2174 u32 proto_down_reason;
2176 struct list_head todo_list;
2178 #ifdef CONFIG_PCPU_DEV_REFCNT
2179 int __percpu *pcpu_refcnt;
2181 refcount_t dev_refcnt;
2184 struct list_head link_watch_list;
2186 enum { NETREG_UNINITIALIZED=0,
2187 NETREG_REGISTERED, /* completed register_netdevice */
2188 NETREG_UNREGISTERING, /* called unregister_netdevice */
2189 NETREG_UNREGISTERED, /* completed unregister todo */
2190 NETREG_RELEASED, /* called free_netdev */
2191 NETREG_DUMMY, /* dummy device for NAPI poll */
2197 RTNL_LINK_INITIALIZED,
2198 RTNL_LINK_INITIALIZING,
2199 } rtnl_link_state:16;
2201 bool needs_free_netdev;
2202 void (*priv_destructor)(struct net_device *dev);
2204 #ifdef CONFIG_NETPOLL
2205 struct netpoll_info __rcu *npinfo;
2208 possible_net_t nd_net;
2210 /* mid-layer private */
2212 enum netdev_ml_priv_type ml_priv_type;
2215 struct pcpu_lstats __percpu *lstats;
2216 struct pcpu_sw_netstats __percpu *tstats;
2217 struct pcpu_dstats __percpu *dstats;
2220 #if IS_ENABLED(CONFIG_GARP)
2221 struct garp_port __rcu *garp_port;
2223 #if IS_ENABLED(CONFIG_MRP)
2224 struct mrp_port __rcu *mrp_port;
2228 const struct attribute_group *sysfs_groups[4];
2229 const struct attribute_group *sysfs_rx_queue_group;
2231 const struct rtnl_link_ops *rtnl_link_ops;
2233 /* for setting kernel sock attribute on TCP connection setup */
2234 #define GSO_MAX_SIZE 65536
2235 unsigned int gso_max_size;
2236 #define GSO_MAX_SEGS 65535
2240 const struct dcbnl_rtnl_ops *dcbnl_ops;
2243 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2244 u8 prio_tc_map[TC_BITMASK + 1];
2246 #if IS_ENABLED(CONFIG_FCOE)
2247 unsigned int fcoe_ddp_xid;
2249 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2250 struct netprio_map __rcu *priomap;
2252 struct phy_device *phydev;
2253 struct sfp_bus *sfp_bus;
2254 struct lock_class_key *qdisc_tx_busylock;
2255 struct lock_class_key *qdisc_running_key;
2257 unsigned wol_enabled:1;
2258 unsigned threaded:1;
2260 struct list_head net_notifier_list;
2262 #if IS_ENABLED(CONFIG_MACSEC)
2263 /* MACsec management functions */
2264 const struct macsec_ops *macsec_ops;
2266 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2267 struct udp_tunnel_nic *udp_tunnel_nic;
2269 /* protected by rtnl_lock */
2270 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2272 #define to_net_dev(d) container_of(d, struct net_device, dev)
2274 static inline bool netif_elide_gro(const struct net_device *dev)
2276 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2281 #define NETDEV_ALIGN 32
2284 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2286 return dev->prio_tc_map[prio & TC_BITMASK];
2290 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2292 if (tc >= dev->num_tc)
2295 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2299 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2300 void netdev_reset_tc(struct net_device *dev);
2301 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2302 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2305 int netdev_get_num_tc(struct net_device *dev)
2310 static inline void net_prefetch(void *p)
2313 #if L1_CACHE_BYTES < 128
2314 prefetch((u8 *)p + L1_CACHE_BYTES);
2318 static inline void net_prefetchw(void *p)
2321 #if L1_CACHE_BYTES < 128
2322 prefetchw((u8 *)p + L1_CACHE_BYTES);
2326 void netdev_unbind_sb_channel(struct net_device *dev,
2327 struct net_device *sb_dev);
2328 int netdev_bind_sb_channel_queue(struct net_device *dev,
2329 struct net_device *sb_dev,
2330 u8 tc, u16 count, u16 offset);
2331 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2332 static inline int netdev_get_sb_channel(struct net_device *dev)
2334 return max_t(int, -dev->num_tc, 0);
2338 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2341 return &dev->_tx[index];
2344 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2345 const struct sk_buff *skb)
2347 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2350 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2351 void (*f)(struct net_device *,
2352 struct netdev_queue *,
2358 for (i = 0; i < dev->num_tx_queues; i++)
2359 f(dev, &dev->_tx[i], arg);
2362 #define netdev_lockdep_set_classes(dev) \
2364 static struct lock_class_key qdisc_tx_busylock_key; \
2365 static struct lock_class_key qdisc_running_key; \
2366 static struct lock_class_key qdisc_xmit_lock_key; \
2367 static struct lock_class_key dev_addr_list_lock_key; \
2370 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2371 (dev)->qdisc_running_key = &qdisc_running_key; \
2372 lockdep_set_class(&(dev)->addr_list_lock, \
2373 &dev_addr_list_lock_key); \
2374 for (i = 0; i < (dev)->num_tx_queues; i++) \
2375 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2376 &qdisc_xmit_lock_key); \
2379 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2380 struct net_device *sb_dev);
2381 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2382 struct sk_buff *skb,
2383 struct net_device *sb_dev);
2385 /* returns the headroom that the master device needs to take in account
2386 * when forwarding to this dev
2388 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2390 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2393 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2395 if (dev->netdev_ops->ndo_set_rx_headroom)
2396 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2399 /* set the device rx headroom to the dev's default */
2400 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2402 netdev_set_rx_headroom(dev, -1);
2405 static inline void *netdev_get_ml_priv(struct net_device *dev,
2406 enum netdev_ml_priv_type type)
2408 if (dev->ml_priv_type != type)
2411 return dev->ml_priv;
2414 static inline void netdev_set_ml_priv(struct net_device *dev,
2416 enum netdev_ml_priv_type type)
2418 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2419 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2420 dev->ml_priv_type, type);
2421 WARN(!dev->ml_priv_type && dev->ml_priv,
2422 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2424 dev->ml_priv = ml_priv;
2425 dev->ml_priv_type = type;
2429 * Net namespace inlines
2432 struct net *dev_net(const struct net_device *dev)
2434 return read_pnet(&dev->nd_net);
2438 void dev_net_set(struct net_device *dev, struct net *net)
2440 write_pnet(&dev->nd_net, net);
2444 * netdev_priv - access network device private data
2445 * @dev: network device
2447 * Get network device private data
2449 static inline void *netdev_priv(const struct net_device *dev)
2451 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2454 /* Set the sysfs physical device reference for the network logical device
2455 * if set prior to registration will cause a symlink during initialization.
2457 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2459 /* Set the sysfs device type for the network logical device to allow
2460 * fine-grained identification of different network device types. For
2461 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2463 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2465 /* Default NAPI poll() weight
2466 * Device drivers are strongly advised to not use bigger value
2468 #define NAPI_POLL_WEIGHT 64
2471 * netif_napi_add - initialize a NAPI context
2472 * @dev: network device
2473 * @napi: NAPI context
2474 * @poll: polling function
2475 * @weight: default weight
2477 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2478 * *any* of the other NAPI-related functions.
2480 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2481 int (*poll)(struct napi_struct *, int), int weight);
2484 * netif_tx_napi_add - initialize a NAPI context
2485 * @dev: network device
2486 * @napi: NAPI context
2487 * @poll: polling function
2488 * @weight: default weight
2490 * This variant of netif_napi_add() should be used from drivers using NAPI
2491 * to exclusively poll a TX queue.
2492 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2494 static inline void netif_tx_napi_add(struct net_device *dev,
2495 struct napi_struct *napi,
2496 int (*poll)(struct napi_struct *, int),
2499 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2500 netif_napi_add(dev, napi, poll, weight);
2504 * __netif_napi_del - remove a NAPI context
2505 * @napi: NAPI context
2507 * Warning: caller must observe RCU grace period before freeing memory
2508 * containing @napi. Drivers might want to call this helper to combine
2509 * all the needed RCU grace periods into a single one.
2511 void __netif_napi_del(struct napi_struct *napi);
2514 * netif_napi_del - remove a NAPI context
2515 * @napi: NAPI context
2517 * netif_napi_del() removes a NAPI context from the network device NAPI list
2519 static inline void netif_napi_del(struct napi_struct *napi)
2521 __netif_napi_del(napi);
2525 struct napi_gro_cb {
2526 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2529 /* Length of frag0. */
2530 unsigned int frag0_len;
2532 /* This indicates where we are processing relative to skb->data. */
2535 /* This is non-zero if the packet cannot be merged with the new skb. */
2538 /* Save the IP ID here and check when we get to the transport layer */
2541 /* Number of segments aggregated. */
2544 /* Start offset for remote checksum offload */
2545 u16 gro_remcsum_start;
2547 /* jiffies when first packet was created/queued */
2550 /* Used in ipv6_gro_receive() and foo-over-udp */
2553 /* This is non-zero if the packet may be of the same flow. */
2556 /* Used in tunnel GRO receive */
2559 /* GRO checksum is valid */
2562 /* Number of checksums via CHECKSUM_UNNECESSARY */
2567 #define NAPI_GRO_FREE 1
2568 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2570 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2573 /* Used in GRE, set in fou/gue_gro_receive */
2576 /* Used to determine if flush_id can be ignored */
2579 /* Number of gro_receive callbacks this packet already went through */
2580 u8 recursion_counter:4;
2582 /* GRO is done by frag_list pointer chaining. */
2585 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2588 /* used in skb_gro_receive() slow path */
2589 struct sk_buff *last;
2592 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2594 #define GRO_RECURSION_LIMIT 15
2595 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2597 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2600 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2601 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2602 struct list_head *head,
2603 struct sk_buff *skb)
2605 if (unlikely(gro_recursion_inc_test(skb))) {
2606 NAPI_GRO_CB(skb)->flush |= 1;
2610 return cb(head, skb);
2613 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2615 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2617 struct list_head *head,
2618 struct sk_buff *skb)
2620 if (unlikely(gro_recursion_inc_test(skb))) {
2621 NAPI_GRO_CB(skb)->flush |= 1;
2625 return cb(sk, head, skb);
2628 struct packet_type {
2629 __be16 type; /* This is really htons(ether_type). */
2630 bool ignore_outgoing;
2631 struct net_device *dev; /* NULL is wildcarded here */
2632 int (*func) (struct sk_buff *,
2633 struct net_device *,
2634 struct packet_type *,
2635 struct net_device *);
2636 void (*list_func) (struct list_head *,
2637 struct packet_type *,
2638 struct net_device *);
2639 bool (*id_match)(struct packet_type *ptype,
2641 void *af_packet_priv;
2642 struct list_head list;
2645 struct offload_callbacks {
2646 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2647 netdev_features_t features);
2648 struct sk_buff *(*gro_receive)(struct list_head *head,
2649 struct sk_buff *skb);
2650 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2653 struct packet_offload {
2654 __be16 type; /* This is really htons(ether_type). */
2656 struct offload_callbacks callbacks;
2657 struct list_head list;
2660 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2661 struct pcpu_sw_netstats {
2666 struct u64_stats_sync syncp;
2667 } __aligned(4 * sizeof(u64));
2669 struct pcpu_lstats {
2670 u64_stats_t packets;
2672 struct u64_stats_sync syncp;
2673 } __aligned(2 * sizeof(u64));
2675 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2677 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2679 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2681 u64_stats_update_begin(&tstats->syncp);
2682 tstats->rx_bytes += len;
2683 tstats->rx_packets++;
2684 u64_stats_update_end(&tstats->syncp);
2687 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2688 unsigned int packets,
2691 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2693 u64_stats_update_begin(&tstats->syncp);
2694 tstats->tx_bytes += len;
2695 tstats->tx_packets += packets;
2696 u64_stats_update_end(&tstats->syncp);
2699 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2701 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2703 u64_stats_update_begin(&lstats->syncp);
2704 u64_stats_add(&lstats->bytes, len);
2705 u64_stats_inc(&lstats->packets);
2706 u64_stats_update_end(&lstats->syncp);
2709 #define __netdev_alloc_pcpu_stats(type, gfp) \
2711 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2714 for_each_possible_cpu(__cpu) { \
2715 typeof(type) *stat; \
2716 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2717 u64_stats_init(&stat->syncp); \
2723 #define netdev_alloc_pcpu_stats(type) \
2724 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2726 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2728 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2731 for_each_possible_cpu(__cpu) { \
2732 typeof(type) *stat; \
2733 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2734 u64_stats_init(&stat->syncp); \
2740 enum netdev_lag_tx_type {
2741 NETDEV_LAG_TX_TYPE_UNKNOWN,
2742 NETDEV_LAG_TX_TYPE_RANDOM,
2743 NETDEV_LAG_TX_TYPE_BROADCAST,
2744 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2745 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2746 NETDEV_LAG_TX_TYPE_HASH,
2749 enum netdev_lag_hash {
2750 NETDEV_LAG_HASH_NONE,
2752 NETDEV_LAG_HASH_L34,
2753 NETDEV_LAG_HASH_L23,
2754 NETDEV_LAG_HASH_E23,
2755 NETDEV_LAG_HASH_E34,
2756 NETDEV_LAG_HASH_VLAN_SRCMAC,
2757 NETDEV_LAG_HASH_UNKNOWN,
2760 struct netdev_lag_upper_info {
2761 enum netdev_lag_tx_type tx_type;
2762 enum netdev_lag_hash hash_type;
2765 struct netdev_lag_lower_state_info {
2770 #include <linux/notifier.h>
2772 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2773 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2777 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2779 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2780 detected a hardware crash and restarted
2781 - we can use this eg to kick tcp sessions
2783 NETDEV_CHANGE, /* Notify device state change */
2786 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2787 NETDEV_CHANGEADDR, /* notify after the address change */
2788 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2792 NETDEV_BONDING_FAILOVER,
2794 NETDEV_PRE_TYPE_CHANGE,
2795 NETDEV_POST_TYPE_CHANGE,
2798 NETDEV_NOTIFY_PEERS,
2802 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2803 NETDEV_CHANGEINFODATA,
2804 NETDEV_BONDING_INFO,
2805 NETDEV_PRECHANGEUPPER,
2806 NETDEV_CHANGELOWERSTATE,
2807 NETDEV_UDP_TUNNEL_PUSH_INFO,
2808 NETDEV_UDP_TUNNEL_DROP_INFO,
2809 NETDEV_CHANGE_TX_QUEUE_LEN,
2810 NETDEV_CVLAN_FILTER_PUSH_INFO,
2811 NETDEV_CVLAN_FILTER_DROP_INFO,
2812 NETDEV_SVLAN_FILTER_PUSH_INFO,
2813 NETDEV_SVLAN_FILTER_DROP_INFO,
2815 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2817 int register_netdevice_notifier(struct notifier_block *nb);
2818 int unregister_netdevice_notifier(struct notifier_block *nb);
2819 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2820 int unregister_netdevice_notifier_net(struct net *net,
2821 struct notifier_block *nb);
2822 int register_netdevice_notifier_dev_net(struct net_device *dev,
2823 struct notifier_block *nb,
2824 struct netdev_net_notifier *nn);
2825 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2826 struct notifier_block *nb,
2827 struct netdev_net_notifier *nn);
2829 struct netdev_notifier_info {
2830 struct net_device *dev;
2831 struct netlink_ext_ack *extack;
2834 struct netdev_notifier_info_ext {
2835 struct netdev_notifier_info info; /* must be first */
2841 struct netdev_notifier_change_info {
2842 struct netdev_notifier_info info; /* must be first */
2843 unsigned int flags_changed;
2846 struct netdev_notifier_changeupper_info {
2847 struct netdev_notifier_info info; /* must be first */
2848 struct net_device *upper_dev; /* new upper dev */
2849 bool master; /* is upper dev master */
2850 bool linking; /* is the notification for link or unlink */
2851 void *upper_info; /* upper dev info */
2854 struct netdev_notifier_changelowerstate_info {
2855 struct netdev_notifier_info info; /* must be first */
2856 void *lower_state_info; /* is lower dev state */
2859 struct netdev_notifier_pre_changeaddr_info {
2860 struct netdev_notifier_info info; /* must be first */
2861 const unsigned char *dev_addr;
2864 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2865 struct net_device *dev)
2868 info->extack = NULL;
2871 static inline struct net_device *
2872 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2877 static inline struct netlink_ext_ack *
2878 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2880 return info->extack;
2883 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2886 extern rwlock_t dev_base_lock; /* Device list lock */
2888 #define for_each_netdev(net, d) \
2889 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2890 #define for_each_netdev_reverse(net, d) \
2891 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2892 #define for_each_netdev_rcu(net, d) \
2893 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2894 #define for_each_netdev_safe(net, d, n) \
2895 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2896 #define for_each_netdev_continue(net, d) \
2897 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2898 #define for_each_netdev_continue_reverse(net, d) \
2899 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2901 #define for_each_netdev_continue_rcu(net, d) \
2902 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2903 #define for_each_netdev_in_bond_rcu(bond, slave) \
2904 for_each_netdev_rcu(&init_net, slave) \
2905 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2906 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2908 static inline struct net_device *next_net_device(struct net_device *dev)
2910 struct list_head *lh;
2914 lh = dev->dev_list.next;
2915 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2918 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2920 struct list_head *lh;
2924 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2925 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2928 static inline struct net_device *first_net_device(struct net *net)
2930 return list_empty(&net->dev_base_head) ? NULL :
2931 net_device_entry(net->dev_base_head.next);
2934 static inline struct net_device *first_net_device_rcu(struct net *net)
2936 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2938 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2941 int netdev_boot_setup_check(struct net_device *dev);
2942 unsigned long netdev_boot_base(const char *prefix, int unit);
2943 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2944 const char *hwaddr);
2945 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2946 void dev_add_pack(struct packet_type *pt);
2947 void dev_remove_pack(struct packet_type *pt);
2948 void __dev_remove_pack(struct packet_type *pt);
2949 void dev_add_offload(struct packet_offload *po);
2950 void dev_remove_offload(struct packet_offload *po);
2952 int dev_get_iflink(const struct net_device *dev);
2953 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2954 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2955 struct net_device_path_stack *stack);
2956 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2957 unsigned short mask);
2958 struct net_device *dev_get_by_name(struct net *net, const char *name);
2959 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2960 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2961 int dev_alloc_name(struct net_device *dev, const char *name);
2962 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2963 void dev_close(struct net_device *dev);
2964 void dev_close_many(struct list_head *head, bool unlink);
2965 void dev_disable_lro(struct net_device *dev);
2966 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2967 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2968 struct net_device *sb_dev);
2969 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2970 struct net_device *sb_dev);
2972 int dev_queue_xmit(struct sk_buff *skb);
2973 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2974 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2976 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2980 ret = __dev_direct_xmit(skb, queue_id);
2981 if (!dev_xmit_complete(ret))
2986 int register_netdevice(struct net_device *dev);
2987 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2988 void unregister_netdevice_many(struct list_head *head);
2989 static inline void unregister_netdevice(struct net_device *dev)
2991 unregister_netdevice_queue(dev, NULL);
2994 int netdev_refcnt_read(const struct net_device *dev);
2995 void free_netdev(struct net_device *dev);
2996 void netdev_freemem(struct net_device *dev);
2997 int init_dummy_netdev(struct net_device *dev);
2999 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3000 struct sk_buff *skb,
3002 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3004 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3005 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3006 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3007 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3008 int netdev_get_name(struct net *net, char *name, int ifindex);
3009 int dev_restart(struct net_device *dev);
3010 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3011 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3013 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3015 return NAPI_GRO_CB(skb)->data_offset;
3018 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3020 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3023 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3025 NAPI_GRO_CB(skb)->data_offset += len;
3028 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3029 unsigned int offset)
3031 return NAPI_GRO_CB(skb)->frag0 + offset;
3034 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3036 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3039 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3041 NAPI_GRO_CB(skb)->frag0 = NULL;
3042 NAPI_GRO_CB(skb)->frag0_len = 0;
3045 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3046 unsigned int offset)
3048 if (!pskb_may_pull(skb, hlen))
3051 skb_gro_frag0_invalidate(skb);
3052 return skb->data + offset;
3055 static inline void *skb_gro_network_header(struct sk_buff *skb)
3057 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3058 skb_network_offset(skb);
3061 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3062 const void *start, unsigned int len)
3064 if (NAPI_GRO_CB(skb)->csum_valid)
3065 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3066 csum_partial(start, len, 0));
3069 /* GRO checksum functions. These are logical equivalents of the normal
3070 * checksum functions (in skbuff.h) except that they operate on the GRO
3071 * offsets and fields in sk_buff.
3074 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3076 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3078 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3081 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3085 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3086 skb_checksum_start_offset(skb) <
3087 skb_gro_offset(skb)) &&
3088 !skb_at_gro_remcsum_start(skb) &&
3089 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3090 (!zero_okay || check));
3093 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3096 if (NAPI_GRO_CB(skb)->csum_valid &&
3097 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3100 NAPI_GRO_CB(skb)->csum = psum;
3102 return __skb_gro_checksum_complete(skb);
3105 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3107 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3108 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3109 NAPI_GRO_CB(skb)->csum_cnt--;
3111 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3112 * verified a new top level checksum or an encapsulated one
3113 * during GRO. This saves work if we fallback to normal path.
3115 __skb_incr_checksum_unnecessary(skb);
3119 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3122 __sum16 __ret = 0; \
3123 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3124 __ret = __skb_gro_checksum_validate_complete(skb, \
3125 compute_pseudo(skb, proto)); \
3127 skb_gro_incr_csum_unnecessary(skb); \
3131 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3132 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3134 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3136 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3138 #define skb_gro_checksum_simple_validate(skb) \
3139 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3141 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3143 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3144 !NAPI_GRO_CB(skb)->csum_valid);
3147 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3150 NAPI_GRO_CB(skb)->csum = ~pseudo;
3151 NAPI_GRO_CB(skb)->csum_valid = 1;
3154 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3156 if (__skb_gro_checksum_convert_check(skb)) \
3157 __skb_gro_checksum_convert(skb, \
3158 compute_pseudo(skb, proto)); \
3161 struct gro_remcsum {
3166 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3172 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3173 unsigned int off, size_t hdrlen,
3174 int start, int offset,
3175 struct gro_remcsum *grc,
3179 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3181 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3184 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3188 ptr = skb_gro_header_fast(skb, off);
3189 if (skb_gro_header_hard(skb, off + plen)) {
3190 ptr = skb_gro_header_slow(skb, off + plen, off);
3195 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3198 /* Adjust skb->csum since we changed the packet */
3199 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3201 grc->offset = off + hdrlen + offset;
3207 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3208 struct gro_remcsum *grc)
3211 size_t plen = grc->offset + sizeof(u16);
3216 ptr = skb_gro_header_fast(skb, grc->offset);
3217 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3218 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3223 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3226 #ifdef CONFIG_XFRM_OFFLOAD
3227 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3229 if (PTR_ERR(pp) != -EINPROGRESS)
3230 NAPI_GRO_CB(skb)->flush |= flush;
3232 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3235 struct gro_remcsum *grc)
3237 if (PTR_ERR(pp) != -EINPROGRESS) {
3238 NAPI_GRO_CB(skb)->flush |= flush;
3239 skb_gro_remcsum_cleanup(skb, grc);
3240 skb->remcsum_offload = 0;
3244 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3246 NAPI_GRO_CB(skb)->flush |= flush;
3248 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3251 struct gro_remcsum *grc)
3253 NAPI_GRO_CB(skb)->flush |= flush;
3254 skb_gro_remcsum_cleanup(skb, grc);
3255 skb->remcsum_offload = 0;
3259 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3260 unsigned short type,
3261 const void *daddr, const void *saddr,
3264 if (!dev->header_ops || !dev->header_ops->create)
3267 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3270 static inline int dev_parse_header(const struct sk_buff *skb,
3271 unsigned char *haddr)
3273 const struct net_device *dev = skb->dev;
3275 if (!dev->header_ops || !dev->header_ops->parse)
3277 return dev->header_ops->parse(skb, haddr);
3280 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3282 const struct net_device *dev = skb->dev;
3284 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3286 return dev->header_ops->parse_protocol(skb);
3289 /* ll_header must have at least hard_header_len allocated */
3290 static inline bool dev_validate_header(const struct net_device *dev,
3291 char *ll_header, int len)
3293 if (likely(len >= dev->hard_header_len))
3295 if (len < dev->min_header_len)
3298 if (capable(CAP_SYS_RAWIO)) {
3299 memset(ll_header + len, 0, dev->hard_header_len - len);
3303 if (dev->header_ops && dev->header_ops->validate)
3304 return dev->header_ops->validate(ll_header, len);
3309 static inline bool dev_has_header(const struct net_device *dev)
3311 return dev->header_ops && dev->header_ops->create;
3314 #ifdef CONFIG_NET_FLOW_LIMIT
3315 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3316 struct sd_flow_limit {
3318 unsigned int num_buckets;
3319 unsigned int history_head;
3320 u16 history[FLOW_LIMIT_HISTORY];
3324 extern int netdev_flow_limit_table_len;
3325 #endif /* CONFIG_NET_FLOW_LIMIT */
3328 * Incoming packets are placed on per-CPU queues
3330 struct softnet_data {
3331 struct list_head poll_list;
3332 struct sk_buff_head process_queue;
3335 unsigned int processed;
3336 unsigned int time_squeeze;
3337 unsigned int received_rps;
3339 struct softnet_data *rps_ipi_list;
3341 #ifdef CONFIG_NET_FLOW_LIMIT
3342 struct sd_flow_limit __rcu *flow_limit;
3344 struct Qdisc *output_queue;
3345 struct Qdisc **output_queue_tailp;
3346 struct sk_buff *completion_queue;
3347 #ifdef CONFIG_XFRM_OFFLOAD
3348 struct sk_buff_head xfrm_backlog;
3350 /* written and read only by owning cpu: */
3356 /* input_queue_head should be written by cpu owning this struct,
3357 * and only read by other cpus. Worth using a cache line.
3359 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3361 /* Elements below can be accessed between CPUs for RPS/RFS */
3362 call_single_data_t csd ____cacheline_aligned_in_smp;
3363 struct softnet_data *rps_ipi_next;
3365 unsigned int input_queue_tail;
3367 unsigned int dropped;
3368 struct sk_buff_head input_pkt_queue;
3369 struct napi_struct backlog;
3373 static inline void input_queue_head_incr(struct softnet_data *sd)
3376 sd->input_queue_head++;
3380 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3381 unsigned int *qtail)
3384 *qtail = ++sd->input_queue_tail;
3388 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3390 static inline int dev_recursion_level(void)
3392 return this_cpu_read(softnet_data.xmit.recursion);
3395 #define XMIT_RECURSION_LIMIT 8
3396 static inline bool dev_xmit_recursion(void)
3398 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3399 XMIT_RECURSION_LIMIT);
3402 static inline void dev_xmit_recursion_inc(void)
3404 __this_cpu_inc(softnet_data.xmit.recursion);
3407 static inline void dev_xmit_recursion_dec(void)
3409 __this_cpu_dec(softnet_data.xmit.recursion);
3412 void __netif_schedule(struct Qdisc *q);
3413 void netif_schedule_queue(struct netdev_queue *txq);
3415 static inline void netif_tx_schedule_all(struct net_device *dev)
3419 for (i = 0; i < dev->num_tx_queues; i++)
3420 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3423 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3425 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3429 * netif_start_queue - allow transmit
3430 * @dev: network device
3432 * Allow upper layers to call the device hard_start_xmit routine.
3434 static inline void netif_start_queue(struct net_device *dev)
3436 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3439 static inline void netif_tx_start_all_queues(struct net_device *dev)
3443 for (i = 0; i < dev->num_tx_queues; i++) {
3444 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3445 netif_tx_start_queue(txq);
3449 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3452 * netif_wake_queue - restart transmit
3453 * @dev: network device
3455 * Allow upper layers to call the device hard_start_xmit routine.
3456 * Used for flow control when transmit resources are available.
3458 static inline void netif_wake_queue(struct net_device *dev)
3460 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3463 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3467 for (i = 0; i < dev->num_tx_queues; i++) {
3468 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3469 netif_tx_wake_queue(txq);
3473 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3475 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3479 * netif_stop_queue - stop transmitted packets
3480 * @dev: network device
3482 * Stop upper layers calling the device hard_start_xmit routine.
3483 * Used for flow control when transmit resources are unavailable.
3485 static inline void netif_stop_queue(struct net_device *dev)
3487 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3490 void netif_tx_stop_all_queues(struct net_device *dev);
3492 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3494 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3498 * netif_queue_stopped - test if transmit queue is flowblocked
3499 * @dev: network device
3501 * Test if transmit queue on device is currently unable to send.
3503 static inline bool netif_queue_stopped(const struct net_device *dev)
3505 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3508 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3510 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3514 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3516 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3520 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3522 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3526 * netdev_queue_set_dql_min_limit - set dql minimum limit
3527 * @dev_queue: pointer to transmit queue
3528 * @min_limit: dql minimum limit
3530 * Forces xmit_more() to return true until the minimum threshold
3531 * defined by @min_limit is reached (or until the tx queue is
3532 * empty). Warning: to be use with care, misuse will impact the
3535 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3536 unsigned int min_limit)
3539 dev_queue->dql.min_limit = min_limit;
3544 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3545 * @dev_queue: pointer to transmit queue
3547 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3548 * to give appropriate hint to the CPU.
3550 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3553 prefetchw(&dev_queue->dql.num_queued);
3558 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3559 * @dev_queue: pointer to transmit queue
3561 * BQL enabled drivers might use this helper in their TX completion path,
3562 * to give appropriate hint to the CPU.
3564 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3567 prefetchw(&dev_queue->dql.limit);
3571 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3575 dql_queued(&dev_queue->dql, bytes);
3577 if (likely(dql_avail(&dev_queue->dql) >= 0))
3580 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3583 * The XOFF flag must be set before checking the dql_avail below,
3584 * because in netdev_tx_completed_queue we update the dql_completed
3585 * before checking the XOFF flag.
3589 /* check again in case another CPU has just made room avail */
3590 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3591 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3595 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3596 * that they should not test BQL status themselves.
3597 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3599 * Returns true if the doorbell must be used to kick the NIC.
3601 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3607 dql_queued(&dev_queue->dql, bytes);
3609 return netif_tx_queue_stopped(dev_queue);
3611 netdev_tx_sent_queue(dev_queue, bytes);
3616 * netdev_sent_queue - report the number of bytes queued to hardware
3617 * @dev: network device
3618 * @bytes: number of bytes queued to the hardware device queue
3620 * Report the number of bytes queued for sending/completion to the network
3621 * device hardware queue. @bytes should be a good approximation and should
3622 * exactly match netdev_completed_queue() @bytes
3624 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3626 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3629 static inline bool __netdev_sent_queue(struct net_device *dev,
3633 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3637 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3638 unsigned int pkts, unsigned int bytes)
3641 if (unlikely(!bytes))
3644 dql_completed(&dev_queue->dql, bytes);
3647 * Without the memory barrier there is a small possiblity that
3648 * netdev_tx_sent_queue will miss the update and cause the queue to
3649 * be stopped forever
3653 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3656 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3657 netif_schedule_queue(dev_queue);
3662 * netdev_completed_queue - report bytes and packets completed by device
3663 * @dev: network device
3664 * @pkts: actual number of packets sent over the medium
3665 * @bytes: actual number of bytes sent over the medium
3667 * Report the number of bytes and packets transmitted by the network device
3668 * hardware queue over the physical medium, @bytes must exactly match the
3669 * @bytes amount passed to netdev_sent_queue()
3671 static inline void netdev_completed_queue(struct net_device *dev,
3672 unsigned int pkts, unsigned int bytes)
3674 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3677 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3680 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3686 * netdev_reset_queue - reset the packets and bytes count of a network device
3687 * @dev_queue: network device
3689 * Reset the bytes and packet count of a network device and clear the
3690 * software flow control OFF bit for this network device
3692 static inline void netdev_reset_queue(struct net_device *dev_queue)
3694 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3698 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3699 * @dev: network device
3700 * @queue_index: given tx queue index
3702 * Returns 0 if given tx queue index >= number of device tx queues,
3703 * otherwise returns the originally passed tx queue index.
3705 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3707 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3708 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3709 dev->name, queue_index,
3710 dev->real_num_tx_queues);
3718 * netif_running - test if up
3719 * @dev: network device
3721 * Test if the device has been brought up.
3723 static inline bool netif_running(const struct net_device *dev)
3725 return test_bit(__LINK_STATE_START, &dev->state);
3729 * Routines to manage the subqueues on a device. We only need start,
3730 * stop, and a check if it's stopped. All other device management is
3731 * done at the overall netdevice level.
3732 * Also test the device if we're multiqueue.
3736 * netif_start_subqueue - allow sending packets on subqueue
3737 * @dev: network device
3738 * @queue_index: sub queue index
3740 * Start individual transmit queue of a device with multiple transmit queues.
3742 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3744 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3746 netif_tx_start_queue(txq);
3750 * netif_stop_subqueue - stop sending packets on subqueue
3751 * @dev: network device
3752 * @queue_index: sub queue index
3754 * Stop individual transmit queue of a device with multiple transmit queues.
3756 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3758 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3759 netif_tx_stop_queue(txq);
3763 * __netif_subqueue_stopped - test status of subqueue
3764 * @dev: network device
3765 * @queue_index: sub queue index
3767 * Check individual transmit queue of a device with multiple transmit queues.
3769 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3772 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3774 return netif_tx_queue_stopped(txq);
3778 * netif_subqueue_stopped - test status of subqueue
3779 * @dev: network device
3780 * @skb: sub queue buffer pointer
3782 * Check individual transmit queue of a device with multiple transmit queues.
3784 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3785 struct sk_buff *skb)
3787 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3791 * netif_wake_subqueue - allow sending packets on subqueue
3792 * @dev: network device
3793 * @queue_index: sub queue index
3795 * Resume individual transmit queue of a device with multiple transmit queues.
3797 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3799 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3801 netif_tx_wake_queue(txq);
3805 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3807 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3808 u16 index, enum xps_map_type type);
3811 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3812 * @j: CPU/Rx queue index
3813 * @mask: bitmask of all cpus/rx queues
3814 * @nr_bits: number of bits in the bitmask
3816 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3818 static inline bool netif_attr_test_mask(unsigned long j,
3819 const unsigned long *mask,
3820 unsigned int nr_bits)
3822 cpu_max_bits_warn(j, nr_bits);
3823 return test_bit(j, mask);
3827 * netif_attr_test_online - Test for online CPU/Rx queue
3828 * @j: CPU/Rx queue index
3829 * @online_mask: bitmask for CPUs/Rx queues that are online
3830 * @nr_bits: number of bits in the bitmask
3832 * Returns true if a CPU/Rx queue is online.
3834 static inline bool netif_attr_test_online(unsigned long j,
3835 const unsigned long *online_mask,
3836 unsigned int nr_bits)
3838 cpu_max_bits_warn(j, nr_bits);
3841 return test_bit(j, online_mask);
3843 return (j < nr_bits);
3847 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3848 * @n: CPU/Rx queue index
3849 * @srcp: the cpumask/Rx queue mask pointer
3850 * @nr_bits: number of bits in the bitmask
3852 * Returns >= nr_bits if no further CPUs/Rx queues set.
3854 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3855 unsigned int nr_bits)
3857 /* -1 is a legal arg here. */
3859 cpu_max_bits_warn(n, nr_bits);
3862 return find_next_bit(srcp, nr_bits, n + 1);
3868 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3869 * @n: CPU/Rx queue index
3870 * @src1p: the first CPUs/Rx queues mask pointer
3871 * @src2p: the second CPUs/Rx queues mask pointer
3872 * @nr_bits: number of bits in the bitmask
3874 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3876 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3877 const unsigned long *src2p,
3878 unsigned int nr_bits)
3880 /* -1 is a legal arg here. */
3882 cpu_max_bits_warn(n, nr_bits);
3885 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3887 return find_next_bit(src1p, nr_bits, n + 1);
3889 return find_next_bit(src2p, nr_bits, n + 1);
3894 static inline int netif_set_xps_queue(struct net_device *dev,
3895 const struct cpumask *mask,
3901 static inline int __netif_set_xps_queue(struct net_device *dev,
3902 const unsigned long *mask,
3903 u16 index, enum xps_map_type type)
3910 * netif_is_multiqueue - test if device has multiple transmit queues
3911 * @dev: network device
3913 * Check if device has multiple transmit queues
3915 static inline bool netif_is_multiqueue(const struct net_device *dev)
3917 return dev->num_tx_queues > 1;
3920 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3923 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3925 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3928 dev->real_num_rx_queues = rxqs;
3933 static inline struct netdev_rx_queue *
3934 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3936 return dev->_rx + rxq;
3940 static inline unsigned int get_netdev_rx_queue_index(
3941 struct netdev_rx_queue *queue)
3943 struct net_device *dev = queue->dev;
3944 int index = queue - dev->_rx;
3946 BUG_ON(index >= dev->num_rx_queues);
3951 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3952 int netif_get_num_default_rss_queues(void);
3954 enum skb_free_reason {
3955 SKB_REASON_CONSUMED,
3959 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3960 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3963 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3964 * interrupt context or with hardware interrupts being disabled.
3965 * (in_irq() || irqs_disabled())
3967 * We provide four helpers that can be used in following contexts :
3969 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3970 * replacing kfree_skb(skb)
3972 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3973 * Typically used in place of consume_skb(skb) in TX completion path
3975 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3976 * replacing kfree_skb(skb)
3978 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3979 * and consumed a packet. Used in place of consume_skb(skb)
3981 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3983 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3986 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3988 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3991 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3993 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3996 static inline void dev_consume_skb_any(struct sk_buff *skb)
3998 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
4001 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
4002 struct bpf_prog *xdp_prog);
4003 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
4004 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
4005 int netif_rx(struct sk_buff *skb);
4006 int netif_rx_ni(struct sk_buff *skb);
4007 int netif_rx_any_context(struct sk_buff *skb);
4008 int netif_receive_skb(struct sk_buff *skb);
4009 int netif_receive_skb_core(struct sk_buff *skb);
4010 void netif_receive_skb_list(struct list_head *head);
4011 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4012 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4013 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4014 gro_result_t napi_gro_frags(struct napi_struct *napi);
4015 struct packet_offload *gro_find_receive_by_type(__be16 type);
4016 struct packet_offload *gro_find_complete_by_type(__be16 type);
4018 static inline void napi_free_frags(struct napi_struct *napi)
4020 kfree_skb(napi->skb);
4024 bool netdev_is_rx_handler_busy(struct net_device *dev);
4025 int netdev_rx_handler_register(struct net_device *dev,
4026 rx_handler_func_t *rx_handler,
4027 void *rx_handler_data);
4028 void netdev_rx_handler_unregister(struct net_device *dev);
4030 bool dev_valid_name(const char *name);
4031 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4032 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4033 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4034 void __user *data, bool *need_copyout);
4035 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4036 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4037 unsigned int dev_get_flags(const struct net_device *);
4038 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4039 struct netlink_ext_ack *extack);
4040 int dev_change_flags(struct net_device *dev, unsigned int flags,
4041 struct netlink_ext_ack *extack);
4042 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4043 unsigned int gchanges);
4044 int dev_change_name(struct net_device *, const char *);
4045 int dev_set_alias(struct net_device *, const char *, size_t);
4046 int dev_get_alias(const struct net_device *, char *, size_t);
4047 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4048 const char *pat, int new_ifindex);
4050 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4053 return __dev_change_net_namespace(dev, net, pat, 0);
4055 int __dev_set_mtu(struct net_device *, int);
4056 int dev_validate_mtu(struct net_device *dev, int mtu,
4057 struct netlink_ext_ack *extack);
4058 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4059 struct netlink_ext_ack *extack);
4060 int dev_set_mtu(struct net_device *, int);
4061 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4062 void dev_set_group(struct net_device *, int);
4063 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4064 struct netlink_ext_ack *extack);
4065 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4066 struct netlink_ext_ack *extack);
4067 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4068 struct netlink_ext_ack *extack);
4069 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4070 int dev_change_carrier(struct net_device *, bool new_carrier);
4071 int dev_get_phys_port_id(struct net_device *dev,
4072 struct netdev_phys_item_id *ppid);
4073 int dev_get_phys_port_name(struct net_device *dev,
4074 char *name, size_t len);
4075 int dev_get_port_parent_id(struct net_device *dev,
4076 struct netdev_phys_item_id *ppid, bool recurse);
4077 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4078 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4079 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4080 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4082 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4083 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4084 struct netdev_queue *txq, int *ret);
4086 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4087 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4088 int fd, int expected_fd, u32 flags);
4089 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4090 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4092 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4093 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4094 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4095 bool is_skb_forwardable(const struct net_device *dev,
4096 const struct sk_buff *skb);
4098 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4099 const struct sk_buff *skb,
4100 const bool check_mtu)
4102 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4105 if (!(dev->flags & IFF_UP))
4111 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4112 if (skb->len <= len)
4115 /* if TSO is enabled, we don't care about the length as the packet
4116 * could be forwarded without being segmented before
4118 if (skb_is_gso(skb))
4124 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4125 struct sk_buff *skb,
4126 const bool check_mtu)
4128 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4129 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4130 atomic_long_inc(&dev->rx_dropped);
4135 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4140 bool dev_nit_active(struct net_device *dev);
4141 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4143 extern int netdev_budget;
4144 extern unsigned int netdev_budget_usecs;
4146 /* Called by rtnetlink.c:rtnl_unlock() */
4147 void netdev_run_todo(void);
4150 * dev_put - release reference to device
4151 * @dev: network device
4153 * Release reference to device to allow it to be freed.
4155 static inline void dev_put(struct net_device *dev)
4157 #ifdef CONFIG_PCPU_DEV_REFCNT
4158 this_cpu_dec(*dev->pcpu_refcnt);
4160 refcount_dec(&dev->dev_refcnt);
4165 * dev_hold - get reference to device
4166 * @dev: network device
4168 * Hold reference to device to keep it from being freed.
4170 static inline void dev_hold(struct net_device *dev)
4172 #ifdef CONFIG_PCPU_DEV_REFCNT
4173 this_cpu_inc(*dev->pcpu_refcnt);
4175 refcount_inc(&dev->dev_refcnt);
4179 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4180 * and _off may be called from IRQ context, but it is caller
4181 * who is responsible for serialization of these calls.
4183 * The name carrier is inappropriate, these functions should really be
4184 * called netif_lowerlayer_*() because they represent the state of any
4185 * kind of lower layer not just hardware media.
4188 void linkwatch_init_dev(struct net_device *dev);
4189 void linkwatch_fire_event(struct net_device *dev);
4190 void linkwatch_forget_dev(struct net_device *dev);
4193 * netif_carrier_ok - test if carrier present
4194 * @dev: network device
4196 * Check if carrier is present on device
4198 static inline bool netif_carrier_ok(const struct net_device *dev)
4200 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4203 unsigned long dev_trans_start(struct net_device *dev);
4205 void __netdev_watchdog_up(struct net_device *dev);
4207 void netif_carrier_on(struct net_device *dev);
4208 void netif_carrier_off(struct net_device *dev);
4209 void netif_carrier_event(struct net_device *dev);
4212 * netif_dormant_on - mark device as dormant.
4213 * @dev: network device
4215 * Mark device as dormant (as per RFC2863).
4217 * The dormant state indicates that the relevant interface is not
4218 * actually in a condition to pass packets (i.e., it is not 'up') but is
4219 * in a "pending" state, waiting for some external event. For "on-
4220 * demand" interfaces, this new state identifies the situation where the
4221 * interface is waiting for events to place it in the up state.
4223 static inline void netif_dormant_on(struct net_device *dev)
4225 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4226 linkwatch_fire_event(dev);
4230 * netif_dormant_off - set device as not dormant.
4231 * @dev: network device
4233 * Device is not in dormant state.
4235 static inline void netif_dormant_off(struct net_device *dev)
4237 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4238 linkwatch_fire_event(dev);
4242 * netif_dormant - test if device is dormant
4243 * @dev: network device
4245 * Check if device is dormant.
4247 static inline bool netif_dormant(const struct net_device *dev)
4249 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4254 * netif_testing_on - mark device as under test.
4255 * @dev: network device
4257 * Mark device as under test (as per RFC2863).
4259 * The testing state indicates that some test(s) must be performed on
4260 * the interface. After completion, of the test, the interface state
4261 * will change to up, dormant, or down, as appropriate.
4263 static inline void netif_testing_on(struct net_device *dev)
4265 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4266 linkwatch_fire_event(dev);
4270 * netif_testing_off - set device as not under test.
4271 * @dev: network device
4273 * Device is not in testing state.
4275 static inline void netif_testing_off(struct net_device *dev)
4277 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4278 linkwatch_fire_event(dev);
4282 * netif_testing - test if device is under test
4283 * @dev: network device
4285 * Check if device is under test
4287 static inline bool netif_testing(const struct net_device *dev)
4289 return test_bit(__LINK_STATE_TESTING, &dev->state);
4294 * netif_oper_up - test if device is operational
4295 * @dev: network device
4297 * Check if carrier is operational
4299 static inline bool netif_oper_up(const struct net_device *dev)
4301 return (dev->operstate == IF_OPER_UP ||
4302 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4306 * netif_device_present - is device available or removed
4307 * @dev: network device
4309 * Check if device has not been removed from system.
4311 static inline bool netif_device_present(const struct net_device *dev)
4313 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4316 void netif_device_detach(struct net_device *dev);
4318 void netif_device_attach(struct net_device *dev);
4321 * Network interface message level settings
4326 NETIF_MSG_PROBE_BIT,
4328 NETIF_MSG_TIMER_BIT,
4329 NETIF_MSG_IFDOWN_BIT,
4331 NETIF_MSG_RX_ERR_BIT,
4332 NETIF_MSG_TX_ERR_BIT,
4333 NETIF_MSG_TX_QUEUED_BIT,
4335 NETIF_MSG_TX_DONE_BIT,
4336 NETIF_MSG_RX_STATUS_BIT,
4337 NETIF_MSG_PKTDATA_BIT,
4341 /* When you add a new bit above, update netif_msg_class_names array
4342 * in net/ethtool/common.c
4344 NETIF_MSG_CLASS_COUNT,
4346 /* Both ethtool_ops interface and internal driver implementation use u32 */
4347 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4349 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4350 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4352 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4353 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4354 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4355 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4356 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4357 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4358 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4359 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4360 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4361 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4362 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4363 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4364 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4365 #define NETIF_MSG_HW __NETIF_MSG(HW)
4366 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4368 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4369 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4370 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4371 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4372 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4373 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4374 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4375 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4376 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4377 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4378 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4379 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4380 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4381 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4382 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4384 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4387 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4388 return default_msg_enable_bits;
4389 if (debug_value == 0) /* no output */
4391 /* set low N bits */
4392 return (1U << debug_value) - 1;
4395 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4397 spin_lock(&txq->_xmit_lock);
4398 txq->xmit_lock_owner = cpu;
4401 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4403 __acquire(&txq->_xmit_lock);
4407 static inline void __netif_tx_release(struct netdev_queue *txq)
4409 __release(&txq->_xmit_lock);
4412 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4414 spin_lock_bh(&txq->_xmit_lock);
4415 txq->xmit_lock_owner = smp_processor_id();
4418 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4420 bool ok = spin_trylock(&txq->_xmit_lock);
4422 txq->xmit_lock_owner = smp_processor_id();
4426 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4428 txq->xmit_lock_owner = -1;
4429 spin_unlock(&txq->_xmit_lock);
4432 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4434 txq->xmit_lock_owner = -1;
4435 spin_unlock_bh(&txq->_xmit_lock);
4438 static inline void txq_trans_update(struct netdev_queue *txq)
4440 if (txq->xmit_lock_owner != -1)
4441 txq->trans_start = jiffies;
4444 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4445 static inline void netif_trans_update(struct net_device *dev)
4447 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4449 if (txq->trans_start != jiffies)
4450 txq->trans_start = jiffies;
4454 * netif_tx_lock - grab network device transmit lock
4455 * @dev: network device
4457 * Get network device transmit lock
4459 static inline void netif_tx_lock(struct net_device *dev)
4464 spin_lock(&dev->tx_global_lock);
4465 cpu = smp_processor_id();
4466 for (i = 0; i < dev->num_tx_queues; i++) {
4467 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4469 /* We are the only thread of execution doing a
4470 * freeze, but we have to grab the _xmit_lock in
4471 * order to synchronize with threads which are in
4472 * the ->hard_start_xmit() handler and already
4473 * checked the frozen bit.
4475 __netif_tx_lock(txq, cpu);
4476 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4477 __netif_tx_unlock(txq);
4481 static inline void netif_tx_lock_bh(struct net_device *dev)
4487 static inline void netif_tx_unlock(struct net_device *dev)
4491 for (i = 0; i < dev->num_tx_queues; i++) {
4492 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4494 /* No need to grab the _xmit_lock here. If the
4495 * queue is not stopped for another reason, we
4498 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4499 netif_schedule_queue(txq);
4501 spin_unlock(&dev->tx_global_lock);
4504 static inline void netif_tx_unlock_bh(struct net_device *dev)
4506 netif_tx_unlock(dev);
4510 #define HARD_TX_LOCK(dev, txq, cpu) { \
4511 if ((dev->features & NETIF_F_LLTX) == 0) { \
4512 __netif_tx_lock(txq, cpu); \
4514 __netif_tx_acquire(txq); \
4518 #define HARD_TX_TRYLOCK(dev, txq) \
4519 (((dev->features & NETIF_F_LLTX) == 0) ? \
4520 __netif_tx_trylock(txq) : \
4521 __netif_tx_acquire(txq))
4523 #define HARD_TX_UNLOCK(dev, txq) { \
4524 if ((dev->features & NETIF_F_LLTX) == 0) { \
4525 __netif_tx_unlock(txq); \
4527 __netif_tx_release(txq); \
4531 static inline void netif_tx_disable(struct net_device *dev)
4537 cpu = smp_processor_id();
4538 spin_lock(&dev->tx_global_lock);
4539 for (i = 0; i < dev->num_tx_queues; i++) {
4540 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4542 __netif_tx_lock(txq, cpu);
4543 netif_tx_stop_queue(txq);
4544 __netif_tx_unlock(txq);
4546 spin_unlock(&dev->tx_global_lock);
4550 static inline void netif_addr_lock(struct net_device *dev)
4552 unsigned char nest_level = 0;
4554 #ifdef CONFIG_LOCKDEP
4555 nest_level = dev->nested_level;
4557 spin_lock_nested(&dev->addr_list_lock, nest_level);
4560 static inline void netif_addr_lock_bh(struct net_device *dev)
4562 unsigned char nest_level = 0;
4564 #ifdef CONFIG_LOCKDEP
4565 nest_level = dev->nested_level;
4568 spin_lock_nested(&dev->addr_list_lock, nest_level);
4571 static inline void netif_addr_unlock(struct net_device *dev)
4573 spin_unlock(&dev->addr_list_lock);
4576 static inline void netif_addr_unlock_bh(struct net_device *dev)
4578 spin_unlock_bh(&dev->addr_list_lock);
4582 * dev_addrs walker. Should be used only for read access. Call with
4583 * rcu_read_lock held.
4585 #define for_each_dev_addr(dev, ha) \
4586 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4588 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4590 void ether_setup(struct net_device *dev);
4592 /* Support for loadable net-drivers */
4593 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4594 unsigned char name_assign_type,
4595 void (*setup)(struct net_device *),
4596 unsigned int txqs, unsigned int rxqs);
4597 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4598 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4600 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4601 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4604 int register_netdev(struct net_device *dev);
4605 void unregister_netdev(struct net_device *dev);
4607 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4609 /* General hardware address lists handling functions */
4610 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4611 struct netdev_hw_addr_list *from_list, int addr_len);
4612 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4613 struct netdev_hw_addr_list *from_list, int addr_len);
4614 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4615 struct net_device *dev,
4616 int (*sync)(struct net_device *, const unsigned char *),
4617 int (*unsync)(struct net_device *,
4618 const unsigned char *));
4619 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4620 struct net_device *dev,
4621 int (*sync)(struct net_device *,
4622 const unsigned char *, int),
4623 int (*unsync)(struct net_device *,
4624 const unsigned char *, int));
4625 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4626 struct net_device *dev,
4627 int (*unsync)(struct net_device *,
4628 const unsigned char *, int));
4629 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4630 struct net_device *dev,
4631 int (*unsync)(struct net_device *,
4632 const unsigned char *));
4633 void __hw_addr_init(struct netdev_hw_addr_list *list);
4635 /* Functions used for device addresses handling */
4636 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4637 unsigned char addr_type);
4638 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4639 unsigned char addr_type);
4640 void dev_addr_flush(struct net_device *dev);
4641 int dev_addr_init(struct net_device *dev);
4643 /* Functions used for unicast addresses handling */
4644 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4645 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4646 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4647 int dev_uc_sync(struct net_device *to, struct net_device *from);
4648 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4649 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4650 void dev_uc_flush(struct net_device *dev);
4651 void dev_uc_init(struct net_device *dev);
4654 * __dev_uc_sync - Synchonize device's unicast list
4655 * @dev: device to sync
4656 * @sync: function to call if address should be added
4657 * @unsync: function to call if address should be removed
4659 * Add newly added addresses to the interface, and release
4660 * addresses that have been deleted.
4662 static inline int __dev_uc_sync(struct net_device *dev,
4663 int (*sync)(struct net_device *,
4664 const unsigned char *),
4665 int (*unsync)(struct net_device *,
4666 const unsigned char *))
4668 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4672 * __dev_uc_unsync - Remove synchronized addresses from device
4673 * @dev: device to sync
4674 * @unsync: function to call if address should be removed
4676 * Remove all addresses that were added to the device by dev_uc_sync().
4678 static inline void __dev_uc_unsync(struct net_device *dev,
4679 int (*unsync)(struct net_device *,
4680 const unsigned char *))
4682 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4685 /* Functions used for multicast addresses handling */
4686 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4687 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4688 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4689 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4690 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4691 int dev_mc_sync(struct net_device *to, struct net_device *from);
4692 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4693 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4694 void dev_mc_flush(struct net_device *dev);
4695 void dev_mc_init(struct net_device *dev);
4698 * __dev_mc_sync - Synchonize device's multicast list
4699 * @dev: device to sync
4700 * @sync: function to call if address should be added
4701 * @unsync: function to call if address should be removed
4703 * Add newly added addresses to the interface, and release
4704 * addresses that have been deleted.
4706 static inline int __dev_mc_sync(struct net_device *dev,
4707 int (*sync)(struct net_device *,
4708 const unsigned char *),
4709 int (*unsync)(struct net_device *,
4710 const unsigned char *))
4712 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4716 * __dev_mc_unsync - Remove synchronized addresses from device
4717 * @dev: device to sync
4718 * @unsync: function to call if address should be removed
4720 * Remove all addresses that were added to the device by dev_mc_sync().
4722 static inline void __dev_mc_unsync(struct net_device *dev,
4723 int (*unsync)(struct net_device *,
4724 const unsigned char *))
4726 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4729 /* Functions used for secondary unicast and multicast support */
4730 void dev_set_rx_mode(struct net_device *dev);
4731 void __dev_set_rx_mode(struct net_device *dev);
4732 int dev_set_promiscuity(struct net_device *dev, int inc);
4733 int dev_set_allmulti(struct net_device *dev, int inc);
4734 void netdev_state_change(struct net_device *dev);
4735 void __netdev_notify_peers(struct net_device *dev);
4736 void netdev_notify_peers(struct net_device *dev);
4737 void netdev_features_change(struct net_device *dev);
4738 /* Load a device via the kmod */
4739 void dev_load(struct net *net, const char *name);
4740 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4741 struct rtnl_link_stats64 *storage);
4742 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4743 const struct net_device_stats *netdev_stats);
4744 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4745 const struct pcpu_sw_netstats __percpu *netstats);
4746 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4748 extern int netdev_max_backlog;
4749 extern int netdev_tstamp_prequeue;
4750 extern int netdev_unregister_timeout_secs;
4751 extern int weight_p;
4752 extern int dev_weight_rx_bias;
4753 extern int dev_weight_tx_bias;
4754 extern int dev_rx_weight;
4755 extern int dev_tx_weight;
4756 extern int gro_normal_batch;
4759 NESTED_SYNC_IMM_BIT,
4760 NESTED_SYNC_TODO_BIT,
4763 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4764 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4766 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4767 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4769 struct netdev_nested_priv {
4770 unsigned char flags;
4774 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4775 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4776 struct list_head **iter);
4777 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4778 struct list_head **iter);
4780 #ifdef CONFIG_LOCKDEP
4781 static LIST_HEAD(net_unlink_list);
4783 static inline void net_unlink_todo(struct net_device *dev)
4785 if (list_empty(&dev->unlink_list))
4786 list_add_tail(&dev->unlink_list, &net_unlink_list);
4790 /* iterate through upper list, must be called under RCU read lock */
4791 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4792 for (iter = &(dev)->adj_list.upper, \
4793 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4795 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4797 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4798 int (*fn)(struct net_device *upper_dev,
4799 struct netdev_nested_priv *priv),
4800 struct netdev_nested_priv *priv);
4802 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4803 struct net_device *upper_dev);
4805 bool netdev_has_any_upper_dev(struct net_device *dev);
4807 void *netdev_lower_get_next_private(struct net_device *dev,
4808 struct list_head **iter);
4809 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4810 struct list_head **iter);
4812 #define netdev_for_each_lower_private(dev, priv, iter) \
4813 for (iter = (dev)->adj_list.lower.next, \
4814 priv = netdev_lower_get_next_private(dev, &(iter)); \
4816 priv = netdev_lower_get_next_private(dev, &(iter)))
4818 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4819 for (iter = &(dev)->adj_list.lower, \
4820 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4822 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4824 void *netdev_lower_get_next(struct net_device *dev,
4825 struct list_head **iter);
4827 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4828 for (iter = (dev)->adj_list.lower.next, \
4829 ldev = netdev_lower_get_next(dev, &(iter)); \
4831 ldev = netdev_lower_get_next(dev, &(iter)))
4833 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4834 struct list_head **iter);
4835 int netdev_walk_all_lower_dev(struct net_device *dev,
4836 int (*fn)(struct net_device *lower_dev,
4837 struct netdev_nested_priv *priv),
4838 struct netdev_nested_priv *priv);
4839 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4840 int (*fn)(struct net_device *lower_dev,
4841 struct netdev_nested_priv *priv),
4842 struct netdev_nested_priv *priv);
4844 void *netdev_adjacent_get_private(struct list_head *adj_list);
4845 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4846 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4847 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4848 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4849 struct netlink_ext_ack *extack);
4850 int netdev_master_upper_dev_link(struct net_device *dev,
4851 struct net_device *upper_dev,
4852 void *upper_priv, void *upper_info,
4853 struct netlink_ext_ack *extack);
4854 void netdev_upper_dev_unlink(struct net_device *dev,
4855 struct net_device *upper_dev);
4856 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4857 struct net_device *new_dev,
4858 struct net_device *dev,
4859 struct netlink_ext_ack *extack);
4860 void netdev_adjacent_change_commit(struct net_device *old_dev,
4861 struct net_device *new_dev,
4862 struct net_device *dev);
4863 void netdev_adjacent_change_abort(struct net_device *old_dev,
4864 struct net_device *new_dev,
4865 struct net_device *dev);
4866 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4867 void *netdev_lower_dev_get_private(struct net_device *dev,
4868 struct net_device *lower_dev);
4869 void netdev_lower_state_changed(struct net_device *lower_dev,
4870 void *lower_state_info);
4872 /* RSS keys are 40 or 52 bytes long */
4873 #define NETDEV_RSS_KEY_LEN 52
4874 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4875 void netdev_rss_key_fill(void *buffer, size_t len);
4877 int skb_checksum_help(struct sk_buff *skb);
4878 int skb_crc32c_csum_help(struct sk_buff *skb);
4879 int skb_csum_hwoffload_help(struct sk_buff *skb,
4880 const netdev_features_t features);
4882 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4883 netdev_features_t features, bool tx_path);
4884 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4885 netdev_features_t features);
4887 struct netdev_bonding_info {
4892 struct netdev_notifier_bonding_info {
4893 struct netdev_notifier_info info; /* must be first */
4894 struct netdev_bonding_info bonding_info;
4897 void netdev_bonding_info_change(struct net_device *dev,
4898 struct netdev_bonding_info *bonding_info);
4900 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4901 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4903 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4910 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4912 return __skb_gso_segment(skb, features, true);
4914 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4916 static inline bool can_checksum_protocol(netdev_features_t features,
4919 if (protocol == htons(ETH_P_FCOE))
4920 return !!(features & NETIF_F_FCOE_CRC);
4922 /* Assume this is an IP checksum (not SCTP CRC) */
4924 if (features & NETIF_F_HW_CSUM) {
4925 /* Can checksum everything */
4930 case htons(ETH_P_IP):
4931 return !!(features & NETIF_F_IP_CSUM);
4932 case htons(ETH_P_IPV6):
4933 return !!(features & NETIF_F_IPV6_CSUM);
4940 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4942 static inline void netdev_rx_csum_fault(struct net_device *dev,
4943 struct sk_buff *skb)
4947 /* rx skb timestamps */
4948 void net_enable_timestamp(void);
4949 void net_disable_timestamp(void);
4951 #ifdef CONFIG_PROC_FS
4952 int __init dev_proc_init(void);
4954 #define dev_proc_init() 0
4957 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4958 struct sk_buff *skb, struct net_device *dev,
4961 __this_cpu_write(softnet_data.xmit.more, more);
4962 return ops->ndo_start_xmit(skb, dev);
4965 static inline bool netdev_xmit_more(void)
4967 return __this_cpu_read(softnet_data.xmit.more);
4970 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4971 struct netdev_queue *txq, bool more)
4973 const struct net_device_ops *ops = dev->netdev_ops;
4976 rc = __netdev_start_xmit(ops, skb, dev, more);
4977 if (rc == NETDEV_TX_OK)
4978 txq_trans_update(txq);
4983 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4985 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4988 extern const struct kobj_ns_type_operations net_ns_type_operations;
4990 const char *netdev_drivername(const struct net_device *dev);
4992 void linkwatch_run_queue(void);
4994 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4995 netdev_features_t f2)
4997 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4998 if (f1 & NETIF_F_HW_CSUM)
4999 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5001 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5007 static inline netdev_features_t netdev_get_wanted_features(
5008 struct net_device *dev)
5010 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5012 netdev_features_t netdev_increment_features(netdev_features_t all,
5013 netdev_features_t one, netdev_features_t mask);
5015 /* Allow TSO being used on stacked device :
5016 * Performing the GSO segmentation before last device
5017 * is a performance improvement.
5019 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5020 netdev_features_t mask)
5022 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5025 int __netdev_update_features(struct net_device *dev);
5026 void netdev_update_features(struct net_device *dev);
5027 void netdev_change_features(struct net_device *dev);
5029 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5030 struct net_device *dev);
5032 netdev_features_t passthru_features_check(struct sk_buff *skb,
5033 struct net_device *dev,
5034 netdev_features_t features);
5035 netdev_features_t netif_skb_features(struct sk_buff *skb);
5037 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5039 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5041 /* check flags correspondence */
5042 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5043 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5044 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5045 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5046 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5047 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5048 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5049 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5050 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5051 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5052 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5053 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5054 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5055 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5056 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5057 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5058 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5059 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5060 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5062 return (features & feature) == feature;
5065 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5067 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5068 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5071 static inline bool netif_needs_gso(struct sk_buff *skb,
5072 netdev_features_t features)
5074 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5075 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5076 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5079 static inline void netif_set_gso_max_size(struct net_device *dev,
5082 dev->gso_max_size = size;
5085 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5086 int pulled_hlen, u16 mac_offset,
5089 skb->protocol = protocol;
5090 skb->encapsulation = 1;
5091 skb_push(skb, pulled_hlen);
5092 skb_reset_transport_header(skb);
5093 skb->mac_header = mac_offset;
5094 skb->network_header = skb->mac_header + mac_len;
5095 skb->mac_len = mac_len;
5098 static inline bool netif_is_macsec(const struct net_device *dev)
5100 return dev->priv_flags & IFF_MACSEC;
5103 static inline bool netif_is_macvlan(const struct net_device *dev)
5105 return dev->priv_flags & IFF_MACVLAN;
5108 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5110 return dev->priv_flags & IFF_MACVLAN_PORT;
5113 static inline bool netif_is_bond_master(const struct net_device *dev)
5115 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5118 static inline bool netif_is_bond_slave(const struct net_device *dev)
5120 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5123 static inline bool netif_supports_nofcs(struct net_device *dev)
5125 return dev->priv_flags & IFF_SUPP_NOFCS;
5128 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5130 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5133 static inline bool netif_is_l3_master(const struct net_device *dev)
5135 return dev->priv_flags & IFF_L3MDEV_MASTER;
5138 static inline bool netif_is_l3_slave(const struct net_device *dev)
5140 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5143 static inline bool netif_is_bridge_master(const struct net_device *dev)
5145 return dev->priv_flags & IFF_EBRIDGE;
5148 static inline bool netif_is_bridge_port(const struct net_device *dev)
5150 return dev->priv_flags & IFF_BRIDGE_PORT;
5153 static inline bool netif_is_ovs_master(const struct net_device *dev)
5155 return dev->priv_flags & IFF_OPENVSWITCH;
5158 static inline bool netif_is_ovs_port(const struct net_device *dev)
5160 return dev->priv_flags & IFF_OVS_DATAPATH;
5163 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5165 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5168 static inline bool netif_is_team_master(const struct net_device *dev)
5170 return dev->priv_flags & IFF_TEAM;
5173 static inline bool netif_is_team_port(const struct net_device *dev)
5175 return dev->priv_flags & IFF_TEAM_PORT;
5178 static inline bool netif_is_lag_master(const struct net_device *dev)
5180 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5183 static inline bool netif_is_lag_port(const struct net_device *dev)
5185 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5188 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5190 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5193 static inline bool netif_is_failover(const struct net_device *dev)
5195 return dev->priv_flags & IFF_FAILOVER;
5198 static inline bool netif_is_failover_slave(const struct net_device *dev)
5200 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5203 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5204 static inline void netif_keep_dst(struct net_device *dev)
5206 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5209 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5210 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5212 /* TODO: reserve and use an additional IFF bit, if we get more users */
5213 return dev->priv_flags & IFF_MACSEC;
5216 extern struct pernet_operations __net_initdata loopback_net_ops;
5218 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5220 /* netdev_printk helpers, similar to dev_printk */
5222 static inline const char *netdev_name(const struct net_device *dev)
5224 if (!dev->name[0] || strchr(dev->name, '%'))
5225 return "(unnamed net_device)";
5229 static inline bool netdev_unregistering(const struct net_device *dev)
5231 return dev->reg_state == NETREG_UNREGISTERING;
5234 static inline const char *netdev_reg_state(const struct net_device *dev)
5236 switch (dev->reg_state) {
5237 case NETREG_UNINITIALIZED: return " (uninitialized)";
5238 case NETREG_REGISTERED: return "";
5239 case NETREG_UNREGISTERING: return " (unregistering)";
5240 case NETREG_UNREGISTERED: return " (unregistered)";
5241 case NETREG_RELEASED: return " (released)";
5242 case NETREG_DUMMY: return " (dummy)";
5245 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5246 return " (unknown)";
5249 __printf(3, 4) __cold
5250 void netdev_printk(const char *level, const struct net_device *dev,
5251 const char *format, ...);
5252 __printf(2, 3) __cold
5253 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5254 __printf(2, 3) __cold
5255 void netdev_alert(const struct net_device *dev, const char *format, ...);
5256 __printf(2, 3) __cold
5257 void netdev_crit(const struct net_device *dev, const char *format, ...);
5258 __printf(2, 3) __cold
5259 void netdev_err(const struct net_device *dev, const char *format, ...);
5260 __printf(2, 3) __cold
5261 void netdev_warn(const struct net_device *dev, const char *format, ...);
5262 __printf(2, 3) __cold
5263 void netdev_notice(const struct net_device *dev, const char *format, ...);
5264 __printf(2, 3) __cold
5265 void netdev_info(const struct net_device *dev, const char *format, ...);
5267 #define netdev_level_once(level, dev, fmt, ...) \
5269 static bool __print_once __read_mostly; \
5271 if (!__print_once) { \
5272 __print_once = true; \
5273 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5277 #define netdev_emerg_once(dev, fmt, ...) \
5278 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5279 #define netdev_alert_once(dev, fmt, ...) \
5280 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5281 #define netdev_crit_once(dev, fmt, ...) \
5282 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5283 #define netdev_err_once(dev, fmt, ...) \
5284 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5285 #define netdev_warn_once(dev, fmt, ...) \
5286 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5287 #define netdev_notice_once(dev, fmt, ...) \
5288 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5289 #define netdev_info_once(dev, fmt, ...) \
5290 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5292 #define MODULE_ALIAS_NETDEV(device) \
5293 MODULE_ALIAS("netdev-" device)
5295 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5296 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5297 #define netdev_dbg(__dev, format, args...) \
5299 dynamic_netdev_dbg(__dev, format, ##args); \
5301 #elif defined(DEBUG)
5302 #define netdev_dbg(__dev, format, args...) \
5303 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5305 #define netdev_dbg(__dev, format, args...) \
5308 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5312 #if defined(VERBOSE_DEBUG)
5313 #define netdev_vdbg netdev_dbg
5316 #define netdev_vdbg(dev, format, args...) \
5319 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5325 * netdev_WARN() acts like dev_printk(), but with the key difference
5326 * of using a WARN/WARN_ON to get the message out, including the
5327 * file/line information and a backtrace.
5329 #define netdev_WARN(dev, format, args...) \
5330 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5331 netdev_reg_state(dev), ##args)
5333 #define netdev_WARN_ONCE(dev, format, args...) \
5334 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5335 netdev_reg_state(dev), ##args)
5337 /* netif printk helpers, similar to netdev_printk */
5339 #define netif_printk(priv, type, level, dev, fmt, args...) \
5341 if (netif_msg_##type(priv)) \
5342 netdev_printk(level, (dev), fmt, ##args); \
5345 #define netif_level(level, priv, type, dev, fmt, args...) \
5347 if (netif_msg_##type(priv)) \
5348 netdev_##level(dev, fmt, ##args); \
5351 #define netif_emerg(priv, type, dev, fmt, args...) \
5352 netif_level(emerg, priv, type, dev, fmt, ##args)
5353 #define netif_alert(priv, type, dev, fmt, args...) \
5354 netif_level(alert, priv, type, dev, fmt, ##args)
5355 #define netif_crit(priv, type, dev, fmt, args...) \
5356 netif_level(crit, priv, type, dev, fmt, ##args)
5357 #define netif_err(priv, type, dev, fmt, args...) \
5358 netif_level(err, priv, type, dev, fmt, ##args)
5359 #define netif_warn(priv, type, dev, fmt, args...) \
5360 netif_level(warn, priv, type, dev, fmt, ##args)
5361 #define netif_notice(priv, type, dev, fmt, args...) \
5362 netif_level(notice, priv, type, dev, fmt, ##args)
5363 #define netif_info(priv, type, dev, fmt, args...) \
5364 netif_level(info, priv, type, dev, fmt, ##args)
5366 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5367 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5368 #define netif_dbg(priv, type, netdev, format, args...) \
5370 if (netif_msg_##type(priv)) \
5371 dynamic_netdev_dbg(netdev, format, ##args); \
5373 #elif defined(DEBUG)
5374 #define netif_dbg(priv, type, dev, format, args...) \
5375 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5377 #define netif_dbg(priv, type, dev, format, args...) \
5380 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5385 /* if @cond then downgrade to debug, else print at @level */
5386 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5389 netif_dbg(priv, type, netdev, fmt, ##args); \
5391 netif_ ## level(priv, type, netdev, fmt, ##args); \
5394 #if defined(VERBOSE_DEBUG)
5395 #define netif_vdbg netif_dbg
5397 #define netif_vdbg(priv, type, dev, format, args...) \
5400 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5406 * The list of packet types we will receive (as opposed to discard)
5407 * and the routines to invoke.
5409 * Why 16. Because with 16 the only overlap we get on a hash of the
5410 * low nibble of the protocol value is RARP/SNAP/X.25.
5424 #define PTYPE_HASH_SIZE (16)
5425 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5427 extern struct list_head ptype_all __read_mostly;
5428 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5430 extern struct net_device *blackhole_netdev;
5432 #endif /* _LINUX_NETDEVICE_H */