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 * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1322 * struct xdp_buff *xdp);
1323 * Get the xmit slave of master device based on the xdp_buff.
1324 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1325 * This function is used to wake up the softirq, ksoftirqd or kthread
1326 * responsible for sending and/or receiving packets on a specific
1327 * queue id bound to an AF_XDP socket. The flags field specifies if
1328 * only RX, only Tx, or both should be woken up using the flags
1329 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1330 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1331 * Get devlink port instance associated with a given netdev.
1332 * Called with a reference on the netdevice and devlink locks only,
1333 * rtnl_lock is not held.
1334 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1336 * Add, change, delete or get information on an IPv4 tunnel.
1337 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1338 * If a device is paired with a peer device, return the peer instance.
1339 * The caller must be under RCU read context.
1340 * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1341 * Get the forwarding path to reach the real device from the HW destination address
1343 struct net_device_ops {
1344 int (*ndo_init)(struct net_device *dev);
1345 void (*ndo_uninit)(struct net_device *dev);
1346 int (*ndo_open)(struct net_device *dev);
1347 int (*ndo_stop)(struct net_device *dev);
1348 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1349 struct net_device *dev);
1350 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1351 struct net_device *dev,
1352 netdev_features_t features);
1353 u16 (*ndo_select_queue)(struct net_device *dev,
1354 struct sk_buff *skb,
1355 struct net_device *sb_dev);
1356 void (*ndo_change_rx_flags)(struct net_device *dev,
1358 void (*ndo_set_rx_mode)(struct net_device *dev);
1359 int (*ndo_set_mac_address)(struct net_device *dev,
1361 int (*ndo_validate_addr)(struct net_device *dev);
1362 int (*ndo_do_ioctl)(struct net_device *dev,
1363 struct ifreq *ifr, int cmd);
1364 int (*ndo_eth_ioctl)(struct net_device *dev,
1365 struct ifreq *ifr, int cmd);
1366 int (*ndo_siocbond)(struct net_device *dev,
1367 struct ifreq *ifr, int cmd);
1368 int (*ndo_siocwandev)(struct net_device *dev,
1369 struct if_settings *ifs);
1370 int (*ndo_siocdevprivate)(struct net_device *dev,
1372 void __user *data, int cmd);
1373 int (*ndo_set_config)(struct net_device *dev,
1375 int (*ndo_change_mtu)(struct net_device *dev,
1377 int (*ndo_neigh_setup)(struct net_device *dev,
1378 struct neigh_parms *);
1379 void (*ndo_tx_timeout) (struct net_device *dev,
1380 unsigned int txqueue);
1382 void (*ndo_get_stats64)(struct net_device *dev,
1383 struct rtnl_link_stats64 *storage);
1384 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1385 int (*ndo_get_offload_stats)(int attr_id,
1386 const struct net_device *dev,
1388 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1390 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1391 __be16 proto, u16 vid);
1392 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1393 __be16 proto, u16 vid);
1394 #ifdef CONFIG_NET_POLL_CONTROLLER
1395 void (*ndo_poll_controller)(struct net_device *dev);
1396 int (*ndo_netpoll_setup)(struct net_device *dev,
1397 struct netpoll_info *info);
1398 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1400 int (*ndo_set_vf_mac)(struct net_device *dev,
1401 int queue, u8 *mac);
1402 int (*ndo_set_vf_vlan)(struct net_device *dev,
1403 int queue, u16 vlan,
1404 u8 qos, __be16 proto);
1405 int (*ndo_set_vf_rate)(struct net_device *dev,
1406 int vf, int min_tx_rate,
1408 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1409 int vf, bool setting);
1410 int (*ndo_set_vf_trust)(struct net_device *dev,
1411 int vf, bool setting);
1412 int (*ndo_get_vf_config)(struct net_device *dev,
1414 struct ifla_vf_info *ivf);
1415 int (*ndo_set_vf_link_state)(struct net_device *dev,
1416 int vf, int link_state);
1417 int (*ndo_get_vf_stats)(struct net_device *dev,
1419 struct ifla_vf_stats
1421 int (*ndo_set_vf_port)(struct net_device *dev,
1423 struct nlattr *port[]);
1424 int (*ndo_get_vf_port)(struct net_device *dev,
1425 int vf, struct sk_buff *skb);
1426 int (*ndo_get_vf_guid)(struct net_device *dev,
1428 struct ifla_vf_guid *node_guid,
1429 struct ifla_vf_guid *port_guid);
1430 int (*ndo_set_vf_guid)(struct net_device *dev,
1433 int (*ndo_set_vf_rss_query_en)(
1434 struct net_device *dev,
1435 int vf, bool setting);
1436 int (*ndo_setup_tc)(struct net_device *dev,
1437 enum tc_setup_type type,
1439 #if IS_ENABLED(CONFIG_FCOE)
1440 int (*ndo_fcoe_enable)(struct net_device *dev);
1441 int (*ndo_fcoe_disable)(struct net_device *dev);
1442 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1444 struct scatterlist *sgl,
1446 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1448 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1450 struct scatterlist *sgl,
1452 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1453 struct netdev_fcoe_hbainfo *hbainfo);
1456 #if IS_ENABLED(CONFIG_LIBFCOE)
1457 #define NETDEV_FCOE_WWNN 0
1458 #define NETDEV_FCOE_WWPN 1
1459 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1460 u64 *wwn, int type);
1463 #ifdef CONFIG_RFS_ACCEL
1464 int (*ndo_rx_flow_steer)(struct net_device *dev,
1465 const struct sk_buff *skb,
1469 int (*ndo_add_slave)(struct net_device *dev,
1470 struct net_device *slave_dev,
1471 struct netlink_ext_ack *extack);
1472 int (*ndo_del_slave)(struct net_device *dev,
1473 struct net_device *slave_dev);
1474 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1475 struct sk_buff *skb,
1477 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1479 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1480 netdev_features_t features);
1481 int (*ndo_set_features)(struct net_device *dev,
1482 netdev_features_t features);
1483 int (*ndo_neigh_construct)(struct net_device *dev,
1484 struct neighbour *n);
1485 void (*ndo_neigh_destroy)(struct net_device *dev,
1486 struct neighbour *n);
1488 int (*ndo_fdb_add)(struct ndmsg *ndm,
1489 struct nlattr *tb[],
1490 struct net_device *dev,
1491 const unsigned char *addr,
1494 struct netlink_ext_ack *extack);
1495 int (*ndo_fdb_del)(struct ndmsg *ndm,
1496 struct nlattr *tb[],
1497 struct net_device *dev,
1498 const unsigned char *addr,
1500 int (*ndo_fdb_dump)(struct sk_buff *skb,
1501 struct netlink_callback *cb,
1502 struct net_device *dev,
1503 struct net_device *filter_dev,
1505 int (*ndo_fdb_get)(struct sk_buff *skb,
1506 struct nlattr *tb[],
1507 struct net_device *dev,
1508 const unsigned char *addr,
1509 u16 vid, u32 portid, u32 seq,
1510 struct netlink_ext_ack *extack);
1511 int (*ndo_bridge_setlink)(struct net_device *dev,
1512 struct nlmsghdr *nlh,
1514 struct netlink_ext_ack *extack);
1515 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1517 struct net_device *dev,
1520 int (*ndo_bridge_dellink)(struct net_device *dev,
1521 struct nlmsghdr *nlh,
1523 int (*ndo_change_carrier)(struct net_device *dev,
1525 int (*ndo_get_phys_port_id)(struct net_device *dev,
1526 struct netdev_phys_item_id *ppid);
1527 int (*ndo_get_port_parent_id)(struct net_device *dev,
1528 struct netdev_phys_item_id *ppid);
1529 int (*ndo_get_phys_port_name)(struct net_device *dev,
1530 char *name, size_t len);
1531 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1532 struct net_device *dev);
1533 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1536 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1539 int (*ndo_get_iflink)(const struct net_device *dev);
1540 int (*ndo_change_proto_down)(struct net_device *dev,
1542 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1543 struct sk_buff *skb);
1544 void (*ndo_set_rx_headroom)(struct net_device *dev,
1545 int needed_headroom);
1546 int (*ndo_bpf)(struct net_device *dev,
1547 struct netdev_bpf *bpf);
1548 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1549 struct xdp_frame **xdp,
1551 struct net_device * (*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1552 struct xdp_buff *xdp);
1553 int (*ndo_xsk_wakeup)(struct net_device *dev,
1554 u32 queue_id, u32 flags);
1555 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1556 int (*ndo_tunnel_ctl)(struct net_device *dev,
1557 struct ip_tunnel_parm *p, int cmd);
1558 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1559 int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1560 struct net_device_path *path);
1564 * enum netdev_priv_flags - &struct net_device priv_flags
1566 * These are the &struct net_device, they are only set internally
1567 * by drivers and used in the kernel. These flags are invisible to
1568 * userspace; this means that the order of these flags can change
1569 * during any kernel release.
1571 * You should have a pretty good reason to be extending these flags.
1573 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1574 * @IFF_EBRIDGE: Ethernet bridging device
1575 * @IFF_BONDING: bonding master or slave
1576 * @IFF_ISATAP: ISATAP interface (RFC4214)
1577 * @IFF_WAN_HDLC: WAN HDLC device
1578 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1580 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1581 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1582 * @IFF_MACVLAN_PORT: device used as macvlan port
1583 * @IFF_BRIDGE_PORT: device used as bridge port
1584 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1585 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1586 * @IFF_UNICAST_FLT: Supports unicast filtering
1587 * @IFF_TEAM_PORT: device used as team port
1588 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1589 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1590 * change when it's running
1591 * @IFF_MACVLAN: Macvlan device
1592 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1593 * underlying stacked devices
1594 * @IFF_L3MDEV_MASTER: device is an L3 master device
1595 * @IFF_NO_QUEUE: device can run without qdisc attached
1596 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1597 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1598 * @IFF_TEAM: device is a team device
1599 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1600 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1601 * entity (i.e. the master device for bridged veth)
1602 * @IFF_MACSEC: device is a MACsec device
1603 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1604 * @IFF_FAILOVER: device is a failover master device
1605 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1606 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1607 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1608 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1609 * skb_headlen(skb) == 0 (data starts from frag0)
1611 enum netdev_priv_flags {
1612 IFF_802_1Q_VLAN = 1<<0,
1616 IFF_WAN_HDLC = 1<<4,
1617 IFF_XMIT_DST_RELEASE = 1<<5,
1618 IFF_DONT_BRIDGE = 1<<6,
1619 IFF_DISABLE_NETPOLL = 1<<7,
1620 IFF_MACVLAN_PORT = 1<<8,
1621 IFF_BRIDGE_PORT = 1<<9,
1622 IFF_OVS_DATAPATH = 1<<10,
1623 IFF_TX_SKB_SHARING = 1<<11,
1624 IFF_UNICAST_FLT = 1<<12,
1625 IFF_TEAM_PORT = 1<<13,
1626 IFF_SUPP_NOFCS = 1<<14,
1627 IFF_LIVE_ADDR_CHANGE = 1<<15,
1628 IFF_MACVLAN = 1<<16,
1629 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1630 IFF_L3MDEV_MASTER = 1<<18,
1631 IFF_NO_QUEUE = 1<<19,
1632 IFF_OPENVSWITCH = 1<<20,
1633 IFF_L3MDEV_SLAVE = 1<<21,
1635 IFF_RXFH_CONFIGURED = 1<<23,
1636 IFF_PHONY_HEADROOM = 1<<24,
1638 IFF_NO_RX_HANDLER = 1<<26,
1639 IFF_FAILOVER = 1<<27,
1640 IFF_FAILOVER_SLAVE = 1<<28,
1641 IFF_L3MDEV_RX_HANDLER = 1<<29,
1642 IFF_LIVE_RENAME_OK = 1<<30,
1643 IFF_TX_SKB_NO_LINEAR = 1<<31,
1646 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1647 #define IFF_EBRIDGE IFF_EBRIDGE
1648 #define IFF_BONDING IFF_BONDING
1649 #define IFF_ISATAP IFF_ISATAP
1650 #define IFF_WAN_HDLC IFF_WAN_HDLC
1651 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1652 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1653 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1654 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1655 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1656 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1657 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1658 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1659 #define IFF_TEAM_PORT IFF_TEAM_PORT
1660 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1661 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1662 #define IFF_MACVLAN IFF_MACVLAN
1663 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1664 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1665 #define IFF_NO_QUEUE IFF_NO_QUEUE
1666 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1667 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1668 #define IFF_TEAM IFF_TEAM
1669 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1670 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1671 #define IFF_MACSEC IFF_MACSEC
1672 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1673 #define IFF_FAILOVER IFF_FAILOVER
1674 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1675 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1676 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1677 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1679 /* Specifies the type of the struct net_device::ml_priv pointer */
1680 enum netdev_ml_priv_type {
1686 * struct net_device - The DEVICE structure.
1688 * Actually, this whole structure is a big mistake. It mixes I/O
1689 * data with strictly "high-level" data, and it has to know about
1690 * almost every data structure used in the INET module.
1692 * @name: This is the first field of the "visible" part of this structure
1693 * (i.e. as seen by users in the "Space.c" file). It is the name
1696 * @name_node: Name hashlist node
1697 * @ifalias: SNMP alias
1698 * @mem_end: Shared memory end
1699 * @mem_start: Shared memory start
1700 * @base_addr: Device I/O address
1701 * @irq: Device IRQ number
1703 * @state: Generic network queuing layer state, see netdev_state_t
1704 * @dev_list: The global list of network devices
1705 * @napi_list: List entry used for polling NAPI devices
1706 * @unreg_list: List entry when we are unregistering the
1707 * device; see the function unregister_netdev
1708 * @close_list: List entry used when we are closing the device
1709 * @ptype_all: Device-specific packet handlers for all protocols
1710 * @ptype_specific: Device-specific, protocol-specific packet handlers
1712 * @adj_list: Directly linked devices, like slaves for bonding
1713 * @features: Currently active device features
1714 * @hw_features: User-changeable features
1716 * @wanted_features: User-requested features
1717 * @vlan_features: Mask of features inheritable by VLAN devices
1719 * @hw_enc_features: Mask of features inherited by encapsulating devices
1720 * This field indicates what encapsulation
1721 * offloads the hardware is capable of doing,
1722 * and drivers will need to set them appropriately.
1724 * @mpls_features: Mask of features inheritable by MPLS
1725 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1727 * @ifindex: interface index
1728 * @group: The group the device belongs to
1730 * @stats: Statistics struct, which was left as a legacy, use
1731 * rtnl_link_stats64 instead
1733 * @rx_dropped: Dropped packets by core network,
1734 * do not use this in drivers
1735 * @tx_dropped: Dropped packets by core network,
1736 * do not use this in drivers
1737 * @rx_nohandler: nohandler dropped packets by core network on
1738 * inactive devices, do not use this in drivers
1739 * @carrier_up_count: Number of times the carrier has been up
1740 * @carrier_down_count: Number of times the carrier has been down
1742 * @wireless_handlers: List of functions to handle Wireless Extensions,
1744 * see <net/iw_handler.h> for details.
1745 * @wireless_data: Instance data managed by the core of wireless extensions
1747 * @netdev_ops: Includes several pointers to callbacks,
1748 * if one wants to override the ndo_*() functions
1749 * @ethtool_ops: Management operations
1750 * @l3mdev_ops: Layer 3 master device operations
1751 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1752 * discovery handling. Necessary for e.g. 6LoWPAN.
1753 * @xfrmdev_ops: Transformation offload operations
1754 * @tlsdev_ops: Transport Layer Security offload operations
1755 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1756 * of Layer 2 headers.
1758 * @flags: Interface flags (a la BSD)
1759 * @priv_flags: Like 'flags' but invisible to userspace,
1760 * see if.h for the definitions
1761 * @gflags: Global flags ( kept as legacy )
1762 * @padded: How much padding added by alloc_netdev()
1763 * @operstate: RFC2863 operstate
1764 * @link_mode: Mapping policy to operstate
1765 * @if_port: Selectable AUI, TP, ...
1767 * @mtu: Interface MTU value
1768 * @min_mtu: Interface Minimum MTU value
1769 * @max_mtu: Interface Maximum MTU value
1770 * @type: Interface hardware type
1771 * @hard_header_len: Maximum hardware header length.
1772 * @min_header_len: Minimum hardware header length
1774 * @needed_headroom: Extra headroom the hardware may need, but not in all
1775 * cases can this be guaranteed
1776 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1777 * cases can this be guaranteed. Some cases also use
1778 * LL_MAX_HEADER instead to allocate the skb
1780 * interface address info:
1782 * @perm_addr: Permanent hw address
1783 * @addr_assign_type: Hw address assignment type
1784 * @addr_len: Hardware address length
1785 * @upper_level: Maximum depth level of upper devices.
1786 * @lower_level: Maximum depth level of lower devices.
1787 * @neigh_priv_len: Used in neigh_alloc()
1788 * @dev_id: Used to differentiate devices that share
1789 * the same link layer address
1790 * @dev_port: Used to differentiate devices that share
1792 * @addr_list_lock: XXX: need comments on this one
1793 * @name_assign_type: network interface name assignment type
1794 * @uc_promisc: Counter that indicates promiscuous mode
1795 * has been enabled due to the need to listen to
1796 * additional unicast addresses in a device that
1797 * does not implement ndo_set_rx_mode()
1798 * @uc: unicast mac addresses
1799 * @mc: multicast mac addresses
1800 * @dev_addrs: list of device hw addresses
1801 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1802 * @promiscuity: Number of times the NIC is told to work in
1803 * promiscuous mode; if it becomes 0 the NIC will
1804 * exit promiscuous mode
1805 * @allmulti: Counter, enables or disables allmulticast mode
1807 * @vlan_info: VLAN info
1808 * @dsa_ptr: dsa specific data
1809 * @tipc_ptr: TIPC specific data
1810 * @atalk_ptr: AppleTalk link
1811 * @ip_ptr: IPv4 specific data
1812 * @dn_ptr: DECnet specific data
1813 * @ip6_ptr: IPv6 specific data
1814 * @ax25_ptr: AX.25 specific data
1815 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1816 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1818 * @mpls_ptr: mpls_dev struct pointer
1819 * @mctp_ptr: MCTP specific data
1821 * @dev_addr: Hw address (before bcast,
1822 * because most packets are unicast)
1824 * @_rx: Array of RX queues
1825 * @num_rx_queues: Number of RX queues
1826 * allocated at register_netdev() time
1827 * @real_num_rx_queues: Number of RX queues currently active in device
1828 * @xdp_prog: XDP sockets filter program pointer
1829 * @gro_flush_timeout: timeout for GRO layer in NAPI
1830 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1831 * allow to avoid NIC hard IRQ, on busy queues.
1833 * @rx_handler: handler for received packets
1834 * @rx_handler_data: XXX: need comments on this one
1835 * @miniq_ingress: ingress/clsact qdisc specific data for
1836 * ingress processing
1837 * @ingress_queue: XXX: need comments on this one
1838 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1839 * @broadcast: hw bcast address
1841 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1842 * indexed by RX queue number. Assigned by driver.
1843 * This must only be set if the ndo_rx_flow_steer
1844 * operation is defined
1845 * @index_hlist: Device index hash chain
1847 * @_tx: Array of TX queues
1848 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1849 * @real_num_tx_queues: Number of TX queues currently active in device
1850 * @qdisc: Root qdisc from userspace point of view
1851 * @tx_queue_len: Max frames per queue allowed
1852 * @tx_global_lock: XXX: need comments on this one
1853 * @xdp_bulkq: XDP device bulk queue
1854 * @xps_maps: all CPUs/RXQs maps for XPS device
1856 * @xps_maps: XXX: need comments on this one
1857 * @miniq_egress: clsact qdisc specific data for
1859 * @qdisc_hash: qdisc hash table
1860 * @watchdog_timeo: Represents the timeout that is used by
1861 * the watchdog (see dev_watchdog())
1862 * @watchdog_timer: List of timers
1864 * @proto_down_reason: reason a netdev interface is held down
1865 * @pcpu_refcnt: Number of references to this device
1866 * @dev_refcnt: Number of references to this device
1867 * @todo_list: Delayed register/unregister
1868 * @link_watch_list: XXX: need comments on this one
1870 * @reg_state: Register/unregister state machine
1871 * @dismantle: Device is going to be freed
1872 * @rtnl_link_state: This enum represents the phases of creating
1875 * @needs_free_netdev: Should unregister perform free_netdev?
1876 * @priv_destructor: Called from unregister
1877 * @npinfo: XXX: need comments on this one
1878 * @nd_net: Network namespace this network device is inside
1880 * @ml_priv: Mid-layer private
1881 * @ml_priv_type: Mid-layer private type
1882 * @lstats: Loopback statistics
1883 * @tstats: Tunnel statistics
1884 * @dstats: Dummy statistics
1885 * @vstats: Virtual ethernet statistics
1890 * @dev: Class/net/name entry
1891 * @sysfs_groups: Space for optional device, statistics and wireless
1894 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1895 * @rtnl_link_ops: Rtnl_link_ops
1897 * @gso_max_size: Maximum size of generic segmentation offload
1898 * @gso_max_segs: Maximum number of segments that can be passed to the
1901 * @dcbnl_ops: Data Center Bridging netlink ops
1902 * @num_tc: Number of traffic classes in the net device
1903 * @tc_to_txq: XXX: need comments on this one
1904 * @prio_tc_map: XXX: need comments on this one
1906 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1908 * @priomap: XXX: need comments on this one
1909 * @phydev: Physical device may attach itself
1910 * for hardware timestamping
1911 * @sfp_bus: attached &struct sfp_bus structure.
1913 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1914 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1916 * @proto_down: protocol port state information can be sent to the
1917 * switch driver and used to set the phys state of the
1920 * @wol_enabled: Wake-on-LAN is enabled
1922 * @threaded: napi threaded mode is enabled
1924 * @net_notifier_list: List of per-net netdev notifier block
1925 * that follow this device when it is moved
1926 * to another network namespace.
1928 * @macsec_ops: MACsec offloading ops
1930 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1931 * offload capabilities of the device
1932 * @udp_tunnel_nic: UDP tunnel offload state
1933 * @xdp_state: stores info on attached XDP BPF programs
1935 * @nested_level: Used as as a parameter of spin_lock_nested() of
1936 * dev->addr_list_lock.
1937 * @unlink_list: As netif_addr_lock() can be called recursively,
1938 * keep a list of interfaces to be deleted.
1940 * FIXME: cleanup struct net_device such that network protocol info
1945 char name[IFNAMSIZ];
1946 struct netdev_name_node *name_node;
1947 struct dev_ifalias __rcu *ifalias;
1949 * I/O specific fields
1950 * FIXME: Merge these and struct ifmap into one
1952 unsigned long mem_end;
1953 unsigned long mem_start;
1954 unsigned long base_addr;
1957 * Some hardware also needs these fields (state,dev_list,
1958 * napi_list,unreg_list,close_list) but they are not
1959 * part of the usual set specified in Space.c.
1962 unsigned long state;
1964 struct list_head dev_list;
1965 struct list_head napi_list;
1966 struct list_head unreg_list;
1967 struct list_head close_list;
1968 struct list_head ptype_all;
1969 struct list_head ptype_specific;
1972 struct list_head upper;
1973 struct list_head lower;
1976 /* Read-mostly cache-line for fast-path access */
1978 unsigned int priv_flags;
1979 const struct net_device_ops *netdev_ops;
1981 unsigned short gflags;
1982 unsigned short hard_header_len;
1984 /* Note : dev->mtu is often read without holding a lock.
1985 * Writers usually hold RTNL.
1986 * It is recommended to use READ_ONCE() to annotate the reads,
1987 * and to use WRITE_ONCE() to annotate the writes.
1990 unsigned short needed_headroom;
1991 unsigned short needed_tailroom;
1993 netdev_features_t features;
1994 netdev_features_t hw_features;
1995 netdev_features_t wanted_features;
1996 netdev_features_t vlan_features;
1997 netdev_features_t hw_enc_features;
1998 netdev_features_t mpls_features;
1999 netdev_features_t gso_partial_features;
2001 unsigned int min_mtu;
2002 unsigned int max_mtu;
2003 unsigned short type;
2004 unsigned char min_header_len;
2005 unsigned char name_assign_type;
2009 struct net_device_stats stats; /* not used by modern drivers */
2011 atomic_long_t rx_dropped;
2012 atomic_long_t tx_dropped;
2013 atomic_long_t rx_nohandler;
2015 /* Stats to monitor link on/off, flapping */
2016 atomic_t carrier_up_count;
2017 atomic_t carrier_down_count;
2019 #ifdef CONFIG_WIRELESS_EXT
2020 const struct iw_handler_def *wireless_handlers;
2021 struct iw_public_data *wireless_data;
2023 const struct ethtool_ops *ethtool_ops;
2024 #ifdef CONFIG_NET_L3_MASTER_DEV
2025 const struct l3mdev_ops *l3mdev_ops;
2027 #if IS_ENABLED(CONFIG_IPV6)
2028 const struct ndisc_ops *ndisc_ops;
2031 #ifdef CONFIG_XFRM_OFFLOAD
2032 const struct xfrmdev_ops *xfrmdev_ops;
2035 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2036 const struct tlsdev_ops *tlsdev_ops;
2039 const struct header_ops *header_ops;
2041 unsigned char operstate;
2042 unsigned char link_mode;
2044 unsigned char if_port;
2047 /* Interface address info. */
2048 unsigned char perm_addr[MAX_ADDR_LEN];
2049 unsigned char addr_assign_type;
2050 unsigned char addr_len;
2051 unsigned char upper_level;
2052 unsigned char lower_level;
2054 unsigned short neigh_priv_len;
2055 unsigned short dev_id;
2056 unsigned short dev_port;
2057 unsigned short padded;
2059 spinlock_t addr_list_lock;
2062 struct netdev_hw_addr_list uc;
2063 struct netdev_hw_addr_list mc;
2064 struct netdev_hw_addr_list dev_addrs;
2067 struct kset *queues_kset;
2069 #ifdef CONFIG_LOCKDEP
2070 struct list_head unlink_list;
2072 unsigned int promiscuity;
2073 unsigned int allmulti;
2075 #ifdef CONFIG_LOCKDEP
2076 unsigned char nested_level;
2080 /* Protocol-specific pointers */
2082 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2083 struct vlan_info __rcu *vlan_info;
2085 #if IS_ENABLED(CONFIG_NET_DSA)
2086 struct dsa_port *dsa_ptr;
2088 #if IS_ENABLED(CONFIG_TIPC)
2089 struct tipc_bearer __rcu *tipc_ptr;
2091 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2094 struct in_device __rcu *ip_ptr;
2095 #if IS_ENABLED(CONFIG_DECNET)
2096 struct dn_dev __rcu *dn_ptr;
2098 struct inet6_dev __rcu *ip6_ptr;
2099 #if IS_ENABLED(CONFIG_AX25)
2102 struct wireless_dev *ieee80211_ptr;
2103 struct wpan_dev *ieee802154_ptr;
2104 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2105 struct mpls_dev __rcu *mpls_ptr;
2107 #if IS_ENABLED(CONFIG_MCTP)
2108 struct mctp_dev __rcu *mctp_ptr;
2112 * Cache lines mostly used on receive path (including eth_type_trans())
2114 /* Interface address info used in eth_type_trans() */
2115 unsigned char *dev_addr;
2117 struct netdev_rx_queue *_rx;
2118 unsigned int num_rx_queues;
2119 unsigned int real_num_rx_queues;
2121 struct bpf_prog __rcu *xdp_prog;
2122 unsigned long gro_flush_timeout;
2123 int napi_defer_hard_irqs;
2124 rx_handler_func_t __rcu *rx_handler;
2125 void __rcu *rx_handler_data;
2127 #ifdef CONFIG_NET_CLS_ACT
2128 struct mini_Qdisc __rcu *miniq_ingress;
2130 struct netdev_queue __rcu *ingress_queue;
2131 #ifdef CONFIG_NETFILTER_INGRESS
2132 struct nf_hook_entries __rcu *nf_hooks_ingress;
2135 unsigned char broadcast[MAX_ADDR_LEN];
2136 #ifdef CONFIG_RFS_ACCEL
2137 struct cpu_rmap *rx_cpu_rmap;
2139 struct hlist_node index_hlist;
2142 * Cache lines mostly used on transmit path
2144 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2145 unsigned int num_tx_queues;
2146 unsigned int real_num_tx_queues;
2147 struct Qdisc *qdisc;
2148 unsigned int tx_queue_len;
2149 spinlock_t tx_global_lock;
2151 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2154 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2156 #ifdef CONFIG_NET_CLS_ACT
2157 struct mini_Qdisc __rcu *miniq_egress;
2160 #ifdef CONFIG_NET_SCHED
2161 DECLARE_HASHTABLE (qdisc_hash, 4);
2163 /* These may be needed for future network-power-down code. */
2164 struct timer_list watchdog_timer;
2167 u32 proto_down_reason;
2169 struct list_head todo_list;
2171 #ifdef CONFIG_PCPU_DEV_REFCNT
2172 int __percpu *pcpu_refcnt;
2174 refcount_t dev_refcnt;
2177 struct list_head link_watch_list;
2179 enum { NETREG_UNINITIALIZED=0,
2180 NETREG_REGISTERED, /* completed register_netdevice */
2181 NETREG_UNREGISTERING, /* called unregister_netdevice */
2182 NETREG_UNREGISTERED, /* completed unregister todo */
2183 NETREG_RELEASED, /* called free_netdev */
2184 NETREG_DUMMY, /* dummy device for NAPI poll */
2190 RTNL_LINK_INITIALIZED,
2191 RTNL_LINK_INITIALIZING,
2192 } rtnl_link_state:16;
2194 bool needs_free_netdev;
2195 void (*priv_destructor)(struct net_device *dev);
2197 #ifdef CONFIG_NETPOLL
2198 struct netpoll_info __rcu *npinfo;
2201 possible_net_t nd_net;
2203 /* mid-layer private */
2205 enum netdev_ml_priv_type ml_priv_type;
2208 struct pcpu_lstats __percpu *lstats;
2209 struct pcpu_sw_netstats __percpu *tstats;
2210 struct pcpu_dstats __percpu *dstats;
2213 #if IS_ENABLED(CONFIG_GARP)
2214 struct garp_port __rcu *garp_port;
2216 #if IS_ENABLED(CONFIG_MRP)
2217 struct mrp_port __rcu *mrp_port;
2221 const struct attribute_group *sysfs_groups[4];
2222 const struct attribute_group *sysfs_rx_queue_group;
2224 const struct rtnl_link_ops *rtnl_link_ops;
2226 /* for setting kernel sock attribute on TCP connection setup */
2227 #define GSO_MAX_SIZE 65536
2228 unsigned int gso_max_size;
2229 #define GSO_MAX_SEGS 65535
2233 const struct dcbnl_rtnl_ops *dcbnl_ops;
2236 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2237 u8 prio_tc_map[TC_BITMASK + 1];
2239 #if IS_ENABLED(CONFIG_FCOE)
2240 unsigned int fcoe_ddp_xid;
2242 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2243 struct netprio_map __rcu *priomap;
2245 struct phy_device *phydev;
2246 struct sfp_bus *sfp_bus;
2247 struct lock_class_key *qdisc_tx_busylock;
2248 struct lock_class_key *qdisc_running_key;
2250 unsigned wol_enabled:1;
2251 unsigned threaded:1;
2253 struct list_head net_notifier_list;
2255 #if IS_ENABLED(CONFIG_MACSEC)
2256 /* MACsec management functions */
2257 const struct macsec_ops *macsec_ops;
2259 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2260 struct udp_tunnel_nic *udp_tunnel_nic;
2262 /* protected by rtnl_lock */
2263 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2265 #define to_net_dev(d) container_of(d, struct net_device, dev)
2267 static inline bool netif_elide_gro(const struct net_device *dev)
2269 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2274 #define NETDEV_ALIGN 32
2277 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2279 return dev->prio_tc_map[prio & TC_BITMASK];
2283 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2285 if (tc >= dev->num_tc)
2288 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2292 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2293 void netdev_reset_tc(struct net_device *dev);
2294 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2295 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2298 int netdev_get_num_tc(struct net_device *dev)
2303 static inline void net_prefetch(void *p)
2306 #if L1_CACHE_BYTES < 128
2307 prefetch((u8 *)p + L1_CACHE_BYTES);
2311 static inline void net_prefetchw(void *p)
2314 #if L1_CACHE_BYTES < 128
2315 prefetchw((u8 *)p + L1_CACHE_BYTES);
2319 void netdev_unbind_sb_channel(struct net_device *dev,
2320 struct net_device *sb_dev);
2321 int netdev_bind_sb_channel_queue(struct net_device *dev,
2322 struct net_device *sb_dev,
2323 u8 tc, u16 count, u16 offset);
2324 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2325 static inline int netdev_get_sb_channel(struct net_device *dev)
2327 return max_t(int, -dev->num_tc, 0);
2331 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2334 return &dev->_tx[index];
2337 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2338 const struct sk_buff *skb)
2340 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2343 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2344 void (*f)(struct net_device *,
2345 struct netdev_queue *,
2351 for (i = 0; i < dev->num_tx_queues; i++)
2352 f(dev, &dev->_tx[i], arg);
2355 #define netdev_lockdep_set_classes(dev) \
2357 static struct lock_class_key qdisc_tx_busylock_key; \
2358 static struct lock_class_key qdisc_running_key; \
2359 static struct lock_class_key qdisc_xmit_lock_key; \
2360 static struct lock_class_key dev_addr_list_lock_key; \
2363 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2364 (dev)->qdisc_running_key = &qdisc_running_key; \
2365 lockdep_set_class(&(dev)->addr_list_lock, \
2366 &dev_addr_list_lock_key); \
2367 for (i = 0; i < (dev)->num_tx_queues; i++) \
2368 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2369 &qdisc_xmit_lock_key); \
2372 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2373 struct net_device *sb_dev);
2374 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2375 struct sk_buff *skb,
2376 struct net_device *sb_dev);
2378 /* returns the headroom that the master device needs to take in account
2379 * when forwarding to this dev
2381 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2383 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2386 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2388 if (dev->netdev_ops->ndo_set_rx_headroom)
2389 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2392 /* set the device rx headroom to the dev's default */
2393 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2395 netdev_set_rx_headroom(dev, -1);
2398 static inline void *netdev_get_ml_priv(struct net_device *dev,
2399 enum netdev_ml_priv_type type)
2401 if (dev->ml_priv_type != type)
2404 return dev->ml_priv;
2407 static inline void netdev_set_ml_priv(struct net_device *dev,
2409 enum netdev_ml_priv_type type)
2411 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2412 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2413 dev->ml_priv_type, type);
2414 WARN(!dev->ml_priv_type && dev->ml_priv,
2415 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2417 dev->ml_priv = ml_priv;
2418 dev->ml_priv_type = type;
2422 * Net namespace inlines
2425 struct net *dev_net(const struct net_device *dev)
2427 return read_pnet(&dev->nd_net);
2431 void dev_net_set(struct net_device *dev, struct net *net)
2433 write_pnet(&dev->nd_net, net);
2437 * netdev_priv - access network device private data
2438 * @dev: network device
2440 * Get network device private data
2442 static inline void *netdev_priv(const struct net_device *dev)
2444 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2447 /* Set the sysfs physical device reference for the network logical device
2448 * if set prior to registration will cause a symlink during initialization.
2450 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2452 /* Set the sysfs device type for the network logical device to allow
2453 * fine-grained identification of different network device types. For
2454 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2456 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2458 /* Default NAPI poll() weight
2459 * Device drivers are strongly advised to not use bigger value
2461 #define NAPI_POLL_WEIGHT 64
2464 * netif_napi_add - initialize a NAPI context
2465 * @dev: network device
2466 * @napi: NAPI context
2467 * @poll: polling function
2468 * @weight: default weight
2470 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2471 * *any* of the other NAPI-related functions.
2473 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2474 int (*poll)(struct napi_struct *, int), int weight);
2477 * netif_tx_napi_add - initialize a NAPI context
2478 * @dev: network device
2479 * @napi: NAPI context
2480 * @poll: polling function
2481 * @weight: default weight
2483 * This variant of netif_napi_add() should be used from drivers using NAPI
2484 * to exclusively poll a TX queue.
2485 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2487 static inline void netif_tx_napi_add(struct net_device *dev,
2488 struct napi_struct *napi,
2489 int (*poll)(struct napi_struct *, int),
2492 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2493 netif_napi_add(dev, napi, poll, weight);
2497 * __netif_napi_del - remove a NAPI context
2498 * @napi: NAPI context
2500 * Warning: caller must observe RCU grace period before freeing memory
2501 * containing @napi. Drivers might want to call this helper to combine
2502 * all the needed RCU grace periods into a single one.
2504 void __netif_napi_del(struct napi_struct *napi);
2507 * netif_napi_del - remove a NAPI context
2508 * @napi: NAPI context
2510 * netif_napi_del() removes a NAPI context from the network device NAPI list
2512 static inline void netif_napi_del(struct napi_struct *napi)
2514 __netif_napi_del(napi);
2518 struct napi_gro_cb {
2519 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2522 /* Length of frag0. */
2523 unsigned int frag0_len;
2525 /* This indicates where we are processing relative to skb->data. */
2528 /* This is non-zero if the packet cannot be merged with the new skb. */
2531 /* Save the IP ID here and check when we get to the transport layer */
2534 /* Number of segments aggregated. */
2537 /* Start offset for remote checksum offload */
2538 u16 gro_remcsum_start;
2540 /* jiffies when first packet was created/queued */
2543 /* Used in ipv6_gro_receive() and foo-over-udp */
2546 /* This is non-zero if the packet may be of the same flow. */
2549 /* Used in tunnel GRO receive */
2552 /* GRO checksum is valid */
2555 /* Number of checksums via CHECKSUM_UNNECESSARY */
2560 #define NAPI_GRO_FREE 1
2561 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2563 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2566 /* Used in GRE, set in fou/gue_gro_receive */
2569 /* Used to determine if flush_id can be ignored */
2572 /* Number of gro_receive callbacks this packet already went through */
2573 u8 recursion_counter:4;
2575 /* GRO is done by frag_list pointer chaining. */
2578 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2581 /* used in skb_gro_receive() slow path */
2582 struct sk_buff *last;
2585 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2587 #define GRO_RECURSION_LIMIT 15
2588 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2590 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2593 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2594 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2595 struct list_head *head,
2596 struct sk_buff *skb)
2598 if (unlikely(gro_recursion_inc_test(skb))) {
2599 NAPI_GRO_CB(skb)->flush |= 1;
2603 return cb(head, skb);
2606 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2608 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2610 struct list_head *head,
2611 struct sk_buff *skb)
2613 if (unlikely(gro_recursion_inc_test(skb))) {
2614 NAPI_GRO_CB(skb)->flush |= 1;
2618 return cb(sk, head, skb);
2621 struct packet_type {
2622 __be16 type; /* This is really htons(ether_type). */
2623 bool ignore_outgoing;
2624 struct net_device *dev; /* NULL is wildcarded here */
2625 int (*func) (struct sk_buff *,
2626 struct net_device *,
2627 struct packet_type *,
2628 struct net_device *);
2629 void (*list_func) (struct list_head *,
2630 struct packet_type *,
2631 struct net_device *);
2632 bool (*id_match)(struct packet_type *ptype,
2634 void *af_packet_priv;
2635 struct list_head list;
2638 struct offload_callbacks {
2639 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2640 netdev_features_t features);
2641 struct sk_buff *(*gro_receive)(struct list_head *head,
2642 struct sk_buff *skb);
2643 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2646 struct packet_offload {
2647 __be16 type; /* This is really htons(ether_type). */
2649 struct offload_callbacks callbacks;
2650 struct list_head list;
2653 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2654 struct pcpu_sw_netstats {
2659 struct u64_stats_sync syncp;
2660 } __aligned(4 * sizeof(u64));
2662 struct pcpu_lstats {
2663 u64_stats_t packets;
2665 struct u64_stats_sync syncp;
2666 } __aligned(2 * sizeof(u64));
2668 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2670 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2672 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2674 u64_stats_update_begin(&tstats->syncp);
2675 tstats->rx_bytes += len;
2676 tstats->rx_packets++;
2677 u64_stats_update_end(&tstats->syncp);
2680 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2681 unsigned int packets,
2684 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2686 u64_stats_update_begin(&tstats->syncp);
2687 tstats->tx_bytes += len;
2688 tstats->tx_packets += packets;
2689 u64_stats_update_end(&tstats->syncp);
2692 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2694 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2696 u64_stats_update_begin(&lstats->syncp);
2697 u64_stats_add(&lstats->bytes, len);
2698 u64_stats_inc(&lstats->packets);
2699 u64_stats_update_end(&lstats->syncp);
2702 #define __netdev_alloc_pcpu_stats(type, gfp) \
2704 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2707 for_each_possible_cpu(__cpu) { \
2708 typeof(type) *stat; \
2709 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2710 u64_stats_init(&stat->syncp); \
2716 #define netdev_alloc_pcpu_stats(type) \
2717 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2719 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2721 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2724 for_each_possible_cpu(__cpu) { \
2725 typeof(type) *stat; \
2726 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2727 u64_stats_init(&stat->syncp); \
2733 enum netdev_lag_tx_type {
2734 NETDEV_LAG_TX_TYPE_UNKNOWN,
2735 NETDEV_LAG_TX_TYPE_RANDOM,
2736 NETDEV_LAG_TX_TYPE_BROADCAST,
2737 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2738 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2739 NETDEV_LAG_TX_TYPE_HASH,
2742 enum netdev_lag_hash {
2743 NETDEV_LAG_HASH_NONE,
2745 NETDEV_LAG_HASH_L34,
2746 NETDEV_LAG_HASH_L23,
2747 NETDEV_LAG_HASH_E23,
2748 NETDEV_LAG_HASH_E34,
2749 NETDEV_LAG_HASH_VLAN_SRCMAC,
2750 NETDEV_LAG_HASH_UNKNOWN,
2753 struct netdev_lag_upper_info {
2754 enum netdev_lag_tx_type tx_type;
2755 enum netdev_lag_hash hash_type;
2758 struct netdev_lag_lower_state_info {
2763 #include <linux/notifier.h>
2765 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2766 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2770 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2772 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2773 detected a hardware crash and restarted
2774 - we can use this eg to kick tcp sessions
2776 NETDEV_CHANGE, /* Notify device state change */
2779 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2780 NETDEV_CHANGEADDR, /* notify after the address change */
2781 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2785 NETDEV_BONDING_FAILOVER,
2787 NETDEV_PRE_TYPE_CHANGE,
2788 NETDEV_POST_TYPE_CHANGE,
2791 NETDEV_NOTIFY_PEERS,
2795 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2796 NETDEV_CHANGEINFODATA,
2797 NETDEV_BONDING_INFO,
2798 NETDEV_PRECHANGEUPPER,
2799 NETDEV_CHANGELOWERSTATE,
2800 NETDEV_UDP_TUNNEL_PUSH_INFO,
2801 NETDEV_UDP_TUNNEL_DROP_INFO,
2802 NETDEV_CHANGE_TX_QUEUE_LEN,
2803 NETDEV_CVLAN_FILTER_PUSH_INFO,
2804 NETDEV_CVLAN_FILTER_DROP_INFO,
2805 NETDEV_SVLAN_FILTER_PUSH_INFO,
2806 NETDEV_SVLAN_FILTER_DROP_INFO,
2808 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2810 int register_netdevice_notifier(struct notifier_block *nb);
2811 int unregister_netdevice_notifier(struct notifier_block *nb);
2812 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2813 int unregister_netdevice_notifier_net(struct net *net,
2814 struct notifier_block *nb);
2815 int register_netdevice_notifier_dev_net(struct net_device *dev,
2816 struct notifier_block *nb,
2817 struct netdev_net_notifier *nn);
2818 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2819 struct notifier_block *nb,
2820 struct netdev_net_notifier *nn);
2822 struct netdev_notifier_info {
2823 struct net_device *dev;
2824 struct netlink_ext_ack *extack;
2827 struct netdev_notifier_info_ext {
2828 struct netdev_notifier_info info; /* must be first */
2834 struct netdev_notifier_change_info {
2835 struct netdev_notifier_info info; /* must be first */
2836 unsigned int flags_changed;
2839 struct netdev_notifier_changeupper_info {
2840 struct netdev_notifier_info info; /* must be first */
2841 struct net_device *upper_dev; /* new upper dev */
2842 bool master; /* is upper dev master */
2843 bool linking; /* is the notification for link or unlink */
2844 void *upper_info; /* upper dev info */
2847 struct netdev_notifier_changelowerstate_info {
2848 struct netdev_notifier_info info; /* must be first */
2849 void *lower_state_info; /* is lower dev state */
2852 struct netdev_notifier_pre_changeaddr_info {
2853 struct netdev_notifier_info info; /* must be first */
2854 const unsigned char *dev_addr;
2857 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2858 struct net_device *dev)
2861 info->extack = NULL;
2864 static inline struct net_device *
2865 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2870 static inline struct netlink_ext_ack *
2871 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2873 return info->extack;
2876 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2879 extern rwlock_t dev_base_lock; /* Device list lock */
2881 #define for_each_netdev(net, d) \
2882 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2883 #define for_each_netdev_reverse(net, d) \
2884 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2885 #define for_each_netdev_rcu(net, d) \
2886 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2887 #define for_each_netdev_safe(net, d, n) \
2888 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2889 #define for_each_netdev_continue(net, d) \
2890 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2891 #define for_each_netdev_continue_reverse(net, d) \
2892 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2894 #define for_each_netdev_continue_rcu(net, d) \
2895 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2896 #define for_each_netdev_in_bond_rcu(bond, slave) \
2897 for_each_netdev_rcu(&init_net, slave) \
2898 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2899 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2901 static inline struct net_device *next_net_device(struct net_device *dev)
2903 struct list_head *lh;
2907 lh = dev->dev_list.next;
2908 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2911 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2913 struct list_head *lh;
2917 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2918 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2921 static inline struct net_device *first_net_device(struct net *net)
2923 return list_empty(&net->dev_base_head) ? NULL :
2924 net_device_entry(net->dev_base_head.next);
2927 static inline struct net_device *first_net_device_rcu(struct net *net)
2929 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2931 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2934 int netdev_boot_setup_check(struct net_device *dev);
2935 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2936 const char *hwaddr);
2937 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2938 void dev_add_pack(struct packet_type *pt);
2939 void dev_remove_pack(struct packet_type *pt);
2940 void __dev_remove_pack(struct packet_type *pt);
2941 void dev_add_offload(struct packet_offload *po);
2942 void dev_remove_offload(struct packet_offload *po);
2944 int dev_get_iflink(const struct net_device *dev);
2945 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2946 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
2947 struct net_device_path_stack *stack);
2948 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2949 unsigned short mask);
2950 struct net_device *dev_get_by_name(struct net *net, const char *name);
2951 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2952 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2953 int dev_alloc_name(struct net_device *dev, const char *name);
2954 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2955 void dev_close(struct net_device *dev);
2956 void dev_close_many(struct list_head *head, bool unlink);
2957 void dev_disable_lro(struct net_device *dev);
2958 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2959 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2960 struct net_device *sb_dev);
2961 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2962 struct net_device *sb_dev);
2964 int dev_queue_xmit(struct sk_buff *skb);
2965 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2966 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2968 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2972 ret = __dev_direct_xmit(skb, queue_id);
2973 if (!dev_xmit_complete(ret))
2978 int register_netdevice(struct net_device *dev);
2979 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2980 void unregister_netdevice_many(struct list_head *head);
2981 static inline void unregister_netdevice(struct net_device *dev)
2983 unregister_netdevice_queue(dev, NULL);
2986 int netdev_refcnt_read(const struct net_device *dev);
2987 void free_netdev(struct net_device *dev);
2988 void netdev_freemem(struct net_device *dev);
2989 int init_dummy_netdev(struct net_device *dev);
2991 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2992 struct sk_buff *skb,
2994 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2996 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2997 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2998 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2999 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3000 int netdev_get_name(struct net *net, char *name, int ifindex);
3001 int dev_restart(struct net_device *dev);
3002 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
3003 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
3005 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
3007 return NAPI_GRO_CB(skb)->data_offset;
3010 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
3012 return skb->len - NAPI_GRO_CB(skb)->data_offset;
3015 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
3017 NAPI_GRO_CB(skb)->data_offset += len;
3020 static inline void *skb_gro_header_fast(struct sk_buff *skb,
3021 unsigned int offset)
3023 return NAPI_GRO_CB(skb)->frag0 + offset;
3026 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
3028 return NAPI_GRO_CB(skb)->frag0_len < hlen;
3031 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
3033 NAPI_GRO_CB(skb)->frag0 = NULL;
3034 NAPI_GRO_CB(skb)->frag0_len = 0;
3037 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
3038 unsigned int offset)
3040 if (!pskb_may_pull(skb, hlen))
3043 skb_gro_frag0_invalidate(skb);
3044 return skb->data + offset;
3047 static inline void *skb_gro_network_header(struct sk_buff *skb)
3049 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
3050 skb_network_offset(skb);
3053 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
3054 const void *start, unsigned int len)
3056 if (NAPI_GRO_CB(skb)->csum_valid)
3057 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
3058 csum_partial(start, len, 0));
3061 /* GRO checksum functions. These are logical equivalents of the normal
3062 * checksum functions (in skbuff.h) except that they operate on the GRO
3063 * offsets and fields in sk_buff.
3066 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
3068 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
3070 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
3073 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3077 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3078 skb_checksum_start_offset(skb) <
3079 skb_gro_offset(skb)) &&
3080 !skb_at_gro_remcsum_start(skb) &&
3081 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3082 (!zero_okay || check));
3085 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3088 if (NAPI_GRO_CB(skb)->csum_valid &&
3089 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3092 NAPI_GRO_CB(skb)->csum = psum;
3094 return __skb_gro_checksum_complete(skb);
3097 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3099 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3100 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3101 NAPI_GRO_CB(skb)->csum_cnt--;
3103 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3104 * verified a new top level checksum or an encapsulated one
3105 * during GRO. This saves work if we fallback to normal path.
3107 __skb_incr_checksum_unnecessary(skb);
3111 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3114 __sum16 __ret = 0; \
3115 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3116 __ret = __skb_gro_checksum_validate_complete(skb, \
3117 compute_pseudo(skb, proto)); \
3119 skb_gro_incr_csum_unnecessary(skb); \
3123 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3124 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3126 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3128 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3130 #define skb_gro_checksum_simple_validate(skb) \
3131 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3133 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3135 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3136 !NAPI_GRO_CB(skb)->csum_valid);
3139 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3142 NAPI_GRO_CB(skb)->csum = ~pseudo;
3143 NAPI_GRO_CB(skb)->csum_valid = 1;
3146 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3148 if (__skb_gro_checksum_convert_check(skb)) \
3149 __skb_gro_checksum_convert(skb, \
3150 compute_pseudo(skb, proto)); \
3153 struct gro_remcsum {
3158 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3164 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3165 unsigned int off, size_t hdrlen,
3166 int start, int offset,
3167 struct gro_remcsum *grc,
3171 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3173 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3176 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3180 ptr = skb_gro_header_fast(skb, off);
3181 if (skb_gro_header_hard(skb, off + plen)) {
3182 ptr = skb_gro_header_slow(skb, off + plen, off);
3187 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3190 /* Adjust skb->csum since we changed the packet */
3191 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3193 grc->offset = off + hdrlen + offset;
3199 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3200 struct gro_remcsum *grc)
3203 size_t plen = grc->offset + sizeof(u16);
3208 ptr = skb_gro_header_fast(skb, grc->offset);
3209 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3210 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3215 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3218 #ifdef CONFIG_XFRM_OFFLOAD
3219 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3221 if (PTR_ERR(pp) != -EINPROGRESS)
3222 NAPI_GRO_CB(skb)->flush |= flush;
3224 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3227 struct gro_remcsum *grc)
3229 if (PTR_ERR(pp) != -EINPROGRESS) {
3230 NAPI_GRO_CB(skb)->flush |= flush;
3231 skb_gro_remcsum_cleanup(skb, grc);
3232 skb->remcsum_offload = 0;
3236 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3238 NAPI_GRO_CB(skb)->flush |= flush;
3240 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3243 struct gro_remcsum *grc)
3245 NAPI_GRO_CB(skb)->flush |= flush;
3246 skb_gro_remcsum_cleanup(skb, grc);
3247 skb->remcsum_offload = 0;
3251 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3252 unsigned short type,
3253 const void *daddr, const void *saddr,
3256 if (!dev->header_ops || !dev->header_ops->create)
3259 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3262 static inline int dev_parse_header(const struct sk_buff *skb,
3263 unsigned char *haddr)
3265 const struct net_device *dev = skb->dev;
3267 if (!dev->header_ops || !dev->header_ops->parse)
3269 return dev->header_ops->parse(skb, haddr);
3272 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3274 const struct net_device *dev = skb->dev;
3276 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3278 return dev->header_ops->parse_protocol(skb);
3281 /* ll_header must have at least hard_header_len allocated */
3282 static inline bool dev_validate_header(const struct net_device *dev,
3283 char *ll_header, int len)
3285 if (likely(len >= dev->hard_header_len))
3287 if (len < dev->min_header_len)
3290 if (capable(CAP_SYS_RAWIO)) {
3291 memset(ll_header + len, 0, dev->hard_header_len - len);
3295 if (dev->header_ops && dev->header_ops->validate)
3296 return dev->header_ops->validate(ll_header, len);
3301 static inline bool dev_has_header(const struct net_device *dev)
3303 return dev->header_ops && dev->header_ops->create;
3306 #ifdef CONFIG_NET_FLOW_LIMIT
3307 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3308 struct sd_flow_limit {
3310 unsigned int num_buckets;
3311 unsigned int history_head;
3312 u16 history[FLOW_LIMIT_HISTORY];
3316 extern int netdev_flow_limit_table_len;
3317 #endif /* CONFIG_NET_FLOW_LIMIT */
3320 * Incoming packets are placed on per-CPU queues
3322 struct softnet_data {
3323 struct list_head poll_list;
3324 struct sk_buff_head process_queue;
3327 unsigned int processed;
3328 unsigned int time_squeeze;
3329 unsigned int received_rps;
3331 struct softnet_data *rps_ipi_list;
3333 #ifdef CONFIG_NET_FLOW_LIMIT
3334 struct sd_flow_limit __rcu *flow_limit;
3336 struct Qdisc *output_queue;
3337 struct Qdisc **output_queue_tailp;
3338 struct sk_buff *completion_queue;
3339 #ifdef CONFIG_XFRM_OFFLOAD
3340 struct sk_buff_head xfrm_backlog;
3342 /* written and read only by owning cpu: */
3348 /* input_queue_head should be written by cpu owning this struct,
3349 * and only read by other cpus. Worth using a cache line.
3351 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3353 /* Elements below can be accessed between CPUs for RPS/RFS */
3354 call_single_data_t csd ____cacheline_aligned_in_smp;
3355 struct softnet_data *rps_ipi_next;
3357 unsigned int input_queue_tail;
3359 unsigned int dropped;
3360 struct sk_buff_head input_pkt_queue;
3361 struct napi_struct backlog;
3365 static inline void input_queue_head_incr(struct softnet_data *sd)
3368 sd->input_queue_head++;
3372 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3373 unsigned int *qtail)
3376 *qtail = ++sd->input_queue_tail;
3380 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3382 static inline int dev_recursion_level(void)
3384 return this_cpu_read(softnet_data.xmit.recursion);
3387 #define XMIT_RECURSION_LIMIT 8
3388 static inline bool dev_xmit_recursion(void)
3390 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3391 XMIT_RECURSION_LIMIT);
3394 static inline void dev_xmit_recursion_inc(void)
3396 __this_cpu_inc(softnet_data.xmit.recursion);
3399 static inline void dev_xmit_recursion_dec(void)
3401 __this_cpu_dec(softnet_data.xmit.recursion);
3404 void __netif_schedule(struct Qdisc *q);
3405 void netif_schedule_queue(struct netdev_queue *txq);
3407 static inline void netif_tx_schedule_all(struct net_device *dev)
3411 for (i = 0; i < dev->num_tx_queues; i++)
3412 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3415 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3417 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3421 * netif_start_queue - allow transmit
3422 * @dev: network device
3424 * Allow upper layers to call the device hard_start_xmit routine.
3426 static inline void netif_start_queue(struct net_device *dev)
3428 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3431 static inline void netif_tx_start_all_queues(struct net_device *dev)
3435 for (i = 0; i < dev->num_tx_queues; i++) {
3436 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3437 netif_tx_start_queue(txq);
3441 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3444 * netif_wake_queue - restart transmit
3445 * @dev: network device
3447 * Allow upper layers to call the device hard_start_xmit routine.
3448 * Used for flow control when transmit resources are available.
3450 static inline void netif_wake_queue(struct net_device *dev)
3452 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3455 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3459 for (i = 0; i < dev->num_tx_queues; i++) {
3460 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3461 netif_tx_wake_queue(txq);
3465 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3467 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3471 * netif_stop_queue - stop transmitted packets
3472 * @dev: network device
3474 * Stop upper layers calling the device hard_start_xmit routine.
3475 * Used for flow control when transmit resources are unavailable.
3477 static inline void netif_stop_queue(struct net_device *dev)
3479 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3482 void netif_tx_stop_all_queues(struct net_device *dev);
3484 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3486 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3490 * netif_queue_stopped - test if transmit queue is flowblocked
3491 * @dev: network device
3493 * Test if transmit queue on device is currently unable to send.
3495 static inline bool netif_queue_stopped(const struct net_device *dev)
3497 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3500 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3502 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3506 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3508 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3512 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3514 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3518 * netdev_queue_set_dql_min_limit - set dql minimum limit
3519 * @dev_queue: pointer to transmit queue
3520 * @min_limit: dql minimum limit
3522 * Forces xmit_more() to return true until the minimum threshold
3523 * defined by @min_limit is reached (or until the tx queue is
3524 * empty). Warning: to be use with care, misuse will impact the
3527 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3528 unsigned int min_limit)
3531 dev_queue->dql.min_limit = min_limit;
3536 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3537 * @dev_queue: pointer to transmit queue
3539 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3540 * to give appropriate hint to the CPU.
3542 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3545 prefetchw(&dev_queue->dql.num_queued);
3550 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3551 * @dev_queue: pointer to transmit queue
3553 * BQL enabled drivers might use this helper in their TX completion path,
3554 * to give appropriate hint to the CPU.
3556 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3559 prefetchw(&dev_queue->dql.limit);
3563 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3567 dql_queued(&dev_queue->dql, bytes);
3569 if (likely(dql_avail(&dev_queue->dql) >= 0))
3572 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3575 * The XOFF flag must be set before checking the dql_avail below,
3576 * because in netdev_tx_completed_queue we update the dql_completed
3577 * before checking the XOFF flag.
3581 /* check again in case another CPU has just made room avail */
3582 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3583 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3587 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3588 * that they should not test BQL status themselves.
3589 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3591 * Returns true if the doorbell must be used to kick the NIC.
3593 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3599 dql_queued(&dev_queue->dql, bytes);
3601 return netif_tx_queue_stopped(dev_queue);
3603 netdev_tx_sent_queue(dev_queue, bytes);
3608 * netdev_sent_queue - report the number of bytes queued to hardware
3609 * @dev: network device
3610 * @bytes: number of bytes queued to the hardware device queue
3612 * Report the number of bytes queued for sending/completion to the network
3613 * device hardware queue. @bytes should be a good approximation and should
3614 * exactly match netdev_completed_queue() @bytes
3616 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3618 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3621 static inline bool __netdev_sent_queue(struct net_device *dev,
3625 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3629 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3630 unsigned int pkts, unsigned int bytes)
3633 if (unlikely(!bytes))
3636 dql_completed(&dev_queue->dql, bytes);
3639 * Without the memory barrier there is a small possiblity that
3640 * netdev_tx_sent_queue will miss the update and cause the queue to
3641 * be stopped forever
3645 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3648 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3649 netif_schedule_queue(dev_queue);
3654 * netdev_completed_queue - report bytes and packets completed by device
3655 * @dev: network device
3656 * @pkts: actual number of packets sent over the medium
3657 * @bytes: actual number of bytes sent over the medium
3659 * Report the number of bytes and packets transmitted by the network device
3660 * hardware queue over the physical medium, @bytes must exactly match the
3661 * @bytes amount passed to netdev_sent_queue()
3663 static inline void netdev_completed_queue(struct net_device *dev,
3664 unsigned int pkts, unsigned int bytes)
3666 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3669 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3672 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3678 * netdev_reset_queue - reset the packets and bytes count of a network device
3679 * @dev_queue: network device
3681 * Reset the bytes and packet count of a network device and clear the
3682 * software flow control OFF bit for this network device
3684 static inline void netdev_reset_queue(struct net_device *dev_queue)
3686 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3690 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3691 * @dev: network device
3692 * @queue_index: given tx queue index
3694 * Returns 0 if given tx queue index >= number of device tx queues,
3695 * otherwise returns the originally passed tx queue index.
3697 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3699 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3700 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3701 dev->name, queue_index,
3702 dev->real_num_tx_queues);
3710 * netif_running - test if up
3711 * @dev: network device
3713 * Test if the device has been brought up.
3715 static inline bool netif_running(const struct net_device *dev)
3717 return test_bit(__LINK_STATE_START, &dev->state);
3721 * Routines to manage the subqueues on a device. We only need start,
3722 * stop, and a check if it's stopped. All other device management is
3723 * done at the overall netdevice level.
3724 * Also test the device if we're multiqueue.
3728 * netif_start_subqueue - allow sending packets on subqueue
3729 * @dev: network device
3730 * @queue_index: sub queue index
3732 * Start individual transmit queue of a device with multiple transmit queues.
3734 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3736 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3738 netif_tx_start_queue(txq);
3742 * netif_stop_subqueue - stop sending packets on subqueue
3743 * @dev: network device
3744 * @queue_index: sub queue index
3746 * Stop individual transmit queue of a device with multiple transmit queues.
3748 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3750 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3751 netif_tx_stop_queue(txq);
3755 * __netif_subqueue_stopped - test status of subqueue
3756 * @dev: network device
3757 * @queue_index: sub queue index
3759 * Check individual transmit queue of a device with multiple transmit queues.
3761 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3764 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3766 return netif_tx_queue_stopped(txq);
3770 * netif_subqueue_stopped - test status of subqueue
3771 * @dev: network device
3772 * @skb: sub queue buffer pointer
3774 * Check individual transmit queue of a device with multiple transmit queues.
3776 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3777 struct sk_buff *skb)
3779 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3783 * netif_wake_subqueue - allow sending packets on subqueue
3784 * @dev: network device
3785 * @queue_index: sub queue index
3787 * Resume individual transmit queue of a device with multiple transmit queues.
3789 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3791 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3793 netif_tx_wake_queue(txq);
3797 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3799 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3800 u16 index, enum xps_map_type type);
3803 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3804 * @j: CPU/Rx queue index
3805 * @mask: bitmask of all cpus/rx queues
3806 * @nr_bits: number of bits in the bitmask
3808 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3810 static inline bool netif_attr_test_mask(unsigned long j,
3811 const unsigned long *mask,
3812 unsigned int nr_bits)
3814 cpu_max_bits_warn(j, nr_bits);
3815 return test_bit(j, mask);
3819 * netif_attr_test_online - Test for online CPU/Rx queue
3820 * @j: CPU/Rx queue index
3821 * @online_mask: bitmask for CPUs/Rx queues that are online
3822 * @nr_bits: number of bits in the bitmask
3824 * Returns true if a CPU/Rx queue is online.
3826 static inline bool netif_attr_test_online(unsigned long j,
3827 const unsigned long *online_mask,
3828 unsigned int nr_bits)
3830 cpu_max_bits_warn(j, nr_bits);
3833 return test_bit(j, online_mask);
3835 return (j < nr_bits);
3839 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3840 * @n: CPU/Rx queue index
3841 * @srcp: the cpumask/Rx queue mask pointer
3842 * @nr_bits: number of bits in the bitmask
3844 * Returns >= nr_bits if no further CPUs/Rx queues set.
3846 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3847 unsigned int nr_bits)
3849 /* -1 is a legal arg here. */
3851 cpu_max_bits_warn(n, nr_bits);
3854 return find_next_bit(srcp, nr_bits, n + 1);
3860 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3861 * @n: CPU/Rx queue index
3862 * @src1p: the first CPUs/Rx queues mask pointer
3863 * @src2p: the second CPUs/Rx queues mask pointer
3864 * @nr_bits: number of bits in the bitmask
3866 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3868 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3869 const unsigned long *src2p,
3870 unsigned int nr_bits)
3872 /* -1 is a legal arg here. */
3874 cpu_max_bits_warn(n, nr_bits);
3877 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3879 return find_next_bit(src1p, nr_bits, n + 1);
3881 return find_next_bit(src2p, nr_bits, n + 1);
3886 static inline int netif_set_xps_queue(struct net_device *dev,
3887 const struct cpumask *mask,
3893 static inline int __netif_set_xps_queue(struct net_device *dev,
3894 const unsigned long *mask,
3895 u16 index, enum xps_map_type type)
3902 * netif_is_multiqueue - test if device has multiple transmit queues
3903 * @dev: network device
3905 * Check if device has multiple transmit queues
3907 static inline bool netif_is_multiqueue(const struct net_device *dev)
3909 return dev->num_tx_queues > 1;
3912 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3915 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3917 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3920 dev->real_num_rx_queues = rxqs;
3924 int netif_set_real_num_queues(struct net_device *dev,
3925 unsigned int txq, unsigned int rxq);
3927 static inline struct netdev_rx_queue *
3928 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3930 return dev->_rx + rxq;
3934 static inline unsigned int get_netdev_rx_queue_index(
3935 struct netdev_rx_queue *queue)
3937 struct net_device *dev = queue->dev;
3938 int index = queue - dev->_rx;
3940 BUG_ON(index >= dev->num_rx_queues);
3945 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3946 int netif_get_num_default_rss_queues(void);
3948 enum skb_free_reason {
3949 SKB_REASON_CONSUMED,
3953 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3954 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3957 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3958 * interrupt context or with hardware interrupts being disabled.
3959 * (in_irq() || irqs_disabled())
3961 * We provide four helpers that can be used in following contexts :
3963 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3964 * replacing kfree_skb(skb)
3966 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3967 * Typically used in place of consume_skb(skb) in TX completion path
3969 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3970 * replacing kfree_skb(skb)
3972 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3973 * and consumed a packet. Used in place of consume_skb(skb)
3975 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3977 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3980 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3982 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3985 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3987 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3990 static inline void dev_consume_skb_any(struct sk_buff *skb)
3992 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3995 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3996 struct bpf_prog *xdp_prog);
3997 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3998 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3999 int netif_rx(struct sk_buff *skb);
4000 int netif_rx_ni(struct sk_buff *skb);
4001 int netif_rx_any_context(struct sk_buff *skb);
4002 int netif_receive_skb(struct sk_buff *skb);
4003 int netif_receive_skb_core(struct sk_buff *skb);
4004 void netif_receive_skb_list(struct list_head *head);
4005 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
4006 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
4007 struct sk_buff *napi_get_frags(struct napi_struct *napi);
4008 gro_result_t napi_gro_frags(struct napi_struct *napi);
4009 struct packet_offload *gro_find_receive_by_type(__be16 type);
4010 struct packet_offload *gro_find_complete_by_type(__be16 type);
4012 static inline void napi_free_frags(struct napi_struct *napi)
4014 kfree_skb(napi->skb);
4018 bool netdev_is_rx_handler_busy(struct net_device *dev);
4019 int netdev_rx_handler_register(struct net_device *dev,
4020 rx_handler_func_t *rx_handler,
4021 void *rx_handler_data);
4022 void netdev_rx_handler_unregister(struct net_device *dev);
4024 bool dev_valid_name(const char *name);
4025 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
4026 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
4027 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
4028 void __user *data, bool *need_copyout);
4029 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
4030 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
4031 unsigned int dev_get_flags(const struct net_device *);
4032 int __dev_change_flags(struct net_device *dev, unsigned int flags,
4033 struct netlink_ext_ack *extack);
4034 int dev_change_flags(struct net_device *dev, unsigned int flags,
4035 struct netlink_ext_ack *extack);
4036 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
4037 unsigned int gchanges);
4038 int dev_change_name(struct net_device *, const char *);
4039 int dev_set_alias(struct net_device *, const char *, size_t);
4040 int dev_get_alias(const struct net_device *, char *, size_t);
4041 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
4042 const char *pat, int new_ifindex);
4044 int dev_change_net_namespace(struct net_device *dev, struct net *net,
4047 return __dev_change_net_namespace(dev, net, pat, 0);
4049 int __dev_set_mtu(struct net_device *, int);
4050 int dev_validate_mtu(struct net_device *dev, int mtu,
4051 struct netlink_ext_ack *extack);
4052 int dev_set_mtu_ext(struct net_device *dev, int mtu,
4053 struct netlink_ext_ack *extack);
4054 int dev_set_mtu(struct net_device *, int);
4055 int dev_change_tx_queue_len(struct net_device *, unsigned long);
4056 void dev_set_group(struct net_device *, int);
4057 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
4058 struct netlink_ext_ack *extack);
4059 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
4060 struct netlink_ext_ack *extack);
4061 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
4062 struct netlink_ext_ack *extack);
4063 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
4064 int dev_change_carrier(struct net_device *, bool new_carrier);
4065 int dev_get_phys_port_id(struct net_device *dev,
4066 struct netdev_phys_item_id *ppid);
4067 int dev_get_phys_port_name(struct net_device *dev,
4068 char *name, size_t len);
4069 int dev_get_port_parent_id(struct net_device *dev,
4070 struct netdev_phys_item_id *ppid, bool recurse);
4071 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
4072 int dev_change_proto_down(struct net_device *dev, bool proto_down);
4073 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
4074 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
4076 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
4077 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
4078 struct netdev_queue *txq, int *ret);
4080 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4081 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4082 int fd, int expected_fd, u32 flags);
4083 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4084 u8 dev_xdp_prog_count(struct net_device *dev);
4085 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4087 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4088 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4089 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4090 bool is_skb_forwardable(const struct net_device *dev,
4091 const struct sk_buff *skb);
4093 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4094 const struct sk_buff *skb,
4095 const bool check_mtu)
4097 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4100 if (!(dev->flags & IFF_UP))
4106 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4107 if (skb->len <= len)
4110 /* if TSO is enabled, we don't care about the length as the packet
4111 * could be forwarded without being segmented before
4113 if (skb_is_gso(skb))
4119 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4120 struct sk_buff *skb,
4121 const bool check_mtu)
4123 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4124 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4125 atomic_long_inc(&dev->rx_dropped);
4130 skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4135 bool dev_nit_active(struct net_device *dev);
4136 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4138 extern int netdev_budget;
4139 extern unsigned int netdev_budget_usecs;
4141 /* Called by rtnetlink.c:rtnl_unlock() */
4142 void netdev_run_todo(void);
4145 * dev_put - release reference to device
4146 * @dev: network device
4148 * Release reference to device to allow it to be freed.
4150 static inline void dev_put(struct net_device *dev)
4153 #ifdef CONFIG_PCPU_DEV_REFCNT
4154 this_cpu_dec(*dev->pcpu_refcnt);
4156 refcount_dec(&dev->dev_refcnt);
4162 * dev_hold - get reference to device
4163 * @dev: network device
4165 * Hold reference to device to keep it from being freed.
4167 static inline void dev_hold(struct net_device *dev)
4170 #ifdef CONFIG_PCPU_DEV_REFCNT
4171 this_cpu_inc(*dev->pcpu_refcnt);
4173 refcount_inc(&dev->dev_refcnt);
4178 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4179 * and _off may be called from IRQ context, but it is caller
4180 * who is responsible for serialization of these calls.
4182 * The name carrier is inappropriate, these functions should really be
4183 * called netif_lowerlayer_*() because they represent the state of any
4184 * kind of lower layer not just hardware media.
4187 void linkwatch_init_dev(struct net_device *dev);
4188 void linkwatch_fire_event(struct net_device *dev);
4189 void linkwatch_forget_dev(struct net_device *dev);
4192 * netif_carrier_ok - test if carrier present
4193 * @dev: network device
4195 * Check if carrier is present on device
4197 static inline bool netif_carrier_ok(const struct net_device *dev)
4199 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4202 unsigned long dev_trans_start(struct net_device *dev);
4204 void __netdev_watchdog_up(struct net_device *dev);
4206 void netif_carrier_on(struct net_device *dev);
4207 void netif_carrier_off(struct net_device *dev);
4208 void netif_carrier_event(struct net_device *dev);
4211 * netif_dormant_on - mark device as dormant.
4212 * @dev: network device
4214 * Mark device as dormant (as per RFC2863).
4216 * The dormant state indicates that the relevant interface is not
4217 * actually in a condition to pass packets (i.e., it is not 'up') but is
4218 * in a "pending" state, waiting for some external event. For "on-
4219 * demand" interfaces, this new state identifies the situation where the
4220 * interface is waiting for events to place it in the up state.
4222 static inline void netif_dormant_on(struct net_device *dev)
4224 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4225 linkwatch_fire_event(dev);
4229 * netif_dormant_off - set device as not dormant.
4230 * @dev: network device
4232 * Device is not in dormant state.
4234 static inline void netif_dormant_off(struct net_device *dev)
4236 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4237 linkwatch_fire_event(dev);
4241 * netif_dormant - test if device is dormant
4242 * @dev: network device
4244 * Check if device is dormant.
4246 static inline bool netif_dormant(const struct net_device *dev)
4248 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4253 * netif_testing_on - mark device as under test.
4254 * @dev: network device
4256 * Mark device as under test (as per RFC2863).
4258 * The testing state indicates that some test(s) must be performed on
4259 * the interface. After completion, of the test, the interface state
4260 * will change to up, dormant, or down, as appropriate.
4262 static inline void netif_testing_on(struct net_device *dev)
4264 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4265 linkwatch_fire_event(dev);
4269 * netif_testing_off - set device as not under test.
4270 * @dev: network device
4272 * Device is not in testing state.
4274 static inline void netif_testing_off(struct net_device *dev)
4276 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4277 linkwatch_fire_event(dev);
4281 * netif_testing - test if device is under test
4282 * @dev: network device
4284 * Check if device is under test
4286 static inline bool netif_testing(const struct net_device *dev)
4288 return test_bit(__LINK_STATE_TESTING, &dev->state);
4293 * netif_oper_up - test if device is operational
4294 * @dev: network device
4296 * Check if carrier is operational
4298 static inline bool netif_oper_up(const struct net_device *dev)
4300 return (dev->operstate == IF_OPER_UP ||
4301 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4305 * netif_device_present - is device available or removed
4306 * @dev: network device
4308 * Check if device has not been removed from system.
4310 static inline bool netif_device_present(const struct net_device *dev)
4312 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4315 void netif_device_detach(struct net_device *dev);
4317 void netif_device_attach(struct net_device *dev);
4320 * Network interface message level settings
4325 NETIF_MSG_PROBE_BIT,
4327 NETIF_MSG_TIMER_BIT,
4328 NETIF_MSG_IFDOWN_BIT,
4330 NETIF_MSG_RX_ERR_BIT,
4331 NETIF_MSG_TX_ERR_BIT,
4332 NETIF_MSG_TX_QUEUED_BIT,
4334 NETIF_MSG_TX_DONE_BIT,
4335 NETIF_MSG_RX_STATUS_BIT,
4336 NETIF_MSG_PKTDATA_BIT,
4340 /* When you add a new bit above, update netif_msg_class_names array
4341 * in net/ethtool/common.c
4343 NETIF_MSG_CLASS_COUNT,
4345 /* Both ethtool_ops interface and internal driver implementation use u32 */
4346 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4348 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4349 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4351 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4352 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4353 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4354 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4355 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4356 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4357 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4358 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4359 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4360 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4361 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4362 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4363 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4364 #define NETIF_MSG_HW __NETIF_MSG(HW)
4365 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4367 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4368 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4369 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4370 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4371 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4372 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4373 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4374 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4375 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4376 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4377 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4378 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4379 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4380 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4381 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4383 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4386 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4387 return default_msg_enable_bits;
4388 if (debug_value == 0) /* no output */
4390 /* set low N bits */
4391 return (1U << debug_value) - 1;
4394 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4396 spin_lock(&txq->_xmit_lock);
4397 txq->xmit_lock_owner = cpu;
4400 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4402 __acquire(&txq->_xmit_lock);
4406 static inline void __netif_tx_release(struct netdev_queue *txq)
4408 __release(&txq->_xmit_lock);
4411 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4413 spin_lock_bh(&txq->_xmit_lock);
4414 txq->xmit_lock_owner = smp_processor_id();
4417 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4419 bool ok = spin_trylock(&txq->_xmit_lock);
4421 txq->xmit_lock_owner = smp_processor_id();
4425 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4427 txq->xmit_lock_owner = -1;
4428 spin_unlock(&txq->_xmit_lock);
4431 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4433 txq->xmit_lock_owner = -1;
4434 spin_unlock_bh(&txq->_xmit_lock);
4437 static inline void txq_trans_update(struct netdev_queue *txq)
4439 if (txq->xmit_lock_owner != -1)
4440 txq->trans_start = jiffies;
4443 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4444 static inline void netif_trans_update(struct net_device *dev)
4446 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4448 if (txq->trans_start != jiffies)
4449 txq->trans_start = jiffies;
4453 * netif_tx_lock - grab network device transmit lock
4454 * @dev: network device
4456 * Get network device transmit lock
4458 static inline void netif_tx_lock(struct net_device *dev)
4463 spin_lock(&dev->tx_global_lock);
4464 cpu = smp_processor_id();
4465 for (i = 0; i < dev->num_tx_queues; i++) {
4466 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4468 /* We are the only thread of execution doing a
4469 * freeze, but we have to grab the _xmit_lock in
4470 * order to synchronize with threads which are in
4471 * the ->hard_start_xmit() handler and already
4472 * checked the frozen bit.
4474 __netif_tx_lock(txq, cpu);
4475 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4476 __netif_tx_unlock(txq);
4480 static inline void netif_tx_lock_bh(struct net_device *dev)
4486 static inline void netif_tx_unlock(struct net_device *dev)
4490 for (i = 0; i < dev->num_tx_queues; i++) {
4491 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4493 /* No need to grab the _xmit_lock here. If the
4494 * queue is not stopped for another reason, we
4497 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4498 netif_schedule_queue(txq);
4500 spin_unlock(&dev->tx_global_lock);
4503 static inline void netif_tx_unlock_bh(struct net_device *dev)
4505 netif_tx_unlock(dev);
4509 #define HARD_TX_LOCK(dev, txq, cpu) { \
4510 if ((dev->features & NETIF_F_LLTX) == 0) { \
4511 __netif_tx_lock(txq, cpu); \
4513 __netif_tx_acquire(txq); \
4517 #define HARD_TX_TRYLOCK(dev, txq) \
4518 (((dev->features & NETIF_F_LLTX) == 0) ? \
4519 __netif_tx_trylock(txq) : \
4520 __netif_tx_acquire(txq))
4522 #define HARD_TX_UNLOCK(dev, txq) { \
4523 if ((dev->features & NETIF_F_LLTX) == 0) { \
4524 __netif_tx_unlock(txq); \
4526 __netif_tx_release(txq); \
4530 static inline void netif_tx_disable(struct net_device *dev)
4536 cpu = smp_processor_id();
4537 spin_lock(&dev->tx_global_lock);
4538 for (i = 0; i < dev->num_tx_queues; i++) {
4539 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4541 __netif_tx_lock(txq, cpu);
4542 netif_tx_stop_queue(txq);
4543 __netif_tx_unlock(txq);
4545 spin_unlock(&dev->tx_global_lock);
4549 static inline void netif_addr_lock(struct net_device *dev)
4551 unsigned char nest_level = 0;
4553 #ifdef CONFIG_LOCKDEP
4554 nest_level = dev->nested_level;
4556 spin_lock_nested(&dev->addr_list_lock, nest_level);
4559 static inline void netif_addr_lock_bh(struct net_device *dev)
4561 unsigned char nest_level = 0;
4563 #ifdef CONFIG_LOCKDEP
4564 nest_level = dev->nested_level;
4567 spin_lock_nested(&dev->addr_list_lock, nest_level);
4570 static inline void netif_addr_unlock(struct net_device *dev)
4572 spin_unlock(&dev->addr_list_lock);
4575 static inline void netif_addr_unlock_bh(struct net_device *dev)
4577 spin_unlock_bh(&dev->addr_list_lock);
4581 * dev_addrs walker. Should be used only for read access. Call with
4582 * rcu_read_lock held.
4584 #define for_each_dev_addr(dev, ha) \
4585 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4587 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4589 void ether_setup(struct net_device *dev);
4591 /* Support for loadable net-drivers */
4592 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4593 unsigned char name_assign_type,
4594 void (*setup)(struct net_device *),
4595 unsigned int txqs, unsigned int rxqs);
4596 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4597 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4599 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4600 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4603 int register_netdev(struct net_device *dev);
4604 void unregister_netdev(struct net_device *dev);
4606 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4608 /* General hardware address lists handling functions */
4609 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4610 struct netdev_hw_addr_list *from_list, int addr_len);
4611 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4612 struct netdev_hw_addr_list *from_list, int addr_len);
4613 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4614 struct net_device *dev,
4615 int (*sync)(struct net_device *, const unsigned char *),
4616 int (*unsync)(struct net_device *,
4617 const unsigned char *));
4618 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4619 struct net_device *dev,
4620 int (*sync)(struct net_device *,
4621 const unsigned char *, int),
4622 int (*unsync)(struct net_device *,
4623 const unsigned char *, int));
4624 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4625 struct net_device *dev,
4626 int (*unsync)(struct net_device *,
4627 const unsigned char *, int));
4628 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4629 struct net_device *dev,
4630 int (*unsync)(struct net_device *,
4631 const unsigned char *));
4632 void __hw_addr_init(struct netdev_hw_addr_list *list);
4634 /* Functions used for device addresses handling */
4635 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4636 unsigned char addr_type);
4637 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4638 unsigned char addr_type);
4639 void dev_addr_flush(struct net_device *dev);
4640 int dev_addr_init(struct net_device *dev);
4642 /* Functions used for unicast addresses handling */
4643 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4644 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4645 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4646 int dev_uc_sync(struct net_device *to, struct net_device *from);
4647 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4648 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4649 void dev_uc_flush(struct net_device *dev);
4650 void dev_uc_init(struct net_device *dev);
4653 * __dev_uc_sync - Synchonize device's unicast list
4654 * @dev: device to sync
4655 * @sync: function to call if address should be added
4656 * @unsync: function to call if address should be removed
4658 * Add newly added addresses to the interface, and release
4659 * addresses that have been deleted.
4661 static inline int __dev_uc_sync(struct net_device *dev,
4662 int (*sync)(struct net_device *,
4663 const unsigned char *),
4664 int (*unsync)(struct net_device *,
4665 const unsigned char *))
4667 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4671 * __dev_uc_unsync - Remove synchronized addresses from device
4672 * @dev: device to sync
4673 * @unsync: function to call if address should be removed
4675 * Remove all addresses that were added to the device by dev_uc_sync().
4677 static inline void __dev_uc_unsync(struct net_device *dev,
4678 int (*unsync)(struct net_device *,
4679 const unsigned char *))
4681 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4684 /* Functions used for multicast addresses handling */
4685 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4686 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4687 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4688 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4689 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4690 int dev_mc_sync(struct net_device *to, struct net_device *from);
4691 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4692 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4693 void dev_mc_flush(struct net_device *dev);
4694 void dev_mc_init(struct net_device *dev);
4697 * __dev_mc_sync - Synchonize device's multicast list
4698 * @dev: device to sync
4699 * @sync: function to call if address should be added
4700 * @unsync: function to call if address should be removed
4702 * Add newly added addresses to the interface, and release
4703 * addresses that have been deleted.
4705 static inline int __dev_mc_sync(struct net_device *dev,
4706 int (*sync)(struct net_device *,
4707 const unsigned char *),
4708 int (*unsync)(struct net_device *,
4709 const unsigned char *))
4711 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4715 * __dev_mc_unsync - Remove synchronized addresses from device
4716 * @dev: device to sync
4717 * @unsync: function to call if address should be removed
4719 * Remove all addresses that were added to the device by dev_mc_sync().
4721 static inline void __dev_mc_unsync(struct net_device *dev,
4722 int (*unsync)(struct net_device *,
4723 const unsigned char *))
4725 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4728 /* Functions used for secondary unicast and multicast support */
4729 void dev_set_rx_mode(struct net_device *dev);
4730 void __dev_set_rx_mode(struct net_device *dev);
4731 int dev_set_promiscuity(struct net_device *dev, int inc);
4732 int dev_set_allmulti(struct net_device *dev, int inc);
4733 void netdev_state_change(struct net_device *dev);
4734 void __netdev_notify_peers(struct net_device *dev);
4735 void netdev_notify_peers(struct net_device *dev);
4736 void netdev_features_change(struct net_device *dev);
4737 /* Load a device via the kmod */
4738 void dev_load(struct net *net, const char *name);
4739 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4740 struct rtnl_link_stats64 *storage);
4741 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4742 const struct net_device_stats *netdev_stats);
4743 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4744 const struct pcpu_sw_netstats __percpu *netstats);
4745 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4747 extern int netdev_max_backlog;
4748 extern int netdev_tstamp_prequeue;
4749 extern int netdev_unregister_timeout_secs;
4750 extern int weight_p;
4751 extern int dev_weight_rx_bias;
4752 extern int dev_weight_tx_bias;
4753 extern int dev_rx_weight;
4754 extern int dev_tx_weight;
4755 extern int gro_normal_batch;
4758 NESTED_SYNC_IMM_BIT,
4759 NESTED_SYNC_TODO_BIT,
4762 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4763 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4765 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4766 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4768 struct netdev_nested_priv {
4769 unsigned char flags;
4773 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4774 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4775 struct list_head **iter);
4776 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4777 struct list_head **iter);
4779 #ifdef CONFIG_LOCKDEP
4780 static LIST_HEAD(net_unlink_list);
4782 static inline void net_unlink_todo(struct net_device *dev)
4784 if (list_empty(&dev->unlink_list))
4785 list_add_tail(&dev->unlink_list, &net_unlink_list);
4789 /* iterate through upper list, must be called under RCU read lock */
4790 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4791 for (iter = &(dev)->adj_list.upper, \
4792 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4794 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4796 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4797 int (*fn)(struct net_device *upper_dev,
4798 struct netdev_nested_priv *priv),
4799 struct netdev_nested_priv *priv);
4801 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4802 struct net_device *upper_dev);
4804 bool netdev_has_any_upper_dev(struct net_device *dev);
4806 void *netdev_lower_get_next_private(struct net_device *dev,
4807 struct list_head **iter);
4808 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4809 struct list_head **iter);
4811 #define netdev_for_each_lower_private(dev, priv, iter) \
4812 for (iter = (dev)->adj_list.lower.next, \
4813 priv = netdev_lower_get_next_private(dev, &(iter)); \
4815 priv = netdev_lower_get_next_private(dev, &(iter)))
4817 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4818 for (iter = &(dev)->adj_list.lower, \
4819 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4821 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4823 void *netdev_lower_get_next(struct net_device *dev,
4824 struct list_head **iter);
4826 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4827 for (iter = (dev)->adj_list.lower.next, \
4828 ldev = netdev_lower_get_next(dev, &(iter)); \
4830 ldev = netdev_lower_get_next(dev, &(iter)))
4832 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4833 struct list_head **iter);
4834 int netdev_walk_all_lower_dev(struct net_device *dev,
4835 int (*fn)(struct net_device *lower_dev,
4836 struct netdev_nested_priv *priv),
4837 struct netdev_nested_priv *priv);
4838 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4839 int (*fn)(struct net_device *lower_dev,
4840 struct netdev_nested_priv *priv),
4841 struct netdev_nested_priv *priv);
4843 void *netdev_adjacent_get_private(struct list_head *adj_list);
4844 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4845 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4846 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4847 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4848 struct netlink_ext_ack *extack);
4849 int netdev_master_upper_dev_link(struct net_device *dev,
4850 struct net_device *upper_dev,
4851 void *upper_priv, void *upper_info,
4852 struct netlink_ext_ack *extack);
4853 void netdev_upper_dev_unlink(struct net_device *dev,
4854 struct net_device *upper_dev);
4855 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4856 struct net_device *new_dev,
4857 struct net_device *dev,
4858 struct netlink_ext_ack *extack);
4859 void netdev_adjacent_change_commit(struct net_device *old_dev,
4860 struct net_device *new_dev,
4861 struct net_device *dev);
4862 void netdev_adjacent_change_abort(struct net_device *old_dev,
4863 struct net_device *new_dev,
4864 struct net_device *dev);
4865 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4866 void *netdev_lower_dev_get_private(struct net_device *dev,
4867 struct net_device *lower_dev);
4868 void netdev_lower_state_changed(struct net_device *lower_dev,
4869 void *lower_state_info);
4871 /* RSS keys are 40 or 52 bytes long */
4872 #define NETDEV_RSS_KEY_LEN 52
4873 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4874 void netdev_rss_key_fill(void *buffer, size_t len);
4876 int skb_checksum_help(struct sk_buff *skb);
4877 int skb_crc32c_csum_help(struct sk_buff *skb);
4878 int skb_csum_hwoffload_help(struct sk_buff *skb,
4879 const netdev_features_t features);
4881 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4882 netdev_features_t features, bool tx_path);
4883 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4884 netdev_features_t features);
4886 struct netdev_bonding_info {
4891 struct netdev_notifier_bonding_info {
4892 struct netdev_notifier_info info; /* must be first */
4893 struct netdev_bonding_info bonding_info;
4896 void netdev_bonding_info_change(struct net_device *dev,
4897 struct netdev_bonding_info *bonding_info);
4899 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4900 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4902 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4909 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4911 return __skb_gso_segment(skb, features, true);
4913 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4915 static inline bool can_checksum_protocol(netdev_features_t features,
4918 if (protocol == htons(ETH_P_FCOE))
4919 return !!(features & NETIF_F_FCOE_CRC);
4921 /* Assume this is an IP checksum (not SCTP CRC) */
4923 if (features & NETIF_F_HW_CSUM) {
4924 /* Can checksum everything */
4929 case htons(ETH_P_IP):
4930 return !!(features & NETIF_F_IP_CSUM);
4931 case htons(ETH_P_IPV6):
4932 return !!(features & NETIF_F_IPV6_CSUM);
4939 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4941 static inline void netdev_rx_csum_fault(struct net_device *dev,
4942 struct sk_buff *skb)
4946 /* rx skb timestamps */
4947 void net_enable_timestamp(void);
4948 void net_disable_timestamp(void);
4950 #ifdef CONFIG_PROC_FS
4951 int __init dev_proc_init(void);
4953 #define dev_proc_init() 0
4956 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4957 struct sk_buff *skb, struct net_device *dev,
4960 __this_cpu_write(softnet_data.xmit.more, more);
4961 return ops->ndo_start_xmit(skb, dev);
4964 static inline bool netdev_xmit_more(void)
4966 return __this_cpu_read(softnet_data.xmit.more);
4969 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4970 struct netdev_queue *txq, bool more)
4972 const struct net_device_ops *ops = dev->netdev_ops;
4975 rc = __netdev_start_xmit(ops, skb, dev, more);
4976 if (rc == NETDEV_TX_OK)
4977 txq_trans_update(txq);
4982 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4984 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4987 extern const struct kobj_ns_type_operations net_ns_type_operations;
4989 const char *netdev_drivername(const struct net_device *dev);
4991 void linkwatch_run_queue(void);
4993 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4994 netdev_features_t f2)
4996 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4997 if (f1 & NETIF_F_HW_CSUM)
4998 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5000 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
5006 static inline netdev_features_t netdev_get_wanted_features(
5007 struct net_device *dev)
5009 return (dev->features & ~dev->hw_features) | dev->wanted_features;
5011 netdev_features_t netdev_increment_features(netdev_features_t all,
5012 netdev_features_t one, netdev_features_t mask);
5014 /* Allow TSO being used on stacked device :
5015 * Performing the GSO segmentation before last device
5016 * is a performance improvement.
5018 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
5019 netdev_features_t mask)
5021 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
5024 int __netdev_update_features(struct net_device *dev);
5025 void netdev_update_features(struct net_device *dev);
5026 void netdev_change_features(struct net_device *dev);
5028 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
5029 struct net_device *dev);
5031 netdev_features_t passthru_features_check(struct sk_buff *skb,
5032 struct net_device *dev,
5033 netdev_features_t features);
5034 netdev_features_t netif_skb_features(struct sk_buff *skb);
5036 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
5038 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
5040 /* check flags correspondence */
5041 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
5042 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
5043 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
5044 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
5045 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
5046 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
5047 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
5048 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
5049 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
5050 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
5051 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5052 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5053 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5054 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5055 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5056 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5057 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5058 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5059 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5061 return (features & feature) == feature;
5064 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5066 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5067 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5070 static inline bool netif_needs_gso(struct sk_buff *skb,
5071 netdev_features_t features)
5073 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5074 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5075 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5078 static inline void netif_set_gso_max_size(struct net_device *dev,
5081 dev->gso_max_size = size;
5084 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
5085 int pulled_hlen, u16 mac_offset,
5088 skb->protocol = protocol;
5089 skb->encapsulation = 1;
5090 skb_push(skb, pulled_hlen);
5091 skb_reset_transport_header(skb);
5092 skb->mac_header = mac_offset;
5093 skb->network_header = skb->mac_header + mac_len;
5094 skb->mac_len = mac_len;
5097 static inline bool netif_is_macsec(const struct net_device *dev)
5099 return dev->priv_flags & IFF_MACSEC;
5102 static inline bool netif_is_macvlan(const struct net_device *dev)
5104 return dev->priv_flags & IFF_MACVLAN;
5107 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5109 return dev->priv_flags & IFF_MACVLAN_PORT;
5112 static inline bool netif_is_bond_master(const struct net_device *dev)
5114 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5117 static inline bool netif_is_bond_slave(const struct net_device *dev)
5119 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5122 static inline bool netif_supports_nofcs(struct net_device *dev)
5124 return dev->priv_flags & IFF_SUPP_NOFCS;
5127 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5129 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5132 static inline bool netif_is_l3_master(const struct net_device *dev)
5134 return dev->priv_flags & IFF_L3MDEV_MASTER;
5137 static inline bool netif_is_l3_slave(const struct net_device *dev)
5139 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5142 static inline bool netif_is_bridge_master(const struct net_device *dev)
5144 return dev->priv_flags & IFF_EBRIDGE;
5147 static inline bool netif_is_bridge_port(const struct net_device *dev)
5149 return dev->priv_flags & IFF_BRIDGE_PORT;
5152 static inline bool netif_is_ovs_master(const struct net_device *dev)
5154 return dev->priv_flags & IFF_OPENVSWITCH;
5157 static inline bool netif_is_ovs_port(const struct net_device *dev)
5159 return dev->priv_flags & IFF_OVS_DATAPATH;
5162 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5164 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5167 static inline bool netif_is_team_master(const struct net_device *dev)
5169 return dev->priv_flags & IFF_TEAM;
5172 static inline bool netif_is_team_port(const struct net_device *dev)
5174 return dev->priv_flags & IFF_TEAM_PORT;
5177 static inline bool netif_is_lag_master(const struct net_device *dev)
5179 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5182 static inline bool netif_is_lag_port(const struct net_device *dev)
5184 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5187 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5189 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5192 static inline bool netif_is_failover(const struct net_device *dev)
5194 return dev->priv_flags & IFF_FAILOVER;
5197 static inline bool netif_is_failover_slave(const struct net_device *dev)
5199 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5202 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5203 static inline void netif_keep_dst(struct net_device *dev)
5205 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5208 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5209 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5211 /* TODO: reserve and use an additional IFF bit, if we get more users */
5212 return dev->priv_flags & IFF_MACSEC;
5215 extern struct pernet_operations __net_initdata loopback_net_ops;
5217 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5219 /* netdev_printk helpers, similar to dev_printk */
5221 static inline const char *netdev_name(const struct net_device *dev)
5223 if (!dev->name[0] || strchr(dev->name, '%'))
5224 return "(unnamed net_device)";
5228 static inline bool netdev_unregistering(const struct net_device *dev)
5230 return dev->reg_state == NETREG_UNREGISTERING;
5233 static inline const char *netdev_reg_state(const struct net_device *dev)
5235 switch (dev->reg_state) {
5236 case NETREG_UNINITIALIZED: return " (uninitialized)";
5237 case NETREG_REGISTERED: return "";
5238 case NETREG_UNREGISTERING: return " (unregistering)";
5239 case NETREG_UNREGISTERED: return " (unregistered)";
5240 case NETREG_RELEASED: return " (released)";
5241 case NETREG_DUMMY: return " (dummy)";
5244 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5245 return " (unknown)";
5248 __printf(3, 4) __cold
5249 void netdev_printk(const char *level, const struct net_device *dev,
5250 const char *format, ...);
5251 __printf(2, 3) __cold
5252 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5253 __printf(2, 3) __cold
5254 void netdev_alert(const struct net_device *dev, const char *format, ...);
5255 __printf(2, 3) __cold
5256 void netdev_crit(const struct net_device *dev, const char *format, ...);
5257 __printf(2, 3) __cold
5258 void netdev_err(const struct net_device *dev, const char *format, ...);
5259 __printf(2, 3) __cold
5260 void netdev_warn(const struct net_device *dev, const char *format, ...);
5261 __printf(2, 3) __cold
5262 void netdev_notice(const struct net_device *dev, const char *format, ...);
5263 __printf(2, 3) __cold
5264 void netdev_info(const struct net_device *dev, const char *format, ...);
5266 #define netdev_level_once(level, dev, fmt, ...) \
5268 static bool __print_once __read_mostly; \
5270 if (!__print_once) { \
5271 __print_once = true; \
5272 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5276 #define netdev_emerg_once(dev, fmt, ...) \
5277 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5278 #define netdev_alert_once(dev, fmt, ...) \
5279 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5280 #define netdev_crit_once(dev, fmt, ...) \
5281 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5282 #define netdev_err_once(dev, fmt, ...) \
5283 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5284 #define netdev_warn_once(dev, fmt, ...) \
5285 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5286 #define netdev_notice_once(dev, fmt, ...) \
5287 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5288 #define netdev_info_once(dev, fmt, ...) \
5289 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5291 #define MODULE_ALIAS_NETDEV(device) \
5292 MODULE_ALIAS("netdev-" device)
5294 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5295 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5296 #define netdev_dbg(__dev, format, args...) \
5298 dynamic_netdev_dbg(__dev, format, ##args); \
5300 #elif defined(DEBUG)
5301 #define netdev_dbg(__dev, format, args...) \
5302 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5304 #define netdev_dbg(__dev, format, args...) \
5307 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5311 #if defined(VERBOSE_DEBUG)
5312 #define netdev_vdbg netdev_dbg
5315 #define netdev_vdbg(dev, format, args...) \
5318 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5324 * netdev_WARN() acts like dev_printk(), but with the key difference
5325 * of using a WARN/WARN_ON to get the message out, including the
5326 * file/line information and a backtrace.
5328 #define netdev_WARN(dev, format, args...) \
5329 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5330 netdev_reg_state(dev), ##args)
5332 #define netdev_WARN_ONCE(dev, format, args...) \
5333 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5334 netdev_reg_state(dev), ##args)
5336 /* netif printk helpers, similar to netdev_printk */
5338 #define netif_printk(priv, type, level, dev, fmt, args...) \
5340 if (netif_msg_##type(priv)) \
5341 netdev_printk(level, (dev), fmt, ##args); \
5344 #define netif_level(level, priv, type, dev, fmt, args...) \
5346 if (netif_msg_##type(priv)) \
5347 netdev_##level(dev, fmt, ##args); \
5350 #define netif_emerg(priv, type, dev, fmt, args...) \
5351 netif_level(emerg, priv, type, dev, fmt, ##args)
5352 #define netif_alert(priv, type, dev, fmt, args...) \
5353 netif_level(alert, priv, type, dev, fmt, ##args)
5354 #define netif_crit(priv, type, dev, fmt, args...) \
5355 netif_level(crit, priv, type, dev, fmt, ##args)
5356 #define netif_err(priv, type, dev, fmt, args...) \
5357 netif_level(err, priv, type, dev, fmt, ##args)
5358 #define netif_warn(priv, type, dev, fmt, args...) \
5359 netif_level(warn, priv, type, dev, fmt, ##args)
5360 #define netif_notice(priv, type, dev, fmt, args...) \
5361 netif_level(notice, priv, type, dev, fmt, ##args)
5362 #define netif_info(priv, type, dev, fmt, args...) \
5363 netif_level(info, priv, type, dev, fmt, ##args)
5365 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5366 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5367 #define netif_dbg(priv, type, netdev, format, args...) \
5369 if (netif_msg_##type(priv)) \
5370 dynamic_netdev_dbg(netdev, format, ##args); \
5372 #elif defined(DEBUG)
5373 #define netif_dbg(priv, type, dev, format, args...) \
5374 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5376 #define netif_dbg(priv, type, dev, format, args...) \
5379 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5384 /* if @cond then downgrade to debug, else print at @level */
5385 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5388 netif_dbg(priv, type, netdev, fmt, ##args); \
5390 netif_ ## level(priv, type, netdev, fmt, ##args); \
5393 #if defined(VERBOSE_DEBUG)
5394 #define netif_vdbg netif_dbg
5396 #define netif_vdbg(priv, type, dev, format, args...) \
5399 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5405 * The list of packet types we will receive (as opposed to discard)
5406 * and the routines to invoke.
5408 * Why 16. Because with 16 the only overlap we get on a hash of the
5409 * low nibble of the protocol value is RARP/SNAP/X.25.
5423 #define PTYPE_HASH_SIZE (16)
5424 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5426 extern struct list_head ptype_all __read_mostly;
5427 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5429 extern struct net_device *blackhole_netdev;
5431 #endif /* _LINUX_NETDEVICE_H */