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 struct list_head list;
304 * size of gro hash buckets, must less than bit number of
305 * napi_struct::gro_bitmask
307 #define GRO_HASH_BUCKETS 8
310 * Structure for NAPI scheduling similar to tasklet but with weighting
313 /* The poll_list must only be managed by the entity which
314 * changes the state of the NAPI_STATE_SCHED bit. This means
315 * whoever atomically sets that bit can add this napi_struct
316 * to the per-CPU poll_list, and whoever clears that bit
317 * can remove from the list right before clearing the bit.
319 struct list_head poll_list;
323 int defer_hard_irqs_count;
324 unsigned long gro_bitmask;
325 int (*poll)(struct napi_struct *, int);
326 #ifdef CONFIG_NETPOLL
329 struct net_device *dev;
330 struct gro_list gro_hash[GRO_HASH_BUCKETS];
332 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
333 int rx_count; /* length of rx_list */
334 struct hrtimer timer;
335 struct list_head dev_list;
336 struct hlist_node napi_hash_node;
337 unsigned int napi_id;
338 struct task_struct *thread;
342 NAPI_STATE_SCHED, /* Poll is scheduled */
343 NAPI_STATE_MISSED, /* reschedule a napi */
344 NAPI_STATE_DISABLE, /* Disable pending */
345 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
346 NAPI_STATE_LISTED, /* NAPI added to system lists */
347 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
348 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
349 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
350 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
351 NAPI_STATE_SCHED_THREADED, /* Napi is currently scheduled in threaded mode */
355 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
356 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
357 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
358 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
359 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
360 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
361 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
362 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
363 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
364 NAPIF_STATE_SCHED_THREADED = BIT(NAPI_STATE_SCHED_THREADED),
374 typedef enum gro_result gro_result_t;
377 * enum rx_handler_result - Possible return values for rx_handlers.
378 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
380 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
381 * case skb->dev was changed by rx_handler.
382 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
383 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
385 * rx_handlers are functions called from inside __netif_receive_skb(), to do
386 * special processing of the skb, prior to delivery to protocol handlers.
388 * Currently, a net_device can only have a single rx_handler registered. Trying
389 * to register a second rx_handler will return -EBUSY.
391 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
392 * To unregister a rx_handler on a net_device, use
393 * netdev_rx_handler_unregister().
395 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
398 * If the rx_handler consumed the skb in some way, it should return
399 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
400 * the skb to be delivered in some other way.
402 * If the rx_handler changed skb->dev, to divert the skb to another
403 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
404 * new device will be called if it exists.
406 * If the rx_handler decides the skb should be ignored, it should return
407 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
408 * are registered on exact device (ptype->dev == skb->dev).
410 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
411 * delivered, it should return RX_HANDLER_PASS.
413 * A device without a registered rx_handler will behave as if rx_handler
414 * returned RX_HANDLER_PASS.
417 enum rx_handler_result {
423 typedef enum rx_handler_result rx_handler_result_t;
424 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
426 void __napi_schedule(struct napi_struct *n);
427 void __napi_schedule_irqoff(struct napi_struct *n);
429 static inline bool napi_disable_pending(struct napi_struct *n)
431 return test_bit(NAPI_STATE_DISABLE, &n->state);
434 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
436 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
439 bool napi_schedule_prep(struct napi_struct *n);
442 * napi_schedule - schedule NAPI poll
445 * Schedule NAPI poll routine to be called if it is not already
448 static inline void napi_schedule(struct napi_struct *n)
450 if (napi_schedule_prep(n))
455 * napi_schedule_irqoff - schedule NAPI poll
458 * Variant of napi_schedule(), assuming hard irqs are masked.
460 static inline void napi_schedule_irqoff(struct napi_struct *n)
462 if (napi_schedule_prep(n))
463 __napi_schedule_irqoff(n);
466 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
467 static inline bool napi_reschedule(struct napi_struct *napi)
469 if (napi_schedule_prep(napi)) {
470 __napi_schedule(napi);
476 bool napi_complete_done(struct napi_struct *n, int work_done);
478 * napi_complete - NAPI processing complete
481 * Mark NAPI processing as complete.
482 * Consider using napi_complete_done() instead.
483 * Return false if device should avoid rearming interrupts.
485 static inline bool napi_complete(struct napi_struct *n)
487 return napi_complete_done(n, 0);
490 int dev_set_threaded(struct net_device *dev, bool threaded);
493 * napi_disable - prevent NAPI from scheduling
496 * Stop NAPI from being scheduled on this context.
497 * Waits till any outstanding processing completes.
499 void napi_disable(struct napi_struct *n);
501 void napi_enable(struct napi_struct *n);
504 * napi_synchronize - wait until NAPI is not running
507 * Wait until NAPI is done being scheduled on this context.
508 * Waits till any outstanding processing completes but
509 * does not disable future activations.
511 static inline void napi_synchronize(const struct napi_struct *n)
513 if (IS_ENABLED(CONFIG_SMP))
514 while (test_bit(NAPI_STATE_SCHED, &n->state))
521 * napi_if_scheduled_mark_missed - if napi is running, set the
525 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
528 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
530 unsigned long val, new;
533 val = READ_ONCE(n->state);
534 if (val & NAPIF_STATE_DISABLE)
537 if (!(val & NAPIF_STATE_SCHED))
540 new = val | NAPIF_STATE_MISSED;
541 } while (cmpxchg(&n->state, val, new) != val);
546 enum netdev_queue_state_t {
547 __QUEUE_STATE_DRV_XOFF,
548 __QUEUE_STATE_STACK_XOFF,
549 __QUEUE_STATE_FROZEN,
552 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
553 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
554 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
556 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
557 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
559 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
563 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
564 * netif_tx_* functions below are used to manipulate this flag. The
565 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
566 * queue independently. The netif_xmit_*stopped functions below are called
567 * to check if the queue has been stopped by the driver or stack (either
568 * of the XOFF bits are set in the state). Drivers should not need to call
569 * netif_xmit*stopped functions, they should only be using netif_tx_*.
572 struct netdev_queue {
576 struct net_device *dev;
577 struct Qdisc __rcu *qdisc;
578 struct Qdisc *qdisc_sleeping;
582 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
585 unsigned long tx_maxrate;
587 * Number of TX timeouts for this queue
588 * (/sys/class/net/DEV/Q/trans_timeout)
590 unsigned long trans_timeout;
592 /* Subordinate device that the queue has been assigned to */
593 struct net_device *sb_dev;
594 #ifdef CONFIG_XDP_SOCKETS
595 struct xsk_buff_pool *pool;
600 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
603 * Time (in jiffies) of last Tx
605 unsigned long trans_start;
612 } ____cacheline_aligned_in_smp;
614 extern int sysctl_fb_tunnels_only_for_init_net;
615 extern int sysctl_devconf_inherit_init_net;
618 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
619 * == 1 : For initns only
622 static inline bool net_has_fallback_tunnels(const struct net *net)
624 return !IS_ENABLED(CONFIG_SYSCTL) ||
625 !sysctl_fb_tunnels_only_for_init_net ||
626 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
629 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
631 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
638 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
640 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
647 * This structure holds an RPS map which can be of variable length. The
648 * map is an array of CPUs.
655 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
658 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
659 * tail pointer for that CPU's input queue at the time of last enqueue, and
660 * a hardware filter index.
662 struct rps_dev_flow {
665 unsigned int last_qtail;
667 #define RPS_NO_FILTER 0xffff
670 * The rps_dev_flow_table structure contains a table of flow mappings.
672 struct rps_dev_flow_table {
675 struct rps_dev_flow flows[];
677 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
678 ((_num) * sizeof(struct rps_dev_flow)))
681 * The rps_sock_flow_table contains mappings of flows to the last CPU
682 * on which they were processed by the application (set in recvmsg).
683 * Each entry is a 32bit value. Upper part is the high-order bits
684 * of flow hash, lower part is CPU number.
685 * rps_cpu_mask is used to partition the space, depending on number of
686 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
687 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
688 * meaning we use 32-6=26 bits for the hash.
690 struct rps_sock_flow_table {
693 u32 ents[] ____cacheline_aligned_in_smp;
695 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
697 #define RPS_NO_CPU 0xffff
699 extern u32 rps_cpu_mask;
700 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
702 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
706 unsigned int index = hash & table->mask;
707 u32 val = hash & ~rps_cpu_mask;
709 /* We only give a hint, preemption can change CPU under us */
710 val |= raw_smp_processor_id();
712 if (table->ents[index] != val)
713 table->ents[index] = val;
717 #ifdef CONFIG_RFS_ACCEL
718 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
721 #endif /* CONFIG_RPS */
723 /* This structure contains an instance of an RX queue. */
724 struct netdev_rx_queue {
726 struct rps_map __rcu *rps_map;
727 struct rps_dev_flow_table __rcu *rps_flow_table;
730 struct net_device *dev;
731 struct xdp_rxq_info xdp_rxq;
732 #ifdef CONFIG_XDP_SOCKETS
733 struct xsk_buff_pool *pool;
735 } ____cacheline_aligned_in_smp;
738 * RX queue sysfs structures and functions.
740 struct rx_queue_attribute {
741 struct attribute attr;
742 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
743 ssize_t (*store)(struct netdev_rx_queue *queue,
744 const char *buf, size_t len);
747 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
756 * This structure holds an XPS map which can be of variable length. The
757 * map is an array of queues.
761 unsigned int alloc_len;
765 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
766 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
767 - sizeof(struct xps_map)) / sizeof(u16))
770 * This structure holds all XPS maps for device. Maps are indexed by CPU.
772 * We keep track of the number of cpus/rxqs used when the struct is allocated,
773 * in nr_ids. This will help not accessing out-of-bound memory.
775 * We keep track of the number of traffic classes used when the struct is
776 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
777 * not crossing its upper bound, as the original dev->num_tc can be updated in
780 struct xps_dev_maps {
784 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
787 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
788 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
790 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
791 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
793 #endif /* CONFIG_XPS */
795 #define TC_MAX_QUEUE 16
796 #define TC_BITMASK 15
797 /* HW offloaded queuing disciplines txq count and offset maps */
798 struct netdev_tc_txq {
803 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
805 * This structure is to hold information about the device
806 * configured to run FCoE protocol stack.
808 struct netdev_fcoe_hbainfo {
809 char manufacturer[64];
810 char serial_number[64];
811 char hardware_version[64];
812 char driver_version[64];
813 char optionrom_version[64];
814 char firmware_version[64];
816 char model_description[256];
820 #define MAX_PHYS_ITEM_ID_LEN 32
822 /* This structure holds a unique identifier to identify some
823 * physical item (port for example) used by a netdevice.
825 struct netdev_phys_item_id {
826 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
827 unsigned char id_len;
830 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
831 struct netdev_phys_item_id *b)
833 return a->id_len == b->id_len &&
834 memcmp(a->id, b->id, a->id_len) == 0;
837 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
839 struct net_device *sb_dev);
841 enum net_device_path_type {
842 DEV_PATH_ETHERNET = 0,
849 struct net_device_path {
850 enum net_device_path_type type;
851 const struct net_device *dev;
860 DEV_PATH_BR_VLAN_KEEP,
861 DEV_PATH_BR_VLAN_TAG,
862 DEV_PATH_BR_VLAN_UNTAG,
863 DEV_PATH_BR_VLAN_UNTAG_HW,
875 #define NET_DEVICE_PATH_STACK_MAX 5
876 #define NET_DEVICE_PATH_VLAN_MAX 2
878 struct net_device_path_stack {
880 struct net_device_path path[NET_DEVICE_PATH_STACK_MAX];
883 struct net_device_path_ctx {
884 const struct net_device *dev;
891 } vlan[NET_DEVICE_PATH_VLAN_MAX];
895 TC_SETUP_QDISC_MQPRIO,
898 TC_SETUP_CLSMATCHALL,
908 TC_SETUP_QDISC_TAPRIO,
916 /* These structures hold the attributes of bpf state that are being passed
917 * to the netdevice through the bpf op.
919 enum bpf_netdev_command {
920 /* Set or clear a bpf program used in the earliest stages of packet
921 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
922 * is responsible for calling bpf_prog_put on any old progs that are
923 * stored. In case of error, the callee need not release the new prog
924 * reference, but on success it takes ownership and must bpf_prog_put
925 * when it is no longer used.
929 /* BPF program for offload callbacks, invoked at program load time. */
930 BPF_OFFLOAD_MAP_ALLOC,
931 BPF_OFFLOAD_MAP_FREE,
935 struct bpf_prog_offload_ops;
936 struct netlink_ext_ack;
938 struct xdp_dev_bulk_queue;
948 struct bpf_xdp_entity {
949 struct bpf_prog *prog;
950 struct bpf_xdp_link *link;
954 enum bpf_netdev_command command;
959 struct bpf_prog *prog;
960 struct netlink_ext_ack *extack;
962 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
964 struct bpf_offloaded_map *offmap;
966 /* XDP_SETUP_XSK_POOL */
968 struct xsk_buff_pool *pool;
974 /* Flags for ndo_xsk_wakeup. */
975 #define XDP_WAKEUP_RX (1 << 0)
976 #define XDP_WAKEUP_TX (1 << 1)
978 #ifdef CONFIG_XFRM_OFFLOAD
980 int (*xdo_dev_state_add) (struct xfrm_state *x);
981 void (*xdo_dev_state_delete) (struct xfrm_state *x);
982 void (*xdo_dev_state_free) (struct xfrm_state *x);
983 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
984 struct xfrm_state *x);
985 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
990 struct rcu_head rcuhead;
997 struct netdev_name_node {
998 struct hlist_node hlist;
999 struct list_head list;
1000 struct net_device *dev;
1004 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
1005 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
1007 struct netdev_net_notifier {
1008 struct list_head list;
1009 struct notifier_block *nb;
1013 * This structure defines the management hooks for network devices.
1014 * The following hooks can be defined; unless noted otherwise, they are
1015 * optional and can be filled with a null pointer.
1017 * int (*ndo_init)(struct net_device *dev);
1018 * This function is called once when a network device is registered.
1019 * The network device can use this for any late stage initialization
1020 * or semantic validation. It can fail with an error code which will
1021 * be propagated back to register_netdev.
1023 * void (*ndo_uninit)(struct net_device *dev);
1024 * This function is called when device is unregistered or when registration
1025 * fails. It is not called if init fails.
1027 * int (*ndo_open)(struct net_device *dev);
1028 * This function is called when a network device transitions to the up
1031 * int (*ndo_stop)(struct net_device *dev);
1032 * This function is called when a network device transitions to the down
1035 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1036 * struct net_device *dev);
1037 * Called when a packet needs to be transmitted.
1038 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
1039 * the queue before that can happen; it's for obsolete devices and weird
1040 * corner cases, but the stack really does a non-trivial amount
1041 * of useless work if you return NETDEV_TX_BUSY.
1042 * Required; cannot be NULL.
1044 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1045 * struct net_device *dev
1046 * netdev_features_t features);
1047 * Called by core transmit path to determine if device is capable of
1048 * performing offload operations on a given packet. This is to give
1049 * the device an opportunity to implement any restrictions that cannot
1050 * be otherwise expressed by feature flags. The check is called with
1051 * the set of features that the stack has calculated and it returns
1052 * those the driver believes to be appropriate.
1054 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1055 * struct net_device *sb_dev);
1056 * Called to decide which queue to use when device supports multiple
1059 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1060 * This function is called to allow device receiver to make
1061 * changes to configuration when multicast or promiscuous is enabled.
1063 * void (*ndo_set_rx_mode)(struct net_device *dev);
1064 * This function is called device changes address list filtering.
1065 * If driver handles unicast address filtering, it should set
1066 * IFF_UNICAST_FLT in its priv_flags.
1068 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1069 * This function is called when the Media Access Control address
1070 * needs to be changed. If this interface is not defined, the
1071 * MAC address can not be changed.
1073 * int (*ndo_validate_addr)(struct net_device *dev);
1074 * Test if Media Access Control address is valid for the device.
1076 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1077 * Old-style ioctl entry point. This is used internally by the
1078 * appletalk and ieee802154 subsystems but is no longer called by
1079 * the device ioctl handler.
1081 * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1082 * Used by the bonding driver for its device specific ioctls:
1083 * SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1084 * SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1086 * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1087 * Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1088 * SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1090 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1091 * Used to set network devices bus interface parameters. This interface
1092 * is retained for legacy reasons; new devices should use the bus
1093 * interface (PCI) for low level management.
1095 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1096 * Called when a user wants to change the Maximum Transfer Unit
1099 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1100 * Callback used when the transmitter has not made any progress
1101 * for dev->watchdog ticks.
1103 * void (*ndo_get_stats64)(struct net_device *dev,
1104 * struct rtnl_link_stats64 *storage);
1105 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1106 * Called when a user wants to get the network device usage
1107 * statistics. Drivers must do one of the following:
1108 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1109 * rtnl_link_stats64 structure passed by the caller.
1110 * 2. Define @ndo_get_stats to update a net_device_stats structure
1111 * (which should normally be dev->stats) and return a pointer to
1112 * it. The structure may be changed asynchronously only if each
1113 * field is written atomically.
1114 * 3. Update dev->stats asynchronously and atomically, and define
1115 * neither operation.
1117 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1118 * Return true if this device supports offload stats of this attr_id.
1120 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1122 * Get statistics for offload operations by attr_id. Write it into the
1123 * attr_data pointer.
1125 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1126 * If device supports VLAN filtering this function is called when a
1127 * VLAN id is registered.
1129 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1130 * If device supports VLAN filtering this function is called when a
1131 * VLAN id is unregistered.
1133 * void (*ndo_poll_controller)(struct net_device *dev);
1135 * SR-IOV management functions.
1136 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1137 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1138 * u8 qos, __be16 proto);
1139 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1141 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1142 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1143 * int (*ndo_get_vf_config)(struct net_device *dev,
1144 * int vf, struct ifla_vf_info *ivf);
1145 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1146 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1147 * struct nlattr *port[]);
1149 * Enable or disable the VF ability to query its RSS Redirection Table and
1150 * Hash Key. This is needed since on some devices VF share this information
1151 * with PF and querying it may introduce a theoretical security risk.
1152 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1153 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1154 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1156 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1157 * This is always called from the stack with the rtnl lock held and netif
1158 * tx queues stopped. This allows the netdevice to perform queue
1159 * management safely.
1161 * Fiber Channel over Ethernet (FCoE) offload functions.
1162 * int (*ndo_fcoe_enable)(struct net_device *dev);
1163 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1164 * so the underlying device can perform whatever needed configuration or
1165 * initialization to support acceleration of FCoE traffic.
1167 * int (*ndo_fcoe_disable)(struct net_device *dev);
1168 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1169 * so the underlying device can perform whatever needed clean-ups to
1170 * stop supporting acceleration of FCoE traffic.
1172 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1173 * struct scatterlist *sgl, unsigned int sgc);
1174 * Called when the FCoE Initiator wants to initialize an I/O that
1175 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1176 * perform necessary setup and returns 1 to indicate the device is set up
1177 * successfully to perform DDP on this I/O, otherwise this returns 0.
1179 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1180 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1181 * indicated by the FC exchange id 'xid', so the underlying device can
1182 * clean up and reuse resources for later DDP requests.
1184 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1185 * struct scatterlist *sgl, unsigned int sgc);
1186 * Called when the FCoE Target wants to initialize an I/O that
1187 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1188 * perform necessary setup and returns 1 to indicate the device is set up
1189 * successfully to perform DDP on this I/O, otherwise this returns 0.
1191 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1192 * struct netdev_fcoe_hbainfo *hbainfo);
1193 * Called when the FCoE Protocol stack wants information on the underlying
1194 * device. This information is utilized by the FCoE protocol stack to
1195 * register attributes with Fiber Channel management service as per the
1196 * FC-GS Fabric Device Management Information(FDMI) specification.
1198 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1199 * Called when the underlying device wants to override default World Wide
1200 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1201 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1202 * protocol stack to use.
1205 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1206 * u16 rxq_index, u32 flow_id);
1207 * Set hardware filter for RFS. rxq_index is the target queue index;
1208 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1209 * Return the filter ID on success, or a negative error code.
1211 * Slave management functions (for bridge, bonding, etc).
1212 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1213 * Called to make another netdev an underling.
1215 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1216 * Called to release previously enslaved netdev.
1218 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1219 * struct sk_buff *skb,
1221 * Get the xmit slave of master device. If all_slaves is true, function
1222 * assume all the slaves can transmit.
1224 * Feature/offload setting functions.
1225 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1226 * netdev_features_t features);
1227 * Adjusts the requested feature flags according to device-specific
1228 * constraints, and returns the resulting flags. Must not modify
1231 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1232 * Called to update device configuration to new features. Passed
1233 * feature set might be less than what was returned by ndo_fix_features()).
1234 * Must return >0 or -errno if it changed dev->features itself.
1236 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1237 * struct net_device *dev,
1238 * const unsigned char *addr, u16 vid, u16 flags,
1239 * struct netlink_ext_ack *extack);
1240 * Adds an FDB entry to dev for addr.
1241 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1242 * struct net_device *dev,
1243 * const unsigned char *addr, u16 vid)
1244 * Deletes the FDB entry from dev coresponding to addr.
1245 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1246 * struct net_device *dev, struct net_device *filter_dev,
1248 * Used to add FDB entries to dump requests. Implementers should add
1249 * entries to skb and update idx with the number of entries.
1251 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1252 * u16 flags, struct netlink_ext_ack *extack)
1253 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1254 * struct net_device *dev, u32 filter_mask,
1256 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1259 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1260 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1261 * which do not represent real hardware may define this to allow their
1262 * userspace components to manage their virtual carrier state. Devices
1263 * that determine carrier state from physical hardware properties (eg
1264 * network cables) or protocol-dependent mechanisms (eg
1265 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1267 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1268 * struct netdev_phys_item_id *ppid);
1269 * Called to get ID of physical port of this device. If driver does
1270 * not implement this, it is assumed that the hw is not able to have
1271 * multiple net devices on single physical port.
1273 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1274 * struct netdev_phys_item_id *ppid)
1275 * Called to get the parent ID of the physical port of this device.
1277 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1278 * struct net_device *dev)
1279 * Called by upper layer devices to accelerate switching or other
1280 * station functionality into hardware. 'pdev is the lowerdev
1281 * to use for the offload and 'dev' is the net device that will
1282 * back the offload. Returns a pointer to the private structure
1283 * the upper layer will maintain.
1284 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1285 * Called by upper layer device to delete the station created
1286 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1287 * the station and priv is the structure returned by the add
1289 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1290 * int queue_index, u32 maxrate);
1291 * Called when a user wants to set a max-rate limitation of specific
1293 * int (*ndo_get_iflink)(const struct net_device *dev);
1294 * Called to get the iflink value of this device.
1295 * void (*ndo_change_proto_down)(struct net_device *dev,
1297 * This function is used to pass protocol port error state information
1298 * to the switch driver. The switch driver can react to the proto_down
1299 * by doing a phys down on the associated switch port.
1300 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1301 * This function is used to get egress tunnel information for given skb.
1302 * This is useful for retrieving outer tunnel header parameters while
1304 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1305 * This function is used to specify the headroom that the skb must
1306 * consider when allocation skb during packet reception. Setting
1307 * appropriate rx headroom value allows avoiding skb head copy on
1308 * forward. Setting a negative value resets the rx headroom to the
1310 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1311 * This function is used to set or query state related to XDP on the
1312 * netdevice and manage BPF offload. See definition of
1313 * enum bpf_netdev_command for details.
1314 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1316 * This function is used to submit @n XDP packets for transmit on a
1317 * netdevice. Returns number of frames successfully transmitted, frames
1318 * that got dropped are freed/returned via xdp_return_frame().
1319 * Returns negative number, means general error invoking ndo, meaning
1320 * no frames were xmit'ed and core-caller will free all frames.
1321 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1322 * This function is used to wake up the softirq, ksoftirqd or kthread
1323 * responsible for sending and/or receiving packets on a specific
1324 * queue id bound to an AF_XDP socket. The flags field specifies if
1325 * only RX, only Tx, or both should be woken up using the flags
1326 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1327 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1328 * Get devlink port instance associated with a given netdev.
1329 * Called with a reference on the netdevice and devlink locks only,
1330 * rtnl_lock is not held.
1331 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1333 * Add, change, delete or get information on an IPv4 tunnel.
1334 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1335 * If a device is paired with a peer device, return the peer instance.
1336 * The caller must be under RCU read context.
1337 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1338 * Get the forwarding path to reach the real device from the HW destination address
1340 struct net_device_ops {
1341 int (*ndo_init)(struct net_device *dev);
1342 void (*ndo_uninit)(struct net_device *dev);
1343 int (*ndo_open)(struct net_device *dev);
1344 int (*ndo_stop)(struct net_device *dev);
1345 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1346 struct net_device *dev);
1347 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1348 struct net_device *dev,
1349 netdev_features_t features);
1350 u16 (*ndo_select_queue)(struct net_device *dev,
1351 struct sk_buff *skb,
1352 struct net_device *sb_dev);
1353 void (*ndo_change_rx_flags)(struct net_device *dev,
1355 void (*ndo_set_rx_mode)(struct net_device *dev);
1356 int (*ndo_set_mac_address)(struct net_device *dev,
1358 int (*ndo_validate_addr)(struct net_device *dev);
1359 int (*ndo_do_ioctl)(struct net_device *dev,
1360 struct ifreq *ifr, int cmd);
1361 int (*ndo_eth_ioctl)(struct net_device *dev,
1362 struct ifreq *ifr, int cmd);
1363 int (*ndo_siocbond)(struct net_device *dev,
1364 struct ifreq *ifr, int cmd);
1365 int (*ndo_siocwandev)(struct net_device *dev,
1366 struct if_settings *ifs);
1367 int (*ndo_siocdevprivate)(struct net_device *dev,
1369 void __user *data, int cmd);
1370 int (*ndo_set_config)(struct net_device *dev,
1372 int (*ndo_change_mtu)(struct net_device *dev,
1374 int (*ndo_neigh_setup)(struct net_device *dev,
1375 struct neigh_parms *);
1376 void (*ndo_tx_timeout) (struct net_device *dev,
1377 unsigned int txqueue);
1379 void (*ndo_get_stats64)(struct net_device *dev,
1380 struct rtnl_link_stats64 *storage);
1381 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1382 int (*ndo_get_offload_stats)(int attr_id,
1383 const struct net_device *dev,
1385 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1387 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1388 __be16 proto, u16 vid);
1389 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1390 __be16 proto, u16 vid);
1391 #ifdef CONFIG_NET_POLL_CONTROLLER
1392 void (*ndo_poll_controller)(struct net_device *dev);
1393 int (*ndo_netpoll_setup)(struct net_device *dev,
1394 struct netpoll_info *info);
1395 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1397 int (*ndo_set_vf_mac)(struct net_device *dev,
1398 int queue, u8 *mac);
1399 int (*ndo_set_vf_vlan)(struct net_device *dev,
1400 int queue, u16 vlan,
1401 u8 qos, __be16 proto);
1402 int (*ndo_set_vf_rate)(struct net_device *dev,
1403 int vf, int min_tx_rate,
1405 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1406 int vf, bool setting);
1407 int (*ndo_set_vf_trust)(struct net_device *dev,
1408 int vf, bool setting);
1409 int (*ndo_get_vf_config)(struct net_device *dev,
1411 struct ifla_vf_info *ivf);
1412 int (*ndo_set_vf_link_state)(struct net_device *dev,
1413 int vf, int link_state);
1414 int (*ndo_get_vf_stats)(struct net_device *dev,
1416 struct ifla_vf_stats
1418 int (*ndo_set_vf_port)(struct net_device *dev,
1420 struct nlattr *port[]);
1421 int (*ndo_get_vf_port)(struct net_device *dev,
1422 int vf, struct sk_buff *skb);
1423 int (*ndo_get_vf_guid)(struct net_device *dev,
1425 struct ifla_vf_guid *node_guid,
1426 struct ifla_vf_guid *port_guid);
1427 int (*ndo_set_vf_guid)(struct net_device *dev,
1430 int (*ndo_set_vf_rss_query_en)(
1431 struct net_device *dev,
1432 int vf, bool setting);
1433 int (*ndo_setup_tc)(struct net_device *dev,
1434 enum tc_setup_type type,
1436 #if IS_ENABLED(CONFIG_FCOE)
1437 int (*ndo_fcoe_enable)(struct net_device *dev);
1438 int (*ndo_fcoe_disable)(struct net_device *dev);
1439 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1441 struct scatterlist *sgl,
1443 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1445 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1447 struct scatterlist *sgl,
1449 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1450 struct netdev_fcoe_hbainfo *hbainfo);
1453 #if IS_ENABLED(CONFIG_LIBFCOE)
1454 #define NETDEV_FCOE_WWNN 0
1455 #define NETDEV_FCOE_WWPN 1
1456 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1457 u64 *wwn, int type);
1460 #ifdef CONFIG_RFS_ACCEL
1461 int (*ndo_rx_flow_steer)(struct net_device *dev,
1462 const struct sk_buff *skb,
1466 int (*ndo_add_slave)(struct net_device *dev,
1467 struct net_device *slave_dev,
1468 struct netlink_ext_ack *extack);
1469 int (*ndo_del_slave)(struct net_device *dev,
1470 struct net_device *slave_dev);
1471 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1472 struct sk_buff *skb,
1474 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1476 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1477 netdev_features_t features);
1478 int (*ndo_set_features)(struct net_device *dev,
1479 netdev_features_t features);
1480 int (*ndo_neigh_construct)(struct net_device *dev,
1481 struct neighbour *n);
1482 void (*ndo_neigh_destroy)(struct net_device *dev,
1483 struct neighbour *n);
1485 int (*ndo_fdb_add)(struct ndmsg *ndm,
1486 struct nlattr *tb[],
1487 struct net_device *dev,
1488 const unsigned char *addr,
1491 struct netlink_ext_ack *extack);
1492 int (*ndo_fdb_del)(struct ndmsg *ndm,
1493 struct nlattr *tb[],
1494 struct net_device *dev,
1495 const unsigned char *addr,
1497 int (*ndo_fdb_dump)(struct sk_buff *skb,
1498 struct netlink_callback *cb,
1499 struct net_device *dev,
1500 struct net_device *filter_dev,
1502 int (*ndo_fdb_get)(struct sk_buff *skb,
1503 struct nlattr *tb[],
1504 struct net_device *dev,
1505 const unsigned char *addr,
1506 u16 vid, u32 portid, u32 seq,
1507 struct netlink_ext_ack *extack);
1508 int (*ndo_bridge_setlink)(struct net_device *dev,
1509 struct nlmsghdr *nlh,
1511 struct netlink_ext_ack *extack);
1512 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1514 struct net_device *dev,
1517 int (*ndo_bridge_dellink)(struct net_device *dev,
1518 struct nlmsghdr *nlh,
1520 int (*ndo_change_carrier)(struct net_device *dev,
1522 int (*ndo_get_phys_port_id)(struct net_device *dev,
1523 struct netdev_phys_item_id *ppid);
1524 int (*ndo_get_port_parent_id)(struct net_device *dev,
1525 struct netdev_phys_item_id *ppid);
1526 int (*ndo_get_phys_port_name)(struct net_device *dev,
1527 char *name, size_t len);
1528 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1529 struct net_device *dev);
1530 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1533 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1536 int (*ndo_get_iflink)(const struct net_device *dev);
1537 int (*ndo_change_proto_down)(struct net_device *dev,
1539 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1540 struct sk_buff *skb);
1541 void (*ndo_set_rx_headroom)(struct net_device *dev,
1542 int needed_headroom);
1543 int (*ndo_bpf)(struct net_device *dev,
1544 struct netdev_bpf *bpf);
1545 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1546 struct xdp_frame **xdp,
1548 int (*ndo_xsk_wakeup)(struct net_device *dev,
1549 u32 queue_id, u32 flags);
1550 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1551 int (*ndo_tunnel_ctl)(struct net_device *dev,
1552 struct ip_tunnel_parm *p, int cmd);
1553 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1554 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1555 struct net_device_path *path);
1559 * enum netdev_priv_flags - &struct net_device priv_flags
1561 * These are the &struct net_device, they are only set internally
1562 * by drivers and used in the kernel. These flags are invisible to
1563 * userspace; this means that the order of these flags can change
1564 * during any kernel release.
1566 * You should have a pretty good reason to be extending these flags.
1568 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1569 * @IFF_EBRIDGE: Ethernet bridging device
1570 * @IFF_BONDING: bonding master or slave
1571 * @IFF_ISATAP: ISATAP interface (RFC4214)
1572 * @IFF_WAN_HDLC: WAN HDLC device
1573 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1575 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1576 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1577 * @IFF_MACVLAN_PORT: device used as macvlan port
1578 * @IFF_BRIDGE_PORT: device used as bridge port
1579 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1580 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1581 * @IFF_UNICAST_FLT: Supports unicast filtering
1582 * @IFF_TEAM_PORT: device used as team port
1583 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1584 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1585 * change when it's running
1586 * @IFF_MACVLAN: Macvlan device
1587 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1588 * underlying stacked devices
1589 * @IFF_L3MDEV_MASTER: device is an L3 master device
1590 * @IFF_NO_QUEUE: device can run without qdisc attached
1591 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1592 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1593 * @IFF_TEAM: device is a team device
1594 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1595 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1596 * entity (i.e. the master device for bridged veth)
1597 * @IFF_MACSEC: device is a MACsec device
1598 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1599 * @IFF_FAILOVER: device is a failover master device
1600 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1601 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1602 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1603 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1604 * skb_headlen(skb) == 0 (data starts from frag0)
1606 enum netdev_priv_flags {
1607 IFF_802_1Q_VLAN = 1<<0,
1611 IFF_WAN_HDLC = 1<<4,
1612 IFF_XMIT_DST_RELEASE = 1<<5,
1613 IFF_DONT_BRIDGE = 1<<6,
1614 IFF_DISABLE_NETPOLL = 1<<7,
1615 IFF_MACVLAN_PORT = 1<<8,
1616 IFF_BRIDGE_PORT = 1<<9,
1617 IFF_OVS_DATAPATH = 1<<10,
1618 IFF_TX_SKB_SHARING = 1<<11,
1619 IFF_UNICAST_FLT = 1<<12,
1620 IFF_TEAM_PORT = 1<<13,
1621 IFF_SUPP_NOFCS = 1<<14,
1622 IFF_LIVE_ADDR_CHANGE = 1<<15,
1623 IFF_MACVLAN = 1<<16,
1624 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1625 IFF_L3MDEV_MASTER = 1<<18,
1626 IFF_NO_QUEUE = 1<<19,
1627 IFF_OPENVSWITCH = 1<<20,
1628 IFF_L3MDEV_SLAVE = 1<<21,
1630 IFF_RXFH_CONFIGURED = 1<<23,
1631 IFF_PHONY_HEADROOM = 1<<24,
1633 IFF_NO_RX_HANDLER = 1<<26,
1634 IFF_FAILOVER = 1<<27,
1635 IFF_FAILOVER_SLAVE = 1<<28,
1636 IFF_L3MDEV_RX_HANDLER = 1<<29,
1637 IFF_LIVE_RENAME_OK = 1<<30,
1638 IFF_TX_SKB_NO_LINEAR = 1<<31,
1641 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1642 #define IFF_EBRIDGE IFF_EBRIDGE
1643 #define IFF_BONDING IFF_BONDING
1644 #define IFF_ISATAP IFF_ISATAP
1645 #define IFF_WAN_HDLC IFF_WAN_HDLC
1646 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1647 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1648 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1649 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1650 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1651 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1652 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1653 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1654 #define IFF_TEAM_PORT IFF_TEAM_PORT
1655 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1656 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1657 #define IFF_MACVLAN IFF_MACVLAN
1658 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1659 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1660 #define IFF_NO_QUEUE IFF_NO_QUEUE
1661 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1662 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1663 #define IFF_TEAM IFF_TEAM
1664 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1665 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1666 #define IFF_MACSEC IFF_MACSEC
1667 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1668 #define IFF_FAILOVER IFF_FAILOVER
1669 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1670 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1671 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1672 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1674 /* Specifies the type of the struct net_device::ml_priv pointer */
1675 enum netdev_ml_priv_type {
1681 * struct net_device - The DEVICE structure.
1683 * Actually, this whole structure is a big mistake. It mixes I/O
1684 * data with strictly "high-level" data, and it has to know about
1685 * almost every data structure used in the INET module.
1687 * @name: This is the first field of the "visible" part of this structure
1688 * (i.e. as seen by users in the "Space.c" file). It is the name
1691 * @name_node: Name hashlist node
1692 * @ifalias: SNMP alias
1693 * @mem_end: Shared memory end
1694 * @mem_start: Shared memory start
1695 * @base_addr: Device I/O address
1696 * @irq: Device IRQ number
1698 * @state: Generic network queuing layer state, see netdev_state_t
1699 * @dev_list: The global list of network devices
1700 * @napi_list: List entry used for polling NAPI devices
1701 * @unreg_list: List entry when we are unregistering the
1702 * device; see the function unregister_netdev
1703 * @close_list: List entry used when we are closing the device
1704 * @ptype_all: Device-specific packet handlers for all protocols
1705 * @ptype_specific: Device-specific, protocol-specific packet handlers
1707 * @adj_list: Directly linked devices, like slaves for bonding
1708 * @features: Currently active device features
1709 * @hw_features: User-changeable features
1711 * @wanted_features: User-requested features
1712 * @vlan_features: Mask of features inheritable by VLAN devices
1714 * @hw_enc_features: Mask of features inherited by encapsulating devices
1715 * This field indicates what encapsulation
1716 * offloads the hardware is capable of doing,
1717 * and drivers will need to set them appropriately.
1719 * @mpls_features: Mask of features inheritable by MPLS
1720 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1722 * @ifindex: interface index
1723 * @group: The group the device belongs to
1725 * @stats: Statistics struct, which was left as a legacy, use
1726 * rtnl_link_stats64 instead
1728 * @rx_dropped: Dropped packets by core network,
1729 * do not use this in drivers
1730 * @tx_dropped: Dropped packets by core network,
1731 * do not use this in drivers
1732 * @rx_nohandler: nohandler dropped packets by core network on
1733 * inactive devices, do not use this in drivers
1734 * @carrier_up_count: Number of times the carrier has been up
1735 * @carrier_down_count: Number of times the carrier has been down
1737 * @wireless_handlers: List of functions to handle Wireless Extensions,
1739 * see <net/iw_handler.h> for details.
1740 * @wireless_data: Instance data managed by the core of wireless extensions
1742 * @netdev_ops: Includes several pointers to callbacks,
1743 * if one wants to override the ndo_*() functions
1744 * @ethtool_ops: Management operations
1745 * @l3mdev_ops: Layer 3 master device operations
1746 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1747 * discovery handling. Necessary for e.g. 6LoWPAN.
1748 * @xfrmdev_ops: Transformation offload operations
1749 * @tlsdev_ops: Transport Layer Security offload operations
1750 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1751 * of Layer 2 headers.
1753 * @flags: Interface flags (a la BSD)
1754 * @priv_flags: Like 'flags' but invisible to userspace,
1755 * see if.h for the definitions
1756 * @gflags: Global flags ( kept as legacy )
1757 * @padded: How much padding added by alloc_netdev()
1758 * @operstate: RFC2863 operstate
1759 * @link_mode: Mapping policy to operstate
1760 * @if_port: Selectable AUI, TP, ...
1762 * @mtu: Interface MTU value
1763 * @min_mtu: Interface Minimum MTU value
1764 * @max_mtu: Interface Maximum MTU value
1765 * @type: Interface hardware type
1766 * @hard_header_len: Maximum hardware header length.
1767 * @min_header_len: Minimum hardware header length
1769 * @needed_headroom: Extra headroom the hardware may need, but not in all
1770 * cases can this be guaranteed
1771 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1772 * cases can this be guaranteed. Some cases also use
1773 * LL_MAX_HEADER instead to allocate the skb
1775 * interface address info:
1777 * @perm_addr: Permanent hw address
1778 * @addr_assign_type: Hw address assignment type
1779 * @addr_len: Hardware address length
1780 * @upper_level: Maximum depth level of upper devices.
1781 * @lower_level: Maximum depth level of lower devices.
1782 * @neigh_priv_len: Used in neigh_alloc()
1783 * @dev_id: Used to differentiate devices that share
1784 * the same link layer address
1785 * @dev_port: Used to differentiate devices that share
1787 * @addr_list_lock: XXX: need comments on this one
1788 * @name_assign_type: network interface name assignment type
1789 * @uc_promisc: Counter that indicates promiscuous mode
1790 * has been enabled due to the need to listen to
1791 * additional unicast addresses in a device that
1792 * does not implement ndo_set_rx_mode()
1793 * @uc: unicast mac addresses
1794 * @mc: multicast mac addresses
1795 * @dev_addrs: list of device hw addresses
1796 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1797 * @promiscuity: Number of times the NIC is told to work in
1798 * promiscuous mode; if it becomes 0 the NIC will
1799 * exit promiscuous mode
1800 * @allmulti: Counter, enables or disables allmulticast mode
1802 * @vlan_info: VLAN info
1803 * @dsa_ptr: dsa specific data
1804 * @tipc_ptr: TIPC specific data
1805 * @atalk_ptr: AppleTalk link
1806 * @ip_ptr: IPv4 specific data
1807 * @dn_ptr: DECnet specific data
1808 * @ip6_ptr: IPv6 specific data
1809 * @ax25_ptr: AX.25 specific data
1810 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1811 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1813 * @mpls_ptr: mpls_dev struct pointer
1814 * @mctp_ptr: MCTP specific data
1816 * @dev_addr: Hw address (before bcast,
1817 * because most packets are unicast)
1819 * @_rx: Array of RX queues
1820 * @num_rx_queues: Number of RX queues
1821 * allocated at register_netdev() time
1822 * @real_num_rx_queues: Number of RX queues currently active in device
1823 * @xdp_prog: XDP sockets filter program pointer
1824 * @gro_flush_timeout: timeout for GRO layer in NAPI
1825 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1826 * allow to avoid NIC hard IRQ, on busy queues.
1828 * @rx_handler: handler for received packets
1829 * @rx_handler_data: XXX: need comments on this one
1830 * @miniq_ingress: ingress/clsact qdisc specific data for
1831 * ingress processing
1832 * @ingress_queue: XXX: need comments on this one
1833 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1834 * @broadcast: hw bcast address
1836 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1837 * indexed by RX queue number. Assigned by driver.
1838 * This must only be set if the ndo_rx_flow_steer
1839 * operation is defined
1840 * @index_hlist: Device index hash chain
1842 * @_tx: Array of TX queues
1843 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1844 * @real_num_tx_queues: Number of TX queues currently active in device
1845 * @qdisc: Root qdisc from userspace point of view
1846 * @tx_queue_len: Max frames per queue allowed
1847 * @tx_global_lock: XXX: need comments on this one
1848 * @xdp_bulkq: XDP device bulk queue
1849 * @xps_maps: all CPUs/RXQs maps for XPS device
1851 * @xps_maps: XXX: need comments on this one
1852 * @miniq_egress: clsact qdisc specific data for
1854 * @qdisc_hash: qdisc hash table
1855 * @watchdog_timeo: Represents the timeout that is used by
1856 * the watchdog (see dev_watchdog())
1857 * @watchdog_timer: List of timers
1859 * @proto_down_reason: reason a netdev interface is held down
1860 * @pcpu_refcnt: Number of references to this device
1861 * @dev_refcnt: Number of references to this device
1862 * @todo_list: Delayed register/unregister
1863 * @link_watch_list: XXX: need comments on this one
1865 * @reg_state: Register/unregister state machine
1866 * @dismantle: Device is going to be freed
1867 * @rtnl_link_state: This enum represents the phases of creating
1870 * @needs_free_netdev: Should unregister perform free_netdev?
1871 * @priv_destructor: Called from unregister
1872 * @npinfo: XXX: need comments on this one
1873 * @nd_net: Network namespace this network device is inside
1875 * @ml_priv: Mid-layer private
1876 * @ml_priv_type: Mid-layer private type
1877 * @lstats: Loopback statistics
1878 * @tstats: Tunnel statistics
1879 * @dstats: Dummy statistics
1880 * @vstats: Virtual ethernet statistics
1885 * @dev: Class/net/name entry
1886 * @sysfs_groups: Space for optional device, statistics and wireless
1889 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1890 * @rtnl_link_ops: Rtnl_link_ops
1892 * @gso_max_size: Maximum size of generic segmentation offload
1893 * @gso_max_segs: Maximum number of segments that can be passed to the
1896 * @dcbnl_ops: Data Center Bridging netlink ops
1897 * @num_tc: Number of traffic classes in the net device
1898 * @tc_to_txq: XXX: need comments on this one
1899 * @prio_tc_map: XXX: need comments on this one
1901 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1903 * @priomap: XXX: need comments on this one
1904 * @phydev: Physical device may attach itself
1905 * for hardware timestamping
1906 * @sfp_bus: attached &struct sfp_bus structure.
1908 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1909 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1911 * @proto_down: protocol port state information can be sent to the
1912 * switch driver and used to set the phys state of the
1915 * @wol_enabled: Wake-on-LAN is enabled
1917 * @threaded: napi threaded mode is enabled
1919 * @net_notifier_list: List of per-net netdev notifier block
1920 * that follow this device when it is moved
1921 * to another network namespace.
1923 * @macsec_ops: MACsec offloading ops
1925 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1926 * offload capabilities of the device
1927 * @udp_tunnel_nic: UDP tunnel offload state
1928 * @xdp_state: stores info on attached XDP BPF programs
1930 * @nested_level: Used as as a parameter of spin_lock_nested() of
1931 * dev->addr_list_lock.
1932 * @unlink_list: As netif_addr_lock() can be called recursively,
1933 * keep a list of interfaces to be deleted.
1935 * FIXME: cleanup struct net_device such that network protocol info
1940 char name[IFNAMSIZ];
1941 struct netdev_name_node *name_node;
1942 struct dev_ifalias __rcu *ifalias;
1944 * I/O specific fields
1945 * FIXME: Merge these and struct ifmap into one
1947 unsigned long mem_end;
1948 unsigned long mem_start;
1949 unsigned long base_addr;
1952 * Some hardware also needs these fields (state,dev_list,
1953 * napi_list,unreg_list,close_list) but they are not
1954 * part of the usual set specified in Space.c.
1957 unsigned long state;
1959 struct list_head dev_list;
1960 struct list_head napi_list;
1961 struct list_head unreg_list;
1962 struct list_head close_list;
1963 struct list_head ptype_all;
1964 struct list_head ptype_specific;
1967 struct list_head upper;
1968 struct list_head lower;
1971 /* Read-mostly cache-line for fast-path access */
1973 unsigned int priv_flags;
1974 const struct net_device_ops *netdev_ops;
1976 unsigned short gflags;
1977 unsigned short hard_header_len;
1979 /* Note : dev->mtu is often read without holding a lock.
1980 * Writers usually hold RTNL.
1981 * It is recommended to use READ_ONCE() to annotate the reads,
1982 * and to use WRITE_ONCE() to annotate the writes.
1985 unsigned short needed_headroom;
1986 unsigned short needed_tailroom;
1988 netdev_features_t features;
1989 netdev_features_t hw_features;
1990 netdev_features_t wanted_features;
1991 netdev_features_t vlan_features;
1992 netdev_features_t hw_enc_features;
1993 netdev_features_t mpls_features;
1994 netdev_features_t gso_partial_features;
1996 unsigned int min_mtu;
1997 unsigned int max_mtu;
1998 unsigned short type;
1999 unsigned char min_header_len;
2000 unsigned char name_assign_type;
2004 struct net_device_stats stats; /* not used by modern drivers */
2006 atomic_long_t rx_dropped;
2007 atomic_long_t tx_dropped;
2008 atomic_long_t rx_nohandler;
2010 /* Stats to monitor link on/off, flapping */
2011 atomic_t carrier_up_count;
2012 atomic_t carrier_down_count;
2014 #ifdef CONFIG_WIRELESS_EXT
2015 const struct iw_handler_def *wireless_handlers;
2016 struct iw_public_data *wireless_data;
2018 const struct ethtool_ops *ethtool_ops;
2019 #ifdef CONFIG_NET_L3_MASTER_DEV
2020 const struct l3mdev_ops *l3mdev_ops;
2022 #if IS_ENABLED(CONFIG_IPV6)
2023 const struct ndisc_ops *ndisc_ops;
2026 #ifdef CONFIG_XFRM_OFFLOAD
2027 const struct xfrmdev_ops *xfrmdev_ops;
2030 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2031 const struct tlsdev_ops *tlsdev_ops;
2034 const struct header_ops *header_ops;
2036 unsigned char operstate;
2037 unsigned char link_mode;
2039 unsigned char if_port;
2042 /* Interface address info. */
2043 unsigned char perm_addr[MAX_ADDR_LEN];
2044 unsigned char addr_assign_type;
2045 unsigned char addr_len;
2046 unsigned char upper_level;
2047 unsigned char lower_level;
2049 unsigned short neigh_priv_len;
2050 unsigned short dev_id;
2051 unsigned short dev_port;
2052 unsigned short padded;
2054 spinlock_t addr_list_lock;
2057 struct netdev_hw_addr_list uc;
2058 struct netdev_hw_addr_list mc;
2059 struct netdev_hw_addr_list dev_addrs;
2062 struct kset *queues_kset;
2064 #ifdef CONFIG_LOCKDEP
2065 struct list_head unlink_list;
2067 unsigned int promiscuity;
2068 unsigned int allmulti;
2070 #ifdef CONFIG_LOCKDEP
2071 unsigned char nested_level;
2075 /* Protocol-specific pointers */
2077 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2078 struct vlan_info __rcu *vlan_info;
2080 #if IS_ENABLED(CONFIG_NET_DSA)
2081 struct dsa_port *dsa_ptr;
2083 #if IS_ENABLED(CONFIG_TIPC)
2084 struct tipc_bearer __rcu *tipc_ptr;
2086 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2089 struct in_device __rcu *ip_ptr;
2090 #if IS_ENABLED(CONFIG_DECNET)
2091 struct dn_dev __rcu *dn_ptr;
2093 struct inet6_dev __rcu *ip6_ptr;
2094 #if IS_ENABLED(CONFIG_AX25)
2097 struct wireless_dev *ieee80211_ptr;
2098 struct wpan_dev *ieee802154_ptr;
2099 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2100 struct mpls_dev __rcu *mpls_ptr;
2102 #if IS_ENABLED(CONFIG_MCTP)
2103 struct mctp_dev __rcu *mctp_ptr;
2107 * Cache lines mostly used on receive path (including eth_type_trans())
2109 /* Interface address info used in eth_type_trans() */
2110 unsigned char *dev_addr;
2112 struct netdev_rx_queue *_rx;
2113 unsigned int num_rx_queues;
2114 unsigned int real_num_rx_queues;
2116 struct bpf_prog __rcu *xdp_prog;
2117 unsigned long gro_flush_timeout;
2118 int napi_defer_hard_irqs;
2119 rx_handler_func_t __rcu *rx_handler;
2120 void __rcu *rx_handler_data;
2122 #ifdef CONFIG_NET_CLS_ACT
2123 struct mini_Qdisc __rcu *miniq_ingress;
2125 struct netdev_queue __rcu *ingress_queue;
2126 #ifdef CONFIG_NETFILTER_INGRESS
2127 struct nf_hook_entries __rcu *nf_hooks_ingress;
2130 unsigned char broadcast[MAX_ADDR_LEN];
2131 #ifdef CONFIG_RFS_ACCEL
2132 struct cpu_rmap *rx_cpu_rmap;
2134 struct hlist_node index_hlist;
2137 * Cache lines mostly used on transmit path
2139 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2140 unsigned int num_tx_queues;
2141 unsigned int real_num_tx_queues;
2142 struct Qdisc *qdisc;
2143 unsigned int tx_queue_len;
2144 spinlock_t tx_global_lock;
2146 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2149 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2151 #ifdef CONFIG_NET_CLS_ACT
2152 struct mini_Qdisc __rcu *miniq_egress;
2155 #ifdef CONFIG_NET_SCHED
2156 DECLARE_HASHTABLE (qdisc_hash, 4);
2158 /* These may be needed for future network-power-down code. */
2159 struct timer_list watchdog_timer;
2162 u32 proto_down_reason;
2164 struct list_head todo_list;
2166 #ifdef CONFIG_PCPU_DEV_REFCNT
2167 int __percpu *pcpu_refcnt;
2169 refcount_t dev_refcnt;
2172 struct list_head link_watch_list;
2174 enum { NETREG_UNINITIALIZED=0,
2175 NETREG_REGISTERED, /* completed register_netdevice */
2176 NETREG_UNREGISTERING, /* called unregister_netdevice */
2177 NETREG_UNREGISTERED, /* completed unregister todo */
2178 NETREG_RELEASED, /* called free_netdev */
2179 NETREG_DUMMY, /* dummy device for NAPI poll */
2185 RTNL_LINK_INITIALIZED,
2186 RTNL_LINK_INITIALIZING,
2187 } rtnl_link_state:16;
2189 bool needs_free_netdev;
2190 void (*priv_destructor)(struct net_device *dev);
2192 #ifdef CONFIG_NETPOLL
2193 struct netpoll_info __rcu *npinfo;
2196 possible_net_t nd_net;
2198 /* mid-layer private */
2200 enum netdev_ml_priv_type ml_priv_type;
2203 struct pcpu_lstats __percpu *lstats;
2204 struct pcpu_sw_netstats __percpu *tstats;
2205 struct pcpu_dstats __percpu *dstats;
2208 #if IS_ENABLED(CONFIG_GARP)
2209 struct garp_port __rcu *garp_port;
2211 #if IS_ENABLED(CONFIG_MRP)
2212 struct mrp_port __rcu *mrp_port;
2216 const struct attribute_group *sysfs_groups[4];
2217 const struct attribute_group *sysfs_rx_queue_group;
2219 const struct rtnl_link_ops *rtnl_link_ops;
2221 /* for setting kernel sock attribute on TCP connection setup */
2222 #define GSO_MAX_SIZE 65536
2223 unsigned int gso_max_size;
2224 #define GSO_MAX_SEGS 65535
2228 const struct dcbnl_rtnl_ops *dcbnl_ops;
2231 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2232 u8 prio_tc_map[TC_BITMASK + 1];
2234 #if IS_ENABLED(CONFIG_FCOE)
2235 unsigned int fcoe_ddp_xid;
2237 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2238 struct netprio_map __rcu *priomap;
2240 struct phy_device *phydev;
2241 struct sfp_bus *sfp_bus;
2242 struct lock_class_key *qdisc_tx_busylock;
2243 struct lock_class_key *qdisc_running_key;
2245 unsigned wol_enabled:1;
2246 unsigned threaded:1;
2248 struct list_head net_notifier_list;
2250 #if IS_ENABLED(CONFIG_MACSEC)
2251 /* MACsec management functions */
2252 const struct macsec_ops *macsec_ops;
2254 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2255 struct udp_tunnel_nic *udp_tunnel_nic;
2257 /* protected by rtnl_lock */
2258 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2260 #define to_net_dev(d) container_of(d, struct net_device, dev)
2262 static inline bool netif_elide_gro(const struct net_device *dev)
2264 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2269 #define NETDEV_ALIGN 32
2272 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2274 return dev->prio_tc_map[prio & TC_BITMASK];
2278 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2280 if (tc >= dev->num_tc)
2283 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2287 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2288 void netdev_reset_tc(struct net_device *dev);
2289 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2290 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2293 int netdev_get_num_tc(struct net_device *dev)
2298 static inline void net_prefetch(void *p)
2301 #if L1_CACHE_BYTES < 128
2302 prefetch((u8 *)p + L1_CACHE_BYTES);
2306 static inline void net_prefetchw(void *p)
2309 #if L1_CACHE_BYTES < 128
2310 prefetchw((u8 *)p + L1_CACHE_BYTES);
2314 void netdev_unbind_sb_channel(struct net_device *dev,
2315 struct net_device *sb_dev);
2316 int netdev_bind_sb_channel_queue(struct net_device *dev,
2317 struct net_device *sb_dev,
2318 u8 tc, u16 count, u16 offset);
2319 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2320 static inline int netdev_get_sb_channel(struct net_device *dev)
2322 return max_t(int, -dev->num_tc, 0);
2326 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2329 return &dev->_tx[index];
2332 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2333 const struct sk_buff *skb)
2335 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2338 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2339 void (*f)(struct net_device *,
2340 struct netdev_queue *,
2346 for (i = 0; i < dev->num_tx_queues; i++)
2347 f(dev, &dev->_tx[i], arg);
2350 #define netdev_lockdep_set_classes(dev) \
2352 static struct lock_class_key qdisc_tx_busylock_key; \
2353 static struct lock_class_key qdisc_running_key; \
2354 static struct lock_class_key qdisc_xmit_lock_key; \
2355 static struct lock_class_key dev_addr_list_lock_key; \
2358 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2359 (dev)->qdisc_running_key = &qdisc_running_key; \
2360 lockdep_set_class(&(dev)->addr_list_lock, \
2361 &dev_addr_list_lock_key); \
2362 for (i = 0; i < (dev)->num_tx_queues; i++) \
2363 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2364 &qdisc_xmit_lock_key); \
2367 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2368 struct net_device *sb_dev);
2369 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2370 struct sk_buff *skb,
2371 struct net_device *sb_dev);
2373 /* returns the headroom that the master device needs to take in account
2374 * when forwarding to this dev
2376 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2378 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2381 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2383 if (dev->netdev_ops->ndo_set_rx_headroom)
2384 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2387 /* set the device rx headroom to the dev's default */
2388 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2390 netdev_set_rx_headroom(dev, -1);
2393 static inline void *netdev_get_ml_priv(struct net_device *dev,
2394 enum netdev_ml_priv_type type)
2396 if (dev->ml_priv_type != type)
2399 return dev->ml_priv;
2402 static inline void netdev_set_ml_priv(struct net_device *dev,
2404 enum netdev_ml_priv_type type)
2406 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2407 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2408 dev->ml_priv_type, type);
2409 WARN(!dev->ml_priv_type && dev->ml_priv,
2410 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2412 dev->ml_priv = ml_priv;
2413 dev->ml_priv_type = type;
2417 * Net namespace inlines
2420 struct net *dev_net(const struct net_device *dev)
2422 return read_pnet(&dev->nd_net);
2426 void dev_net_set(struct net_device *dev, struct net *net)
2428 write_pnet(&dev->nd_net, net);
2432 * netdev_priv - access network device private data
2433 * @dev: network device
2435 * Get network device private data
2437 static inline void *netdev_priv(const struct net_device *dev)
2439 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2442 /* Set the sysfs physical device reference for the network logical device
2443 * if set prior to registration will cause a symlink during initialization.
2445 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2447 /* Set the sysfs device type for the network logical device to allow
2448 * fine-grained identification of different network device types. For
2449 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2451 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2453 /* Default NAPI poll() weight
2454 * Device drivers are strongly advised to not use bigger value
2456 #define NAPI_POLL_WEIGHT 64
2459 * netif_napi_add - initialize a NAPI context
2460 * @dev: network device
2461 * @napi: NAPI context
2462 * @poll: polling function
2463 * @weight: default weight
2465 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2466 * *any* of the other NAPI-related functions.
2468 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2469 int (*poll)(struct napi_struct *, int), int weight);
2472 * netif_tx_napi_add - initialize a NAPI context
2473 * @dev: network device
2474 * @napi: NAPI context
2475 * @poll: polling function
2476 * @weight: default weight
2478 * This variant of netif_napi_add() should be used from drivers using NAPI
2479 * to exclusively poll a TX queue.
2480 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2482 static inline void netif_tx_napi_add(struct net_device *dev,
2483 struct napi_struct *napi,
2484 int (*poll)(struct napi_struct *, int),
2487 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2488 netif_napi_add(dev, napi, poll, weight);
2492 * __netif_napi_del - remove a NAPI context
2493 * @napi: NAPI context
2495 * Warning: caller must observe RCU grace period before freeing memory
2496 * containing @napi. Drivers might want to call this helper to combine
2497 * all the needed RCU grace periods into a single one.
2499 void __netif_napi_del(struct napi_struct *napi);
2502 * netif_napi_del - remove a NAPI context
2503 * @napi: NAPI context
2505 * netif_napi_del() removes a NAPI context from the network device NAPI list
2507 static inline void netif_napi_del(struct napi_struct *napi)
2509 __netif_napi_del(napi);
2513 struct napi_gro_cb {
2514 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2517 /* Length of frag0. */
2518 unsigned int frag0_len;
2520 /* This indicates where we are processing relative to skb->data. */
2523 /* This is non-zero if the packet cannot be merged with the new skb. */
2526 /* Save the IP ID here and check when we get to the transport layer */
2529 /* Number of segments aggregated. */
2532 /* Start offset for remote checksum offload */
2533 u16 gro_remcsum_start;
2535 /* jiffies when first packet was created/queued */
2538 /* Used in ipv6_gro_receive() and foo-over-udp */
2541 /* This is non-zero if the packet may be of the same flow. */
2544 /* Used in tunnel GRO receive */
2547 /* GRO checksum is valid */
2550 /* Number of checksums via CHECKSUM_UNNECESSARY */
2555 #define NAPI_GRO_FREE 1
2556 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2558 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2561 /* Used in GRE, set in fou/gue_gro_receive */
2564 /* Used to determine if flush_id can be ignored */
2567 /* Number of gro_receive callbacks this packet already went through */
2568 u8 recursion_counter:4;
2570 /* GRO is done by frag_list pointer chaining. */
2573 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2576 /* used in skb_gro_receive() slow path */
2577 struct sk_buff *last;
2580 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2582 #define GRO_RECURSION_LIMIT 15
2583 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2585 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2588 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2589 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2590 struct list_head *head,
2591 struct sk_buff *skb)
2593 if (unlikely(gro_recursion_inc_test(skb))) {
2594 NAPI_GRO_CB(skb)->flush |= 1;
2598 return cb(head, skb);
2601 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2603 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2605 struct list_head *head,
2606 struct sk_buff *skb)
2608 if (unlikely(gro_recursion_inc_test(skb))) {
2609 NAPI_GRO_CB(skb)->flush |= 1;
2613 return cb(sk, head, skb);
2616 struct packet_type {
2617 __be16 type; /* This is really htons(ether_type). */
2618 bool ignore_outgoing;
2619 struct net_device *dev; /* NULL is wildcarded here */
2620 int (*func) (struct sk_buff *,
2621 struct net_device *,
2622 struct packet_type *,
2623 struct net_device *);
2624 void (*list_func) (struct list_head *,
2625 struct packet_type *,
2626 struct net_device *);
2627 bool (*id_match)(struct packet_type *ptype,
2629 void *af_packet_priv;
2630 struct list_head list;
2633 struct offload_callbacks {
2634 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2635 netdev_features_t features);
2636 struct sk_buff *(*gro_receive)(struct list_head *head,
2637 struct sk_buff *skb);
2638 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2641 struct packet_offload {
2642 __be16 type; /* This is really htons(ether_type). */
2644 struct offload_callbacks callbacks;
2645 struct list_head list;
2648 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2649 struct pcpu_sw_netstats {
2654 struct u64_stats_sync syncp;
2655 } __aligned(4 * sizeof(u64));
2657 struct pcpu_lstats {
2658 u64_stats_t packets;
2660 struct u64_stats_sync syncp;
2661 } __aligned(2 * sizeof(u64));
2663 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2665 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2667 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2669 u64_stats_update_begin(&tstats->syncp);
2670 tstats->rx_bytes += len;
2671 tstats->rx_packets++;
2672 u64_stats_update_end(&tstats->syncp);
2675 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2676 unsigned int packets,
2679 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2681 u64_stats_update_begin(&tstats->syncp);
2682 tstats->tx_bytes += len;
2683 tstats->tx_packets += packets;
2684 u64_stats_update_end(&tstats->syncp);
2687 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2689 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2691 u64_stats_update_begin(&lstats->syncp);
2692 u64_stats_add(&lstats->bytes, len);
2693 u64_stats_inc(&lstats->packets);
2694 u64_stats_update_end(&lstats->syncp);
2697 #define __netdev_alloc_pcpu_stats(type, gfp) \
2699 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2702 for_each_possible_cpu(__cpu) { \
2703 typeof(type) *stat; \
2704 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2705 u64_stats_init(&stat->syncp); \
2711 #define netdev_alloc_pcpu_stats(type) \
2712 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2714 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2716 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2719 for_each_possible_cpu(__cpu) { \
2720 typeof(type) *stat; \
2721 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2722 u64_stats_init(&stat->syncp); \
2728 enum netdev_lag_tx_type {
2729 NETDEV_LAG_TX_TYPE_UNKNOWN,
2730 NETDEV_LAG_TX_TYPE_RANDOM,
2731 NETDEV_LAG_TX_TYPE_BROADCAST,
2732 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2733 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2734 NETDEV_LAG_TX_TYPE_HASH,
2737 enum netdev_lag_hash {
2738 NETDEV_LAG_HASH_NONE,
2740 NETDEV_LAG_HASH_L34,
2741 NETDEV_LAG_HASH_L23,
2742 NETDEV_LAG_HASH_E23,
2743 NETDEV_LAG_HASH_E34,
2744 NETDEV_LAG_HASH_VLAN_SRCMAC,
2745 NETDEV_LAG_HASH_UNKNOWN,
2748 struct netdev_lag_upper_info {
2749 enum netdev_lag_tx_type tx_type;
2750 enum netdev_lag_hash hash_type;
2753 struct netdev_lag_lower_state_info {
2758 #include <linux/notifier.h>
2760 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2761 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2765 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2767 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2768 detected a hardware crash and restarted
2769 - we can use this eg to kick tcp sessions
2771 NETDEV_CHANGE, /* Notify device state change */
2774 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2775 NETDEV_CHANGEADDR, /* notify after the address change */
2776 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2780 NETDEV_BONDING_FAILOVER,
2782 NETDEV_PRE_TYPE_CHANGE,
2783 NETDEV_POST_TYPE_CHANGE,
2786 NETDEV_NOTIFY_PEERS,
2790 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2791 NETDEV_CHANGEINFODATA,
2792 NETDEV_BONDING_INFO,
2793 NETDEV_PRECHANGEUPPER,
2794 NETDEV_CHANGELOWERSTATE,
2795 NETDEV_UDP_TUNNEL_PUSH_INFO,
2796 NETDEV_UDP_TUNNEL_DROP_INFO,
2797 NETDEV_CHANGE_TX_QUEUE_LEN,
2798 NETDEV_CVLAN_FILTER_PUSH_INFO,
2799 NETDEV_CVLAN_FILTER_DROP_INFO,
2800 NETDEV_SVLAN_FILTER_PUSH_INFO,
2801 NETDEV_SVLAN_FILTER_DROP_INFO,
2803 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2805 int register_netdevice_notifier(struct notifier_block *nb);
2806 int unregister_netdevice_notifier(struct notifier_block *nb);
2807 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2808 int unregister_netdevice_notifier_net(struct net *net,
2809 struct notifier_block *nb);
2810 int register_netdevice_notifier_dev_net(struct net_device *dev,
2811 struct notifier_block *nb,
2812 struct netdev_net_notifier *nn);
2813 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2814 struct notifier_block *nb,
2815 struct netdev_net_notifier *nn);
2817 struct netdev_notifier_info {
2818 struct net_device *dev;
2819 struct netlink_ext_ack *extack;
2822 struct netdev_notifier_info_ext {
2823 struct netdev_notifier_info info; /* must be first */
2829 struct netdev_notifier_change_info {
2830 struct netdev_notifier_info info; /* must be first */
2831 unsigned int flags_changed;
2834 struct netdev_notifier_changeupper_info {
2835 struct netdev_notifier_info info; /* must be first */
2836 struct net_device *upper_dev; /* new upper dev */
2837 bool master; /* is upper dev master */
2838 bool linking; /* is the notification for link or unlink */
2839 void *upper_info; /* upper dev info */
2842 struct netdev_notifier_changelowerstate_info {
2843 struct netdev_notifier_info info; /* must be first */
2844 void *lower_state_info; /* is lower dev state */
2847 struct netdev_notifier_pre_changeaddr_info {
2848 struct netdev_notifier_info info; /* must be first */
2849 const unsigned char *dev_addr;
2852 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2853 struct net_device *dev)
2856 info->extack = NULL;
2859 static inline struct net_device *
2860 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2865 static inline struct netlink_ext_ack *
2866 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2868 return info->extack;
2871 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2874 extern rwlock_t dev_base_lock; /* Device list lock */
2876 #define for_each_netdev(net, d) \
2877 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2878 #define for_each_netdev_reverse(net, d) \
2879 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2880 #define for_each_netdev_rcu(net, d) \
2881 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2882 #define for_each_netdev_safe(net, d, n) \
2883 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2884 #define for_each_netdev_continue(net, d) \
2885 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2886 #define for_each_netdev_continue_reverse(net, d) \
2887 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2889 #define for_each_netdev_continue_rcu(net, d) \
2890 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2891 #define for_each_netdev_in_bond_rcu(bond, slave) \
2892 for_each_netdev_rcu(&init_net, slave) \
2893 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2894 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2896 static inline struct net_device *next_net_device(struct net_device *dev)
2898 struct list_head *lh;
2902 lh = dev->dev_list.next;
2903 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2906 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2908 struct list_head *lh;
2912 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2913 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2916 static inline struct net_device *first_net_device(struct net *net)
2918 return list_empty(&net->dev_base_head) ? NULL :
2919 net_device_entry(net->dev_base_head.next);
2922 static inline struct net_device *first_net_device_rcu(struct net *net)
2924 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2926 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2929 int netdev_boot_setup_check(struct net_device *dev);
2930 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2931 const char *hwaddr);
2932 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2933 void dev_add_pack(struct packet_type *pt);
2934 void dev_remove_pack(struct packet_type *pt);
2935 void __dev_remove_pack(struct packet_type *pt);
2936 void dev_add_offload(struct packet_offload *po);
2937 void dev_remove_offload(struct packet_offload *po);
2939 int dev_get_iflink(const struct net_device *dev);
2940 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2941 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2942 struct net_device_path_stack *stack);
2943 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2944 unsigned short mask);
2945 struct net_device *dev_get_by_name(struct net *net, const char *name);
2946 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2947 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2948 int dev_alloc_name(struct net_device *dev, const char *name);
2949 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2950 void dev_close(struct net_device *dev);
2951 void dev_close_many(struct list_head *head, bool unlink);
2952 void dev_disable_lro(struct net_device *dev);
2953 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2954 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2955 struct net_device *sb_dev);
2956 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2957 struct net_device *sb_dev);
2959 int dev_queue_xmit(struct sk_buff *skb);
2960 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2961 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2963 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2967 ret = __dev_direct_xmit(skb, queue_id);
2968 if (!dev_xmit_complete(ret))
2973 int register_netdevice(struct net_device *dev);
2974 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2975 void unregister_netdevice_many(struct list_head *head);
2976 static inline void unregister_netdevice(struct net_device *dev)
2978 unregister_netdevice_queue(dev, NULL);
2981 int netdev_refcnt_read(const struct net_device *dev);
2982 void free_netdev(struct net_device *dev);
2983 void netdev_freemem(struct net_device *dev);
2984 int init_dummy_netdev(struct net_device *dev);
2986 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2987 struct sk_buff *skb,
2989 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2991 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2992 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2993 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2994 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2995 int netdev_get_name(struct net *net, char *name, int ifindex);
2996 int dev_restart(struct net_device *dev);
2997 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2998 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3000 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3002 return NAPI_GRO_CB(skb)->data_offset;
3005 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3007 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3010 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3012 NAPI_GRO_CB(skb)->data_offset += len;
3015 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3016 unsigned int offset)
3018 return NAPI_GRO_CB(skb)->frag0 + offset;
3021 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3023 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3026 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3028 NAPI_GRO_CB(skb)->frag0 = NULL;
3029 NAPI_GRO_CB(skb)->frag0_len = 0;
3032 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3033 unsigned int offset)
3035 if (!pskb_may_pull(skb, hlen))
3038 skb_gro_frag0_invalidate(skb);
3039 return skb->data + offset;
3042 static inline void *skb_gro_network_header(struct sk_buff *skb)
3044 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3045 skb_network_offset(skb);
3048 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3049 const void *start, unsigned int len)
3051 if (NAPI_GRO_CB(skb)->csum_valid)
3052 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3053 csum_partial(start, len, 0));
3056 /* GRO checksum functions. These are logical equivalents of the normal
3057 * checksum functions (in skbuff.h) except that they operate on the GRO
3058 * offsets and fields in sk_buff.
3061 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3063 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3065 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3068 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3072 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3073 skb_checksum_start_offset(skb) <
3074 skb_gro_offset(skb)) &&
3075 !skb_at_gro_remcsum_start(skb) &&
3076 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3077 (!zero_okay || check));
3080 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3083 if (NAPI_GRO_CB(skb)->csum_valid &&
3084 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3087 NAPI_GRO_CB(skb)->csum = psum;
3089 return __skb_gro_checksum_complete(skb);
3092 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3094 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3095 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3096 NAPI_GRO_CB(skb)->csum_cnt--;
3098 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3099 * verified a new top level checksum or an encapsulated one
3100 * during GRO. This saves work if we fallback to normal path.
3102 __skb_incr_checksum_unnecessary(skb);
3106 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3109 __sum16 __ret = 0; \
3110 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3111 __ret = __skb_gro_checksum_validate_complete(skb, \
3112 compute_pseudo(skb, proto)); \
3114 skb_gro_incr_csum_unnecessary(skb); \
3118 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3119 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3121 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3123 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3125 #define skb_gro_checksum_simple_validate(skb) \
3126 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3128 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3130 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3131 !NAPI_GRO_CB(skb)->csum_valid);
3134 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3137 NAPI_GRO_CB(skb)->csum = ~pseudo;
3138 NAPI_GRO_CB(skb)->csum_valid = 1;
3141 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3143 if (__skb_gro_checksum_convert_check(skb)) \
3144 __skb_gro_checksum_convert(skb, \
3145 compute_pseudo(skb, proto)); \
3148 struct gro_remcsum {
3153 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3159 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3160 unsigned int off, size_t hdrlen,
3161 int start, int offset,
3162 struct gro_remcsum *grc,
3166 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3168 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3171 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3175 ptr = skb_gro_header_fast(skb, off);
3176 if (skb_gro_header_hard(skb, off + plen)) {
3177 ptr = skb_gro_header_slow(skb, off + plen, off);
3182 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3185 /* Adjust skb->csum since we changed the packet */
3186 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3188 grc->offset = off + hdrlen + offset;
3194 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3195 struct gro_remcsum *grc)
3198 size_t plen = grc->offset + sizeof(u16);
3203 ptr = skb_gro_header_fast(skb, grc->offset);
3204 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3205 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3210 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3213 #ifdef CONFIG_XFRM_OFFLOAD
3214 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3216 if (PTR_ERR(pp) != -EINPROGRESS)
3217 NAPI_GRO_CB(skb)->flush |= flush;
3219 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3222 struct gro_remcsum *grc)
3224 if (PTR_ERR(pp) != -EINPROGRESS) {
3225 NAPI_GRO_CB(skb)->flush |= flush;
3226 skb_gro_remcsum_cleanup(skb, grc);
3227 skb->remcsum_offload = 0;
3231 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3233 NAPI_GRO_CB(skb)->flush |= flush;
3235 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3238 struct gro_remcsum *grc)
3240 NAPI_GRO_CB(skb)->flush |= flush;
3241 skb_gro_remcsum_cleanup(skb, grc);
3242 skb->remcsum_offload = 0;
3246 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3247 unsigned short type,
3248 const void *daddr, const void *saddr,
3251 if (!dev->header_ops || !dev->header_ops->create)
3254 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3257 static inline int dev_parse_header(const struct sk_buff *skb,
3258 unsigned char *haddr)
3260 const struct net_device *dev = skb->dev;
3262 if (!dev->header_ops || !dev->header_ops->parse)
3264 return dev->header_ops->parse(skb, haddr);
3267 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3269 const struct net_device *dev = skb->dev;
3271 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3273 return dev->header_ops->parse_protocol(skb);
3276 /* ll_header must have at least hard_header_len allocated */
3277 static inline bool dev_validate_header(const struct net_device *dev,
3278 char *ll_header, int len)
3280 if (likely(len >= dev->hard_header_len))
3282 if (len < dev->min_header_len)
3285 if (capable(CAP_SYS_RAWIO)) {
3286 memset(ll_header + len, 0, dev->hard_header_len - len);
3290 if (dev->header_ops && dev->header_ops->validate)
3291 return dev->header_ops->validate(ll_header, len);
3296 static inline bool dev_has_header(const struct net_device *dev)
3298 return dev->header_ops && dev->header_ops->create;
3301 #ifdef CONFIG_NET_FLOW_LIMIT
3302 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3303 struct sd_flow_limit {
3305 unsigned int num_buckets;
3306 unsigned int history_head;
3307 u16 history[FLOW_LIMIT_HISTORY];
3311 extern int netdev_flow_limit_table_len;
3312 #endif /* CONFIG_NET_FLOW_LIMIT */
3315 * Incoming packets are placed on per-CPU queues
3317 struct softnet_data {
3318 struct list_head poll_list;
3319 struct sk_buff_head process_queue;
3322 unsigned int processed;
3323 unsigned int time_squeeze;
3324 unsigned int received_rps;
3326 struct softnet_data *rps_ipi_list;
3328 #ifdef CONFIG_NET_FLOW_LIMIT
3329 struct sd_flow_limit __rcu *flow_limit;
3331 struct Qdisc *output_queue;
3332 struct Qdisc **output_queue_tailp;
3333 struct sk_buff *completion_queue;
3334 #ifdef CONFIG_XFRM_OFFLOAD
3335 struct sk_buff_head xfrm_backlog;
3337 /* written and read only by owning cpu: */
3343 /* input_queue_head should be written by cpu owning this struct,
3344 * and only read by other cpus. Worth using a cache line.
3346 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3348 /* Elements below can be accessed between CPUs for RPS/RFS */
3349 call_single_data_t csd ____cacheline_aligned_in_smp;
3350 struct softnet_data *rps_ipi_next;
3352 unsigned int input_queue_tail;
3354 unsigned int dropped;
3355 struct sk_buff_head input_pkt_queue;
3356 struct napi_struct backlog;
3360 static inline void input_queue_head_incr(struct softnet_data *sd)
3363 sd->input_queue_head++;
3367 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3368 unsigned int *qtail)
3371 *qtail = ++sd->input_queue_tail;
3375 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3377 static inline int dev_recursion_level(void)
3379 return this_cpu_read(softnet_data.xmit.recursion);
3382 #define XMIT_RECURSION_LIMIT 8
3383 static inline bool dev_xmit_recursion(void)
3385 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3386 XMIT_RECURSION_LIMIT);
3389 static inline void dev_xmit_recursion_inc(void)
3391 __this_cpu_inc(softnet_data.xmit.recursion);
3394 static inline void dev_xmit_recursion_dec(void)
3396 __this_cpu_dec(softnet_data.xmit.recursion);
3399 void __netif_schedule(struct Qdisc *q);
3400 void netif_schedule_queue(struct netdev_queue *txq);
3402 static inline void netif_tx_schedule_all(struct net_device *dev)
3406 for (i = 0; i < dev->num_tx_queues; i++)
3407 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3410 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3412 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3416 * netif_start_queue - allow transmit
3417 * @dev: network device
3419 * Allow upper layers to call the device hard_start_xmit routine.
3421 static inline void netif_start_queue(struct net_device *dev)
3423 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3426 static inline void netif_tx_start_all_queues(struct net_device *dev)
3430 for (i = 0; i < dev->num_tx_queues; i++) {
3431 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3432 netif_tx_start_queue(txq);
3436 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3439 * netif_wake_queue - restart transmit
3440 * @dev: network device
3442 * Allow upper layers to call the device hard_start_xmit routine.
3443 * Used for flow control when transmit resources are available.
3445 static inline void netif_wake_queue(struct net_device *dev)
3447 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3450 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3454 for (i = 0; i < dev->num_tx_queues; i++) {
3455 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3456 netif_tx_wake_queue(txq);
3460 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3462 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3466 * netif_stop_queue - stop transmitted packets
3467 * @dev: network device
3469 * Stop upper layers calling the device hard_start_xmit routine.
3470 * Used for flow control when transmit resources are unavailable.
3472 static inline void netif_stop_queue(struct net_device *dev)
3474 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3477 void netif_tx_stop_all_queues(struct net_device *dev);
3479 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3481 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3485 * netif_queue_stopped - test if transmit queue is flowblocked
3486 * @dev: network device
3488 * Test if transmit queue on device is currently unable to send.
3490 static inline bool netif_queue_stopped(const struct net_device *dev)
3492 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3495 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3497 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3501 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3503 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3507 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3509 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3513 * netdev_queue_set_dql_min_limit - set dql minimum limit
3514 * @dev_queue: pointer to transmit queue
3515 * @min_limit: dql minimum limit
3517 * Forces xmit_more() to return true until the minimum threshold
3518 * defined by @min_limit is reached (or until the tx queue is
3519 * empty). Warning: to be use with care, misuse will impact the
3522 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3523 unsigned int min_limit)
3526 dev_queue->dql.min_limit = min_limit;
3531 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3532 * @dev_queue: pointer to transmit queue
3534 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3535 * to give appropriate hint to the CPU.
3537 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3540 prefetchw(&dev_queue->dql.num_queued);
3545 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3546 * @dev_queue: pointer to transmit queue
3548 * BQL enabled drivers might use this helper in their TX completion path,
3549 * to give appropriate hint to the CPU.
3551 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3554 prefetchw(&dev_queue->dql.limit);
3558 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3562 dql_queued(&dev_queue->dql, bytes);
3564 if (likely(dql_avail(&dev_queue->dql) >= 0))
3567 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3570 * The XOFF flag must be set before checking the dql_avail below,
3571 * because in netdev_tx_completed_queue we update the dql_completed
3572 * before checking the XOFF flag.
3576 /* check again in case another CPU has just made room avail */
3577 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3578 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3582 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3583 * that they should not test BQL status themselves.
3584 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3586 * Returns true if the doorbell must be used to kick the NIC.
3588 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3594 dql_queued(&dev_queue->dql, bytes);
3596 return netif_tx_queue_stopped(dev_queue);
3598 netdev_tx_sent_queue(dev_queue, bytes);
3603 * netdev_sent_queue - report the number of bytes queued to hardware
3604 * @dev: network device
3605 * @bytes: number of bytes queued to the hardware device queue
3607 * Report the number of bytes queued for sending/completion to the network
3608 * device hardware queue. @bytes should be a good approximation and should
3609 * exactly match netdev_completed_queue() @bytes
3611 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3613 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3616 static inline bool __netdev_sent_queue(struct net_device *dev,
3620 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3624 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3625 unsigned int pkts, unsigned int bytes)
3628 if (unlikely(!bytes))
3631 dql_completed(&dev_queue->dql, bytes);
3634 * Without the memory barrier there is a small possiblity that
3635 * netdev_tx_sent_queue will miss the update and cause the queue to
3636 * be stopped forever
3640 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3643 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3644 netif_schedule_queue(dev_queue);
3649 * netdev_completed_queue - report bytes and packets completed by device
3650 * @dev: network device
3651 * @pkts: actual number of packets sent over the medium
3652 * @bytes: actual number of bytes sent over the medium
3654 * Report the number of bytes and packets transmitted by the network device
3655 * hardware queue over the physical medium, @bytes must exactly match the
3656 * @bytes amount passed to netdev_sent_queue()
3658 static inline void netdev_completed_queue(struct net_device *dev,
3659 unsigned int pkts, unsigned int bytes)
3661 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3664 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3667 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3673 * netdev_reset_queue - reset the packets and bytes count of a network device
3674 * @dev_queue: network device
3676 * Reset the bytes and packet count of a network device and clear the
3677 * software flow control OFF bit for this network device
3679 static inline void netdev_reset_queue(struct net_device *dev_queue)
3681 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3685 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3686 * @dev: network device
3687 * @queue_index: given tx queue index
3689 * Returns 0 if given tx queue index >= number of device tx queues,
3690 * otherwise returns the originally passed tx queue index.
3692 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3694 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3695 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3696 dev->name, queue_index,
3697 dev->real_num_tx_queues);
3705 * netif_running - test if up
3706 * @dev: network device
3708 * Test if the device has been brought up.
3710 static inline bool netif_running(const struct net_device *dev)
3712 return test_bit(__LINK_STATE_START, &dev->state);
3716 * Routines to manage the subqueues on a device. We only need start,
3717 * stop, and a check if it's stopped. All other device management is
3718 * done at the overall netdevice level.
3719 * Also test the device if we're multiqueue.
3723 * netif_start_subqueue - allow sending packets on subqueue
3724 * @dev: network device
3725 * @queue_index: sub queue index
3727 * Start individual transmit queue of a device with multiple transmit queues.
3729 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3731 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3733 netif_tx_start_queue(txq);
3737 * netif_stop_subqueue - stop sending packets on subqueue
3738 * @dev: network device
3739 * @queue_index: sub queue index
3741 * Stop individual transmit queue of a device with multiple transmit queues.
3743 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3745 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3746 netif_tx_stop_queue(txq);
3750 * __netif_subqueue_stopped - test status of subqueue
3751 * @dev: network device
3752 * @queue_index: sub queue index
3754 * Check individual transmit queue of a device with multiple transmit queues.
3756 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3759 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3761 return netif_tx_queue_stopped(txq);
3765 * netif_subqueue_stopped - test status of subqueue
3766 * @dev: network device
3767 * @skb: sub queue buffer pointer
3769 * Check individual transmit queue of a device with multiple transmit queues.
3771 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3772 struct sk_buff *skb)
3774 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3778 * netif_wake_subqueue - allow sending packets on subqueue
3779 * @dev: network device
3780 * @queue_index: sub queue index
3782 * Resume individual transmit queue of a device with multiple transmit queues.
3784 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3786 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3788 netif_tx_wake_queue(txq);
3792 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3794 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3795 u16 index, enum xps_map_type type);
3798 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3799 * @j: CPU/Rx queue index
3800 * @mask: bitmask of all cpus/rx queues
3801 * @nr_bits: number of bits in the bitmask
3803 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3805 static inline bool netif_attr_test_mask(unsigned long j,
3806 const unsigned long *mask,
3807 unsigned int nr_bits)
3809 cpu_max_bits_warn(j, nr_bits);
3810 return test_bit(j, mask);
3814 * netif_attr_test_online - Test for online CPU/Rx queue
3815 * @j: CPU/Rx queue index
3816 * @online_mask: bitmask for CPUs/Rx queues that are online
3817 * @nr_bits: number of bits in the bitmask
3819 * Returns true if a CPU/Rx queue is online.
3821 static inline bool netif_attr_test_online(unsigned long j,
3822 const unsigned long *online_mask,
3823 unsigned int nr_bits)
3825 cpu_max_bits_warn(j, nr_bits);
3828 return test_bit(j, online_mask);
3830 return (j < nr_bits);
3834 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3835 * @n: CPU/Rx queue index
3836 * @srcp: the cpumask/Rx queue mask pointer
3837 * @nr_bits: number of bits in the bitmask
3839 * Returns >= nr_bits if no further CPUs/Rx queues set.
3841 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3842 unsigned int nr_bits)
3844 /* -1 is a legal arg here. */
3846 cpu_max_bits_warn(n, nr_bits);
3849 return find_next_bit(srcp, nr_bits, n + 1);
3855 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3856 * @n: CPU/Rx queue index
3857 * @src1p: the first CPUs/Rx queues mask pointer
3858 * @src2p: the second CPUs/Rx queues mask pointer
3859 * @nr_bits: number of bits in the bitmask
3861 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3863 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3864 const unsigned long *src2p,
3865 unsigned int nr_bits)
3867 /* -1 is a legal arg here. */
3869 cpu_max_bits_warn(n, nr_bits);
3872 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3874 return find_next_bit(src1p, nr_bits, n + 1);
3876 return find_next_bit(src2p, nr_bits, n + 1);
3881 static inline int netif_set_xps_queue(struct net_device *dev,
3882 const struct cpumask *mask,
3888 static inline int __netif_set_xps_queue(struct net_device *dev,
3889 const unsigned long *mask,
3890 u16 index, enum xps_map_type type)
3897 * netif_is_multiqueue - test if device has multiple transmit queues
3898 * @dev: network device
3900 * Check if device has multiple transmit queues
3902 static inline bool netif_is_multiqueue(const struct net_device *dev)
3904 return dev->num_tx_queues > 1;
3907 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3910 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3912 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3915 dev->real_num_rx_queues = rxqs;
3919 int netif_set_real_num_queues(struct net_device *dev,
3920 unsigned int txq, unsigned int rxq);
3922 static inline struct netdev_rx_queue *
3923 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3925 return dev->_rx + rxq;
3929 static inline unsigned int get_netdev_rx_queue_index(
3930 struct netdev_rx_queue *queue)
3932 struct net_device *dev = queue->dev;
3933 int index = queue - dev->_rx;
3935 BUG_ON(index >= dev->num_rx_queues);
3940 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3941 int netif_get_num_default_rss_queues(void);
3943 enum skb_free_reason {
3944 SKB_REASON_CONSUMED,
3948 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3949 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3952 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3953 * interrupt context or with hardware interrupts being disabled.
3954 * (in_irq() || irqs_disabled())
3956 * We provide four helpers that can be used in following contexts :
3958 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3959 * replacing kfree_skb(skb)
3961 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3962 * Typically used in place of consume_skb(skb) in TX completion path
3964 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3965 * replacing kfree_skb(skb)
3967 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3968 * and consumed a packet. Used in place of consume_skb(skb)
3970 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3972 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3975 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3977 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3980 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3982 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3985 static inline void dev_consume_skb_any(struct sk_buff *skb)
3987 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3990 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3991 struct bpf_prog *xdp_prog);
3992 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3993 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3994 int netif_rx(struct sk_buff *skb);
3995 int netif_rx_ni(struct sk_buff *skb);
3996 int netif_rx_any_context(struct sk_buff *skb);
3997 int netif_receive_skb(struct sk_buff *skb);
3998 int netif_receive_skb_core(struct sk_buff *skb);
3999 void netif_receive_skb_list(struct list_head *head);
4000 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4001 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4002 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4003 gro_result_t napi_gro_frags(struct napi_struct *napi);
4004 struct packet_offload *gro_find_receive_by_type(__be16 type);
4005 struct packet_offload *gro_find_complete_by_type(__be16 type);
4007 static inline void napi_free_frags(struct napi_struct *napi)
4009 kfree_skb(napi->skb);
4013 bool netdev_is_rx_handler_busy(struct net_device *dev);
4014 int netdev_rx_handler_register(struct net_device *dev,
4015 rx_handler_func_t *rx_handler,
4016 void *rx_handler_data);
4017 void netdev_rx_handler_unregister(struct net_device *dev);
4019 bool dev_valid_name(const char *name);
4020 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4021 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4022 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4023 void __user *data, bool *need_copyout);
4024 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4025 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4026 unsigned int dev_get_flags(const struct net_device *);
4027 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4028 struct netlink_ext_ack *extack);
4029 int dev_change_flags(struct net_device *dev, unsigned int flags,
4030 struct netlink_ext_ack *extack);
4031 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4032 unsigned int gchanges);
4033 int dev_change_name(struct net_device *, const char *);
4034 int dev_set_alias(struct net_device *, const char *, size_t);
4035 int dev_get_alias(const struct net_device *, char *, size_t);
4036 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4037 const char *pat, int new_ifindex);
4039 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4042 return __dev_change_net_namespace(dev, net, pat, 0);
4044 int __dev_set_mtu(struct net_device *, int);
4045 int dev_validate_mtu(struct net_device *dev, int mtu,
4046 struct netlink_ext_ack *extack);
4047 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4048 struct netlink_ext_ack *extack);
4049 int dev_set_mtu(struct net_device *, int);
4050 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4051 void dev_set_group(struct net_device *, int);
4052 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4053 struct netlink_ext_ack *extack);
4054 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4055 struct netlink_ext_ack *extack);
4056 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4057 struct netlink_ext_ack *extack);
4058 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4059 int dev_change_carrier(struct net_device *, bool new_carrier);
4060 int dev_get_phys_port_id(struct net_device *dev,
4061 struct netdev_phys_item_id *ppid);
4062 int dev_get_phys_port_name(struct net_device *dev,
4063 char *name, size_t len);
4064 int dev_get_port_parent_id(struct net_device *dev,
4065 struct netdev_phys_item_id *ppid, bool recurse);
4066 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4067 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4068 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4069 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4071 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4072 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4073 struct netdev_queue *txq, int *ret);
4075 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4076 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4077 int fd, int expected_fd, u32 flags);
4078 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4079 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4081 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4082 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4083 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4084 bool is_skb_forwardable(const struct net_device *dev,
4085 const struct sk_buff *skb);
4087 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4088 const struct sk_buff *skb,
4089 const bool check_mtu)
4091 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4094 if (!(dev->flags & IFF_UP))
4100 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4101 if (skb->len <= len)
4104 /* if TSO is enabled, we don't care about the length as the packet
4105 * could be forwarded without being segmented before
4107 if (skb_is_gso(skb))
4113 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4114 struct sk_buff *skb,
4115 const bool check_mtu)
4117 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4118 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4119 atomic_long_inc(&dev->rx_dropped);
4124 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4129 bool dev_nit_active(struct net_device *dev);
4130 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4132 extern int netdev_budget;
4133 extern unsigned int netdev_budget_usecs;
4135 /* Called by rtnetlink.c:rtnl_unlock() */
4136 void netdev_run_todo(void);
4139 * dev_put - release reference to device
4140 * @dev: network device
4142 * Release reference to device to allow it to be freed.
4144 static inline void dev_put(struct net_device *dev)
4146 #ifdef CONFIG_PCPU_DEV_REFCNT
4147 this_cpu_dec(*dev->pcpu_refcnt);
4149 refcount_dec(&dev->dev_refcnt);
4154 * dev_hold - get reference to device
4155 * @dev: network device
4157 * Hold reference to device to keep it from being freed.
4159 static inline void dev_hold(struct net_device *dev)
4161 #ifdef CONFIG_PCPU_DEV_REFCNT
4162 this_cpu_inc(*dev->pcpu_refcnt);
4164 refcount_inc(&dev->dev_refcnt);
4168 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4169 * and _off may be called from IRQ context, but it is caller
4170 * who is responsible for serialization of these calls.
4172 * The name carrier is inappropriate, these functions should really be
4173 * called netif_lowerlayer_*() because they represent the state of any
4174 * kind of lower layer not just hardware media.
4177 void linkwatch_init_dev(struct net_device *dev);
4178 void linkwatch_fire_event(struct net_device *dev);
4179 void linkwatch_forget_dev(struct net_device *dev);
4182 * netif_carrier_ok - test if carrier present
4183 * @dev: network device
4185 * Check if carrier is present on device
4187 static inline bool netif_carrier_ok(const struct net_device *dev)
4189 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4192 unsigned long dev_trans_start(struct net_device *dev);
4194 void __netdev_watchdog_up(struct net_device *dev);
4196 void netif_carrier_on(struct net_device *dev);
4197 void netif_carrier_off(struct net_device *dev);
4198 void netif_carrier_event(struct net_device *dev);
4201 * netif_dormant_on - mark device as dormant.
4202 * @dev: network device
4204 * Mark device as dormant (as per RFC2863).
4206 * The dormant state indicates that the relevant interface is not
4207 * actually in a condition to pass packets (i.e., it is not 'up') but is
4208 * in a "pending" state, waiting for some external event. For "on-
4209 * demand" interfaces, this new state identifies the situation where the
4210 * interface is waiting for events to place it in the up state.
4212 static inline void netif_dormant_on(struct net_device *dev)
4214 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4215 linkwatch_fire_event(dev);
4219 * netif_dormant_off - set device as not dormant.
4220 * @dev: network device
4222 * Device is not in dormant state.
4224 static inline void netif_dormant_off(struct net_device *dev)
4226 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4227 linkwatch_fire_event(dev);
4231 * netif_dormant - test if device is dormant
4232 * @dev: network device
4234 * Check if device is dormant.
4236 static inline bool netif_dormant(const struct net_device *dev)
4238 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4243 * netif_testing_on - mark device as under test.
4244 * @dev: network device
4246 * Mark device as under test (as per RFC2863).
4248 * The testing state indicates that some test(s) must be performed on
4249 * the interface. After completion, of the test, the interface state
4250 * will change to up, dormant, or down, as appropriate.
4252 static inline void netif_testing_on(struct net_device *dev)
4254 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4255 linkwatch_fire_event(dev);
4259 * netif_testing_off - set device as not under test.
4260 * @dev: network device
4262 * Device is not in testing state.
4264 static inline void netif_testing_off(struct net_device *dev)
4266 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4267 linkwatch_fire_event(dev);
4271 * netif_testing - test if device is under test
4272 * @dev: network device
4274 * Check if device is under test
4276 static inline bool netif_testing(const struct net_device *dev)
4278 return test_bit(__LINK_STATE_TESTING, &dev->state);
4283 * netif_oper_up - test if device is operational
4284 * @dev: network device
4286 * Check if carrier is operational
4288 static inline bool netif_oper_up(const struct net_device *dev)
4290 return (dev->operstate == IF_OPER_UP ||
4291 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4295 * netif_device_present - is device available or removed
4296 * @dev: network device
4298 * Check if device has not been removed from system.
4300 static inline bool netif_device_present(const struct net_device *dev)
4302 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4305 void netif_device_detach(struct net_device *dev);
4307 void netif_device_attach(struct net_device *dev);
4310 * Network interface message level settings
4315 NETIF_MSG_PROBE_BIT,
4317 NETIF_MSG_TIMER_BIT,
4318 NETIF_MSG_IFDOWN_BIT,
4320 NETIF_MSG_RX_ERR_BIT,
4321 NETIF_MSG_TX_ERR_BIT,
4322 NETIF_MSG_TX_QUEUED_BIT,
4324 NETIF_MSG_TX_DONE_BIT,
4325 NETIF_MSG_RX_STATUS_BIT,
4326 NETIF_MSG_PKTDATA_BIT,
4330 /* When you add a new bit above, update netif_msg_class_names array
4331 * in net/ethtool/common.c
4333 NETIF_MSG_CLASS_COUNT,
4335 /* Both ethtool_ops interface and internal driver implementation use u32 */
4336 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4338 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4339 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4341 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4342 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4343 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4344 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4345 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4346 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4347 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4348 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4349 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4350 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4351 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4352 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4353 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4354 #define NETIF_MSG_HW __NETIF_MSG(HW)
4355 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4357 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4358 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4359 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4360 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4361 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4362 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4363 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4364 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4365 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4366 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4367 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4368 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4369 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4370 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4371 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4373 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4376 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4377 return default_msg_enable_bits;
4378 if (debug_value == 0) /* no output */
4380 /* set low N bits */
4381 return (1U << debug_value) - 1;
4384 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4386 spin_lock(&txq->_xmit_lock);
4387 txq->xmit_lock_owner = cpu;
4390 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4392 __acquire(&txq->_xmit_lock);
4396 static inline void __netif_tx_release(struct netdev_queue *txq)
4398 __release(&txq->_xmit_lock);
4401 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4403 spin_lock_bh(&txq->_xmit_lock);
4404 txq->xmit_lock_owner = smp_processor_id();
4407 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4409 bool ok = spin_trylock(&txq->_xmit_lock);
4411 txq->xmit_lock_owner = smp_processor_id();
4415 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4417 txq->xmit_lock_owner = -1;
4418 spin_unlock(&txq->_xmit_lock);
4421 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4423 txq->xmit_lock_owner = -1;
4424 spin_unlock_bh(&txq->_xmit_lock);
4427 static inline void txq_trans_update(struct netdev_queue *txq)
4429 if (txq->xmit_lock_owner != -1)
4430 txq->trans_start = jiffies;
4433 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4434 static inline void netif_trans_update(struct net_device *dev)
4436 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4438 if (txq->trans_start != jiffies)
4439 txq->trans_start = jiffies;
4443 * netif_tx_lock - grab network device transmit lock
4444 * @dev: network device
4446 * Get network device transmit lock
4448 static inline void netif_tx_lock(struct net_device *dev)
4453 spin_lock(&dev->tx_global_lock);
4454 cpu = smp_processor_id();
4455 for (i = 0; i < dev->num_tx_queues; i++) {
4456 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4458 /* We are the only thread of execution doing a
4459 * freeze, but we have to grab the _xmit_lock in
4460 * order to synchronize with threads which are in
4461 * the ->hard_start_xmit() handler and already
4462 * checked the frozen bit.
4464 __netif_tx_lock(txq, cpu);
4465 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4466 __netif_tx_unlock(txq);
4470 static inline void netif_tx_lock_bh(struct net_device *dev)
4476 static inline void netif_tx_unlock(struct net_device *dev)
4480 for (i = 0; i < dev->num_tx_queues; i++) {
4481 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4483 /* No need to grab the _xmit_lock here. If the
4484 * queue is not stopped for another reason, we
4487 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4488 netif_schedule_queue(txq);
4490 spin_unlock(&dev->tx_global_lock);
4493 static inline void netif_tx_unlock_bh(struct net_device *dev)
4495 netif_tx_unlock(dev);
4499 #define HARD_TX_LOCK(dev, txq, cpu) { \
4500 if ((dev->features & NETIF_F_LLTX) == 0) { \
4501 __netif_tx_lock(txq, cpu); \
4503 __netif_tx_acquire(txq); \
4507 #define HARD_TX_TRYLOCK(dev, txq) \
4508 (((dev->features & NETIF_F_LLTX) == 0) ? \
4509 __netif_tx_trylock(txq) : \
4510 __netif_tx_acquire(txq))
4512 #define HARD_TX_UNLOCK(dev, txq) { \
4513 if ((dev->features & NETIF_F_LLTX) == 0) { \
4514 __netif_tx_unlock(txq); \
4516 __netif_tx_release(txq); \
4520 static inline void netif_tx_disable(struct net_device *dev)
4526 cpu = smp_processor_id();
4527 spin_lock(&dev->tx_global_lock);
4528 for (i = 0; i < dev->num_tx_queues; i++) {
4529 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4531 __netif_tx_lock(txq, cpu);
4532 netif_tx_stop_queue(txq);
4533 __netif_tx_unlock(txq);
4535 spin_unlock(&dev->tx_global_lock);
4539 static inline void netif_addr_lock(struct net_device *dev)
4541 unsigned char nest_level = 0;
4543 #ifdef CONFIG_LOCKDEP
4544 nest_level = dev->nested_level;
4546 spin_lock_nested(&dev->addr_list_lock, nest_level);
4549 static inline void netif_addr_lock_bh(struct net_device *dev)
4551 unsigned char nest_level = 0;
4553 #ifdef CONFIG_LOCKDEP
4554 nest_level = dev->nested_level;
4557 spin_lock_nested(&dev->addr_list_lock, nest_level);
4560 static inline void netif_addr_unlock(struct net_device *dev)
4562 spin_unlock(&dev->addr_list_lock);
4565 static inline void netif_addr_unlock_bh(struct net_device *dev)
4567 spin_unlock_bh(&dev->addr_list_lock);
4571 * dev_addrs walker. Should be used only for read access. Call with
4572 * rcu_read_lock held.
4574 #define for_each_dev_addr(dev, ha) \
4575 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4577 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4579 void ether_setup(struct net_device *dev);
4581 /* Support for loadable net-drivers */
4582 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4583 unsigned char name_assign_type,
4584 void (*setup)(struct net_device *),
4585 unsigned int txqs, unsigned int rxqs);
4586 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4587 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4589 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4590 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4593 int register_netdev(struct net_device *dev);
4594 void unregister_netdev(struct net_device *dev);
4596 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4598 /* General hardware address lists handling functions */
4599 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4600 struct netdev_hw_addr_list *from_list, int addr_len);
4601 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4602 struct netdev_hw_addr_list *from_list, int addr_len);
4603 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4604 struct net_device *dev,
4605 int (*sync)(struct net_device *, const unsigned char *),
4606 int (*unsync)(struct net_device *,
4607 const unsigned char *));
4608 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4609 struct net_device *dev,
4610 int (*sync)(struct net_device *,
4611 const unsigned char *, int),
4612 int (*unsync)(struct net_device *,
4613 const unsigned char *, int));
4614 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4615 struct net_device *dev,
4616 int (*unsync)(struct net_device *,
4617 const unsigned char *, int));
4618 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4619 struct net_device *dev,
4620 int (*unsync)(struct net_device *,
4621 const unsigned char *));
4622 void __hw_addr_init(struct netdev_hw_addr_list *list);
4624 /* Functions used for device addresses handling */
4625 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4626 unsigned char addr_type);
4627 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4628 unsigned char addr_type);
4629 void dev_addr_flush(struct net_device *dev);
4630 int dev_addr_init(struct net_device *dev);
4632 /* Functions used for unicast addresses handling */
4633 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4634 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4635 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4636 int dev_uc_sync(struct net_device *to, struct net_device *from);
4637 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4638 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4639 void dev_uc_flush(struct net_device *dev);
4640 void dev_uc_init(struct net_device *dev);
4643 * __dev_uc_sync - Synchonize device's unicast list
4644 * @dev: device to sync
4645 * @sync: function to call if address should be added
4646 * @unsync: function to call if address should be removed
4648 * Add newly added addresses to the interface, and release
4649 * addresses that have been deleted.
4651 static inline int __dev_uc_sync(struct net_device *dev,
4652 int (*sync)(struct net_device *,
4653 const unsigned char *),
4654 int (*unsync)(struct net_device *,
4655 const unsigned char *))
4657 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4661 * __dev_uc_unsync - Remove synchronized addresses from device
4662 * @dev: device to sync
4663 * @unsync: function to call if address should be removed
4665 * Remove all addresses that were added to the device by dev_uc_sync().
4667 static inline void __dev_uc_unsync(struct net_device *dev,
4668 int (*unsync)(struct net_device *,
4669 const unsigned char *))
4671 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4674 /* Functions used for multicast addresses handling */
4675 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4676 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4677 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4678 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4679 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4680 int dev_mc_sync(struct net_device *to, struct net_device *from);
4681 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4682 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4683 void dev_mc_flush(struct net_device *dev);
4684 void dev_mc_init(struct net_device *dev);
4687 * __dev_mc_sync - Synchonize device's multicast list
4688 * @dev: device to sync
4689 * @sync: function to call if address should be added
4690 * @unsync: function to call if address should be removed
4692 * Add newly added addresses to the interface, and release
4693 * addresses that have been deleted.
4695 static inline int __dev_mc_sync(struct net_device *dev,
4696 int (*sync)(struct net_device *,
4697 const unsigned char *),
4698 int (*unsync)(struct net_device *,
4699 const unsigned char *))
4701 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4705 * __dev_mc_unsync - Remove synchronized addresses from device
4706 * @dev: device to sync
4707 * @unsync: function to call if address should be removed
4709 * Remove all addresses that were added to the device by dev_mc_sync().
4711 static inline void __dev_mc_unsync(struct net_device *dev,
4712 int (*unsync)(struct net_device *,
4713 const unsigned char *))
4715 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4718 /* Functions used for secondary unicast and multicast support */
4719 void dev_set_rx_mode(struct net_device *dev);
4720 void __dev_set_rx_mode(struct net_device *dev);
4721 int dev_set_promiscuity(struct net_device *dev, int inc);
4722 int dev_set_allmulti(struct net_device *dev, int inc);
4723 void netdev_state_change(struct net_device *dev);
4724 void __netdev_notify_peers(struct net_device *dev);
4725 void netdev_notify_peers(struct net_device *dev);
4726 void netdev_features_change(struct net_device *dev);
4727 /* Load a device via the kmod */
4728 void dev_load(struct net *net, const char *name);
4729 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4730 struct rtnl_link_stats64 *storage);
4731 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4732 const struct net_device_stats *netdev_stats);
4733 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4734 const struct pcpu_sw_netstats __percpu *netstats);
4735 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4737 extern int netdev_max_backlog;
4738 extern int netdev_tstamp_prequeue;
4739 extern int netdev_unregister_timeout_secs;
4740 extern int weight_p;
4741 extern int dev_weight_rx_bias;
4742 extern int dev_weight_tx_bias;
4743 extern int dev_rx_weight;
4744 extern int dev_tx_weight;
4745 extern int gro_normal_batch;
4748 NESTED_SYNC_IMM_BIT,
4749 NESTED_SYNC_TODO_BIT,
4752 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4753 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4755 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4756 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4758 struct netdev_nested_priv {
4759 unsigned char flags;
4763 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4764 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4765 struct list_head **iter);
4766 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4767 struct list_head **iter);
4769 #ifdef CONFIG_LOCKDEP
4770 static LIST_HEAD(net_unlink_list);
4772 static inline void net_unlink_todo(struct net_device *dev)
4774 if (list_empty(&dev->unlink_list))
4775 list_add_tail(&dev->unlink_list, &net_unlink_list);
4779 /* iterate through upper list, must be called under RCU read lock */
4780 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4781 for (iter = &(dev)->adj_list.upper, \
4782 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4784 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4786 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4787 int (*fn)(struct net_device *upper_dev,
4788 struct netdev_nested_priv *priv),
4789 struct netdev_nested_priv *priv);
4791 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4792 struct net_device *upper_dev);
4794 bool netdev_has_any_upper_dev(struct net_device *dev);
4796 void *netdev_lower_get_next_private(struct net_device *dev,
4797 struct list_head **iter);
4798 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4799 struct list_head **iter);
4801 #define netdev_for_each_lower_private(dev, priv, iter) \
4802 for (iter = (dev)->adj_list.lower.next, \
4803 priv = netdev_lower_get_next_private(dev, &(iter)); \
4805 priv = netdev_lower_get_next_private(dev, &(iter)))
4807 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4808 for (iter = &(dev)->adj_list.lower, \
4809 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4811 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4813 void *netdev_lower_get_next(struct net_device *dev,
4814 struct list_head **iter);
4816 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4817 for (iter = (dev)->adj_list.lower.next, \
4818 ldev = netdev_lower_get_next(dev, &(iter)); \
4820 ldev = netdev_lower_get_next(dev, &(iter)))
4822 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4823 struct list_head **iter);
4824 int netdev_walk_all_lower_dev(struct net_device *dev,
4825 int (*fn)(struct net_device *lower_dev,
4826 struct netdev_nested_priv *priv),
4827 struct netdev_nested_priv *priv);
4828 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4829 int (*fn)(struct net_device *lower_dev,
4830 struct netdev_nested_priv *priv),
4831 struct netdev_nested_priv *priv);
4833 void *netdev_adjacent_get_private(struct list_head *adj_list);
4834 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4835 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4836 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4837 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4838 struct netlink_ext_ack *extack);
4839 int netdev_master_upper_dev_link(struct net_device *dev,
4840 struct net_device *upper_dev,
4841 void *upper_priv, void *upper_info,
4842 struct netlink_ext_ack *extack);
4843 void netdev_upper_dev_unlink(struct net_device *dev,
4844 struct net_device *upper_dev);
4845 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4846 struct net_device *new_dev,
4847 struct net_device *dev,
4848 struct netlink_ext_ack *extack);
4849 void netdev_adjacent_change_commit(struct net_device *old_dev,
4850 struct net_device *new_dev,
4851 struct net_device *dev);
4852 void netdev_adjacent_change_abort(struct net_device *old_dev,
4853 struct net_device *new_dev,
4854 struct net_device *dev);
4855 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4856 void *netdev_lower_dev_get_private(struct net_device *dev,
4857 struct net_device *lower_dev);
4858 void netdev_lower_state_changed(struct net_device *lower_dev,
4859 void *lower_state_info);
4861 /* RSS keys are 40 or 52 bytes long */
4862 #define NETDEV_RSS_KEY_LEN 52
4863 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4864 void netdev_rss_key_fill(void *buffer, size_t len);
4866 int skb_checksum_help(struct sk_buff *skb);
4867 int skb_crc32c_csum_help(struct sk_buff *skb);
4868 int skb_csum_hwoffload_help(struct sk_buff *skb,
4869 const netdev_features_t features);
4871 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4872 netdev_features_t features, bool tx_path);
4873 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4874 netdev_features_t features);
4876 struct netdev_bonding_info {
4881 struct netdev_notifier_bonding_info {
4882 struct netdev_notifier_info info; /* must be first */
4883 struct netdev_bonding_info bonding_info;
4886 void netdev_bonding_info_change(struct net_device *dev,
4887 struct netdev_bonding_info *bonding_info);
4889 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4890 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4892 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4899 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4901 return __skb_gso_segment(skb, features, true);
4903 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4905 static inline bool can_checksum_protocol(netdev_features_t features,
4908 if (protocol == htons(ETH_P_FCOE))
4909 return !!(features & NETIF_F_FCOE_CRC);
4911 /* Assume this is an IP checksum (not SCTP CRC) */
4913 if (features & NETIF_F_HW_CSUM) {
4914 /* Can checksum everything */
4919 case htons(ETH_P_IP):
4920 return !!(features & NETIF_F_IP_CSUM);
4921 case htons(ETH_P_IPV6):
4922 return !!(features & NETIF_F_IPV6_CSUM);
4929 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4931 static inline void netdev_rx_csum_fault(struct net_device *dev,
4932 struct sk_buff *skb)
4936 /* rx skb timestamps */
4937 void net_enable_timestamp(void);
4938 void net_disable_timestamp(void);
4940 #ifdef CONFIG_PROC_FS
4941 int __init dev_proc_init(void);
4943 #define dev_proc_init() 0
4946 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4947 struct sk_buff *skb, struct net_device *dev,
4950 __this_cpu_write(softnet_data.xmit.more, more);
4951 return ops->ndo_start_xmit(skb, dev);
4954 static inline bool netdev_xmit_more(void)
4956 return __this_cpu_read(softnet_data.xmit.more);
4959 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4960 struct netdev_queue *txq, bool more)
4962 const struct net_device_ops *ops = dev->netdev_ops;
4965 rc = __netdev_start_xmit(ops, skb, dev, more);
4966 if (rc == NETDEV_TX_OK)
4967 txq_trans_update(txq);
4972 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4974 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4977 extern const struct kobj_ns_type_operations net_ns_type_operations;
4979 const char *netdev_drivername(const struct net_device *dev);
4981 void linkwatch_run_queue(void);
4983 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4984 netdev_features_t f2)
4986 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4987 if (f1 & NETIF_F_HW_CSUM)
4988 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4990 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4996 static inline netdev_features_t netdev_get_wanted_features(
4997 struct net_device *dev)
4999 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5001 netdev_features_t netdev_increment_features(netdev_features_t all,
5002 netdev_features_t one, netdev_features_t mask);
5004 /* Allow TSO being used on stacked device :
5005 * Performing the GSO segmentation before last device
5006 * is a performance improvement.
5008 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5009 netdev_features_t mask)
5011 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5014 int __netdev_update_features(struct net_device *dev);
5015 void netdev_update_features(struct net_device *dev);
5016 void netdev_change_features(struct net_device *dev);
5018 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5019 struct net_device *dev);
5021 netdev_features_t passthru_features_check(struct sk_buff *skb,
5022 struct net_device *dev,
5023 netdev_features_t features);
5024 netdev_features_t netif_skb_features(struct sk_buff *skb);
5026 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5028 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5030 /* check flags correspondence */
5031 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5032 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5033 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5034 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5035 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5036 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5037 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5038 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5039 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5040 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5041 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5042 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5043 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5044 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5045 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5046 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5047 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5048 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5049 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5051 return (features & feature) == feature;
5054 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5056 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5057 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5060 static inline bool netif_needs_gso(struct sk_buff *skb,
5061 netdev_features_t features)
5063 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5064 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5065 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5068 static inline void netif_set_gso_max_size(struct net_device *dev,
5071 dev->gso_max_size = size;
5074 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5075 int pulled_hlen, u16 mac_offset,
5078 skb->protocol = protocol;
5079 skb->encapsulation = 1;
5080 skb_push(skb, pulled_hlen);
5081 skb_reset_transport_header(skb);
5082 skb->mac_header = mac_offset;
5083 skb->network_header = skb->mac_header + mac_len;
5084 skb->mac_len = mac_len;
5087 static inline bool netif_is_macsec(const struct net_device *dev)
5089 return dev->priv_flags & IFF_MACSEC;
5092 static inline bool netif_is_macvlan(const struct net_device *dev)
5094 return dev->priv_flags & IFF_MACVLAN;
5097 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5099 return dev->priv_flags & IFF_MACVLAN_PORT;
5102 static inline bool netif_is_bond_master(const struct net_device *dev)
5104 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5107 static inline bool netif_is_bond_slave(const struct net_device *dev)
5109 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5112 static inline bool netif_supports_nofcs(struct net_device *dev)
5114 return dev->priv_flags & IFF_SUPP_NOFCS;
5117 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5119 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5122 static inline bool netif_is_l3_master(const struct net_device *dev)
5124 return dev->priv_flags & IFF_L3MDEV_MASTER;
5127 static inline bool netif_is_l3_slave(const struct net_device *dev)
5129 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5132 static inline bool netif_is_bridge_master(const struct net_device *dev)
5134 return dev->priv_flags & IFF_EBRIDGE;
5137 static inline bool netif_is_bridge_port(const struct net_device *dev)
5139 return dev->priv_flags & IFF_BRIDGE_PORT;
5142 static inline bool netif_is_ovs_master(const struct net_device *dev)
5144 return dev->priv_flags & IFF_OPENVSWITCH;
5147 static inline bool netif_is_ovs_port(const struct net_device *dev)
5149 return dev->priv_flags & IFF_OVS_DATAPATH;
5152 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5154 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5157 static inline bool netif_is_team_master(const struct net_device *dev)
5159 return dev->priv_flags & IFF_TEAM;
5162 static inline bool netif_is_team_port(const struct net_device *dev)
5164 return dev->priv_flags & IFF_TEAM_PORT;
5167 static inline bool netif_is_lag_master(const struct net_device *dev)
5169 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5172 static inline bool netif_is_lag_port(const struct net_device *dev)
5174 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5177 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5179 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5182 static inline bool netif_is_failover(const struct net_device *dev)
5184 return dev->priv_flags & IFF_FAILOVER;
5187 static inline bool netif_is_failover_slave(const struct net_device *dev)
5189 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5192 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5193 static inline void netif_keep_dst(struct net_device *dev)
5195 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5198 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5199 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5201 /* TODO: reserve and use an additional IFF bit, if we get more users */
5202 return dev->priv_flags & IFF_MACSEC;
5205 extern struct pernet_operations __net_initdata loopback_net_ops;
5207 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5209 /* netdev_printk helpers, similar to dev_printk */
5211 static inline const char *netdev_name(const struct net_device *dev)
5213 if (!dev->name[0] || strchr(dev->name, '%'))
5214 return "(unnamed net_device)";
5218 static inline bool netdev_unregistering(const struct net_device *dev)
5220 return dev->reg_state == NETREG_UNREGISTERING;
5223 static inline const char *netdev_reg_state(const struct net_device *dev)
5225 switch (dev->reg_state) {
5226 case NETREG_UNINITIALIZED: return " (uninitialized)";
5227 case NETREG_REGISTERED: return "";
5228 case NETREG_UNREGISTERING: return " (unregistering)";
5229 case NETREG_UNREGISTERED: return " (unregistered)";
5230 case NETREG_RELEASED: return " (released)";
5231 case NETREG_DUMMY: return " (dummy)";
5234 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5235 return " (unknown)";
5238 __printf(3, 4) __cold
5239 void netdev_printk(const char *level, const struct net_device *dev,
5240 const char *format, ...);
5241 __printf(2, 3) __cold
5242 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5243 __printf(2, 3) __cold
5244 void netdev_alert(const struct net_device *dev, const char *format, ...);
5245 __printf(2, 3) __cold
5246 void netdev_crit(const struct net_device *dev, const char *format, ...);
5247 __printf(2, 3) __cold
5248 void netdev_err(const struct net_device *dev, const char *format, ...);
5249 __printf(2, 3) __cold
5250 void netdev_warn(const struct net_device *dev, const char *format, ...);
5251 __printf(2, 3) __cold
5252 void netdev_notice(const struct net_device *dev, const char *format, ...);
5253 __printf(2, 3) __cold
5254 void netdev_info(const struct net_device *dev, const char *format, ...);
5256 #define netdev_level_once(level, dev, fmt, ...) \
5258 static bool __print_once __read_mostly; \
5260 if (!__print_once) { \
5261 __print_once = true; \
5262 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5266 #define netdev_emerg_once(dev, fmt, ...) \
5267 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5268 #define netdev_alert_once(dev, fmt, ...) \
5269 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5270 #define netdev_crit_once(dev, fmt, ...) \
5271 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5272 #define netdev_err_once(dev, fmt, ...) \
5273 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5274 #define netdev_warn_once(dev, fmt, ...) \
5275 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5276 #define netdev_notice_once(dev, fmt, ...) \
5277 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5278 #define netdev_info_once(dev, fmt, ...) \
5279 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5281 #define MODULE_ALIAS_NETDEV(device) \
5282 MODULE_ALIAS("netdev-" device)
5284 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5285 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5286 #define netdev_dbg(__dev, format, args...) \
5288 dynamic_netdev_dbg(__dev, format, ##args); \
5290 #elif defined(DEBUG)
5291 #define netdev_dbg(__dev, format, args...) \
5292 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5294 #define netdev_dbg(__dev, format, args...) \
5297 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5301 #if defined(VERBOSE_DEBUG)
5302 #define netdev_vdbg netdev_dbg
5305 #define netdev_vdbg(dev, format, args...) \
5308 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5314 * netdev_WARN() acts like dev_printk(), but with the key difference
5315 * of using a WARN/WARN_ON to get the message out, including the
5316 * file/line information and a backtrace.
5318 #define netdev_WARN(dev, format, args...) \
5319 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5320 netdev_reg_state(dev), ##args)
5322 #define netdev_WARN_ONCE(dev, format, args...) \
5323 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5324 netdev_reg_state(dev), ##args)
5326 /* netif printk helpers, similar to netdev_printk */
5328 #define netif_printk(priv, type, level, dev, fmt, args...) \
5330 if (netif_msg_##type(priv)) \
5331 netdev_printk(level, (dev), fmt, ##args); \
5334 #define netif_level(level, priv, type, dev, fmt, args...) \
5336 if (netif_msg_##type(priv)) \
5337 netdev_##level(dev, fmt, ##args); \
5340 #define netif_emerg(priv, type, dev, fmt, args...) \
5341 netif_level(emerg, priv, type, dev, fmt, ##args)
5342 #define netif_alert(priv, type, dev, fmt, args...) \
5343 netif_level(alert, priv, type, dev, fmt, ##args)
5344 #define netif_crit(priv, type, dev, fmt, args...) \
5345 netif_level(crit, priv, type, dev, fmt, ##args)
5346 #define netif_err(priv, type, dev, fmt, args...) \
5347 netif_level(err, priv, type, dev, fmt, ##args)
5348 #define netif_warn(priv, type, dev, fmt, args...) \
5349 netif_level(warn, priv, type, dev, fmt, ##args)
5350 #define netif_notice(priv, type, dev, fmt, args...) \
5351 netif_level(notice, priv, type, dev, fmt, ##args)
5352 #define netif_info(priv, type, dev, fmt, args...) \
5353 netif_level(info, priv, type, dev, fmt, ##args)
5355 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5356 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5357 #define netif_dbg(priv, type, netdev, format, args...) \
5359 if (netif_msg_##type(priv)) \
5360 dynamic_netdev_dbg(netdev, format, ##args); \
5362 #elif defined(DEBUG)
5363 #define netif_dbg(priv, type, dev, format, args...) \
5364 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5366 #define netif_dbg(priv, type, dev, format, args...) \
5369 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5374 /* if @cond then downgrade to debug, else print at @level */
5375 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5378 netif_dbg(priv, type, netdev, fmt, ##args); \
5380 netif_ ## level(priv, type, netdev, fmt, ##args); \
5383 #if defined(VERBOSE_DEBUG)
5384 #define netif_vdbg netif_dbg
5386 #define netif_vdbg(priv, type, dev, format, args...) \
5389 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5395 * The list of packet types we will receive (as opposed to discard)
5396 * and the routines to invoke.
5398 * Why 16. Because with 16 the only overlap we get on a hash of the
5399 * low nibble of the protocol value is RARP/SNAP/X.25.
5413 #define PTYPE_HASH_SIZE (16)
5414 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5416 extern struct list_head ptype_all __read_mostly;
5417 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5419 extern struct net_device *blackhole_netdev;
5421 #endif /* _LINUX_NETDEVICE_H */