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*/
366 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
367 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
368 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
369 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
370 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
371 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
372 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
373 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
374 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
384 typedef enum gro_result gro_result_t;
387 * enum rx_handler_result - Possible return values for rx_handlers.
388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
391 * case skb->dev was changed by rx_handler.
392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
395 * rx_handlers are functions called from inside __netif_receive_skb(), to do
396 * special processing of the skb, prior to delivery to protocol handlers.
398 * Currently, a net_device can only have a single rx_handler registered. Trying
399 * to register a second rx_handler will return -EBUSY.
401 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
402 * To unregister a rx_handler on a net_device, use
403 * netdev_rx_handler_unregister().
405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
408 * If the rx_handler consumed the skb in some way, it should return
409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
410 * the skb to be delivered in some other way.
412 * If the rx_handler changed skb->dev, to divert the skb to another
413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
414 * new device will be called if it exists.
416 * If the rx_handler decides the skb should be ignored, it should return
417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
418 * are registered on exact device (ptype->dev == skb->dev).
420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
421 * delivered, it should return RX_HANDLER_PASS.
423 * A device without a registered rx_handler will behave as if rx_handler
424 * returned RX_HANDLER_PASS.
427 enum rx_handler_result {
433 typedef enum rx_handler_result rx_handler_result_t;
434 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
436 void __napi_schedule(struct napi_struct *n);
437 void __napi_schedule_irqoff(struct napi_struct *n);
439 static inline bool napi_disable_pending(struct napi_struct *n)
441 return test_bit(NAPI_STATE_DISABLE, &n->state);
444 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
446 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
449 bool napi_schedule_prep(struct napi_struct *n);
452 * napi_schedule - schedule NAPI poll
455 * Schedule NAPI poll routine to be called if it is not already
458 static inline void napi_schedule(struct napi_struct *n)
460 if (napi_schedule_prep(n))
465 * napi_schedule_irqoff - schedule NAPI poll
468 * Variant of napi_schedule(), assuming hard irqs are masked.
470 static inline void napi_schedule_irqoff(struct napi_struct *n)
472 if (napi_schedule_prep(n))
473 __napi_schedule_irqoff(n);
476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
477 static inline bool napi_reschedule(struct napi_struct *napi)
479 if (napi_schedule_prep(napi)) {
480 __napi_schedule(napi);
486 bool napi_complete_done(struct napi_struct *n, int work_done);
488 * napi_complete - NAPI processing complete
491 * Mark NAPI processing as complete.
492 * Consider using napi_complete_done() instead.
493 * Return false if device should avoid rearming interrupts.
495 static inline bool napi_complete(struct napi_struct *n)
497 return napi_complete_done(n, 0);
500 int dev_set_threaded(struct net_device *dev, bool threaded);
503 * napi_disable - prevent NAPI from scheduling
506 * Stop NAPI from being scheduled on this context.
507 * Waits till any outstanding processing completes.
509 void napi_disable(struct napi_struct *n);
511 void napi_enable(struct napi_struct *n);
514 * napi_synchronize - wait until NAPI is not running
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
521 static inline void napi_synchronize(const struct napi_struct *n)
523 if (IS_ENABLED(CONFIG_SMP))
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 * napi_if_scheduled_mark_missed - if napi is running, set the
535 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
538 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
540 unsigned long val, new;
543 val = READ_ONCE(n->state);
544 if (val & NAPIF_STATE_DISABLE)
547 if (!(val & NAPIF_STATE_SCHED))
550 new = val | NAPIF_STATE_MISSED;
551 } while (cmpxchg(&n->state, val, new) != val);
556 enum netdev_queue_state_t {
557 __QUEUE_STATE_DRV_XOFF,
558 __QUEUE_STATE_STACK_XOFF,
559 __QUEUE_STATE_FROZEN,
562 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
563 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
564 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
566 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
567 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
569 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
573 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
574 * netif_tx_* functions below are used to manipulate this flag. The
575 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
576 * queue independently. The netif_xmit_*stopped functions below are called
577 * to check if the queue has been stopped by the driver or stack (either
578 * of the XOFF bits are set in the state). Drivers should not need to call
579 * netif_xmit*stopped functions, they should only be using netif_tx_*.
582 struct netdev_queue {
586 struct net_device *dev;
587 struct Qdisc __rcu *qdisc;
588 struct Qdisc *qdisc_sleeping;
592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
595 unsigned long tx_maxrate;
597 * Number of TX timeouts for this queue
598 * (/sys/class/net/DEV/Q/trans_timeout)
600 unsigned long trans_timeout;
602 /* Subordinate device that the queue has been assigned to */
603 struct net_device *sb_dev;
604 #ifdef CONFIG_XDP_SOCKETS
605 struct xsk_buff_pool *pool;
610 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
613 * Time (in jiffies) of last Tx
615 unsigned long trans_start;
622 } ____cacheline_aligned_in_smp;
624 extern int sysctl_fb_tunnels_only_for_init_net;
625 extern int sysctl_devconf_inherit_init_net;
628 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
629 * == 1 : For initns only
632 static inline bool net_has_fallback_tunnels(const struct net *net)
634 return !IS_ENABLED(CONFIG_SYSCTL) ||
635 !sysctl_fb_tunnels_only_for_init_net ||
636 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
639 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
641 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
648 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 * This structure holds an RPS map which can be of variable length. The
658 * map is an array of CPUs.
665 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
669 * tail pointer for that CPU's input queue at the time of last enqueue, and
670 * a hardware filter index.
672 struct rps_dev_flow {
675 unsigned int last_qtail;
677 #define RPS_NO_FILTER 0xffff
680 * The rps_dev_flow_table structure contains a table of flow mappings.
682 struct rps_dev_flow_table {
685 struct rps_dev_flow flows[];
687 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
688 ((_num) * sizeof(struct rps_dev_flow)))
691 * The rps_sock_flow_table contains mappings of flows to the last CPU
692 * on which they were processed by the application (set in recvmsg).
693 * Each entry is a 32bit value. Upper part is the high-order bits
694 * of flow hash, lower part is CPU number.
695 * rps_cpu_mask is used to partition the space, depending on number of
696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
698 * meaning we use 32-6=26 bits for the hash.
700 struct rps_sock_flow_table {
703 u32 ents[] ____cacheline_aligned_in_smp;
705 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
707 #define RPS_NO_CPU 0xffff
709 extern u32 rps_cpu_mask;
710 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
712 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
716 unsigned int index = hash & table->mask;
717 u32 val = hash & ~rps_cpu_mask;
719 /* We only give a hint, preemption can change CPU under us */
720 val |= raw_smp_processor_id();
722 if (table->ents[index] != val)
723 table->ents[index] = val;
727 #ifdef CONFIG_RFS_ACCEL
728 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 #endif /* CONFIG_RPS */
733 /* This structure contains an instance of an RX queue. */
734 struct netdev_rx_queue {
736 struct rps_map __rcu *rps_map;
737 struct rps_dev_flow_table __rcu *rps_flow_table;
740 struct net_device *dev;
741 struct xdp_rxq_info xdp_rxq;
742 #ifdef CONFIG_XDP_SOCKETS
743 struct xsk_buff_pool *pool;
745 } ____cacheline_aligned_in_smp;
748 * RX queue sysfs structures and functions.
750 struct rx_queue_attribute {
751 struct attribute attr;
752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
753 ssize_t (*store)(struct netdev_rx_queue *queue,
754 const char *buf, size_t len);
757 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
766 * This structure holds an XPS map which can be of variable length. The
767 * map is an array of queues.
771 unsigned int alloc_len;
775 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
776 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
777 - sizeof(struct xps_map)) / sizeof(u16))
780 * This structure holds all XPS maps for device. Maps are indexed by CPU.
782 * We keep track of the number of cpus/rxqs used when the struct is allocated,
783 * in nr_ids. This will help not accessing out-of-bound memory.
785 * We keep track of the number of traffic classes used when the struct is
786 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
787 * not crossing its upper bound, as the original dev->num_tc can be updated in
790 struct xps_dev_maps {
794 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
797 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
798 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
800 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
801 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
803 #endif /* CONFIG_XPS */
805 #define TC_MAX_QUEUE 16
806 #define TC_BITMASK 15
807 /* HW offloaded queuing disciplines txq count and offset maps */
808 struct netdev_tc_txq {
813 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
815 * This structure is to hold information about the device
816 * configured to run FCoE protocol stack.
818 struct netdev_fcoe_hbainfo {
819 char manufacturer[64];
820 char serial_number[64];
821 char hardware_version[64];
822 char driver_version[64];
823 char optionrom_version[64];
824 char firmware_version[64];
826 char model_description[256];
830 #define MAX_PHYS_ITEM_ID_LEN 32
832 /* This structure holds a unique identifier to identify some
833 * physical item (port for example) used by a netdevice.
835 struct netdev_phys_item_id {
836 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
837 unsigned char id_len;
840 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
841 struct netdev_phys_item_id *b)
843 return a->id_len == b->id_len &&
844 memcmp(a->id, b->id, a->id_len) == 0;
847 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
849 struct net_device *sb_dev);
851 enum net_device_path_type {
852 DEV_PATH_ETHERNET = 0,
858 struct net_device_path {
859 enum net_device_path_type type;
860 const struct net_device *dev;
869 DEV_PATH_BR_VLAN_KEEP,
870 DEV_PATH_BR_VLAN_TAG,
871 DEV_PATH_BR_VLAN_UNTAG,
879 #define NET_DEVICE_PATH_STACK_MAX 5
880 #define NET_DEVICE_PATH_VLAN_MAX 2
882 struct net_device_path_stack {
884 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
887 struct net_device_path_ctx {
888 const struct net_device *dev;
895 } vlan[NET_DEVICE_PATH_VLAN_MAX];
899 TC_SETUP_QDISC_MQPRIO,
902 TC_SETUP_CLSMATCHALL,
912 TC_SETUP_QDISC_TAPRIO,
920 /* These structures hold the attributes of bpf state that are being passed
921 * to the netdevice through the bpf op.
923 enum bpf_netdev_command {
924 /* Set or clear a bpf program used in the earliest stages of packet
925 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
926 * is responsible for calling bpf_prog_put on any old progs that are
927 * stored. In case of error, the callee need not release the new prog
928 * reference, but on success it takes ownership and must bpf_prog_put
929 * when it is no longer used.
933 /* BPF program for offload callbacks, invoked at program load time. */
934 BPF_OFFLOAD_MAP_ALLOC,
935 BPF_OFFLOAD_MAP_FREE,
939 struct bpf_prog_offload_ops;
940 struct netlink_ext_ack;
942 struct xdp_dev_bulk_queue;
952 struct bpf_xdp_entity {
953 struct bpf_prog *prog;
954 struct bpf_xdp_link *link;
958 enum bpf_netdev_command command;
963 struct bpf_prog *prog;
964 struct netlink_ext_ack *extack;
966 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
968 struct bpf_offloaded_map *offmap;
970 /* XDP_SETUP_XSK_POOL */
972 struct xsk_buff_pool *pool;
978 /* Flags for ndo_xsk_wakeup. */
979 #define XDP_WAKEUP_RX (1 << 0)
980 #define XDP_WAKEUP_TX (1 << 1)
982 #ifdef CONFIG_XFRM_OFFLOAD
984 int (*xdo_dev_state_add) (struct xfrm_state *x);
985 void (*xdo_dev_state_delete) (struct xfrm_state *x);
986 void (*xdo_dev_state_free) (struct xfrm_state *x);
987 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
988 struct xfrm_state *x);
989 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
994 struct rcu_head rcuhead;
1001 struct netdev_name_node {
1002 struct hlist_node hlist;
1003 struct list_head list;
1004 struct net_device *dev;
1008 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1009 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1011 struct netdev_net_notifier {
1012 struct list_head list;
1013 struct notifier_block *nb;
1017 * This structure defines the management hooks for network devices.
1018 * The following hooks can be defined; unless noted otherwise, they are
1019 * optional and can be filled with a null pointer.
1021 * int (*ndo_init)(struct net_device *dev);
1022 * This function is called once when a network device is registered.
1023 * The network device can use this for any late stage initialization
1024 * or semantic validation. It can fail with an error code which will
1025 * be propagated back to register_netdev.
1027 * void (*ndo_uninit)(struct net_device *dev);
1028 * This function is called when device is unregistered or when registration
1029 * fails. It is not called if init fails.
1031 * int (*ndo_open)(struct net_device *dev);
1032 * This function is called when a network device transitions to the up
1035 * int (*ndo_stop)(struct net_device *dev);
1036 * This function is called when a network device transitions to the down
1039 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1040 * struct net_device *dev);
1041 * Called when a packet needs to be transmitted.
1042 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1043 * the queue before that can happen; it's for obsolete devices and weird
1044 * corner cases, but the stack really does a non-trivial amount
1045 * of useless work if you return NETDEV_TX_BUSY.
1046 * Required; cannot be NULL.
1048 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1049 * struct net_device *dev
1050 * netdev_features_t features);
1051 * Called by core transmit path to determine if device is capable of
1052 * performing offload operations on a given packet. This is to give
1053 * the device an opportunity to implement any restrictions that cannot
1054 * be otherwise expressed by feature flags. The check is called with
1055 * the set of features that the stack has calculated and it returns
1056 * those the driver believes to be appropriate.
1058 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1059 * struct net_device *sb_dev);
1060 * Called to decide which queue to use when device supports multiple
1063 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1064 * This function is called to allow device receiver to make
1065 * changes to configuration when multicast or promiscuous is enabled.
1067 * void (*ndo_set_rx_mode)(struct net_device *dev);
1068 * This function is called device changes address list filtering.
1069 * If driver handles unicast address filtering, it should set
1070 * IFF_UNICAST_FLT in its priv_flags.
1072 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1073 * This function is called when the Media Access Control address
1074 * needs to be changed. If this interface is not defined, the
1075 * MAC address can not be changed.
1077 * int (*ndo_validate_addr)(struct net_device *dev);
1078 * Test if Media Access Control address is valid for the device.
1080 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1081 * Called when a user requests an ioctl which can't be handled by
1082 * the generic interface code. If not defined ioctls return
1083 * not supported error code.
1085 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1086 * Used to set network devices bus interface parameters. This interface
1087 * is retained for legacy reasons; new devices should use the bus
1088 * interface (PCI) for low level management.
1090 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1091 * Called when a user wants to change the Maximum Transfer Unit
1094 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1095 * Callback used when the transmitter has not made any progress
1096 * for dev->watchdog ticks.
1098 * void (*ndo_get_stats64)(struct net_device *dev,
1099 * struct rtnl_link_stats64 *storage);
1100 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1101 * Called when a user wants to get the network device usage
1102 * statistics. Drivers must do one of the following:
1103 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1104 * rtnl_link_stats64 structure passed by the caller.
1105 * 2. Define @ndo_get_stats to update a net_device_stats structure
1106 * (which should normally be dev->stats) and return a pointer to
1107 * it. The structure may be changed asynchronously only if each
1108 * field is written atomically.
1109 * 3. Update dev->stats asynchronously and atomically, and define
1110 * neither operation.
1112 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1113 * Return true if this device supports offload stats of this attr_id.
1115 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1117 * Get statistics for offload operations by attr_id. Write it into the
1118 * attr_data pointer.
1120 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1121 * If device supports VLAN filtering this function is called when a
1122 * VLAN id is registered.
1124 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1125 * If device supports VLAN filtering this function is called when a
1126 * VLAN id is unregistered.
1128 * void (*ndo_poll_controller)(struct net_device *dev);
1130 * SR-IOV management functions.
1131 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1132 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1133 * u8 qos, __be16 proto);
1134 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1136 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1137 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1138 * int (*ndo_get_vf_config)(struct net_device *dev,
1139 * int vf, struct ifla_vf_info *ivf);
1140 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1141 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1142 * struct nlattr *port[]);
1144 * Enable or disable the VF ability to query its RSS Redirection Table and
1145 * Hash Key. This is needed since on some devices VF share this information
1146 * with PF and querying it may introduce a theoretical security risk.
1147 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1148 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1149 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1151 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1152 * This is always called from the stack with the rtnl lock held and netif
1153 * tx queues stopped. This allows the netdevice to perform queue
1154 * management safely.
1156 * Fiber Channel over Ethernet (FCoE) offload functions.
1157 * int (*ndo_fcoe_enable)(struct net_device *dev);
1158 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1159 * so the underlying device can perform whatever needed configuration or
1160 * initialization to support acceleration of FCoE traffic.
1162 * int (*ndo_fcoe_disable)(struct net_device *dev);
1163 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1164 * so the underlying device can perform whatever needed clean-ups to
1165 * stop supporting acceleration of FCoE traffic.
1167 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1168 * struct scatterlist *sgl, unsigned int sgc);
1169 * Called when the FCoE Initiator wants to initialize an I/O that
1170 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1171 * perform necessary setup and returns 1 to indicate the device is set up
1172 * successfully to perform DDP on this I/O, otherwise this returns 0.
1174 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1175 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1176 * indicated by the FC exchange id 'xid', so the underlying device can
1177 * clean up and reuse resources for later DDP requests.
1179 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1180 * struct scatterlist *sgl, unsigned int sgc);
1181 * Called when the FCoE Target wants to initialize an I/O that
1182 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1183 * perform necessary setup and returns 1 to indicate the device is set up
1184 * successfully to perform DDP on this I/O, otherwise this returns 0.
1186 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1187 * struct netdev_fcoe_hbainfo *hbainfo);
1188 * Called when the FCoE Protocol stack wants information on the underlying
1189 * device. This information is utilized by the FCoE protocol stack to
1190 * register attributes with Fiber Channel management service as per the
1191 * FC-GS Fabric Device Management Information(FDMI) specification.
1193 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1194 * Called when the underlying device wants to override default World Wide
1195 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1196 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1197 * protocol stack to use.
1200 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1201 * u16 rxq_index, u32 flow_id);
1202 * Set hardware filter for RFS. rxq_index is the target queue index;
1203 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1204 * Return the filter ID on success, or a negative error code.
1206 * Slave management functions (for bridge, bonding, etc).
1207 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1208 * Called to make another netdev an underling.
1210 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1211 * Called to release previously enslaved netdev.
1213 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1214 * struct sk_buff *skb,
1216 * Get the xmit slave of master device. If all_slaves is true, function
1217 * assume all the slaves can transmit.
1219 * Feature/offload setting functions.
1220 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1221 * netdev_features_t features);
1222 * Adjusts the requested feature flags according to device-specific
1223 * constraints, and returns the resulting flags. Must not modify
1226 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1227 * Called to update device configuration to new features. Passed
1228 * feature set might be less than what was returned by ndo_fix_features()).
1229 * Must return >0 or -errno if it changed dev->features itself.
1231 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1232 * struct net_device *dev,
1233 * const unsigned char *addr, u16 vid, u16 flags,
1234 * struct netlink_ext_ack *extack);
1235 * Adds an FDB entry to dev for addr.
1236 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1237 * struct net_device *dev,
1238 * const unsigned char *addr, u16 vid)
1239 * Deletes the FDB entry from dev coresponding to addr.
1240 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1241 * struct net_device *dev, struct net_device *filter_dev,
1243 * Used to add FDB entries to dump requests. Implementers should add
1244 * entries to skb and update idx with the number of entries.
1246 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1247 * u16 flags, struct netlink_ext_ack *extack)
1248 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1249 * struct net_device *dev, u32 filter_mask,
1251 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1254 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1255 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1256 * which do not represent real hardware may define this to allow their
1257 * userspace components to manage their virtual carrier state. Devices
1258 * that determine carrier state from physical hardware properties (eg
1259 * network cables) or protocol-dependent mechanisms (eg
1260 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1262 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1263 * struct netdev_phys_item_id *ppid);
1264 * Called to get ID of physical port of this device. If driver does
1265 * not implement this, it is assumed that the hw is not able to have
1266 * multiple net devices on single physical port.
1268 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1269 * struct netdev_phys_item_id *ppid)
1270 * Called to get the parent ID of the physical port of this device.
1272 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1273 * struct net_device *dev)
1274 * Called by upper layer devices to accelerate switching or other
1275 * station functionality into hardware. 'pdev is the lowerdev
1276 * to use for the offload and 'dev' is the net device that will
1277 * back the offload. Returns a pointer to the private structure
1278 * the upper layer will maintain.
1279 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1280 * Called by upper layer device to delete the station created
1281 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1282 * the station and priv is the structure returned by the add
1284 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1285 * int queue_index, u32 maxrate);
1286 * Called when a user wants to set a max-rate limitation of specific
1288 * int (*ndo_get_iflink)(const struct net_device *dev);
1289 * Called to get the iflink value of this device.
1290 * void (*ndo_change_proto_down)(struct net_device *dev,
1292 * This function is used to pass protocol port error state information
1293 * to the switch driver. The switch driver can react to the proto_down
1294 * by doing a phys down on the associated switch port.
1295 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1296 * This function is used to get egress tunnel information for given skb.
1297 * This is useful for retrieving outer tunnel header parameters while
1299 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1300 * This function is used to specify the headroom that the skb must
1301 * consider when allocation skb during packet reception. Setting
1302 * appropriate rx headroom value allows avoiding skb head copy on
1303 * forward. Setting a negative value resets the rx headroom to the
1305 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1306 * This function is used to set or query state related to XDP on the
1307 * netdevice and manage BPF offload. See definition of
1308 * enum bpf_netdev_command for details.
1309 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1311 * This function is used to submit @n XDP packets for transmit on a
1312 * netdevice. Returns number of frames successfully transmitted, frames
1313 * that got dropped are freed/returned via xdp_return_frame().
1314 * Returns negative number, means general error invoking ndo, meaning
1315 * no frames were xmit'ed and core-caller will free all frames.
1316 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1317 * This function is used to wake up the softirq, ksoftirqd or kthread
1318 * responsible for sending and/or receiving packets on a specific
1319 * queue id bound to an AF_XDP socket. The flags field specifies if
1320 * only RX, only Tx, or both should be woken up using the flags
1321 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1322 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1323 * Get devlink port instance associated with a given netdev.
1324 * Called with a reference on the netdevice and devlink locks only,
1325 * rtnl_lock is not held.
1326 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1328 * Add, change, delete or get information on an IPv4 tunnel.
1329 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1330 * If a device is paired with a peer device, return the peer instance.
1331 * The caller must be under RCU read context.
1332 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1333 * Get the forwarding path to reach the real device from the HW destination address
1335 struct net_device_ops {
1336 int (*ndo_init)(struct net_device *dev);
1337 void (*ndo_uninit)(struct net_device *dev);
1338 int (*ndo_open)(struct net_device *dev);
1339 int (*ndo_stop)(struct net_device *dev);
1340 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1341 struct net_device *dev);
1342 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1343 struct net_device *dev,
1344 netdev_features_t features);
1345 u16 (*ndo_select_queue)(struct net_device *dev,
1346 struct sk_buff *skb,
1347 struct net_device *sb_dev);
1348 void (*ndo_change_rx_flags)(struct net_device *dev,
1350 void (*ndo_set_rx_mode)(struct net_device *dev);
1351 int (*ndo_set_mac_address)(struct net_device *dev,
1353 int (*ndo_validate_addr)(struct net_device *dev);
1354 int (*ndo_do_ioctl)(struct net_device *dev,
1355 struct ifreq *ifr, int cmd);
1356 int (*ndo_set_config)(struct net_device *dev,
1358 int (*ndo_change_mtu)(struct net_device *dev,
1360 int (*ndo_neigh_setup)(struct net_device *dev,
1361 struct neigh_parms *);
1362 void (*ndo_tx_timeout) (struct net_device *dev,
1363 unsigned int txqueue);
1365 void (*ndo_get_stats64)(struct net_device *dev,
1366 struct rtnl_link_stats64 *storage);
1367 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1368 int (*ndo_get_offload_stats)(int attr_id,
1369 const struct net_device *dev,
1371 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1373 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1374 __be16 proto, u16 vid);
1375 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1376 __be16 proto, u16 vid);
1377 #ifdef CONFIG_NET_POLL_CONTROLLER
1378 void (*ndo_poll_controller)(struct net_device *dev);
1379 int (*ndo_netpoll_setup)(struct net_device *dev,
1380 struct netpoll_info *info);
1381 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1383 int (*ndo_set_vf_mac)(struct net_device *dev,
1384 int queue, u8 *mac);
1385 int (*ndo_set_vf_vlan)(struct net_device *dev,
1386 int queue, u16 vlan,
1387 u8 qos, __be16 proto);
1388 int (*ndo_set_vf_rate)(struct net_device *dev,
1389 int vf, int min_tx_rate,
1391 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1392 int vf, bool setting);
1393 int (*ndo_set_vf_trust)(struct net_device *dev,
1394 int vf, bool setting);
1395 int (*ndo_get_vf_config)(struct net_device *dev,
1397 struct ifla_vf_info *ivf);
1398 int (*ndo_set_vf_link_state)(struct net_device *dev,
1399 int vf, int link_state);
1400 int (*ndo_get_vf_stats)(struct net_device *dev,
1402 struct ifla_vf_stats
1404 int (*ndo_set_vf_port)(struct net_device *dev,
1406 struct nlattr *port[]);
1407 int (*ndo_get_vf_port)(struct net_device *dev,
1408 int vf, struct sk_buff *skb);
1409 int (*ndo_get_vf_guid)(struct net_device *dev,
1411 struct ifla_vf_guid *node_guid,
1412 struct ifla_vf_guid *port_guid);
1413 int (*ndo_set_vf_guid)(struct net_device *dev,
1416 int (*ndo_set_vf_rss_query_en)(
1417 struct net_device *dev,
1418 int vf, bool setting);
1419 int (*ndo_setup_tc)(struct net_device *dev,
1420 enum tc_setup_type type,
1422 #if IS_ENABLED(CONFIG_FCOE)
1423 int (*ndo_fcoe_enable)(struct net_device *dev);
1424 int (*ndo_fcoe_disable)(struct net_device *dev);
1425 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1427 struct scatterlist *sgl,
1429 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1431 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1433 struct scatterlist *sgl,
1435 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1436 struct netdev_fcoe_hbainfo *hbainfo);
1439 #if IS_ENABLED(CONFIG_LIBFCOE)
1440 #define NETDEV_FCOE_WWNN 0
1441 #define NETDEV_FCOE_WWPN 1
1442 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1443 u64 *wwn, int type);
1446 #ifdef CONFIG_RFS_ACCEL
1447 int (*ndo_rx_flow_steer)(struct net_device *dev,
1448 const struct sk_buff *skb,
1452 int (*ndo_add_slave)(struct net_device *dev,
1453 struct net_device *slave_dev,
1454 struct netlink_ext_ack *extack);
1455 int (*ndo_del_slave)(struct net_device *dev,
1456 struct net_device *slave_dev);
1457 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1458 struct sk_buff *skb,
1460 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1462 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1463 netdev_features_t features);
1464 int (*ndo_set_features)(struct net_device *dev,
1465 netdev_features_t features);
1466 int (*ndo_neigh_construct)(struct net_device *dev,
1467 struct neighbour *n);
1468 void (*ndo_neigh_destroy)(struct net_device *dev,
1469 struct neighbour *n);
1471 int (*ndo_fdb_add)(struct ndmsg *ndm,
1472 struct nlattr *tb[],
1473 struct net_device *dev,
1474 const unsigned char *addr,
1477 struct netlink_ext_ack *extack);
1478 int (*ndo_fdb_del)(struct ndmsg *ndm,
1479 struct nlattr *tb[],
1480 struct net_device *dev,
1481 const unsigned char *addr,
1483 int (*ndo_fdb_dump)(struct sk_buff *skb,
1484 struct netlink_callback *cb,
1485 struct net_device *dev,
1486 struct net_device *filter_dev,
1488 int (*ndo_fdb_get)(struct sk_buff *skb,
1489 struct nlattr *tb[],
1490 struct net_device *dev,
1491 const unsigned char *addr,
1492 u16 vid, u32 portid, u32 seq,
1493 struct netlink_ext_ack *extack);
1494 int (*ndo_bridge_setlink)(struct net_device *dev,
1495 struct nlmsghdr *nlh,
1497 struct netlink_ext_ack *extack);
1498 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1500 struct net_device *dev,
1503 int (*ndo_bridge_dellink)(struct net_device *dev,
1504 struct nlmsghdr *nlh,
1506 int (*ndo_change_carrier)(struct net_device *dev,
1508 int (*ndo_get_phys_port_id)(struct net_device *dev,
1509 struct netdev_phys_item_id *ppid);
1510 int (*ndo_get_port_parent_id)(struct net_device *dev,
1511 struct netdev_phys_item_id *ppid);
1512 int (*ndo_get_phys_port_name)(struct net_device *dev,
1513 char *name, size_t len);
1514 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1515 struct net_device *dev);
1516 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1519 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1522 int (*ndo_get_iflink)(const struct net_device *dev);
1523 int (*ndo_change_proto_down)(struct net_device *dev,
1525 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1526 struct sk_buff *skb);
1527 void (*ndo_set_rx_headroom)(struct net_device *dev,
1528 int needed_headroom);
1529 int (*ndo_bpf)(struct net_device *dev,
1530 struct netdev_bpf *bpf);
1531 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1532 struct xdp_frame **xdp,
1534 int (*ndo_xsk_wakeup)(struct net_device *dev,
1535 u32 queue_id, u32 flags);
1536 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1537 int (*ndo_tunnel_ctl)(struct net_device *dev,
1538 struct ip_tunnel_parm *p, int cmd);
1539 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1540 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1541 struct net_device_path *path);
1545 * enum netdev_priv_flags - &struct net_device priv_flags
1547 * These are the &struct net_device, they are only set internally
1548 * by drivers and used in the kernel. These flags are invisible to
1549 * userspace; this means that the order of these flags can change
1550 * during any kernel release.
1552 * You should have a pretty good reason to be extending these flags.
1554 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1555 * @IFF_EBRIDGE: Ethernet bridging device
1556 * @IFF_BONDING: bonding master or slave
1557 * @IFF_ISATAP: ISATAP interface (RFC4214)
1558 * @IFF_WAN_HDLC: WAN HDLC device
1559 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1561 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1562 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1563 * @IFF_MACVLAN_PORT: device used as macvlan port
1564 * @IFF_BRIDGE_PORT: device used as bridge port
1565 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1566 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1567 * @IFF_UNICAST_FLT: Supports unicast filtering
1568 * @IFF_TEAM_PORT: device used as team port
1569 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1570 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1571 * change when it's running
1572 * @IFF_MACVLAN: Macvlan device
1573 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1574 * underlying stacked devices
1575 * @IFF_L3MDEV_MASTER: device is an L3 master device
1576 * @IFF_NO_QUEUE: device can run without qdisc attached
1577 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1578 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1579 * @IFF_TEAM: device is a team device
1580 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1581 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1582 * entity (i.e. the master device for bridged veth)
1583 * @IFF_MACSEC: device is a MACsec device
1584 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1585 * @IFF_FAILOVER: device is a failover master device
1586 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1587 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1588 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1589 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1590 * skb_headlen(skb) == 0 (data starts from frag0)
1592 enum netdev_priv_flags {
1593 IFF_802_1Q_VLAN = 1<<0,
1597 IFF_WAN_HDLC = 1<<4,
1598 IFF_XMIT_DST_RELEASE = 1<<5,
1599 IFF_DONT_BRIDGE = 1<<6,
1600 IFF_DISABLE_NETPOLL = 1<<7,
1601 IFF_MACVLAN_PORT = 1<<8,
1602 IFF_BRIDGE_PORT = 1<<9,
1603 IFF_OVS_DATAPATH = 1<<10,
1604 IFF_TX_SKB_SHARING = 1<<11,
1605 IFF_UNICAST_FLT = 1<<12,
1606 IFF_TEAM_PORT = 1<<13,
1607 IFF_SUPP_NOFCS = 1<<14,
1608 IFF_LIVE_ADDR_CHANGE = 1<<15,
1609 IFF_MACVLAN = 1<<16,
1610 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1611 IFF_L3MDEV_MASTER = 1<<18,
1612 IFF_NO_QUEUE = 1<<19,
1613 IFF_OPENVSWITCH = 1<<20,
1614 IFF_L3MDEV_SLAVE = 1<<21,
1616 IFF_RXFH_CONFIGURED = 1<<23,
1617 IFF_PHONY_HEADROOM = 1<<24,
1619 IFF_NO_RX_HANDLER = 1<<26,
1620 IFF_FAILOVER = 1<<27,
1621 IFF_FAILOVER_SLAVE = 1<<28,
1622 IFF_L3MDEV_RX_HANDLER = 1<<29,
1623 IFF_LIVE_RENAME_OK = 1<<30,
1624 IFF_TX_SKB_NO_LINEAR = 1<<31,
1627 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1628 #define IFF_EBRIDGE IFF_EBRIDGE
1629 #define IFF_BONDING IFF_BONDING
1630 #define IFF_ISATAP IFF_ISATAP
1631 #define IFF_WAN_HDLC IFF_WAN_HDLC
1632 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1633 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1634 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1635 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1636 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1637 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1638 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1639 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1640 #define IFF_TEAM_PORT IFF_TEAM_PORT
1641 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1642 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1643 #define IFF_MACVLAN IFF_MACVLAN
1644 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1645 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1646 #define IFF_NO_QUEUE IFF_NO_QUEUE
1647 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1648 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1649 #define IFF_TEAM IFF_TEAM
1650 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1651 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1652 #define IFF_MACSEC IFF_MACSEC
1653 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1654 #define IFF_FAILOVER IFF_FAILOVER
1655 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1656 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1657 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1658 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1660 /* Specifies the type of the struct net_device::ml_priv pointer */
1661 enum netdev_ml_priv_type {
1667 * struct net_device - The DEVICE structure.
1669 * Actually, this whole structure is a big mistake. It mixes I/O
1670 * data with strictly "high-level" data, and it has to know about
1671 * almost every data structure used in the INET module.
1673 * @name: This is the first field of the "visible" part of this structure
1674 * (i.e. as seen by users in the "Space.c" file). It is the name
1677 * @name_node: Name hashlist node
1678 * @ifalias: SNMP alias
1679 * @mem_end: Shared memory end
1680 * @mem_start: Shared memory start
1681 * @base_addr: Device I/O address
1682 * @irq: Device IRQ number
1684 * @state: Generic network queuing layer state, see netdev_state_t
1685 * @dev_list: The global list of network devices
1686 * @napi_list: List entry used for polling NAPI devices
1687 * @unreg_list: List entry when we are unregistering the
1688 * device; see the function unregister_netdev
1689 * @close_list: List entry used when we are closing the device
1690 * @ptype_all: Device-specific packet handlers for all protocols
1691 * @ptype_specific: Device-specific, protocol-specific packet handlers
1693 * @adj_list: Directly linked devices, like slaves for bonding
1694 * @features: Currently active device features
1695 * @hw_features: User-changeable features
1697 * @wanted_features: User-requested features
1698 * @vlan_features: Mask of features inheritable by VLAN devices
1700 * @hw_enc_features: Mask of features inherited by encapsulating devices
1701 * This field indicates what encapsulation
1702 * offloads the hardware is capable of doing,
1703 * and drivers will need to set them appropriately.
1705 * @mpls_features: Mask of features inheritable by MPLS
1706 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1708 * @ifindex: interface index
1709 * @group: The group the device belongs to
1711 * @stats: Statistics struct, which was left as a legacy, use
1712 * rtnl_link_stats64 instead
1714 * @rx_dropped: Dropped packets by core network,
1715 * do not use this in drivers
1716 * @tx_dropped: Dropped packets by core network,
1717 * do not use this in drivers
1718 * @rx_nohandler: nohandler dropped packets by core network on
1719 * inactive devices, do not use this in drivers
1720 * @carrier_up_count: Number of times the carrier has been up
1721 * @carrier_down_count: Number of times the carrier has been down
1723 * @wireless_handlers: List of functions to handle Wireless Extensions,
1725 * see <net/iw_handler.h> for details.
1726 * @wireless_data: Instance data managed by the core of wireless extensions
1728 * @netdev_ops: Includes several pointers to callbacks,
1729 * if one wants to override the ndo_*() functions
1730 * @ethtool_ops: Management operations
1731 * @l3mdev_ops: Layer 3 master device operations
1732 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1733 * discovery handling. Necessary for e.g. 6LoWPAN.
1734 * @xfrmdev_ops: Transformation offload operations
1735 * @tlsdev_ops: Transport Layer Security offload operations
1736 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1737 * of Layer 2 headers.
1739 * @flags: Interface flags (a la BSD)
1740 * @priv_flags: Like 'flags' but invisible to userspace,
1741 * see if.h for the definitions
1742 * @gflags: Global flags ( kept as legacy )
1743 * @padded: How much padding added by alloc_netdev()
1744 * @operstate: RFC2863 operstate
1745 * @link_mode: Mapping policy to operstate
1746 * @if_port: Selectable AUI, TP, ...
1748 * @mtu: Interface MTU value
1749 * @min_mtu: Interface Minimum MTU value
1750 * @max_mtu: Interface Maximum MTU value
1751 * @type: Interface hardware type
1752 * @hard_header_len: Maximum hardware header length.
1753 * @min_header_len: Minimum hardware header length
1755 * @needed_headroom: Extra headroom the hardware may need, but not in all
1756 * cases can this be guaranteed
1757 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1758 * cases can this be guaranteed. Some cases also use
1759 * LL_MAX_HEADER instead to allocate the skb
1761 * interface address info:
1763 * @perm_addr: Permanent hw address
1764 * @addr_assign_type: Hw address assignment type
1765 * @addr_len: Hardware address length
1766 * @upper_level: Maximum depth level of upper devices.
1767 * @lower_level: Maximum depth level of lower devices.
1768 * @neigh_priv_len: Used in neigh_alloc()
1769 * @dev_id: Used to differentiate devices that share
1770 * the same link layer address
1771 * @dev_port: Used to differentiate devices that share
1773 * @addr_list_lock: XXX: need comments on this one
1774 * @name_assign_type: network interface name assignment type
1775 * @uc_promisc: Counter that indicates promiscuous mode
1776 * has been enabled due to the need to listen to
1777 * additional unicast addresses in a device that
1778 * does not implement ndo_set_rx_mode()
1779 * @uc: unicast mac addresses
1780 * @mc: multicast mac addresses
1781 * @dev_addrs: list of device hw addresses
1782 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1783 * @promiscuity: Number of times the NIC is told to work in
1784 * promiscuous mode; if it becomes 0 the NIC will
1785 * exit promiscuous mode
1786 * @allmulti: Counter, enables or disables allmulticast mode
1788 * @vlan_info: VLAN info
1789 * @dsa_ptr: dsa specific data
1790 * @tipc_ptr: TIPC specific data
1791 * @atalk_ptr: AppleTalk link
1792 * @ip_ptr: IPv4 specific data
1793 * @dn_ptr: DECnet specific data
1794 * @ip6_ptr: IPv6 specific data
1795 * @ax25_ptr: AX.25 specific data
1796 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1797 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1799 * @mpls_ptr: mpls_dev struct pointer
1801 * @dev_addr: Hw address (before bcast,
1802 * because most packets are unicast)
1804 * @_rx: Array of RX queues
1805 * @num_rx_queues: Number of RX queues
1806 * allocated at register_netdev() time
1807 * @real_num_rx_queues: Number of RX queues currently active in device
1808 * @xdp_prog: XDP sockets filter program pointer
1809 * @gro_flush_timeout: timeout for GRO layer in NAPI
1810 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1811 * allow to avoid NIC hard IRQ, on busy queues.
1813 * @rx_handler: handler for received packets
1814 * @rx_handler_data: XXX: need comments on this one
1815 * @miniq_ingress: ingress/clsact qdisc specific data for
1816 * ingress processing
1817 * @ingress_queue: XXX: need comments on this one
1818 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1819 * @broadcast: hw bcast address
1821 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1822 * indexed by RX queue number. Assigned by driver.
1823 * This must only be set if the ndo_rx_flow_steer
1824 * operation is defined
1825 * @index_hlist: Device index hash chain
1827 * @_tx: Array of TX queues
1828 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1829 * @real_num_tx_queues: Number of TX queues currently active in device
1830 * @qdisc: Root qdisc from userspace point of view
1831 * @tx_queue_len: Max frames per queue allowed
1832 * @tx_global_lock: XXX: need comments on this one
1833 * @xdp_bulkq: XDP device bulk queue
1834 * @xps_maps: all CPUs/RXQs maps for XPS device
1836 * @xps_maps: XXX: need comments on this one
1837 * @miniq_egress: clsact qdisc specific data for
1839 * @qdisc_hash: qdisc hash table
1840 * @watchdog_timeo: Represents the timeout that is used by
1841 * the watchdog (see dev_watchdog())
1842 * @watchdog_timer: List of timers
1844 * @proto_down_reason: reason a netdev interface is held down
1845 * @pcpu_refcnt: Number of references to this device
1846 * @dev_refcnt: Number of references to this device
1847 * @todo_list: Delayed register/unregister
1848 * @link_watch_list: XXX: need comments on this one
1850 * @reg_state: Register/unregister state machine
1851 * @dismantle: Device is going to be freed
1852 * @rtnl_link_state: This enum represents the phases of creating
1855 * @needs_free_netdev: Should unregister perform free_netdev?
1856 * @priv_destructor: Called from unregister
1857 * @npinfo: XXX: need comments on this one
1858 * @nd_net: Network namespace this network device is inside
1860 * @ml_priv: Mid-layer private
1861 * @ml_priv_type: Mid-layer private type
1862 * @lstats: Loopback statistics
1863 * @tstats: Tunnel statistics
1864 * @dstats: Dummy statistics
1865 * @vstats: Virtual ethernet statistics
1870 * @dev: Class/net/name entry
1871 * @sysfs_groups: Space for optional device, statistics and wireless
1874 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1875 * @rtnl_link_ops: Rtnl_link_ops
1877 * @gso_max_size: Maximum size of generic segmentation offload
1878 * @gso_max_segs: Maximum number of segments that can be passed to the
1881 * @dcbnl_ops: Data Center Bridging netlink ops
1882 * @num_tc: Number of traffic classes in the net device
1883 * @tc_to_txq: XXX: need comments on this one
1884 * @prio_tc_map: XXX: need comments on this one
1886 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1888 * @priomap: XXX: need comments on this one
1889 * @phydev: Physical device may attach itself
1890 * for hardware timestamping
1891 * @sfp_bus: attached &struct sfp_bus structure.
1893 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1894 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1896 * @proto_down: protocol port state information can be sent to the
1897 * switch driver and used to set the phys state of the
1900 * @wol_enabled: Wake-on-LAN is enabled
1902 * @threaded: napi threaded mode is enabled
1904 * @net_notifier_list: List of per-net netdev notifier block
1905 * that follow this device when it is moved
1906 * to another network namespace.
1908 * @macsec_ops: MACsec offloading ops
1910 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1911 * offload capabilities of the device
1912 * @udp_tunnel_nic: UDP tunnel offload state
1913 * @xdp_state: stores info on attached XDP BPF programs
1915 * @nested_level: Used as as a parameter of spin_lock_nested() of
1916 * dev->addr_list_lock.
1917 * @unlink_list: As netif_addr_lock() can be called recursively,
1918 * keep a list of interfaces to be deleted.
1920 * FIXME: cleanup struct net_device such that network protocol info
1925 char name[IFNAMSIZ];
1926 struct netdev_name_node *name_node;
1927 struct dev_ifalias __rcu *ifalias;
1929 * I/O specific fields
1930 * FIXME: Merge these and struct ifmap into one
1932 unsigned long mem_end;
1933 unsigned long mem_start;
1934 unsigned long base_addr;
1937 * Some hardware also needs these fields (state,dev_list,
1938 * napi_list,unreg_list,close_list) but they are not
1939 * part of the usual set specified in Space.c.
1942 unsigned long state;
1944 struct list_head dev_list;
1945 struct list_head napi_list;
1946 struct list_head unreg_list;
1947 struct list_head close_list;
1948 struct list_head ptype_all;
1949 struct list_head ptype_specific;
1952 struct list_head upper;
1953 struct list_head lower;
1956 /* Read-mostly cache-line for fast-path access */
1958 unsigned int priv_flags;
1959 const struct net_device_ops *netdev_ops;
1961 unsigned short gflags;
1962 unsigned short hard_header_len;
1964 /* Note : dev->mtu is often read without holding a lock.
1965 * Writers usually hold RTNL.
1966 * It is recommended to use READ_ONCE() to annotate the reads,
1967 * and to use WRITE_ONCE() to annotate the writes.
1970 unsigned short needed_headroom;
1971 unsigned short needed_tailroom;
1973 netdev_features_t features;
1974 netdev_features_t hw_features;
1975 netdev_features_t wanted_features;
1976 netdev_features_t vlan_features;
1977 netdev_features_t hw_enc_features;
1978 netdev_features_t mpls_features;
1979 netdev_features_t gso_partial_features;
1981 unsigned int min_mtu;
1982 unsigned int max_mtu;
1983 unsigned short type;
1984 unsigned char min_header_len;
1985 unsigned char name_assign_type;
1989 struct net_device_stats stats; /* not used by modern drivers */
1991 atomic_long_t rx_dropped;
1992 atomic_long_t tx_dropped;
1993 atomic_long_t rx_nohandler;
1995 /* Stats to monitor link on/off, flapping */
1996 atomic_t carrier_up_count;
1997 atomic_t carrier_down_count;
1999 #ifdef CONFIG_WIRELESS_EXT
2000 const struct iw_handler_def *wireless_handlers;
2001 struct iw_public_data *wireless_data;
2003 const struct ethtool_ops *ethtool_ops;
2004 #ifdef CONFIG_NET_L3_MASTER_DEV
2005 const struct l3mdev_ops *l3mdev_ops;
2007 #if IS_ENABLED(CONFIG_IPV6)
2008 const struct ndisc_ops *ndisc_ops;
2011 #ifdef CONFIG_XFRM_OFFLOAD
2012 const struct xfrmdev_ops *xfrmdev_ops;
2015 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2016 const struct tlsdev_ops *tlsdev_ops;
2019 const struct header_ops *header_ops;
2021 unsigned char operstate;
2022 unsigned char link_mode;
2024 unsigned char if_port;
2027 /* Interface address info. */
2028 unsigned char perm_addr[MAX_ADDR_LEN];
2029 unsigned char addr_assign_type;
2030 unsigned char addr_len;
2031 unsigned char upper_level;
2032 unsigned char lower_level;
2034 unsigned short neigh_priv_len;
2035 unsigned short dev_id;
2036 unsigned short dev_port;
2037 unsigned short padded;
2039 spinlock_t addr_list_lock;
2042 struct netdev_hw_addr_list uc;
2043 struct netdev_hw_addr_list mc;
2044 struct netdev_hw_addr_list dev_addrs;
2047 struct kset *queues_kset;
2049 #ifdef CONFIG_LOCKDEP
2050 struct list_head unlink_list;
2052 unsigned int promiscuity;
2053 unsigned int allmulti;
2055 #ifdef CONFIG_LOCKDEP
2056 unsigned char nested_level;
2060 /* Protocol-specific pointers */
2062 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2063 struct vlan_info __rcu *vlan_info;
2065 #if IS_ENABLED(CONFIG_NET_DSA)
2066 struct dsa_port *dsa_ptr;
2068 #if IS_ENABLED(CONFIG_TIPC)
2069 struct tipc_bearer __rcu *tipc_ptr;
2071 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2074 struct in_device __rcu *ip_ptr;
2075 #if IS_ENABLED(CONFIG_DECNET)
2076 struct dn_dev __rcu *dn_ptr;
2078 struct inet6_dev __rcu *ip6_ptr;
2079 #if IS_ENABLED(CONFIG_AX25)
2082 struct wireless_dev *ieee80211_ptr;
2083 struct wpan_dev *ieee802154_ptr;
2084 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2085 struct mpls_dev __rcu *mpls_ptr;
2089 * Cache lines mostly used on receive path (including eth_type_trans())
2091 /* Interface address info used in eth_type_trans() */
2092 unsigned char *dev_addr;
2094 struct netdev_rx_queue *_rx;
2095 unsigned int num_rx_queues;
2096 unsigned int real_num_rx_queues;
2098 struct bpf_prog __rcu *xdp_prog;
2099 unsigned long gro_flush_timeout;
2100 int napi_defer_hard_irqs;
2101 rx_handler_func_t __rcu *rx_handler;
2102 void __rcu *rx_handler_data;
2104 #ifdef CONFIG_NET_CLS_ACT
2105 struct mini_Qdisc __rcu *miniq_ingress;
2107 struct netdev_queue __rcu *ingress_queue;
2108 #ifdef CONFIG_NETFILTER_INGRESS
2109 struct nf_hook_entries __rcu *nf_hooks_ingress;
2112 unsigned char broadcast[MAX_ADDR_LEN];
2113 #ifdef CONFIG_RFS_ACCEL
2114 struct cpu_rmap *rx_cpu_rmap;
2116 struct hlist_node index_hlist;
2119 * Cache lines mostly used on transmit path
2121 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2122 unsigned int num_tx_queues;
2123 unsigned int real_num_tx_queues;
2124 struct Qdisc *qdisc;
2125 unsigned int tx_queue_len;
2126 spinlock_t tx_global_lock;
2128 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2131 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2133 #ifdef CONFIG_NET_CLS_ACT
2134 struct mini_Qdisc __rcu *miniq_egress;
2137 #ifdef CONFIG_NET_SCHED
2138 DECLARE_HASHTABLE (qdisc_hash, 4);
2140 /* These may be needed for future network-power-down code. */
2141 struct timer_list watchdog_timer;
2144 u32 proto_down_reason;
2146 struct list_head todo_list;
2148 #ifdef CONFIG_PCPU_DEV_REFCNT
2149 int __percpu *pcpu_refcnt;
2151 refcount_t dev_refcnt;
2154 struct list_head link_watch_list;
2156 enum { NETREG_UNINITIALIZED=0,
2157 NETREG_REGISTERED, /* completed register_netdevice */
2158 NETREG_UNREGISTERING, /* called unregister_netdevice */
2159 NETREG_UNREGISTERED, /* completed unregister todo */
2160 NETREG_RELEASED, /* called free_netdev */
2161 NETREG_DUMMY, /* dummy device for NAPI poll */
2167 RTNL_LINK_INITIALIZED,
2168 RTNL_LINK_INITIALIZING,
2169 } rtnl_link_state:16;
2171 bool needs_free_netdev;
2172 void (*priv_destructor)(struct net_device *dev);
2174 #ifdef CONFIG_NETPOLL
2175 struct netpoll_info __rcu *npinfo;
2178 possible_net_t nd_net;
2180 /* mid-layer private */
2182 enum netdev_ml_priv_type ml_priv_type;
2185 struct pcpu_lstats __percpu *lstats;
2186 struct pcpu_sw_netstats __percpu *tstats;
2187 struct pcpu_dstats __percpu *dstats;
2190 #if IS_ENABLED(CONFIG_GARP)
2191 struct garp_port __rcu *garp_port;
2193 #if IS_ENABLED(CONFIG_MRP)
2194 struct mrp_port __rcu *mrp_port;
2198 const struct attribute_group *sysfs_groups[4];
2199 const struct attribute_group *sysfs_rx_queue_group;
2201 const struct rtnl_link_ops *rtnl_link_ops;
2203 /* for setting kernel sock attribute on TCP connection setup */
2204 #define GSO_MAX_SIZE 65536
2205 unsigned int gso_max_size;
2206 #define GSO_MAX_SEGS 65535
2210 const struct dcbnl_rtnl_ops *dcbnl_ops;
2213 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2214 u8 prio_tc_map[TC_BITMASK + 1];
2216 #if IS_ENABLED(CONFIG_FCOE)
2217 unsigned int fcoe_ddp_xid;
2219 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2220 struct netprio_map __rcu *priomap;
2222 struct phy_device *phydev;
2223 struct sfp_bus *sfp_bus;
2224 struct lock_class_key *qdisc_tx_busylock;
2225 struct lock_class_key *qdisc_running_key;
2227 unsigned wol_enabled:1;
2228 unsigned threaded:1;
2230 struct list_head net_notifier_list;
2232 #if IS_ENABLED(CONFIG_MACSEC)
2233 /* MACsec management functions */
2234 const struct macsec_ops *macsec_ops;
2236 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2237 struct udp_tunnel_nic *udp_tunnel_nic;
2239 /* protected by rtnl_lock */
2240 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2242 #define to_net_dev(d) container_of(d, struct net_device, dev)
2244 static inline bool netif_elide_gro(const struct net_device *dev)
2246 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2251 #define NETDEV_ALIGN 32
2254 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2256 return dev->prio_tc_map[prio & TC_BITMASK];
2260 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2262 if (tc >= dev->num_tc)
2265 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2269 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2270 void netdev_reset_tc(struct net_device *dev);
2271 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2272 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2275 int netdev_get_num_tc(struct net_device *dev)
2280 static inline void net_prefetch(void *p)
2283 #if L1_CACHE_BYTES < 128
2284 prefetch((u8 *)p + L1_CACHE_BYTES);
2288 static inline void net_prefetchw(void *p)
2291 #if L1_CACHE_BYTES < 128
2292 prefetchw((u8 *)p + L1_CACHE_BYTES);
2296 void netdev_unbind_sb_channel(struct net_device *dev,
2297 struct net_device *sb_dev);
2298 int netdev_bind_sb_channel_queue(struct net_device *dev,
2299 struct net_device *sb_dev,
2300 u8 tc, u16 count, u16 offset);
2301 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2302 static inline int netdev_get_sb_channel(struct net_device *dev)
2304 return max_t(int, -dev->num_tc, 0);
2308 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2311 return &dev->_tx[index];
2314 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2315 const struct sk_buff *skb)
2317 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2320 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2321 void (*f)(struct net_device *,
2322 struct netdev_queue *,
2328 for (i = 0; i < dev->num_tx_queues; i++)
2329 f(dev, &dev->_tx[i], arg);
2332 #define netdev_lockdep_set_classes(dev) \
2334 static struct lock_class_key qdisc_tx_busylock_key; \
2335 static struct lock_class_key qdisc_running_key; \
2336 static struct lock_class_key qdisc_xmit_lock_key; \
2337 static struct lock_class_key dev_addr_list_lock_key; \
2340 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2341 (dev)->qdisc_running_key = &qdisc_running_key; \
2342 lockdep_set_class(&(dev)->addr_list_lock, \
2343 &dev_addr_list_lock_key); \
2344 for (i = 0; i < (dev)->num_tx_queues; i++) \
2345 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2346 &qdisc_xmit_lock_key); \
2349 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2350 struct net_device *sb_dev);
2351 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2352 struct sk_buff *skb,
2353 struct net_device *sb_dev);
2355 /* returns the headroom that the master device needs to take in account
2356 * when forwarding to this dev
2358 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2360 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2363 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2365 if (dev->netdev_ops->ndo_set_rx_headroom)
2366 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2369 /* set the device rx headroom to the dev's default */
2370 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2372 netdev_set_rx_headroom(dev, -1);
2375 static inline void *netdev_get_ml_priv(struct net_device *dev,
2376 enum netdev_ml_priv_type type)
2378 if (dev->ml_priv_type != type)
2381 return dev->ml_priv;
2384 static inline void netdev_set_ml_priv(struct net_device *dev,
2386 enum netdev_ml_priv_type type)
2388 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2389 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2390 dev->ml_priv_type, type);
2391 WARN(!dev->ml_priv_type && dev->ml_priv,
2392 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2394 dev->ml_priv = ml_priv;
2395 dev->ml_priv_type = type;
2399 * Net namespace inlines
2402 struct net *dev_net(const struct net_device *dev)
2404 return read_pnet(&dev->nd_net);
2408 void dev_net_set(struct net_device *dev, struct net *net)
2410 write_pnet(&dev->nd_net, net);
2414 * netdev_priv - access network device private data
2415 * @dev: network device
2417 * Get network device private data
2419 static inline void *netdev_priv(const struct net_device *dev)
2421 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2424 /* Set the sysfs physical device reference for the network logical device
2425 * if set prior to registration will cause a symlink during initialization.
2427 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2429 /* Set the sysfs device type for the network logical device to allow
2430 * fine-grained identification of different network device types. For
2431 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2433 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2435 /* Default NAPI poll() weight
2436 * Device drivers are strongly advised to not use bigger value
2438 #define NAPI_POLL_WEIGHT 64
2441 * netif_napi_add - initialize a NAPI context
2442 * @dev: network device
2443 * @napi: NAPI context
2444 * @poll: polling function
2445 * @weight: default weight
2447 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2448 * *any* of the other NAPI-related functions.
2450 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2451 int (*poll)(struct napi_struct *, int), int weight);
2454 * netif_tx_napi_add - initialize a NAPI context
2455 * @dev: network device
2456 * @napi: NAPI context
2457 * @poll: polling function
2458 * @weight: default weight
2460 * This variant of netif_napi_add() should be used from drivers using NAPI
2461 * to exclusively poll a TX queue.
2462 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2464 static inline void netif_tx_napi_add(struct net_device *dev,
2465 struct napi_struct *napi,
2466 int (*poll)(struct napi_struct *, int),
2469 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2470 netif_napi_add(dev, napi, poll, weight);
2474 * __netif_napi_del - remove a NAPI context
2475 * @napi: NAPI context
2477 * Warning: caller must observe RCU grace period before freeing memory
2478 * containing @napi. Drivers might want to call this helper to combine
2479 * all the needed RCU grace periods into a single one.
2481 void __netif_napi_del(struct napi_struct *napi);
2484 * netif_napi_del - remove a NAPI context
2485 * @napi: NAPI context
2487 * netif_napi_del() removes a NAPI context from the network device NAPI list
2489 static inline void netif_napi_del(struct napi_struct *napi)
2491 __netif_napi_del(napi);
2495 struct napi_gro_cb {
2496 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2499 /* Length of frag0. */
2500 unsigned int frag0_len;
2502 /* This indicates where we are processing relative to skb->data. */
2505 /* This is non-zero if the packet cannot be merged with the new skb. */
2508 /* Save the IP ID here and check when we get to the transport layer */
2511 /* Number of segments aggregated. */
2514 /* Start offset for remote checksum offload */
2515 u16 gro_remcsum_start;
2517 /* jiffies when first packet was created/queued */
2520 /* Used in ipv6_gro_receive() and foo-over-udp */
2523 /* This is non-zero if the packet may be of the same flow. */
2526 /* Used in tunnel GRO receive */
2529 /* GRO checksum is valid */
2532 /* Number of checksums via CHECKSUM_UNNECESSARY */
2537 #define NAPI_GRO_FREE 1
2538 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2540 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2543 /* Used in GRE, set in fou/gue_gro_receive */
2546 /* Used to determine if flush_id can be ignored */
2549 /* Number of gro_receive callbacks this packet already went through */
2550 u8 recursion_counter:4;
2552 /* GRO is done by frag_list pointer chaining. */
2555 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2558 /* used in skb_gro_receive() slow path */
2559 struct sk_buff *last;
2562 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2564 #define GRO_RECURSION_LIMIT 15
2565 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2567 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2570 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2571 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2572 struct list_head *head,
2573 struct sk_buff *skb)
2575 if (unlikely(gro_recursion_inc_test(skb))) {
2576 NAPI_GRO_CB(skb)->flush |= 1;
2580 return cb(head, skb);
2583 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2585 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2587 struct list_head *head,
2588 struct sk_buff *skb)
2590 if (unlikely(gro_recursion_inc_test(skb))) {
2591 NAPI_GRO_CB(skb)->flush |= 1;
2595 return cb(sk, head, skb);
2598 struct packet_type {
2599 __be16 type; /* This is really htons(ether_type). */
2600 bool ignore_outgoing;
2601 struct net_device *dev; /* NULL is wildcarded here */
2602 int (*func) (struct sk_buff *,
2603 struct net_device *,
2604 struct packet_type *,
2605 struct net_device *);
2606 void (*list_func) (struct list_head *,
2607 struct packet_type *,
2608 struct net_device *);
2609 bool (*id_match)(struct packet_type *ptype,
2611 void *af_packet_priv;
2612 struct list_head list;
2615 struct offload_callbacks {
2616 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2617 netdev_features_t features);
2618 struct sk_buff *(*gro_receive)(struct list_head *head,
2619 struct sk_buff *skb);
2620 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2623 struct packet_offload {
2624 __be16 type; /* This is really htons(ether_type). */
2626 struct offload_callbacks callbacks;
2627 struct list_head list;
2630 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2631 struct pcpu_sw_netstats {
2636 struct u64_stats_sync syncp;
2637 } __aligned(4 * sizeof(u64));
2639 struct pcpu_lstats {
2640 u64_stats_t packets;
2642 struct u64_stats_sync syncp;
2643 } __aligned(2 * sizeof(u64));
2645 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2647 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2649 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2651 u64_stats_update_begin(&tstats->syncp);
2652 tstats->rx_bytes += len;
2653 tstats->rx_packets++;
2654 u64_stats_update_end(&tstats->syncp);
2657 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2658 unsigned int packets,
2661 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2663 u64_stats_update_begin(&tstats->syncp);
2664 tstats->tx_bytes += len;
2665 tstats->tx_packets += packets;
2666 u64_stats_update_end(&tstats->syncp);
2669 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2671 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2673 u64_stats_update_begin(&lstats->syncp);
2674 u64_stats_add(&lstats->bytes, len);
2675 u64_stats_inc(&lstats->packets);
2676 u64_stats_update_end(&lstats->syncp);
2679 #define __netdev_alloc_pcpu_stats(type, gfp) \
2681 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2684 for_each_possible_cpu(__cpu) { \
2685 typeof(type) *stat; \
2686 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2687 u64_stats_init(&stat->syncp); \
2693 #define netdev_alloc_pcpu_stats(type) \
2694 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2696 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2698 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2701 for_each_possible_cpu(__cpu) { \
2702 typeof(type) *stat; \
2703 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2704 u64_stats_init(&stat->syncp); \
2710 enum netdev_lag_tx_type {
2711 NETDEV_LAG_TX_TYPE_UNKNOWN,
2712 NETDEV_LAG_TX_TYPE_RANDOM,
2713 NETDEV_LAG_TX_TYPE_BROADCAST,
2714 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2715 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2716 NETDEV_LAG_TX_TYPE_HASH,
2719 enum netdev_lag_hash {
2720 NETDEV_LAG_HASH_NONE,
2722 NETDEV_LAG_HASH_L34,
2723 NETDEV_LAG_HASH_L23,
2724 NETDEV_LAG_HASH_E23,
2725 NETDEV_LAG_HASH_E34,
2726 NETDEV_LAG_HASH_VLAN_SRCMAC,
2727 NETDEV_LAG_HASH_UNKNOWN,
2730 struct netdev_lag_upper_info {
2731 enum netdev_lag_tx_type tx_type;
2732 enum netdev_lag_hash hash_type;
2735 struct netdev_lag_lower_state_info {
2740 #include <linux/notifier.h>
2742 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2743 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2747 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2749 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2750 detected a hardware crash and restarted
2751 - we can use this eg to kick tcp sessions
2753 NETDEV_CHANGE, /* Notify device state change */
2756 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2757 NETDEV_CHANGEADDR, /* notify after the address change */
2758 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2762 NETDEV_BONDING_FAILOVER,
2764 NETDEV_PRE_TYPE_CHANGE,
2765 NETDEV_POST_TYPE_CHANGE,
2768 NETDEV_NOTIFY_PEERS,
2772 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2773 NETDEV_CHANGEINFODATA,
2774 NETDEV_BONDING_INFO,
2775 NETDEV_PRECHANGEUPPER,
2776 NETDEV_CHANGELOWERSTATE,
2777 NETDEV_UDP_TUNNEL_PUSH_INFO,
2778 NETDEV_UDP_TUNNEL_DROP_INFO,
2779 NETDEV_CHANGE_TX_QUEUE_LEN,
2780 NETDEV_CVLAN_FILTER_PUSH_INFO,
2781 NETDEV_CVLAN_FILTER_DROP_INFO,
2782 NETDEV_SVLAN_FILTER_PUSH_INFO,
2783 NETDEV_SVLAN_FILTER_DROP_INFO,
2785 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2787 int register_netdevice_notifier(struct notifier_block *nb);
2788 int unregister_netdevice_notifier(struct notifier_block *nb);
2789 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2790 int unregister_netdevice_notifier_net(struct net *net,
2791 struct notifier_block *nb);
2792 int register_netdevice_notifier_dev_net(struct net_device *dev,
2793 struct notifier_block *nb,
2794 struct netdev_net_notifier *nn);
2795 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2796 struct notifier_block *nb,
2797 struct netdev_net_notifier *nn);
2799 struct netdev_notifier_info {
2800 struct net_device *dev;
2801 struct netlink_ext_ack *extack;
2804 struct netdev_notifier_info_ext {
2805 struct netdev_notifier_info info; /* must be first */
2811 struct netdev_notifier_change_info {
2812 struct netdev_notifier_info info; /* must be first */
2813 unsigned int flags_changed;
2816 struct netdev_notifier_changeupper_info {
2817 struct netdev_notifier_info info; /* must be first */
2818 struct net_device *upper_dev; /* new upper dev */
2819 bool master; /* is upper dev master */
2820 bool linking; /* is the notification for link or unlink */
2821 void *upper_info; /* upper dev info */
2824 struct netdev_notifier_changelowerstate_info {
2825 struct netdev_notifier_info info; /* must be first */
2826 void *lower_state_info; /* is lower dev state */
2829 struct netdev_notifier_pre_changeaddr_info {
2830 struct netdev_notifier_info info; /* must be first */
2831 const unsigned char *dev_addr;
2834 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2835 struct net_device *dev)
2838 info->extack = NULL;
2841 static inline struct net_device *
2842 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2847 static inline struct netlink_ext_ack *
2848 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2850 return info->extack;
2853 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2856 extern rwlock_t dev_base_lock; /* Device list lock */
2858 #define for_each_netdev(net, d) \
2859 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2860 #define for_each_netdev_reverse(net, d) \
2861 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2862 #define for_each_netdev_rcu(net, d) \
2863 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2864 #define for_each_netdev_safe(net, d, n) \
2865 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2866 #define for_each_netdev_continue(net, d) \
2867 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2868 #define for_each_netdev_continue_reverse(net, d) \
2869 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2871 #define for_each_netdev_continue_rcu(net, d) \
2872 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2873 #define for_each_netdev_in_bond_rcu(bond, slave) \
2874 for_each_netdev_rcu(&init_net, slave) \
2875 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2876 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2878 static inline struct net_device *next_net_device(struct net_device *dev)
2880 struct list_head *lh;
2884 lh = dev->dev_list.next;
2885 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2888 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2890 struct list_head *lh;
2894 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2895 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2898 static inline struct net_device *first_net_device(struct net *net)
2900 return list_empty(&net->dev_base_head) ? NULL :
2901 net_device_entry(net->dev_base_head.next);
2904 static inline struct net_device *first_net_device_rcu(struct net *net)
2906 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2908 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2911 int netdev_boot_setup_check(struct net_device *dev);
2912 unsigned long netdev_boot_base(const char *prefix, int unit);
2913 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2914 const char *hwaddr);
2915 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2916 void dev_add_pack(struct packet_type *pt);
2917 void dev_remove_pack(struct packet_type *pt);
2918 void __dev_remove_pack(struct packet_type *pt);
2919 void dev_add_offload(struct packet_offload *po);
2920 void dev_remove_offload(struct packet_offload *po);
2922 int dev_get_iflink(const struct net_device *dev);
2923 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2924 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2925 struct net_device_path_stack *stack);
2926 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2927 unsigned short mask);
2928 struct net_device *dev_get_by_name(struct net *net, const char *name);
2929 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2930 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2931 int dev_alloc_name(struct net_device *dev, const char *name);
2932 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2933 void dev_close(struct net_device *dev);
2934 void dev_close_many(struct list_head *head, bool unlink);
2935 void dev_disable_lro(struct net_device *dev);
2936 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2937 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2938 struct net_device *sb_dev);
2939 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2940 struct net_device *sb_dev);
2942 int dev_queue_xmit(struct sk_buff *skb);
2943 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2944 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2946 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2950 ret = __dev_direct_xmit(skb, queue_id);
2951 if (!dev_xmit_complete(ret))
2956 int register_netdevice(struct net_device *dev);
2957 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2958 void unregister_netdevice_many(struct list_head *head);
2959 static inline void unregister_netdevice(struct net_device *dev)
2961 unregister_netdevice_queue(dev, NULL);
2964 int netdev_refcnt_read(const struct net_device *dev);
2965 void free_netdev(struct net_device *dev);
2966 void netdev_freemem(struct net_device *dev);
2967 int init_dummy_netdev(struct net_device *dev);
2969 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2970 struct sk_buff *skb,
2972 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2974 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2975 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2976 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2977 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2978 int netdev_get_name(struct net *net, char *name, int ifindex);
2979 int dev_restart(struct net_device *dev);
2980 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2981 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2983 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2985 return NAPI_GRO_CB(skb)->data_offset;
2988 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2990 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2993 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2995 NAPI_GRO_CB(skb)->data_offset += len;
2998 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2999 unsigned int offset)
3001 return NAPI_GRO_CB(skb)->frag0 + offset;
3004 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3006 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3009 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3011 NAPI_GRO_CB(skb)->frag0 = NULL;
3012 NAPI_GRO_CB(skb)->frag0_len = 0;
3015 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3016 unsigned int offset)
3018 if (!pskb_may_pull(skb, hlen))
3021 skb_gro_frag0_invalidate(skb);
3022 return skb->data + offset;
3025 static inline void *skb_gro_network_header(struct sk_buff *skb)
3027 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3028 skb_network_offset(skb);
3031 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3032 const void *start, unsigned int len)
3034 if (NAPI_GRO_CB(skb)->csum_valid)
3035 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3036 csum_partial(start, len, 0));
3039 /* GRO checksum functions. These are logical equivalents of the normal
3040 * checksum functions (in skbuff.h) except that they operate on the GRO
3041 * offsets and fields in sk_buff.
3044 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3046 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3048 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3051 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3055 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3056 skb_checksum_start_offset(skb) <
3057 skb_gro_offset(skb)) &&
3058 !skb_at_gro_remcsum_start(skb) &&
3059 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3060 (!zero_okay || check));
3063 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3066 if (NAPI_GRO_CB(skb)->csum_valid &&
3067 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3070 NAPI_GRO_CB(skb)->csum = psum;
3072 return __skb_gro_checksum_complete(skb);
3075 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3077 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3078 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3079 NAPI_GRO_CB(skb)->csum_cnt--;
3081 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3082 * verified a new top level checksum or an encapsulated one
3083 * during GRO. This saves work if we fallback to normal path.
3085 __skb_incr_checksum_unnecessary(skb);
3089 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3092 __sum16 __ret = 0; \
3093 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3094 __ret = __skb_gro_checksum_validate_complete(skb, \
3095 compute_pseudo(skb, proto)); \
3097 skb_gro_incr_csum_unnecessary(skb); \
3101 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3102 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3104 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3106 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3108 #define skb_gro_checksum_simple_validate(skb) \
3109 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3111 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3113 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3114 !NAPI_GRO_CB(skb)->csum_valid);
3117 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3120 NAPI_GRO_CB(skb)->csum = ~pseudo;
3121 NAPI_GRO_CB(skb)->csum_valid = 1;
3124 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3126 if (__skb_gro_checksum_convert_check(skb)) \
3127 __skb_gro_checksum_convert(skb, \
3128 compute_pseudo(skb, proto)); \
3131 struct gro_remcsum {
3136 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3142 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3143 unsigned int off, size_t hdrlen,
3144 int start, int offset,
3145 struct gro_remcsum *grc,
3149 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3151 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3154 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3158 ptr = skb_gro_header_fast(skb, off);
3159 if (skb_gro_header_hard(skb, off + plen)) {
3160 ptr = skb_gro_header_slow(skb, off + plen, off);
3165 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3168 /* Adjust skb->csum since we changed the packet */
3169 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3171 grc->offset = off + hdrlen + offset;
3177 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3178 struct gro_remcsum *grc)
3181 size_t plen = grc->offset + sizeof(u16);
3186 ptr = skb_gro_header_fast(skb, grc->offset);
3187 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3188 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3193 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3196 #ifdef CONFIG_XFRM_OFFLOAD
3197 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3199 if (PTR_ERR(pp) != -EINPROGRESS)
3200 NAPI_GRO_CB(skb)->flush |= flush;
3202 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3205 struct gro_remcsum *grc)
3207 if (PTR_ERR(pp) != -EINPROGRESS) {
3208 NAPI_GRO_CB(skb)->flush |= flush;
3209 skb_gro_remcsum_cleanup(skb, grc);
3210 skb->remcsum_offload = 0;
3214 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3216 NAPI_GRO_CB(skb)->flush |= flush;
3218 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3221 struct gro_remcsum *grc)
3223 NAPI_GRO_CB(skb)->flush |= flush;
3224 skb_gro_remcsum_cleanup(skb, grc);
3225 skb->remcsum_offload = 0;
3229 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3230 unsigned short type,
3231 const void *daddr, const void *saddr,
3234 if (!dev->header_ops || !dev->header_ops->create)
3237 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3240 static inline int dev_parse_header(const struct sk_buff *skb,
3241 unsigned char *haddr)
3243 const struct net_device *dev = skb->dev;
3245 if (!dev->header_ops || !dev->header_ops->parse)
3247 return dev->header_ops->parse(skb, haddr);
3250 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3252 const struct net_device *dev = skb->dev;
3254 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3256 return dev->header_ops->parse_protocol(skb);
3259 /* ll_header must have at least hard_header_len allocated */
3260 static inline bool dev_validate_header(const struct net_device *dev,
3261 char *ll_header, int len)
3263 if (likely(len >= dev->hard_header_len))
3265 if (len < dev->min_header_len)
3268 if (capable(CAP_SYS_RAWIO)) {
3269 memset(ll_header + len, 0, dev->hard_header_len - len);
3273 if (dev->header_ops && dev->header_ops->validate)
3274 return dev->header_ops->validate(ll_header, len);
3279 static inline bool dev_has_header(const struct net_device *dev)
3281 return dev->header_ops && dev->header_ops->create;
3284 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3286 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3287 static inline int unregister_gifconf(unsigned int family)
3289 return register_gifconf(family, NULL);
3292 #ifdef CONFIG_NET_FLOW_LIMIT
3293 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3294 struct sd_flow_limit {
3296 unsigned int num_buckets;
3297 unsigned int history_head;
3298 u16 history[FLOW_LIMIT_HISTORY];
3302 extern int netdev_flow_limit_table_len;
3303 #endif /* CONFIG_NET_FLOW_LIMIT */
3306 * Incoming packets are placed on per-CPU queues
3308 struct softnet_data {
3309 struct list_head poll_list;
3310 struct sk_buff_head process_queue;
3313 unsigned int processed;
3314 unsigned int time_squeeze;
3315 unsigned int received_rps;
3317 struct softnet_data *rps_ipi_list;
3319 #ifdef CONFIG_NET_FLOW_LIMIT
3320 struct sd_flow_limit __rcu *flow_limit;
3322 struct Qdisc *output_queue;
3323 struct Qdisc **output_queue_tailp;
3324 struct sk_buff *completion_queue;
3325 #ifdef CONFIG_XFRM_OFFLOAD
3326 struct sk_buff_head xfrm_backlog;
3328 /* written and read only by owning cpu: */
3334 /* input_queue_head should be written by cpu owning this struct,
3335 * and only read by other cpus. Worth using a cache line.
3337 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3339 /* Elements below can be accessed between CPUs for RPS/RFS */
3340 call_single_data_t csd ____cacheline_aligned_in_smp;
3341 struct softnet_data *rps_ipi_next;
3343 unsigned int input_queue_tail;
3345 unsigned int dropped;
3346 struct sk_buff_head input_pkt_queue;
3347 struct napi_struct backlog;
3351 static inline void input_queue_head_incr(struct softnet_data *sd)
3354 sd->input_queue_head++;
3358 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3359 unsigned int *qtail)
3362 *qtail = ++sd->input_queue_tail;
3366 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3368 static inline int dev_recursion_level(void)
3370 return this_cpu_read(softnet_data.xmit.recursion);
3373 #define XMIT_RECURSION_LIMIT 8
3374 static inline bool dev_xmit_recursion(void)
3376 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3377 XMIT_RECURSION_LIMIT);
3380 static inline void dev_xmit_recursion_inc(void)
3382 __this_cpu_inc(softnet_data.xmit.recursion);
3385 static inline void dev_xmit_recursion_dec(void)
3387 __this_cpu_dec(softnet_data.xmit.recursion);
3390 void __netif_schedule(struct Qdisc *q);
3391 void netif_schedule_queue(struct netdev_queue *txq);
3393 static inline void netif_tx_schedule_all(struct net_device *dev)
3397 for (i = 0; i < dev->num_tx_queues; i++)
3398 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3401 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3403 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3407 * netif_start_queue - allow transmit
3408 * @dev: network device
3410 * Allow upper layers to call the device hard_start_xmit routine.
3412 static inline void netif_start_queue(struct net_device *dev)
3414 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3417 static inline void netif_tx_start_all_queues(struct net_device *dev)
3421 for (i = 0; i < dev->num_tx_queues; i++) {
3422 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3423 netif_tx_start_queue(txq);
3427 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3430 * netif_wake_queue - restart transmit
3431 * @dev: network device
3433 * Allow upper layers to call the device hard_start_xmit routine.
3434 * Used for flow control when transmit resources are available.
3436 static inline void netif_wake_queue(struct net_device *dev)
3438 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3441 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3445 for (i = 0; i < dev->num_tx_queues; i++) {
3446 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3447 netif_tx_wake_queue(txq);
3451 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3453 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3457 * netif_stop_queue - stop transmitted packets
3458 * @dev: network device
3460 * Stop upper layers calling the device hard_start_xmit routine.
3461 * Used for flow control when transmit resources are unavailable.
3463 static inline void netif_stop_queue(struct net_device *dev)
3465 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3468 void netif_tx_stop_all_queues(struct net_device *dev);
3470 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3472 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3476 * netif_queue_stopped - test if transmit queue is flowblocked
3477 * @dev: network device
3479 * Test if transmit queue on device is currently unable to send.
3481 static inline bool netif_queue_stopped(const struct net_device *dev)
3483 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3486 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3488 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3492 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3494 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3498 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3500 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3504 * netdev_queue_set_dql_min_limit - set dql minimum limit
3505 * @dev_queue: pointer to transmit queue
3506 * @min_limit: dql minimum limit
3508 * Forces xmit_more() to return true until the minimum threshold
3509 * defined by @min_limit is reached (or until the tx queue is
3510 * empty). Warning: to be use with care, misuse will impact the
3513 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3514 unsigned int min_limit)
3517 dev_queue->dql.min_limit = min_limit;
3522 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3523 * @dev_queue: pointer to transmit queue
3525 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3526 * to give appropriate hint to the CPU.
3528 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3531 prefetchw(&dev_queue->dql.num_queued);
3536 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3537 * @dev_queue: pointer to transmit queue
3539 * BQL enabled drivers might use this helper in their TX completion path,
3540 * to give appropriate hint to the CPU.
3542 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3545 prefetchw(&dev_queue->dql.limit);
3549 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3553 dql_queued(&dev_queue->dql, bytes);
3555 if (likely(dql_avail(&dev_queue->dql) >= 0))
3558 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3561 * The XOFF flag must be set before checking the dql_avail below,
3562 * because in netdev_tx_completed_queue we update the dql_completed
3563 * before checking the XOFF flag.
3567 /* check again in case another CPU has just made room avail */
3568 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3569 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3573 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3574 * that they should not test BQL status themselves.
3575 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3577 * Returns true if the doorbell must be used to kick the NIC.
3579 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3585 dql_queued(&dev_queue->dql, bytes);
3587 return netif_tx_queue_stopped(dev_queue);
3589 netdev_tx_sent_queue(dev_queue, bytes);
3594 * netdev_sent_queue - report the number of bytes queued to hardware
3595 * @dev: network device
3596 * @bytes: number of bytes queued to the hardware device queue
3598 * Report the number of bytes queued for sending/completion to the network
3599 * device hardware queue. @bytes should be a good approximation and should
3600 * exactly match netdev_completed_queue() @bytes
3602 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3604 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3607 static inline bool __netdev_sent_queue(struct net_device *dev,
3611 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3615 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3616 unsigned int pkts, unsigned int bytes)
3619 if (unlikely(!bytes))
3622 dql_completed(&dev_queue->dql, bytes);
3625 * Without the memory barrier there is a small possiblity that
3626 * netdev_tx_sent_queue will miss the update and cause the queue to
3627 * be stopped forever
3631 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3634 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3635 netif_schedule_queue(dev_queue);
3640 * netdev_completed_queue - report bytes and packets completed by device
3641 * @dev: network device
3642 * @pkts: actual number of packets sent over the medium
3643 * @bytes: actual number of bytes sent over the medium
3645 * Report the number of bytes and packets transmitted by the network device
3646 * hardware queue over the physical medium, @bytes must exactly match the
3647 * @bytes amount passed to netdev_sent_queue()
3649 static inline void netdev_completed_queue(struct net_device *dev,
3650 unsigned int pkts, unsigned int bytes)
3652 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3655 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3658 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3664 * netdev_reset_queue - reset the packets and bytes count of a network device
3665 * @dev_queue: network device
3667 * Reset the bytes and packet count of a network device and clear the
3668 * software flow control OFF bit for this network device
3670 static inline void netdev_reset_queue(struct net_device *dev_queue)
3672 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3676 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3677 * @dev: network device
3678 * @queue_index: given tx queue index
3680 * Returns 0 if given tx queue index >= number of device tx queues,
3681 * otherwise returns the originally passed tx queue index.
3683 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3685 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3686 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3687 dev->name, queue_index,
3688 dev->real_num_tx_queues);
3696 * netif_running - test if up
3697 * @dev: network device
3699 * Test if the device has been brought up.
3701 static inline bool netif_running(const struct net_device *dev)
3703 return test_bit(__LINK_STATE_START, &dev->state);
3707 * Routines to manage the subqueues on a device. We only need start,
3708 * stop, and a check if it's stopped. All other device management is
3709 * done at the overall netdevice level.
3710 * Also test the device if we're multiqueue.
3714 * netif_start_subqueue - allow sending packets on subqueue
3715 * @dev: network device
3716 * @queue_index: sub queue index
3718 * Start individual transmit queue of a device with multiple transmit queues.
3720 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3722 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3724 netif_tx_start_queue(txq);
3728 * netif_stop_subqueue - stop sending packets on subqueue
3729 * @dev: network device
3730 * @queue_index: sub queue index
3732 * Stop individual transmit queue of a device with multiple transmit queues.
3734 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3736 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3737 netif_tx_stop_queue(txq);
3741 * __netif_subqueue_stopped - test status of subqueue
3742 * @dev: network device
3743 * @queue_index: sub queue index
3745 * Check individual transmit queue of a device with multiple transmit queues.
3747 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3750 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3752 return netif_tx_queue_stopped(txq);
3756 * netif_subqueue_stopped - test status of subqueue
3757 * @dev: network device
3758 * @skb: sub queue buffer pointer
3760 * Check individual transmit queue of a device with multiple transmit queues.
3762 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3763 struct sk_buff *skb)
3765 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3769 * netif_wake_subqueue - allow sending packets on subqueue
3770 * @dev: network device
3771 * @queue_index: sub queue index
3773 * Resume individual transmit queue of a device with multiple transmit queues.
3775 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3777 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3779 netif_tx_wake_queue(txq);
3783 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3785 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3786 u16 index, enum xps_map_type type);
3789 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3790 * @j: CPU/Rx queue index
3791 * @mask: bitmask of all cpus/rx queues
3792 * @nr_bits: number of bits in the bitmask
3794 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3796 static inline bool netif_attr_test_mask(unsigned long j,
3797 const unsigned long *mask,
3798 unsigned int nr_bits)
3800 cpu_max_bits_warn(j, nr_bits);
3801 return test_bit(j, mask);
3805 * netif_attr_test_online - Test for online CPU/Rx queue
3806 * @j: CPU/Rx queue index
3807 * @online_mask: bitmask for CPUs/Rx queues that are online
3808 * @nr_bits: number of bits in the bitmask
3810 * Returns true if a CPU/Rx queue is online.
3812 static inline bool netif_attr_test_online(unsigned long j,
3813 const unsigned long *online_mask,
3814 unsigned int nr_bits)
3816 cpu_max_bits_warn(j, nr_bits);
3819 return test_bit(j, online_mask);
3821 return (j < nr_bits);
3825 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3826 * @n: CPU/Rx queue index
3827 * @srcp: the cpumask/Rx queue mask pointer
3828 * @nr_bits: number of bits in the bitmask
3830 * Returns >= nr_bits if no further CPUs/Rx queues set.
3832 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3833 unsigned int nr_bits)
3835 /* -1 is a legal arg here. */
3837 cpu_max_bits_warn(n, nr_bits);
3840 return find_next_bit(srcp, nr_bits, n + 1);
3846 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3847 * @n: CPU/Rx queue index
3848 * @src1p: the first CPUs/Rx queues mask pointer
3849 * @src2p: the second CPUs/Rx queues mask pointer
3850 * @nr_bits: number of bits in the bitmask
3852 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3854 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3855 const unsigned long *src2p,
3856 unsigned int nr_bits)
3858 /* -1 is a legal arg here. */
3860 cpu_max_bits_warn(n, nr_bits);
3863 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3865 return find_next_bit(src1p, nr_bits, n + 1);
3867 return find_next_bit(src2p, nr_bits, n + 1);
3872 static inline int netif_set_xps_queue(struct net_device *dev,
3873 const struct cpumask *mask,
3879 static inline int __netif_set_xps_queue(struct net_device *dev,
3880 const unsigned long *mask,
3881 u16 index, enum xps_map_type type)
3888 * netif_is_multiqueue - test if device has multiple transmit queues
3889 * @dev: network device
3891 * Check if device has multiple transmit queues
3893 static inline bool netif_is_multiqueue(const struct net_device *dev)
3895 return dev->num_tx_queues > 1;
3898 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3901 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3903 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3906 dev->real_num_rx_queues = rxqs;
3911 static inline struct netdev_rx_queue *
3912 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3914 return dev->_rx + rxq;
3918 static inline unsigned int get_netdev_rx_queue_index(
3919 struct netdev_rx_queue *queue)
3921 struct net_device *dev = queue->dev;
3922 int index = queue - dev->_rx;
3924 BUG_ON(index >= dev->num_rx_queues);
3929 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3930 int netif_get_num_default_rss_queues(void);
3932 enum skb_free_reason {
3933 SKB_REASON_CONSUMED,
3937 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3938 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3941 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3942 * interrupt context or with hardware interrupts being disabled.
3943 * (in_irq() || irqs_disabled())
3945 * We provide four helpers that can be used in following contexts :
3947 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3948 * replacing kfree_skb(skb)
3950 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3951 * Typically used in place of consume_skb(skb) in TX completion path
3953 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3954 * replacing kfree_skb(skb)
3956 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3957 * and consumed a packet. Used in place of consume_skb(skb)
3959 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3961 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3964 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3966 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3969 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3971 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3974 static inline void dev_consume_skb_any(struct sk_buff *skb)
3976 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3979 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3980 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3981 int netif_rx(struct sk_buff *skb);
3982 int netif_rx_ni(struct sk_buff *skb);
3983 int netif_rx_any_context(struct sk_buff *skb);
3984 int netif_receive_skb(struct sk_buff *skb);
3985 int netif_receive_skb_core(struct sk_buff *skb);
3986 void netif_receive_skb_list(struct list_head *head);
3987 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3988 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3989 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3990 gro_result_t napi_gro_frags(struct napi_struct *napi);
3991 struct packet_offload *gro_find_receive_by_type(__be16 type);
3992 struct packet_offload *gro_find_complete_by_type(__be16 type);
3994 static inline void napi_free_frags(struct napi_struct *napi)
3996 kfree_skb(napi->skb);
4000 bool netdev_is_rx_handler_busy(struct net_device *dev);
4001 int netdev_rx_handler_register(struct net_device *dev,
4002 rx_handler_func_t *rx_handler,
4003 void *rx_handler_data);
4004 void netdev_rx_handler_unregister(struct net_device *dev);
4006 bool dev_valid_name(const char *name);
4007 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4008 bool *need_copyout);
4009 int dev_ifconf(struct net *net, struct ifconf *, int);
4010 int dev_ethtool(struct net *net, struct ifreq *);
4011 unsigned int dev_get_flags(const struct net_device *);
4012 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4013 struct netlink_ext_ack *extack);
4014 int dev_change_flags(struct net_device *dev, unsigned int flags,
4015 struct netlink_ext_ack *extack);
4016 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4017 unsigned int gchanges);
4018 int dev_change_name(struct net_device *, const char *);
4019 int dev_set_alias(struct net_device *, const char *, size_t);
4020 int dev_get_alias(const struct net_device *, char *, size_t);
4021 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
4022 int __dev_set_mtu(struct net_device *, int);
4023 int dev_validate_mtu(struct net_device *dev, int mtu,
4024 struct netlink_ext_ack *extack);
4025 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4026 struct netlink_ext_ack *extack);
4027 int dev_set_mtu(struct net_device *, int);
4028 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4029 void dev_set_group(struct net_device *, int);
4030 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4031 struct netlink_ext_ack *extack);
4032 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4033 struct netlink_ext_ack *extack);
4034 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4035 struct netlink_ext_ack *extack);
4036 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4037 int dev_change_carrier(struct net_device *, bool new_carrier);
4038 int dev_get_phys_port_id(struct net_device *dev,
4039 struct netdev_phys_item_id *ppid);
4040 int dev_get_phys_port_name(struct net_device *dev,
4041 char *name, size_t len);
4042 int dev_get_port_parent_id(struct net_device *dev,
4043 struct netdev_phys_item_id *ppid, bool recurse);
4044 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4045 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4046 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4047 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4049 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4050 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4051 struct netdev_queue *txq, int *ret);
4053 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4054 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4055 int fd, int expected_fd, u32 flags);
4056 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4057 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4059 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4060 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4061 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4062 bool is_skb_forwardable(const struct net_device *dev,
4063 const struct sk_buff *skb);
4065 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4066 const struct sk_buff *skb,
4067 const bool check_mtu)
4069 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4072 if (!(dev->flags & IFF_UP))
4078 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4079 if (skb->len <= len)
4082 /* if TSO is enabled, we don't care about the length as the packet
4083 * could be forwarded without being segmented before
4085 if (skb_is_gso(skb))
4091 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4092 struct sk_buff *skb,
4093 const bool check_mtu)
4095 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4096 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4097 atomic_long_inc(&dev->rx_dropped);
4102 skb_scrub_packet(skb, true);
4107 bool dev_nit_active(struct net_device *dev);
4108 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4110 extern int netdev_budget;
4111 extern unsigned int netdev_budget_usecs;
4113 /* Called by rtnetlink.c:rtnl_unlock() */
4114 void netdev_run_todo(void);
4117 * dev_put - release reference to device
4118 * @dev: network device
4120 * Release reference to device to allow it to be freed.
4122 static inline void dev_put(struct net_device *dev)
4124 #ifdef CONFIG_PCPU_DEV_REFCNT
4125 this_cpu_dec(*dev->pcpu_refcnt);
4127 refcount_dec(&dev->dev_refcnt);
4132 * dev_hold - get reference to device
4133 * @dev: network device
4135 * Hold reference to device to keep it from being freed.
4137 static inline void dev_hold(struct net_device *dev)
4139 #ifdef CONFIG_PCPU_DEV_REFCNT
4140 this_cpu_inc(*dev->pcpu_refcnt);
4142 refcount_inc(&dev->dev_refcnt);
4146 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4147 * and _off may be called from IRQ context, but it is caller
4148 * who is responsible for serialization of these calls.
4150 * The name carrier is inappropriate, these functions should really be
4151 * called netif_lowerlayer_*() because they represent the state of any
4152 * kind of lower layer not just hardware media.
4155 void linkwatch_init_dev(struct net_device *dev);
4156 void linkwatch_fire_event(struct net_device *dev);
4157 void linkwatch_forget_dev(struct net_device *dev);
4160 * netif_carrier_ok - test if carrier present
4161 * @dev: network device
4163 * Check if carrier is present on device
4165 static inline bool netif_carrier_ok(const struct net_device *dev)
4167 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4170 unsigned long dev_trans_start(struct net_device *dev);
4172 void __netdev_watchdog_up(struct net_device *dev);
4174 void netif_carrier_on(struct net_device *dev);
4176 void netif_carrier_off(struct net_device *dev);
4179 * netif_dormant_on - mark device as dormant.
4180 * @dev: network device
4182 * Mark device as dormant (as per RFC2863).
4184 * The dormant state indicates that the relevant interface is not
4185 * actually in a condition to pass packets (i.e., it is not 'up') but is
4186 * in a "pending" state, waiting for some external event. For "on-
4187 * demand" interfaces, this new state identifies the situation where the
4188 * interface is waiting for events to place it in the up state.
4190 static inline void netif_dormant_on(struct net_device *dev)
4192 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4193 linkwatch_fire_event(dev);
4197 * netif_dormant_off - set device as not dormant.
4198 * @dev: network device
4200 * Device is not in dormant state.
4202 static inline void netif_dormant_off(struct net_device *dev)
4204 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4205 linkwatch_fire_event(dev);
4209 * netif_dormant - test if device is dormant
4210 * @dev: network device
4212 * Check if device is dormant.
4214 static inline bool netif_dormant(const struct net_device *dev)
4216 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4221 * netif_testing_on - mark device as under test.
4222 * @dev: network device
4224 * Mark device as under test (as per RFC2863).
4226 * The testing state indicates that some test(s) must be performed on
4227 * the interface. After completion, of the test, the interface state
4228 * will change to up, dormant, or down, as appropriate.
4230 static inline void netif_testing_on(struct net_device *dev)
4232 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4233 linkwatch_fire_event(dev);
4237 * netif_testing_off - set device as not under test.
4238 * @dev: network device
4240 * Device is not in testing state.
4242 static inline void netif_testing_off(struct net_device *dev)
4244 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4245 linkwatch_fire_event(dev);
4249 * netif_testing - test if device is under test
4250 * @dev: network device
4252 * Check if device is under test
4254 static inline bool netif_testing(const struct net_device *dev)
4256 return test_bit(__LINK_STATE_TESTING, &dev->state);
4261 * netif_oper_up - test if device is operational
4262 * @dev: network device
4264 * Check if carrier is operational
4266 static inline bool netif_oper_up(const struct net_device *dev)
4268 return (dev->operstate == IF_OPER_UP ||
4269 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4273 * netif_device_present - is device available or removed
4274 * @dev: network device
4276 * Check if device has not been removed from system.
4278 static inline bool netif_device_present(const struct net_device *dev)
4280 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4283 void netif_device_detach(struct net_device *dev);
4285 void netif_device_attach(struct net_device *dev);
4288 * Network interface message level settings
4293 NETIF_MSG_PROBE_BIT,
4295 NETIF_MSG_TIMER_BIT,
4296 NETIF_MSG_IFDOWN_BIT,
4298 NETIF_MSG_RX_ERR_BIT,
4299 NETIF_MSG_TX_ERR_BIT,
4300 NETIF_MSG_TX_QUEUED_BIT,
4302 NETIF_MSG_TX_DONE_BIT,
4303 NETIF_MSG_RX_STATUS_BIT,
4304 NETIF_MSG_PKTDATA_BIT,
4308 /* When you add a new bit above, update netif_msg_class_names array
4309 * in net/ethtool/common.c
4311 NETIF_MSG_CLASS_COUNT,
4313 /* Both ethtool_ops interface and internal driver implementation use u32 */
4314 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4316 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4317 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4319 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4320 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4321 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4322 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4323 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4324 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4325 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4326 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4327 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4328 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4329 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4330 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4331 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4332 #define NETIF_MSG_HW __NETIF_MSG(HW)
4333 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4335 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4336 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4337 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4338 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4339 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4340 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4341 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4342 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4343 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4344 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4345 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4346 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4347 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4348 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4349 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4351 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4354 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4355 return default_msg_enable_bits;
4356 if (debug_value == 0) /* no output */
4358 /* set low N bits */
4359 return (1U << debug_value) - 1;
4362 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4364 spin_lock(&txq->_xmit_lock);
4365 txq->xmit_lock_owner = cpu;
4368 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4370 __acquire(&txq->_xmit_lock);
4374 static inline void __netif_tx_release(struct netdev_queue *txq)
4376 __release(&txq->_xmit_lock);
4379 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4381 spin_lock_bh(&txq->_xmit_lock);
4382 txq->xmit_lock_owner = smp_processor_id();
4385 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4387 bool ok = spin_trylock(&txq->_xmit_lock);
4389 txq->xmit_lock_owner = smp_processor_id();
4393 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4395 txq->xmit_lock_owner = -1;
4396 spin_unlock(&txq->_xmit_lock);
4399 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4401 txq->xmit_lock_owner = -1;
4402 spin_unlock_bh(&txq->_xmit_lock);
4405 static inline void txq_trans_update(struct netdev_queue *txq)
4407 if (txq->xmit_lock_owner != -1)
4408 txq->trans_start = jiffies;
4411 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4412 static inline void netif_trans_update(struct net_device *dev)
4414 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4416 if (txq->trans_start != jiffies)
4417 txq->trans_start = jiffies;
4421 * netif_tx_lock - grab network device transmit lock
4422 * @dev: network device
4424 * Get network device transmit lock
4426 static inline void netif_tx_lock(struct net_device *dev)
4431 spin_lock(&dev->tx_global_lock);
4432 cpu = smp_processor_id();
4433 for (i = 0; i < dev->num_tx_queues; i++) {
4434 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4436 /* We are the only thread of execution doing a
4437 * freeze, but we have to grab the _xmit_lock in
4438 * order to synchronize with threads which are in
4439 * the ->hard_start_xmit() handler and already
4440 * checked the frozen bit.
4442 __netif_tx_lock(txq, cpu);
4443 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4444 __netif_tx_unlock(txq);
4448 static inline void netif_tx_lock_bh(struct net_device *dev)
4454 static inline void netif_tx_unlock(struct net_device *dev)
4458 for (i = 0; i < dev->num_tx_queues; i++) {
4459 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4461 /* No need to grab the _xmit_lock here. If the
4462 * queue is not stopped for another reason, we
4465 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4466 netif_schedule_queue(txq);
4468 spin_unlock(&dev->tx_global_lock);
4471 static inline void netif_tx_unlock_bh(struct net_device *dev)
4473 netif_tx_unlock(dev);
4477 #define HARD_TX_LOCK(dev, txq, cpu) { \
4478 if ((dev->features & NETIF_F_LLTX) == 0) { \
4479 __netif_tx_lock(txq, cpu); \
4481 __netif_tx_acquire(txq); \
4485 #define HARD_TX_TRYLOCK(dev, txq) \
4486 (((dev->features & NETIF_F_LLTX) == 0) ? \
4487 __netif_tx_trylock(txq) : \
4488 __netif_tx_acquire(txq))
4490 #define HARD_TX_UNLOCK(dev, txq) { \
4491 if ((dev->features & NETIF_F_LLTX) == 0) { \
4492 __netif_tx_unlock(txq); \
4494 __netif_tx_release(txq); \
4498 static inline void netif_tx_disable(struct net_device *dev)
4504 cpu = smp_processor_id();
4505 spin_lock(&dev->tx_global_lock);
4506 for (i = 0; i < dev->num_tx_queues; i++) {
4507 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4509 __netif_tx_lock(txq, cpu);
4510 netif_tx_stop_queue(txq);
4511 __netif_tx_unlock(txq);
4513 spin_unlock(&dev->tx_global_lock);
4517 static inline void netif_addr_lock(struct net_device *dev)
4519 unsigned char nest_level = 0;
4521 #ifdef CONFIG_LOCKDEP
4522 nest_level = dev->nested_level;
4524 spin_lock_nested(&dev->addr_list_lock, nest_level);
4527 static inline void netif_addr_lock_bh(struct net_device *dev)
4529 unsigned char nest_level = 0;
4531 #ifdef CONFIG_LOCKDEP
4532 nest_level = dev->nested_level;
4535 spin_lock_nested(&dev->addr_list_lock, nest_level);
4538 static inline void netif_addr_unlock(struct net_device *dev)
4540 spin_unlock(&dev->addr_list_lock);
4543 static inline void netif_addr_unlock_bh(struct net_device *dev)
4545 spin_unlock_bh(&dev->addr_list_lock);
4549 * dev_addrs walker. Should be used only for read access. Call with
4550 * rcu_read_lock held.
4552 #define for_each_dev_addr(dev, ha) \
4553 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4555 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4557 void ether_setup(struct net_device *dev);
4559 /* Support for loadable net-drivers */
4560 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4561 unsigned char name_assign_type,
4562 void (*setup)(struct net_device *),
4563 unsigned int txqs, unsigned int rxqs);
4564 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4565 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4567 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4568 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4571 int register_netdev(struct net_device *dev);
4572 void unregister_netdev(struct net_device *dev);
4574 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4576 /* General hardware address lists handling functions */
4577 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4578 struct netdev_hw_addr_list *from_list, int addr_len);
4579 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4580 struct netdev_hw_addr_list *from_list, int addr_len);
4581 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4582 struct net_device *dev,
4583 int (*sync)(struct net_device *, const unsigned char *),
4584 int (*unsync)(struct net_device *,
4585 const unsigned char *));
4586 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4587 struct net_device *dev,
4588 int (*sync)(struct net_device *,
4589 const unsigned char *, int),
4590 int (*unsync)(struct net_device *,
4591 const unsigned char *, int));
4592 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4593 struct net_device *dev,
4594 int (*unsync)(struct net_device *,
4595 const unsigned char *, int));
4596 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4597 struct net_device *dev,
4598 int (*unsync)(struct net_device *,
4599 const unsigned char *));
4600 void __hw_addr_init(struct netdev_hw_addr_list *list);
4602 /* Functions used for device addresses handling */
4603 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4604 unsigned char addr_type);
4605 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4606 unsigned char addr_type);
4607 void dev_addr_flush(struct net_device *dev);
4608 int dev_addr_init(struct net_device *dev);
4610 /* Functions used for unicast addresses handling */
4611 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4612 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4613 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4614 int dev_uc_sync(struct net_device *to, struct net_device *from);
4615 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4616 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4617 void dev_uc_flush(struct net_device *dev);
4618 void dev_uc_init(struct net_device *dev);
4621 * __dev_uc_sync - Synchonize device's unicast list
4622 * @dev: device to sync
4623 * @sync: function to call if address should be added
4624 * @unsync: function to call if address should be removed
4626 * Add newly added addresses to the interface, and release
4627 * addresses that have been deleted.
4629 static inline int __dev_uc_sync(struct net_device *dev,
4630 int (*sync)(struct net_device *,
4631 const unsigned char *),
4632 int (*unsync)(struct net_device *,
4633 const unsigned char *))
4635 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4639 * __dev_uc_unsync - Remove synchronized addresses from device
4640 * @dev: device to sync
4641 * @unsync: function to call if address should be removed
4643 * Remove all addresses that were added to the device by dev_uc_sync().
4645 static inline void __dev_uc_unsync(struct net_device *dev,
4646 int (*unsync)(struct net_device *,
4647 const unsigned char *))
4649 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4652 /* Functions used for multicast addresses handling */
4653 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4654 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4655 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4656 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4657 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4658 int dev_mc_sync(struct net_device *to, struct net_device *from);
4659 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4660 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4661 void dev_mc_flush(struct net_device *dev);
4662 void dev_mc_init(struct net_device *dev);
4665 * __dev_mc_sync - Synchonize device's multicast list
4666 * @dev: device to sync
4667 * @sync: function to call if address should be added
4668 * @unsync: function to call if address should be removed
4670 * Add newly added addresses to the interface, and release
4671 * addresses that have been deleted.
4673 static inline int __dev_mc_sync(struct net_device *dev,
4674 int (*sync)(struct net_device *,
4675 const unsigned char *),
4676 int (*unsync)(struct net_device *,
4677 const unsigned char *))
4679 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4683 * __dev_mc_unsync - Remove synchronized addresses from device
4684 * @dev: device to sync
4685 * @unsync: function to call if address should be removed
4687 * Remove all addresses that were added to the device by dev_mc_sync().
4689 static inline void __dev_mc_unsync(struct net_device *dev,
4690 int (*unsync)(struct net_device *,
4691 const unsigned char *))
4693 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4696 /* Functions used for secondary unicast and multicast support */
4697 void dev_set_rx_mode(struct net_device *dev);
4698 void __dev_set_rx_mode(struct net_device *dev);
4699 int dev_set_promiscuity(struct net_device *dev, int inc);
4700 int dev_set_allmulti(struct net_device *dev, int inc);
4701 void netdev_state_change(struct net_device *dev);
4702 void __netdev_notify_peers(struct net_device *dev);
4703 void netdev_notify_peers(struct net_device *dev);
4704 void netdev_features_change(struct net_device *dev);
4705 /* Load a device via the kmod */
4706 void dev_load(struct net *net, const char *name);
4707 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4708 struct rtnl_link_stats64 *storage);
4709 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4710 const struct net_device_stats *netdev_stats);
4711 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4712 const struct pcpu_sw_netstats __percpu *netstats);
4713 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4715 extern int netdev_max_backlog;
4716 extern int netdev_tstamp_prequeue;
4717 extern int netdev_unregister_timeout_secs;
4718 extern int weight_p;
4719 extern int dev_weight_rx_bias;
4720 extern int dev_weight_tx_bias;
4721 extern int dev_rx_weight;
4722 extern int dev_tx_weight;
4723 extern int gro_normal_batch;
4726 NESTED_SYNC_IMM_BIT,
4727 NESTED_SYNC_TODO_BIT,
4730 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4731 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4733 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4734 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4736 struct netdev_nested_priv {
4737 unsigned char flags;
4741 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4742 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4743 struct list_head **iter);
4744 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4745 struct list_head **iter);
4747 #ifdef CONFIG_LOCKDEP
4748 static LIST_HEAD(net_unlink_list);
4750 static inline void net_unlink_todo(struct net_device *dev)
4752 if (list_empty(&dev->unlink_list))
4753 list_add_tail(&dev->unlink_list, &net_unlink_list);
4757 /* iterate through upper list, must be called under RCU read lock */
4758 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4759 for (iter = &(dev)->adj_list.upper, \
4760 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4762 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4764 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4765 int (*fn)(struct net_device *upper_dev,
4766 struct netdev_nested_priv *priv),
4767 struct netdev_nested_priv *priv);
4769 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4770 struct net_device *upper_dev);
4772 bool netdev_has_any_upper_dev(struct net_device *dev);
4774 void *netdev_lower_get_next_private(struct net_device *dev,
4775 struct list_head **iter);
4776 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4777 struct list_head **iter);
4779 #define netdev_for_each_lower_private(dev, priv, iter) \
4780 for (iter = (dev)->adj_list.lower.next, \
4781 priv = netdev_lower_get_next_private(dev, &(iter)); \
4783 priv = netdev_lower_get_next_private(dev, &(iter)))
4785 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4786 for (iter = &(dev)->adj_list.lower, \
4787 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4789 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4791 void *netdev_lower_get_next(struct net_device *dev,
4792 struct list_head **iter);
4794 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4795 for (iter = (dev)->adj_list.lower.next, \
4796 ldev = netdev_lower_get_next(dev, &(iter)); \
4798 ldev = netdev_lower_get_next(dev, &(iter)))
4800 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4801 struct list_head **iter);
4802 int netdev_walk_all_lower_dev(struct net_device *dev,
4803 int (*fn)(struct net_device *lower_dev,
4804 struct netdev_nested_priv *priv),
4805 struct netdev_nested_priv *priv);
4806 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4807 int (*fn)(struct net_device *lower_dev,
4808 struct netdev_nested_priv *priv),
4809 struct netdev_nested_priv *priv);
4811 void *netdev_adjacent_get_private(struct list_head *adj_list);
4812 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4813 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4814 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4815 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4816 struct netlink_ext_ack *extack);
4817 int netdev_master_upper_dev_link(struct net_device *dev,
4818 struct net_device *upper_dev,
4819 void *upper_priv, void *upper_info,
4820 struct netlink_ext_ack *extack);
4821 void netdev_upper_dev_unlink(struct net_device *dev,
4822 struct net_device *upper_dev);
4823 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4824 struct net_device *new_dev,
4825 struct net_device *dev,
4826 struct netlink_ext_ack *extack);
4827 void netdev_adjacent_change_commit(struct net_device *old_dev,
4828 struct net_device *new_dev,
4829 struct net_device *dev);
4830 void netdev_adjacent_change_abort(struct net_device *old_dev,
4831 struct net_device *new_dev,
4832 struct net_device *dev);
4833 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4834 void *netdev_lower_dev_get_private(struct net_device *dev,
4835 struct net_device *lower_dev);
4836 void netdev_lower_state_changed(struct net_device *lower_dev,
4837 void *lower_state_info);
4839 /* RSS keys are 40 or 52 bytes long */
4840 #define NETDEV_RSS_KEY_LEN 52
4841 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4842 void netdev_rss_key_fill(void *buffer, size_t len);
4844 int skb_checksum_help(struct sk_buff *skb);
4845 int skb_crc32c_csum_help(struct sk_buff *skb);
4846 int skb_csum_hwoffload_help(struct sk_buff *skb,
4847 const netdev_features_t features);
4849 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4850 netdev_features_t features, bool tx_path);
4851 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4852 netdev_features_t features);
4854 struct netdev_bonding_info {
4859 struct netdev_notifier_bonding_info {
4860 struct netdev_notifier_info info; /* must be first */
4861 struct netdev_bonding_info bonding_info;
4864 void netdev_bonding_info_change(struct net_device *dev,
4865 struct netdev_bonding_info *bonding_info);
4867 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4868 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4870 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4877 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4879 return __skb_gso_segment(skb, features, true);
4881 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4883 static inline bool can_checksum_protocol(netdev_features_t features,
4886 if (protocol == htons(ETH_P_FCOE))
4887 return !!(features & NETIF_F_FCOE_CRC);
4889 /* Assume this is an IP checksum (not SCTP CRC) */
4891 if (features & NETIF_F_HW_CSUM) {
4892 /* Can checksum everything */
4897 case htons(ETH_P_IP):
4898 return !!(features & NETIF_F_IP_CSUM);
4899 case htons(ETH_P_IPV6):
4900 return !!(features & NETIF_F_IPV6_CSUM);
4907 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4909 static inline void netdev_rx_csum_fault(struct net_device *dev,
4910 struct sk_buff *skb)
4914 /* rx skb timestamps */
4915 void net_enable_timestamp(void);
4916 void net_disable_timestamp(void);
4918 #ifdef CONFIG_PROC_FS
4919 int __init dev_proc_init(void);
4921 #define dev_proc_init() 0
4924 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4925 struct sk_buff *skb, struct net_device *dev,
4928 __this_cpu_write(softnet_data.xmit.more, more);
4929 return ops->ndo_start_xmit(skb, dev);
4932 static inline bool netdev_xmit_more(void)
4934 return __this_cpu_read(softnet_data.xmit.more);
4937 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4938 struct netdev_queue *txq, bool more)
4940 const struct net_device_ops *ops = dev->netdev_ops;
4943 rc = __netdev_start_xmit(ops, skb, dev, more);
4944 if (rc == NETDEV_TX_OK)
4945 txq_trans_update(txq);
4950 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4952 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4955 extern const struct kobj_ns_type_operations net_ns_type_operations;
4957 const char *netdev_drivername(const struct net_device *dev);
4959 void linkwatch_run_queue(void);
4961 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4962 netdev_features_t f2)
4964 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4965 if (f1 & NETIF_F_HW_CSUM)
4966 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4968 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4974 static inline netdev_features_t netdev_get_wanted_features(
4975 struct net_device *dev)
4977 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4979 netdev_features_t netdev_increment_features(netdev_features_t all,
4980 netdev_features_t one, netdev_features_t mask);
4982 /* Allow TSO being used on stacked device :
4983 * Performing the GSO segmentation before last device
4984 * is a performance improvement.
4986 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4987 netdev_features_t mask)
4989 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4992 int __netdev_update_features(struct net_device *dev);
4993 void netdev_update_features(struct net_device *dev);
4994 void netdev_change_features(struct net_device *dev);
4996 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4997 struct net_device *dev);
4999 netdev_features_t passthru_features_check(struct sk_buff *skb,
5000 struct net_device *dev,
5001 netdev_features_t features);
5002 netdev_features_t netif_skb_features(struct sk_buff *skb);
5004 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5006 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5008 /* check flags correspondence */
5009 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5010 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5011 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5012 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5013 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5014 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5015 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5016 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5017 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5018 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5019 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5020 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5021 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5022 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5023 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5024 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5025 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5026 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5027 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5029 return (features & feature) == feature;
5032 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5034 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5035 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5038 static inline bool netif_needs_gso(struct sk_buff *skb,
5039 netdev_features_t features)
5041 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5042 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5043 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5046 static inline void netif_set_gso_max_size(struct net_device *dev,
5049 dev->gso_max_size = size;
5052 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5053 int pulled_hlen, u16 mac_offset,
5056 skb->protocol = protocol;
5057 skb->encapsulation = 1;
5058 skb_push(skb, pulled_hlen);
5059 skb_reset_transport_header(skb);
5060 skb->mac_header = mac_offset;
5061 skb->network_header = skb->mac_header + mac_len;
5062 skb->mac_len = mac_len;
5065 static inline bool netif_is_macsec(const struct net_device *dev)
5067 return dev->priv_flags & IFF_MACSEC;
5070 static inline bool netif_is_macvlan(const struct net_device *dev)
5072 return dev->priv_flags & IFF_MACVLAN;
5075 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5077 return dev->priv_flags & IFF_MACVLAN_PORT;
5080 static inline bool netif_is_bond_master(const struct net_device *dev)
5082 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5085 static inline bool netif_is_bond_slave(const struct net_device *dev)
5087 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5090 static inline bool netif_supports_nofcs(struct net_device *dev)
5092 return dev->priv_flags & IFF_SUPP_NOFCS;
5095 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5097 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5100 static inline bool netif_is_l3_master(const struct net_device *dev)
5102 return dev->priv_flags & IFF_L3MDEV_MASTER;
5105 static inline bool netif_is_l3_slave(const struct net_device *dev)
5107 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5110 static inline bool netif_is_bridge_master(const struct net_device *dev)
5112 return dev->priv_flags & IFF_EBRIDGE;
5115 static inline bool netif_is_bridge_port(const struct net_device *dev)
5117 return dev->priv_flags & IFF_BRIDGE_PORT;
5120 static inline bool netif_is_ovs_master(const struct net_device *dev)
5122 return dev->priv_flags & IFF_OPENVSWITCH;
5125 static inline bool netif_is_ovs_port(const struct net_device *dev)
5127 return dev->priv_flags & IFF_OVS_DATAPATH;
5130 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5132 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5135 static inline bool netif_is_team_master(const struct net_device *dev)
5137 return dev->priv_flags & IFF_TEAM;
5140 static inline bool netif_is_team_port(const struct net_device *dev)
5142 return dev->priv_flags & IFF_TEAM_PORT;
5145 static inline bool netif_is_lag_master(const struct net_device *dev)
5147 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5150 static inline bool netif_is_lag_port(const struct net_device *dev)
5152 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5155 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5157 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5160 static inline bool netif_is_failover(const struct net_device *dev)
5162 return dev->priv_flags & IFF_FAILOVER;
5165 static inline bool netif_is_failover_slave(const struct net_device *dev)
5167 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5170 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5171 static inline void netif_keep_dst(struct net_device *dev)
5173 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5176 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5177 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5179 /* TODO: reserve and use an additional IFF bit, if we get more users */
5180 return dev->priv_flags & IFF_MACSEC;
5183 extern struct pernet_operations __net_initdata loopback_net_ops;
5185 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5187 /* netdev_printk helpers, similar to dev_printk */
5189 static inline const char *netdev_name(const struct net_device *dev)
5191 if (!dev->name[0] || strchr(dev->name, '%'))
5192 return "(unnamed net_device)";
5196 static inline bool netdev_unregistering(const struct net_device *dev)
5198 return dev->reg_state == NETREG_UNREGISTERING;
5201 static inline const char *netdev_reg_state(const struct net_device *dev)
5203 switch (dev->reg_state) {
5204 case NETREG_UNINITIALIZED: return " (uninitialized)";
5205 case NETREG_REGISTERED: return "";
5206 case NETREG_UNREGISTERING: return " (unregistering)";
5207 case NETREG_UNREGISTERED: return " (unregistered)";
5208 case NETREG_RELEASED: return " (released)";
5209 case NETREG_DUMMY: return " (dummy)";
5212 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5213 return " (unknown)";
5216 __printf(3, 4) __cold
5217 void netdev_printk(const char *level, const struct net_device *dev,
5218 const char *format, ...);
5219 __printf(2, 3) __cold
5220 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5221 __printf(2, 3) __cold
5222 void netdev_alert(const struct net_device *dev, const char *format, ...);
5223 __printf(2, 3) __cold
5224 void netdev_crit(const struct net_device *dev, const char *format, ...);
5225 __printf(2, 3) __cold
5226 void netdev_err(const struct net_device *dev, const char *format, ...);
5227 __printf(2, 3) __cold
5228 void netdev_warn(const struct net_device *dev, const char *format, ...);
5229 __printf(2, 3) __cold
5230 void netdev_notice(const struct net_device *dev, const char *format, ...);
5231 __printf(2, 3) __cold
5232 void netdev_info(const struct net_device *dev, const char *format, ...);
5234 #define netdev_level_once(level, dev, fmt, ...) \
5236 static bool __print_once __read_mostly; \
5238 if (!__print_once) { \
5239 __print_once = true; \
5240 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5244 #define netdev_emerg_once(dev, fmt, ...) \
5245 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5246 #define netdev_alert_once(dev, fmt, ...) \
5247 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5248 #define netdev_crit_once(dev, fmt, ...) \
5249 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5250 #define netdev_err_once(dev, fmt, ...) \
5251 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5252 #define netdev_warn_once(dev, fmt, ...) \
5253 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5254 #define netdev_notice_once(dev, fmt, ...) \
5255 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5256 #define netdev_info_once(dev, fmt, ...) \
5257 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5259 #define MODULE_ALIAS_NETDEV(device) \
5260 MODULE_ALIAS("netdev-" device)
5262 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5263 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5264 #define netdev_dbg(__dev, format, args...) \
5266 dynamic_netdev_dbg(__dev, format, ##args); \
5268 #elif defined(DEBUG)
5269 #define netdev_dbg(__dev, format, args...) \
5270 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5272 #define netdev_dbg(__dev, format, args...) \
5275 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5279 #if defined(VERBOSE_DEBUG)
5280 #define netdev_vdbg netdev_dbg
5283 #define netdev_vdbg(dev, format, args...) \
5286 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5292 * netdev_WARN() acts like dev_printk(), but with the key difference
5293 * of using a WARN/WARN_ON to get the message out, including the
5294 * file/line information and a backtrace.
5296 #define netdev_WARN(dev, format, args...) \
5297 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5298 netdev_reg_state(dev), ##args)
5300 #define netdev_WARN_ONCE(dev, format, args...) \
5301 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5302 netdev_reg_state(dev), ##args)
5304 /* netif printk helpers, similar to netdev_printk */
5306 #define netif_printk(priv, type, level, dev, fmt, args...) \
5308 if (netif_msg_##type(priv)) \
5309 netdev_printk(level, (dev), fmt, ##args); \
5312 #define netif_level(level, priv, type, dev, fmt, args...) \
5314 if (netif_msg_##type(priv)) \
5315 netdev_##level(dev, fmt, ##args); \
5318 #define netif_emerg(priv, type, dev, fmt, args...) \
5319 netif_level(emerg, priv, type, dev, fmt, ##args)
5320 #define netif_alert(priv, type, dev, fmt, args...) \
5321 netif_level(alert, priv, type, dev, fmt, ##args)
5322 #define netif_crit(priv, type, dev, fmt, args...) \
5323 netif_level(crit, priv, type, dev, fmt, ##args)
5324 #define netif_err(priv, type, dev, fmt, args...) \
5325 netif_level(err, priv, type, dev, fmt, ##args)
5326 #define netif_warn(priv, type, dev, fmt, args...) \
5327 netif_level(warn, priv, type, dev, fmt, ##args)
5328 #define netif_notice(priv, type, dev, fmt, args...) \
5329 netif_level(notice, priv, type, dev, fmt, ##args)
5330 #define netif_info(priv, type, dev, fmt, args...) \
5331 netif_level(info, priv, type, dev, fmt, ##args)
5333 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5334 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5335 #define netif_dbg(priv, type, netdev, format, args...) \
5337 if (netif_msg_##type(priv)) \
5338 dynamic_netdev_dbg(netdev, format, ##args); \
5340 #elif defined(DEBUG)
5341 #define netif_dbg(priv, type, dev, format, args...) \
5342 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5344 #define netif_dbg(priv, type, dev, format, args...) \
5347 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5352 /* if @cond then downgrade to debug, else print at @level */
5353 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5356 netif_dbg(priv, type, netdev, fmt, ##args); \
5358 netif_ ## level(priv, type, netdev, fmt, ##args); \
5361 #if defined(VERBOSE_DEBUG)
5362 #define netif_vdbg netif_dbg
5364 #define netif_vdbg(priv, type, dev, format, args...) \
5367 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5373 * The list of packet types we will receive (as opposed to discard)
5374 * and the routines to invoke.
5376 * Why 16. Because with 16 the only overlap we get on a hash of the
5377 * low nibble of the protocol value is RARP/SNAP/X.25.
5391 #define PTYPE_HASH_SIZE (16)
5392 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5394 extern struct list_head ptype_all __read_mostly;
5395 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5397 extern struct net_device *blackhole_netdev;
5399 #endif /* _LINUX_NETDEVICE_H */