1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Definitions for the Interfaces handler.
9 * Version: @(#)dev.h 1.0.10 08/12/93
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Corey Minyard <wf-rch!minyard@relay.EU.net>
14 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
15 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
16 * Bjorn Ekwall. <bj0rn@blox.se>
17 * Pekka Riikonen <priikone@poseidon.pspt.fi>
19 * Moved to /usr/include/linux for NET3
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
37 #include <net/net_namespace.h>
39 #include <net/dcbnl.h>
41 #include <net/netprio_cgroup.h>
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <linux/hashtable.h>
56 struct ip_tunnel_parm;
57 struct macsec_context;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info;
68 struct udp_tunnel_nic_info;
69 struct udp_tunnel_nic;
73 void synchronize_net(void);
74 void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
77 /* Backlog congestion levels */
78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79 #define NET_RX_DROP 1 /* packet dropped */
81 #define MAX_NEST_DEV 8
84 * Transmit return codes: transmit return codes originate from three different
87 * - qdisc return codes
88 * - driver transmit return codes
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
100 /* qdisc ->enqueue() return codes. */
101 #define NET_XMIT_SUCCESS 0x00
102 #define NET_XMIT_DROP 0x01 /* skb dropped */
103 #define NET_XMIT_CN 0x02 /* congestion notification */
104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
112 /* Driver transmit return codes */
113 #define NETDEV_TX_MASK 0xf0
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
120 typedef enum netdev_tx netdev_tx_t;
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
126 static inline bool dev_xmit_complete(int rc)
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
134 if (likely(rc < NET_XMIT_MASK))
141 * Compute the worst-case header length according to the protocols
145 #if defined(CONFIG_HYPERV_NET)
146 # define LL_MAX_HEADER 128
147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
151 # define LL_MAX_HEADER 96
154 # define LL_MAX_HEADER 32
157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159 #define MAX_HEADER LL_MAX_HEADER
161 #define MAX_HEADER (LL_MAX_HEADER + 48)
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
169 struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
196 #include <linux/cache.h>
197 #include <linux/skbuff.h>
200 #include <linux/static_key.h>
201 extern struct static_key_false rps_needed;
202 extern struct static_key_false rfs_needed;
209 struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
213 #define NETDEV_HW_ADDR_T_LAN 1
214 #define NETDEV_HW_ADDR_T_SAN 2
215 #define NETDEV_HW_ADDR_T_UNICAST 3
216 #define NETDEV_HW_ADDR_T_MULTICAST 4
221 struct rcu_head rcu_head;
224 struct netdev_hw_addr_list {
225 struct list_head list;
229 #define netdev_hw_addr_list_count(l) ((l)->count)
230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231 #define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236 #define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241 #define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
248 /* cached hardware header; allow for machine alignment needs. */
249 #define HH_DATA_MOD 16
250 #define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252 #define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
265 #define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
283 /* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
288 enum netdev_state_t {
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
302 struct netdev_boot_setup {
306 #define NETDEV_BOOT_SETUP_MAX 8
308 int __init netdev_boot_setup(char *str);
311 struct list_head list;
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
319 #define GRO_HASH_BUCKETS 8
322 * Structure for NAPI scheduling similar to tasklet but with weighting
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
331 struct list_head poll_list;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338 #ifdef CONFIG_NETPOLL
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
350 struct task_struct *thread;
354 NAPI_STATE_SCHED, /* Poll is scheduled */
355 NAPI_STATE_MISSED, /* reschedule a napi */
356 NAPI_STATE_DISABLE, /* Disable pending */
357 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
358 NAPI_STATE_LISTED, /* NAPI added to system lists */
359 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
360 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
361 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
362 NAPI_STATE_THREADED, /* The poll is performed inside its own thread*/
366 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
367 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
368 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
369 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
370 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
371 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
372 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
373 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
374 NAPIF_STATE_THREADED = BIT(NAPI_STATE_THREADED),
384 typedef enum gro_result gro_result_t;
387 * enum rx_handler_result - Possible return values for rx_handlers.
388 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
390 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
391 * case skb->dev was changed by rx_handler.
392 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
393 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
395 * rx_handlers are functions called from inside __netif_receive_skb(), to do
396 * special processing of the skb, prior to delivery to protocol handlers.
398 * Currently, a net_device can only have a single rx_handler registered. Trying
399 * to register a second rx_handler will return -EBUSY.
401 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
402 * To unregister a rx_handler on a net_device, use
403 * netdev_rx_handler_unregister().
405 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
408 * If the rx_handler consumed the skb in some way, it should return
409 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
410 * the skb to be delivered in some other way.
412 * If the rx_handler changed skb->dev, to divert the skb to another
413 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
414 * new device will be called if it exists.
416 * If the rx_handler decides the skb should be ignored, it should return
417 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
418 * are registered on exact device (ptype->dev == skb->dev).
420 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
421 * delivered, it should return RX_HANDLER_PASS.
423 * A device without a registered rx_handler will behave as if rx_handler
424 * returned RX_HANDLER_PASS.
427 enum rx_handler_result {
433 typedef enum rx_handler_result rx_handler_result_t;
434 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
436 void __napi_schedule(struct napi_struct *n);
437 void __napi_schedule_irqoff(struct napi_struct *n);
439 static inline bool napi_disable_pending(struct napi_struct *n)
441 return test_bit(NAPI_STATE_DISABLE, &n->state);
444 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
446 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
449 bool napi_schedule_prep(struct napi_struct *n);
452 * napi_schedule - schedule NAPI poll
455 * Schedule NAPI poll routine to be called if it is not already
458 static inline void napi_schedule(struct napi_struct *n)
460 if (napi_schedule_prep(n))
465 * napi_schedule_irqoff - schedule NAPI poll
468 * Variant of napi_schedule(), assuming hard irqs are masked.
470 static inline void napi_schedule_irqoff(struct napi_struct *n)
472 if (napi_schedule_prep(n))
473 __napi_schedule_irqoff(n);
476 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
477 static inline bool napi_reschedule(struct napi_struct *napi)
479 if (napi_schedule_prep(napi)) {
480 __napi_schedule(napi);
486 bool napi_complete_done(struct napi_struct *n, int work_done);
488 * napi_complete - NAPI processing complete
491 * Mark NAPI processing as complete.
492 * Consider using napi_complete_done() instead.
493 * Return false if device should avoid rearming interrupts.
495 static inline bool napi_complete(struct napi_struct *n)
497 return napi_complete_done(n, 0);
500 int dev_set_threaded(struct net_device *dev, bool threaded);
503 * napi_disable - prevent NAPI from scheduling
506 * Stop NAPI from being scheduled on this context.
507 * Waits till any outstanding processing completes.
509 void napi_disable(struct napi_struct *n);
511 void napi_enable(struct napi_struct *n);
514 * napi_synchronize - wait until NAPI is not running
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
521 static inline void napi_synchronize(const struct napi_struct *n)
523 if (IS_ENABLED(CONFIG_SMP))
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
531 * napi_if_scheduled_mark_missed - if napi is running, set the
535 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
538 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
540 unsigned long val, new;
543 val = READ_ONCE(n->state);
544 if (val & NAPIF_STATE_DISABLE)
547 if (!(val & NAPIF_STATE_SCHED))
550 new = val | NAPIF_STATE_MISSED;
551 } while (cmpxchg(&n->state, val, new) != val);
556 enum netdev_queue_state_t {
557 __QUEUE_STATE_DRV_XOFF,
558 __QUEUE_STATE_STACK_XOFF,
559 __QUEUE_STATE_FROZEN,
562 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
563 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
564 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
566 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
567 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
569 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
573 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
574 * netif_tx_* functions below are used to manipulate this flag. The
575 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
576 * queue independently. The netif_xmit_*stopped functions below are called
577 * to check if the queue has been stopped by the driver or stack (either
578 * of the XOFF bits are set in the state). Drivers should not need to call
579 * netif_xmit*stopped functions, they should only be using netif_tx_*.
582 struct netdev_queue {
586 struct net_device *dev;
587 struct Qdisc __rcu *qdisc;
588 struct Qdisc *qdisc_sleeping;
592 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
595 unsigned long tx_maxrate;
597 * Number of TX timeouts for this queue
598 * (/sys/class/net/DEV/Q/trans_timeout)
600 unsigned long trans_timeout;
602 /* Subordinate device that the queue has been assigned to */
603 struct net_device *sb_dev;
604 #ifdef CONFIG_XDP_SOCKETS
605 struct xsk_buff_pool *pool;
610 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
613 * Time (in jiffies) of last Tx
615 unsigned long trans_start;
622 } ____cacheline_aligned_in_smp;
624 extern int sysctl_fb_tunnels_only_for_init_net;
625 extern int sysctl_devconf_inherit_init_net;
628 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
629 * == 1 : For initns only
632 static inline bool net_has_fallback_tunnels(const struct net *net)
634 return !IS_ENABLED(CONFIG_SYSCTL) ||
635 !sysctl_fb_tunnels_only_for_init_net ||
636 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
639 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
641 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
648 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
650 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
657 * This structure holds an RPS map which can be of variable length. The
658 * map is an array of CPUs.
665 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
669 * tail pointer for that CPU's input queue at the time of last enqueue, and
670 * a hardware filter index.
672 struct rps_dev_flow {
675 unsigned int last_qtail;
677 #define RPS_NO_FILTER 0xffff
680 * The rps_dev_flow_table structure contains a table of flow mappings.
682 struct rps_dev_flow_table {
685 struct rps_dev_flow flows[];
687 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
688 ((_num) * sizeof(struct rps_dev_flow)))
691 * The rps_sock_flow_table contains mappings of flows to the last CPU
692 * on which they were processed by the application (set in recvmsg).
693 * Each entry is a 32bit value. Upper part is the high-order bits
694 * of flow hash, lower part is CPU number.
695 * rps_cpu_mask is used to partition the space, depending on number of
696 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
697 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
698 * meaning we use 32-6=26 bits for the hash.
700 struct rps_sock_flow_table {
703 u32 ents[] ____cacheline_aligned_in_smp;
705 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
707 #define RPS_NO_CPU 0xffff
709 extern u32 rps_cpu_mask;
710 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
712 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
716 unsigned int index = hash & table->mask;
717 u32 val = hash & ~rps_cpu_mask;
719 /* We only give a hint, preemption can change CPU under us */
720 val |= raw_smp_processor_id();
722 if (table->ents[index] != val)
723 table->ents[index] = val;
727 #ifdef CONFIG_RFS_ACCEL
728 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 #endif /* CONFIG_RPS */
733 /* This structure contains an instance of an RX queue. */
734 struct netdev_rx_queue {
736 struct rps_map __rcu *rps_map;
737 struct rps_dev_flow_table __rcu *rps_flow_table;
740 struct net_device *dev;
741 struct xdp_rxq_info xdp_rxq;
742 #ifdef CONFIG_XDP_SOCKETS
743 struct xsk_buff_pool *pool;
745 } ____cacheline_aligned_in_smp;
748 * RX queue sysfs structures and functions.
750 struct rx_queue_attribute {
751 struct attribute attr;
752 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
753 ssize_t (*store)(struct netdev_rx_queue *queue,
754 const char *buf, size_t len);
757 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
766 * This structure holds an XPS map which can be of variable length. The
767 * map is an array of queues.
771 unsigned int alloc_len;
775 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
776 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
777 - sizeof(struct xps_map)) / sizeof(u16))
780 * This structure holds all XPS maps for device. Maps are indexed by CPU.
782 * We keep track of the number of cpus/rxqs used when the struct is allocated,
783 * in nr_ids. This will help not accessing out-of-bound memory.
785 * We keep track of the number of traffic classes used when the struct is
786 * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
787 * not crossing its upper bound, as the original dev->num_tc can be updated in
790 struct xps_dev_maps {
794 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
797 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
798 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
800 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
801 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
803 #endif /* CONFIG_XPS */
805 #define TC_MAX_QUEUE 16
806 #define TC_BITMASK 15
807 /* HW offloaded queuing disciplines txq count and offset maps */
808 struct netdev_tc_txq {
813 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
815 * This structure is to hold information about the device
816 * configured to run FCoE protocol stack.
818 struct netdev_fcoe_hbainfo {
819 char manufacturer[64];
820 char serial_number[64];
821 char hardware_version[64];
822 char driver_version[64];
823 char optionrom_version[64];
824 char firmware_version[64];
826 char model_description[256];
830 #define MAX_PHYS_ITEM_ID_LEN 32
832 /* This structure holds a unique identifier to identify some
833 * physical item (port for example) used by a netdevice.
835 struct netdev_phys_item_id {
836 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
837 unsigned char id_len;
840 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
841 struct netdev_phys_item_id *b)
843 return a->id_len == b->id_len &&
844 memcmp(a->id, b->id, a->id_len) == 0;
847 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
849 struct net_device *sb_dev);
852 TC_SETUP_QDISC_MQPRIO,
855 TC_SETUP_CLSMATCHALL,
865 TC_SETUP_QDISC_TAPRIO,
873 /* These structures hold the attributes of bpf state that are being passed
874 * to the netdevice through the bpf op.
876 enum bpf_netdev_command {
877 /* Set or clear a bpf program used in the earliest stages of packet
878 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
879 * is responsible for calling bpf_prog_put on any old progs that are
880 * stored. In case of error, the callee need not release the new prog
881 * reference, but on success it takes ownership and must bpf_prog_put
882 * when it is no longer used.
886 /* BPF program for offload callbacks, invoked at program load time. */
887 BPF_OFFLOAD_MAP_ALLOC,
888 BPF_OFFLOAD_MAP_FREE,
892 struct bpf_prog_offload_ops;
893 struct netlink_ext_ack;
895 struct xdp_dev_bulk_queue;
905 struct bpf_xdp_entity {
906 struct bpf_prog *prog;
907 struct bpf_xdp_link *link;
911 enum bpf_netdev_command command;
916 struct bpf_prog *prog;
917 struct netlink_ext_ack *extack;
919 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
921 struct bpf_offloaded_map *offmap;
923 /* XDP_SETUP_XSK_POOL */
925 struct xsk_buff_pool *pool;
931 /* Flags for ndo_xsk_wakeup. */
932 #define XDP_WAKEUP_RX (1 << 0)
933 #define XDP_WAKEUP_TX (1 << 1)
935 #ifdef CONFIG_XFRM_OFFLOAD
937 int (*xdo_dev_state_add) (struct xfrm_state *x);
938 void (*xdo_dev_state_delete) (struct xfrm_state *x);
939 void (*xdo_dev_state_free) (struct xfrm_state *x);
940 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
941 struct xfrm_state *x);
942 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
947 struct rcu_head rcuhead;
954 struct netdev_name_node {
955 struct hlist_node hlist;
956 struct list_head list;
957 struct net_device *dev;
961 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
962 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
964 struct netdev_net_notifier {
965 struct list_head list;
966 struct notifier_block *nb;
970 * This structure defines the management hooks for network devices.
971 * The following hooks can be defined; unless noted otherwise, they are
972 * optional and can be filled with a null pointer.
974 * int (*ndo_init)(struct net_device *dev);
975 * This function is called once when a network device is registered.
976 * The network device can use this for any late stage initialization
977 * or semantic validation. It can fail with an error code which will
978 * be propagated back to register_netdev.
980 * void (*ndo_uninit)(struct net_device *dev);
981 * This function is called when device is unregistered or when registration
982 * fails. It is not called if init fails.
984 * int (*ndo_open)(struct net_device *dev);
985 * This function is called when a network device transitions to the up
988 * int (*ndo_stop)(struct net_device *dev);
989 * This function is called when a network device transitions to the down
992 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
993 * struct net_device *dev);
994 * Called when a packet needs to be transmitted.
995 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
996 * the queue before that can happen; it's for obsolete devices and weird
997 * corner cases, but the stack really does a non-trivial amount
998 * of useless work if you return NETDEV_TX_BUSY.
999 * Required; cannot be NULL.
1001 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1002 * struct net_device *dev
1003 * netdev_features_t features);
1004 * Called by core transmit path to determine if device is capable of
1005 * performing offload operations on a given packet. This is to give
1006 * the device an opportunity to implement any restrictions that cannot
1007 * be otherwise expressed by feature flags. The check is called with
1008 * the set of features that the stack has calculated and it returns
1009 * those the driver believes to be appropriate.
1011 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1012 * struct net_device *sb_dev);
1013 * Called to decide which queue to use when device supports multiple
1016 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1017 * This function is called to allow device receiver to make
1018 * changes to configuration when multicast or promiscuous is enabled.
1020 * void (*ndo_set_rx_mode)(struct net_device *dev);
1021 * This function is called device changes address list filtering.
1022 * If driver handles unicast address filtering, it should set
1023 * IFF_UNICAST_FLT in its priv_flags.
1025 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1026 * This function is called when the Media Access Control address
1027 * needs to be changed. If this interface is not defined, the
1028 * MAC address can not be changed.
1030 * int (*ndo_validate_addr)(struct net_device *dev);
1031 * Test if Media Access Control address is valid for the device.
1033 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1034 * Called when a user requests an ioctl which can't be handled by
1035 * the generic interface code. If not defined ioctls return
1036 * not supported error code.
1038 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1039 * Used to set network devices bus interface parameters. This interface
1040 * is retained for legacy reasons; new devices should use the bus
1041 * interface (PCI) for low level management.
1043 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1044 * Called when a user wants to change the Maximum Transfer Unit
1047 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1048 * Callback used when the transmitter has not made any progress
1049 * for dev->watchdog ticks.
1051 * void (*ndo_get_stats64)(struct net_device *dev,
1052 * struct rtnl_link_stats64 *storage);
1053 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1054 * Called when a user wants to get the network device usage
1055 * statistics. Drivers must do one of the following:
1056 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1057 * rtnl_link_stats64 structure passed by the caller.
1058 * 2. Define @ndo_get_stats to update a net_device_stats structure
1059 * (which should normally be dev->stats) and return a pointer to
1060 * it. The structure may be changed asynchronously only if each
1061 * field is written atomically.
1062 * 3. Update dev->stats asynchronously and atomically, and define
1063 * neither operation.
1065 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1066 * Return true if this device supports offload stats of this attr_id.
1068 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1070 * Get statistics for offload operations by attr_id. Write it into the
1071 * attr_data pointer.
1073 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1074 * If device supports VLAN filtering this function is called when a
1075 * VLAN id is registered.
1077 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1078 * If device supports VLAN filtering this function is called when a
1079 * VLAN id is unregistered.
1081 * void (*ndo_poll_controller)(struct net_device *dev);
1083 * SR-IOV management functions.
1084 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1085 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1086 * u8 qos, __be16 proto);
1087 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1089 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1090 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1091 * int (*ndo_get_vf_config)(struct net_device *dev,
1092 * int vf, struct ifla_vf_info *ivf);
1093 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1094 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1095 * struct nlattr *port[]);
1097 * Enable or disable the VF ability to query its RSS Redirection Table and
1098 * Hash Key. This is needed since on some devices VF share this information
1099 * with PF and querying it may introduce a theoretical security risk.
1100 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1101 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1102 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1104 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1105 * This is always called from the stack with the rtnl lock held and netif
1106 * tx queues stopped. This allows the netdevice to perform queue
1107 * management safely.
1109 * Fiber Channel over Ethernet (FCoE) offload functions.
1110 * int (*ndo_fcoe_enable)(struct net_device *dev);
1111 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1112 * so the underlying device can perform whatever needed configuration or
1113 * initialization to support acceleration of FCoE traffic.
1115 * int (*ndo_fcoe_disable)(struct net_device *dev);
1116 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1117 * so the underlying device can perform whatever needed clean-ups to
1118 * stop supporting acceleration of FCoE traffic.
1120 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1121 * struct scatterlist *sgl, unsigned int sgc);
1122 * Called when the FCoE Initiator wants to initialize an I/O that
1123 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1124 * perform necessary setup and returns 1 to indicate the device is set up
1125 * successfully to perform DDP on this I/O, otherwise this returns 0.
1127 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1128 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1129 * indicated by the FC exchange id 'xid', so the underlying device can
1130 * clean up and reuse resources for later DDP requests.
1132 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1133 * struct scatterlist *sgl, unsigned int sgc);
1134 * Called when the FCoE Target wants to initialize an I/O that
1135 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1136 * perform necessary setup and returns 1 to indicate the device is set up
1137 * successfully to perform DDP on this I/O, otherwise this returns 0.
1139 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1140 * struct netdev_fcoe_hbainfo *hbainfo);
1141 * Called when the FCoE Protocol stack wants information on the underlying
1142 * device. This information is utilized by the FCoE protocol stack to
1143 * register attributes with Fiber Channel management service as per the
1144 * FC-GS Fabric Device Management Information(FDMI) specification.
1146 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1147 * Called when the underlying device wants to override default World Wide
1148 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1149 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1150 * protocol stack to use.
1153 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1154 * u16 rxq_index, u32 flow_id);
1155 * Set hardware filter for RFS. rxq_index is the target queue index;
1156 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1157 * Return the filter ID on success, or a negative error code.
1159 * Slave management functions (for bridge, bonding, etc).
1160 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1161 * Called to make another netdev an underling.
1163 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1164 * Called to release previously enslaved netdev.
1166 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1167 * struct sk_buff *skb,
1169 * Get the xmit slave of master device. If all_slaves is true, function
1170 * assume all the slaves can transmit.
1172 * Feature/offload setting functions.
1173 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1174 * netdev_features_t features);
1175 * Adjusts the requested feature flags according to device-specific
1176 * constraints, and returns the resulting flags. Must not modify
1179 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1180 * Called to update device configuration to new features. Passed
1181 * feature set might be less than what was returned by ndo_fix_features()).
1182 * Must return >0 or -errno if it changed dev->features itself.
1184 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1185 * struct net_device *dev,
1186 * const unsigned char *addr, u16 vid, u16 flags,
1187 * struct netlink_ext_ack *extack);
1188 * Adds an FDB entry to dev for addr.
1189 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1190 * struct net_device *dev,
1191 * const unsigned char *addr, u16 vid)
1192 * Deletes the FDB entry from dev coresponding to addr.
1193 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1194 * struct net_device *dev, struct net_device *filter_dev,
1196 * Used to add FDB entries to dump requests. Implementers should add
1197 * entries to skb and update idx with the number of entries.
1199 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1200 * u16 flags, struct netlink_ext_ack *extack)
1201 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1202 * struct net_device *dev, u32 filter_mask,
1204 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1207 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1208 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1209 * which do not represent real hardware may define this to allow their
1210 * userspace components to manage their virtual carrier state. Devices
1211 * that determine carrier state from physical hardware properties (eg
1212 * network cables) or protocol-dependent mechanisms (eg
1213 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1215 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1216 * struct netdev_phys_item_id *ppid);
1217 * Called to get ID of physical port of this device. If driver does
1218 * not implement this, it is assumed that the hw is not able to have
1219 * multiple net devices on single physical port.
1221 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1222 * struct netdev_phys_item_id *ppid)
1223 * Called to get the parent ID of the physical port of this device.
1225 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1226 * struct net_device *dev)
1227 * Called by upper layer devices to accelerate switching or other
1228 * station functionality into hardware. 'pdev is the lowerdev
1229 * to use for the offload and 'dev' is the net device that will
1230 * back the offload. Returns a pointer to the private structure
1231 * the upper layer will maintain.
1232 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1233 * Called by upper layer device to delete the station created
1234 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1235 * the station and priv is the structure returned by the add
1237 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1238 * int queue_index, u32 maxrate);
1239 * Called when a user wants to set a max-rate limitation of specific
1241 * int (*ndo_get_iflink)(const struct net_device *dev);
1242 * Called to get the iflink value of this device.
1243 * void (*ndo_change_proto_down)(struct net_device *dev,
1245 * This function is used to pass protocol port error state information
1246 * to the switch driver. The switch driver can react to the proto_down
1247 * by doing a phys down on the associated switch port.
1248 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1249 * This function is used to get egress tunnel information for given skb.
1250 * This is useful for retrieving outer tunnel header parameters while
1252 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1253 * This function is used to specify the headroom that the skb must
1254 * consider when allocation skb during packet reception. Setting
1255 * appropriate rx headroom value allows avoiding skb head copy on
1256 * forward. Setting a negative value resets the rx headroom to the
1258 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1259 * This function is used to set or query state related to XDP on the
1260 * netdevice and manage BPF offload. See definition of
1261 * enum bpf_netdev_command for details.
1262 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1264 * This function is used to submit @n XDP packets for transmit on a
1265 * netdevice. Returns number of frames successfully transmitted, frames
1266 * that got dropped are freed/returned via xdp_return_frame().
1267 * Returns negative number, means general error invoking ndo, meaning
1268 * no frames were xmit'ed and core-caller will free all frames.
1269 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1270 * This function is used to wake up the softirq, ksoftirqd or kthread
1271 * responsible for sending and/or receiving packets on a specific
1272 * queue id bound to an AF_XDP socket. The flags field specifies if
1273 * only RX, only Tx, or both should be woken up using the flags
1274 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1275 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1276 * Get devlink port instance associated with a given netdev.
1277 * Called with a reference on the netdevice and devlink locks only,
1278 * rtnl_lock is not held.
1279 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1281 * Add, change, delete or get information on an IPv4 tunnel.
1282 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1283 * If a device is paired with a peer device, return the peer instance.
1284 * The caller must be under RCU read context.
1286 struct net_device_ops {
1287 int (*ndo_init)(struct net_device *dev);
1288 void (*ndo_uninit)(struct net_device *dev);
1289 int (*ndo_open)(struct net_device *dev);
1290 int (*ndo_stop)(struct net_device *dev);
1291 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1292 struct net_device *dev);
1293 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1294 struct net_device *dev,
1295 netdev_features_t features);
1296 u16 (*ndo_select_queue)(struct net_device *dev,
1297 struct sk_buff *skb,
1298 struct net_device *sb_dev);
1299 void (*ndo_change_rx_flags)(struct net_device *dev,
1301 void (*ndo_set_rx_mode)(struct net_device *dev);
1302 int (*ndo_set_mac_address)(struct net_device *dev,
1304 int (*ndo_validate_addr)(struct net_device *dev);
1305 int (*ndo_do_ioctl)(struct net_device *dev,
1306 struct ifreq *ifr, int cmd);
1307 int (*ndo_set_config)(struct net_device *dev,
1309 int (*ndo_change_mtu)(struct net_device *dev,
1311 int (*ndo_neigh_setup)(struct net_device *dev,
1312 struct neigh_parms *);
1313 void (*ndo_tx_timeout) (struct net_device *dev,
1314 unsigned int txqueue);
1316 void (*ndo_get_stats64)(struct net_device *dev,
1317 struct rtnl_link_stats64 *storage);
1318 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1319 int (*ndo_get_offload_stats)(int attr_id,
1320 const struct net_device *dev,
1322 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1324 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1325 __be16 proto, u16 vid);
1326 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1327 __be16 proto, u16 vid);
1328 #ifdef CONFIG_NET_POLL_CONTROLLER
1329 void (*ndo_poll_controller)(struct net_device *dev);
1330 int (*ndo_netpoll_setup)(struct net_device *dev,
1331 struct netpoll_info *info);
1332 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1334 int (*ndo_set_vf_mac)(struct net_device *dev,
1335 int queue, u8 *mac);
1336 int (*ndo_set_vf_vlan)(struct net_device *dev,
1337 int queue, u16 vlan,
1338 u8 qos, __be16 proto);
1339 int (*ndo_set_vf_rate)(struct net_device *dev,
1340 int vf, int min_tx_rate,
1342 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1343 int vf, bool setting);
1344 int (*ndo_set_vf_trust)(struct net_device *dev,
1345 int vf, bool setting);
1346 int (*ndo_get_vf_config)(struct net_device *dev,
1348 struct ifla_vf_info *ivf);
1349 int (*ndo_set_vf_link_state)(struct net_device *dev,
1350 int vf, int link_state);
1351 int (*ndo_get_vf_stats)(struct net_device *dev,
1353 struct ifla_vf_stats
1355 int (*ndo_set_vf_port)(struct net_device *dev,
1357 struct nlattr *port[]);
1358 int (*ndo_get_vf_port)(struct net_device *dev,
1359 int vf, struct sk_buff *skb);
1360 int (*ndo_get_vf_guid)(struct net_device *dev,
1362 struct ifla_vf_guid *node_guid,
1363 struct ifla_vf_guid *port_guid);
1364 int (*ndo_set_vf_guid)(struct net_device *dev,
1367 int (*ndo_set_vf_rss_query_en)(
1368 struct net_device *dev,
1369 int vf, bool setting);
1370 int (*ndo_setup_tc)(struct net_device *dev,
1371 enum tc_setup_type type,
1373 #if IS_ENABLED(CONFIG_FCOE)
1374 int (*ndo_fcoe_enable)(struct net_device *dev);
1375 int (*ndo_fcoe_disable)(struct net_device *dev);
1376 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1378 struct scatterlist *sgl,
1380 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1382 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1384 struct scatterlist *sgl,
1386 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1387 struct netdev_fcoe_hbainfo *hbainfo);
1390 #if IS_ENABLED(CONFIG_LIBFCOE)
1391 #define NETDEV_FCOE_WWNN 0
1392 #define NETDEV_FCOE_WWPN 1
1393 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1394 u64 *wwn, int type);
1397 #ifdef CONFIG_RFS_ACCEL
1398 int (*ndo_rx_flow_steer)(struct net_device *dev,
1399 const struct sk_buff *skb,
1403 int (*ndo_add_slave)(struct net_device *dev,
1404 struct net_device *slave_dev,
1405 struct netlink_ext_ack *extack);
1406 int (*ndo_del_slave)(struct net_device *dev,
1407 struct net_device *slave_dev);
1408 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1409 struct sk_buff *skb,
1411 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1413 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1414 netdev_features_t features);
1415 int (*ndo_set_features)(struct net_device *dev,
1416 netdev_features_t features);
1417 int (*ndo_neigh_construct)(struct net_device *dev,
1418 struct neighbour *n);
1419 void (*ndo_neigh_destroy)(struct net_device *dev,
1420 struct neighbour *n);
1422 int (*ndo_fdb_add)(struct ndmsg *ndm,
1423 struct nlattr *tb[],
1424 struct net_device *dev,
1425 const unsigned char *addr,
1428 struct netlink_ext_ack *extack);
1429 int (*ndo_fdb_del)(struct ndmsg *ndm,
1430 struct nlattr *tb[],
1431 struct net_device *dev,
1432 const unsigned char *addr,
1434 int (*ndo_fdb_dump)(struct sk_buff *skb,
1435 struct netlink_callback *cb,
1436 struct net_device *dev,
1437 struct net_device *filter_dev,
1439 int (*ndo_fdb_get)(struct sk_buff *skb,
1440 struct nlattr *tb[],
1441 struct net_device *dev,
1442 const unsigned char *addr,
1443 u16 vid, u32 portid, u32 seq,
1444 struct netlink_ext_ack *extack);
1445 int (*ndo_bridge_setlink)(struct net_device *dev,
1446 struct nlmsghdr *nlh,
1448 struct netlink_ext_ack *extack);
1449 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1451 struct net_device *dev,
1454 int (*ndo_bridge_dellink)(struct net_device *dev,
1455 struct nlmsghdr *nlh,
1457 int (*ndo_change_carrier)(struct net_device *dev,
1459 int (*ndo_get_phys_port_id)(struct net_device *dev,
1460 struct netdev_phys_item_id *ppid);
1461 int (*ndo_get_port_parent_id)(struct net_device *dev,
1462 struct netdev_phys_item_id *ppid);
1463 int (*ndo_get_phys_port_name)(struct net_device *dev,
1464 char *name, size_t len);
1465 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1466 struct net_device *dev);
1467 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1470 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1473 int (*ndo_get_iflink)(const struct net_device *dev);
1474 int (*ndo_change_proto_down)(struct net_device *dev,
1476 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1477 struct sk_buff *skb);
1478 void (*ndo_set_rx_headroom)(struct net_device *dev,
1479 int needed_headroom);
1480 int (*ndo_bpf)(struct net_device *dev,
1481 struct netdev_bpf *bpf);
1482 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1483 struct xdp_frame **xdp,
1485 int (*ndo_xsk_wakeup)(struct net_device *dev,
1486 u32 queue_id, u32 flags);
1487 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1488 int (*ndo_tunnel_ctl)(struct net_device *dev,
1489 struct ip_tunnel_parm *p, int cmd);
1490 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1494 * enum netdev_priv_flags - &struct net_device priv_flags
1496 * These are the &struct net_device, they are only set internally
1497 * by drivers and used in the kernel. These flags are invisible to
1498 * userspace; this means that the order of these flags can change
1499 * during any kernel release.
1501 * You should have a pretty good reason to be extending these flags.
1503 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1504 * @IFF_EBRIDGE: Ethernet bridging device
1505 * @IFF_BONDING: bonding master or slave
1506 * @IFF_ISATAP: ISATAP interface (RFC4214)
1507 * @IFF_WAN_HDLC: WAN HDLC device
1508 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1510 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1511 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1512 * @IFF_MACVLAN_PORT: device used as macvlan port
1513 * @IFF_BRIDGE_PORT: device used as bridge port
1514 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1515 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1516 * @IFF_UNICAST_FLT: Supports unicast filtering
1517 * @IFF_TEAM_PORT: device used as team port
1518 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1519 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1520 * change when it's running
1521 * @IFF_MACVLAN: Macvlan device
1522 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1523 * underlying stacked devices
1524 * @IFF_L3MDEV_MASTER: device is an L3 master device
1525 * @IFF_NO_QUEUE: device can run without qdisc attached
1526 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1527 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1528 * @IFF_TEAM: device is a team device
1529 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1530 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1531 * entity (i.e. the master device for bridged veth)
1532 * @IFF_MACSEC: device is a MACsec device
1533 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1534 * @IFF_FAILOVER: device is a failover master device
1535 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1536 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1537 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1538 * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1539 * skb_headlen(skb) == 0 (data starts from frag0)
1541 enum netdev_priv_flags {
1542 IFF_802_1Q_VLAN = 1<<0,
1546 IFF_WAN_HDLC = 1<<4,
1547 IFF_XMIT_DST_RELEASE = 1<<5,
1548 IFF_DONT_BRIDGE = 1<<6,
1549 IFF_DISABLE_NETPOLL = 1<<7,
1550 IFF_MACVLAN_PORT = 1<<8,
1551 IFF_BRIDGE_PORT = 1<<9,
1552 IFF_OVS_DATAPATH = 1<<10,
1553 IFF_TX_SKB_SHARING = 1<<11,
1554 IFF_UNICAST_FLT = 1<<12,
1555 IFF_TEAM_PORT = 1<<13,
1556 IFF_SUPP_NOFCS = 1<<14,
1557 IFF_LIVE_ADDR_CHANGE = 1<<15,
1558 IFF_MACVLAN = 1<<16,
1559 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1560 IFF_L3MDEV_MASTER = 1<<18,
1561 IFF_NO_QUEUE = 1<<19,
1562 IFF_OPENVSWITCH = 1<<20,
1563 IFF_L3MDEV_SLAVE = 1<<21,
1565 IFF_RXFH_CONFIGURED = 1<<23,
1566 IFF_PHONY_HEADROOM = 1<<24,
1568 IFF_NO_RX_HANDLER = 1<<26,
1569 IFF_FAILOVER = 1<<27,
1570 IFF_FAILOVER_SLAVE = 1<<28,
1571 IFF_L3MDEV_RX_HANDLER = 1<<29,
1572 IFF_LIVE_RENAME_OK = 1<<30,
1573 IFF_TX_SKB_NO_LINEAR = 1<<31,
1576 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1577 #define IFF_EBRIDGE IFF_EBRIDGE
1578 #define IFF_BONDING IFF_BONDING
1579 #define IFF_ISATAP IFF_ISATAP
1580 #define IFF_WAN_HDLC IFF_WAN_HDLC
1581 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1582 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1583 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1584 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1585 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1586 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1587 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1588 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1589 #define IFF_TEAM_PORT IFF_TEAM_PORT
1590 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1591 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1592 #define IFF_MACVLAN IFF_MACVLAN
1593 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1594 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1595 #define IFF_NO_QUEUE IFF_NO_QUEUE
1596 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1597 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1598 #define IFF_TEAM IFF_TEAM
1599 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1600 #define IFF_PHONY_HEADROOM IFF_PHONY_HEADROOM
1601 #define IFF_MACSEC IFF_MACSEC
1602 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1603 #define IFF_FAILOVER IFF_FAILOVER
1604 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1605 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1606 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1607 #define IFF_TX_SKB_NO_LINEAR IFF_TX_SKB_NO_LINEAR
1609 /* Specifies the type of the struct net_device::ml_priv pointer */
1610 enum netdev_ml_priv_type {
1616 * struct net_device - The DEVICE structure.
1618 * Actually, this whole structure is a big mistake. It mixes I/O
1619 * data with strictly "high-level" data, and it has to know about
1620 * almost every data structure used in the INET module.
1622 * @name: This is the first field of the "visible" part of this structure
1623 * (i.e. as seen by users in the "Space.c" file). It is the name
1626 * @name_node: Name hashlist node
1627 * @ifalias: SNMP alias
1628 * @mem_end: Shared memory end
1629 * @mem_start: Shared memory start
1630 * @base_addr: Device I/O address
1631 * @irq: Device IRQ number
1633 * @state: Generic network queuing layer state, see netdev_state_t
1634 * @dev_list: The global list of network devices
1635 * @napi_list: List entry used for polling NAPI devices
1636 * @unreg_list: List entry when we are unregistering the
1637 * device; see the function unregister_netdev
1638 * @close_list: List entry used when we are closing the device
1639 * @ptype_all: Device-specific packet handlers for all protocols
1640 * @ptype_specific: Device-specific, protocol-specific packet handlers
1642 * @adj_list: Directly linked devices, like slaves for bonding
1643 * @features: Currently active device features
1644 * @hw_features: User-changeable features
1646 * @wanted_features: User-requested features
1647 * @vlan_features: Mask of features inheritable by VLAN devices
1649 * @hw_enc_features: Mask of features inherited by encapsulating devices
1650 * This field indicates what encapsulation
1651 * offloads the hardware is capable of doing,
1652 * and drivers will need to set them appropriately.
1654 * @mpls_features: Mask of features inheritable by MPLS
1655 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1657 * @ifindex: interface index
1658 * @group: The group the device belongs to
1660 * @stats: Statistics struct, which was left as a legacy, use
1661 * rtnl_link_stats64 instead
1663 * @rx_dropped: Dropped packets by core network,
1664 * do not use this in drivers
1665 * @tx_dropped: Dropped packets by core network,
1666 * do not use this in drivers
1667 * @rx_nohandler: nohandler dropped packets by core network on
1668 * inactive devices, do not use this in drivers
1669 * @carrier_up_count: Number of times the carrier has been up
1670 * @carrier_down_count: Number of times the carrier has been down
1672 * @wireless_handlers: List of functions to handle Wireless Extensions,
1674 * see <net/iw_handler.h> for details.
1675 * @wireless_data: Instance data managed by the core of wireless extensions
1677 * @netdev_ops: Includes several pointers to callbacks,
1678 * if one wants to override the ndo_*() functions
1679 * @ethtool_ops: Management operations
1680 * @l3mdev_ops: Layer 3 master device operations
1681 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1682 * discovery handling. Necessary for e.g. 6LoWPAN.
1683 * @xfrmdev_ops: Transformation offload operations
1684 * @tlsdev_ops: Transport Layer Security offload operations
1685 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1686 * of Layer 2 headers.
1688 * @flags: Interface flags (a la BSD)
1689 * @priv_flags: Like 'flags' but invisible to userspace,
1690 * see if.h for the definitions
1691 * @gflags: Global flags ( kept as legacy )
1692 * @padded: How much padding added by alloc_netdev()
1693 * @operstate: RFC2863 operstate
1694 * @link_mode: Mapping policy to operstate
1695 * @if_port: Selectable AUI, TP, ...
1697 * @mtu: Interface MTU value
1698 * @min_mtu: Interface Minimum MTU value
1699 * @max_mtu: Interface Maximum MTU value
1700 * @type: Interface hardware type
1701 * @hard_header_len: Maximum hardware header length.
1702 * @min_header_len: Minimum hardware header length
1704 * @needed_headroom: Extra headroom the hardware may need, but not in all
1705 * cases can this be guaranteed
1706 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1707 * cases can this be guaranteed. Some cases also use
1708 * LL_MAX_HEADER instead to allocate the skb
1710 * interface address info:
1712 * @perm_addr: Permanent hw address
1713 * @addr_assign_type: Hw address assignment type
1714 * @addr_len: Hardware address length
1715 * @upper_level: Maximum depth level of upper devices.
1716 * @lower_level: Maximum depth level of lower devices.
1717 * @neigh_priv_len: Used in neigh_alloc()
1718 * @dev_id: Used to differentiate devices that share
1719 * the same link layer address
1720 * @dev_port: Used to differentiate devices that share
1722 * @addr_list_lock: XXX: need comments on this one
1723 * @name_assign_type: network interface name assignment type
1724 * @uc_promisc: Counter that indicates promiscuous mode
1725 * has been enabled due to the need to listen to
1726 * additional unicast addresses in a device that
1727 * does not implement ndo_set_rx_mode()
1728 * @uc: unicast mac addresses
1729 * @mc: multicast mac addresses
1730 * @dev_addrs: list of device hw addresses
1731 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1732 * @promiscuity: Number of times the NIC is told to work in
1733 * promiscuous mode; if it becomes 0 the NIC will
1734 * exit promiscuous mode
1735 * @allmulti: Counter, enables or disables allmulticast mode
1737 * @vlan_info: VLAN info
1738 * @dsa_ptr: dsa specific data
1739 * @tipc_ptr: TIPC specific data
1740 * @atalk_ptr: AppleTalk link
1741 * @ip_ptr: IPv4 specific data
1742 * @dn_ptr: DECnet specific data
1743 * @ip6_ptr: IPv6 specific data
1744 * @ax25_ptr: AX.25 specific data
1745 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1746 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1748 * @mpls_ptr: mpls_dev struct pointer
1750 * @dev_addr: Hw address (before bcast,
1751 * because most packets are unicast)
1753 * @_rx: Array of RX queues
1754 * @num_rx_queues: Number of RX queues
1755 * allocated at register_netdev() time
1756 * @real_num_rx_queues: Number of RX queues currently active in device
1757 * @xdp_prog: XDP sockets filter program pointer
1758 * @gro_flush_timeout: timeout for GRO layer in NAPI
1759 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1760 * allow to avoid NIC hard IRQ, on busy queues.
1762 * @rx_handler: handler for received packets
1763 * @rx_handler_data: XXX: need comments on this one
1764 * @miniq_ingress: ingress/clsact qdisc specific data for
1765 * ingress processing
1766 * @ingress_queue: XXX: need comments on this one
1767 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1768 * @broadcast: hw bcast address
1770 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1771 * indexed by RX queue number. Assigned by driver.
1772 * This must only be set if the ndo_rx_flow_steer
1773 * operation is defined
1774 * @index_hlist: Device index hash chain
1776 * @_tx: Array of TX queues
1777 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1778 * @real_num_tx_queues: Number of TX queues currently active in device
1779 * @qdisc: Root qdisc from userspace point of view
1780 * @tx_queue_len: Max frames per queue allowed
1781 * @tx_global_lock: XXX: need comments on this one
1782 * @xdp_bulkq: XDP device bulk queue
1783 * @xps_maps: all CPUs/RXQs maps for XPS device
1785 * @xps_maps: XXX: need comments on this one
1786 * @miniq_egress: clsact qdisc specific data for
1788 * @qdisc_hash: qdisc hash table
1789 * @watchdog_timeo: Represents the timeout that is used by
1790 * the watchdog (see dev_watchdog())
1791 * @watchdog_timer: List of timers
1793 * @proto_down_reason: reason a netdev interface is held down
1794 * @pcpu_refcnt: Number of references to this device
1795 * @todo_list: Delayed register/unregister
1796 * @link_watch_list: XXX: need comments on this one
1798 * @reg_state: Register/unregister state machine
1799 * @dismantle: Device is going to be freed
1800 * @rtnl_link_state: This enum represents the phases of creating
1803 * @needs_free_netdev: Should unregister perform free_netdev?
1804 * @priv_destructor: Called from unregister
1805 * @npinfo: XXX: need comments on this one
1806 * @nd_net: Network namespace this network device is inside
1808 * @ml_priv: Mid-layer private
1809 * @ml_priv_type: Mid-layer private type
1810 * @lstats: Loopback statistics
1811 * @tstats: Tunnel statistics
1812 * @dstats: Dummy statistics
1813 * @vstats: Virtual ethernet statistics
1818 * @dev: Class/net/name entry
1819 * @sysfs_groups: Space for optional device, statistics and wireless
1822 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1823 * @rtnl_link_ops: Rtnl_link_ops
1825 * @gso_max_size: Maximum size of generic segmentation offload
1826 * @gso_max_segs: Maximum number of segments that can be passed to the
1829 * @dcbnl_ops: Data Center Bridging netlink ops
1830 * @num_tc: Number of traffic classes in the net device
1831 * @tc_to_txq: XXX: need comments on this one
1832 * @prio_tc_map: XXX: need comments on this one
1834 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1836 * @priomap: XXX: need comments on this one
1837 * @phydev: Physical device may attach itself
1838 * for hardware timestamping
1839 * @sfp_bus: attached &struct sfp_bus structure.
1841 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1842 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1844 * @proto_down: protocol port state information can be sent to the
1845 * switch driver and used to set the phys state of the
1848 * @wol_enabled: Wake-on-LAN is enabled
1850 * @threaded: napi threaded mode is enabled
1852 * @net_notifier_list: List of per-net netdev notifier block
1853 * that follow this device when it is moved
1854 * to another network namespace.
1856 * @macsec_ops: MACsec offloading ops
1858 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1859 * offload capabilities of the device
1860 * @udp_tunnel_nic: UDP tunnel offload state
1861 * @xdp_state: stores info on attached XDP BPF programs
1863 * @nested_level: Used as as a parameter of spin_lock_nested() of
1864 * dev->addr_list_lock.
1865 * @unlink_list: As netif_addr_lock() can be called recursively,
1866 * keep a list of interfaces to be deleted.
1868 * FIXME: cleanup struct net_device such that network protocol info
1873 char name[IFNAMSIZ];
1874 struct netdev_name_node *name_node;
1875 struct dev_ifalias __rcu *ifalias;
1877 * I/O specific fields
1878 * FIXME: Merge these and struct ifmap into one
1880 unsigned long mem_end;
1881 unsigned long mem_start;
1882 unsigned long base_addr;
1885 * Some hardware also needs these fields (state,dev_list,
1886 * napi_list,unreg_list,close_list) but they are not
1887 * part of the usual set specified in Space.c.
1890 unsigned long state;
1892 struct list_head dev_list;
1893 struct list_head napi_list;
1894 struct list_head unreg_list;
1895 struct list_head close_list;
1896 struct list_head ptype_all;
1897 struct list_head ptype_specific;
1900 struct list_head upper;
1901 struct list_head lower;
1904 /* Read-mostly cache-line for fast-path access */
1906 unsigned int priv_flags;
1907 const struct net_device_ops *netdev_ops;
1909 unsigned short gflags;
1910 unsigned short hard_header_len;
1912 /* Note : dev->mtu is often read without holding a lock.
1913 * Writers usually hold RTNL.
1914 * It is recommended to use READ_ONCE() to annotate the reads,
1915 * and to use WRITE_ONCE() to annotate the writes.
1918 unsigned short needed_headroom;
1919 unsigned short needed_tailroom;
1921 netdev_features_t features;
1922 netdev_features_t hw_features;
1923 netdev_features_t wanted_features;
1924 netdev_features_t vlan_features;
1925 netdev_features_t hw_enc_features;
1926 netdev_features_t mpls_features;
1927 netdev_features_t gso_partial_features;
1929 unsigned int min_mtu;
1930 unsigned int max_mtu;
1931 unsigned short type;
1932 unsigned char min_header_len;
1933 unsigned char name_assign_type;
1937 struct net_device_stats stats; /* not used by modern drivers */
1939 atomic_long_t rx_dropped;
1940 atomic_long_t tx_dropped;
1941 atomic_long_t rx_nohandler;
1943 /* Stats to monitor link on/off, flapping */
1944 atomic_t carrier_up_count;
1945 atomic_t carrier_down_count;
1947 #ifdef CONFIG_WIRELESS_EXT
1948 const struct iw_handler_def *wireless_handlers;
1949 struct iw_public_data *wireless_data;
1951 const struct ethtool_ops *ethtool_ops;
1952 #ifdef CONFIG_NET_L3_MASTER_DEV
1953 const struct l3mdev_ops *l3mdev_ops;
1955 #if IS_ENABLED(CONFIG_IPV6)
1956 const struct ndisc_ops *ndisc_ops;
1959 #ifdef CONFIG_XFRM_OFFLOAD
1960 const struct xfrmdev_ops *xfrmdev_ops;
1963 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1964 const struct tlsdev_ops *tlsdev_ops;
1967 const struct header_ops *header_ops;
1969 unsigned char operstate;
1970 unsigned char link_mode;
1972 unsigned char if_port;
1975 /* Interface address info. */
1976 unsigned char perm_addr[MAX_ADDR_LEN];
1977 unsigned char addr_assign_type;
1978 unsigned char addr_len;
1979 unsigned char upper_level;
1980 unsigned char lower_level;
1982 unsigned short neigh_priv_len;
1983 unsigned short dev_id;
1984 unsigned short dev_port;
1985 unsigned short padded;
1987 spinlock_t addr_list_lock;
1990 struct netdev_hw_addr_list uc;
1991 struct netdev_hw_addr_list mc;
1992 struct netdev_hw_addr_list dev_addrs;
1995 struct kset *queues_kset;
1997 #ifdef CONFIG_LOCKDEP
1998 struct list_head unlink_list;
2000 unsigned int promiscuity;
2001 unsigned int allmulti;
2003 #ifdef CONFIG_LOCKDEP
2004 unsigned char nested_level;
2008 /* Protocol-specific pointers */
2010 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2011 struct vlan_info __rcu *vlan_info;
2013 #if IS_ENABLED(CONFIG_NET_DSA)
2014 struct dsa_port *dsa_ptr;
2016 #if IS_ENABLED(CONFIG_TIPC)
2017 struct tipc_bearer __rcu *tipc_ptr;
2019 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2022 struct in_device __rcu *ip_ptr;
2023 #if IS_ENABLED(CONFIG_DECNET)
2024 struct dn_dev __rcu *dn_ptr;
2026 struct inet6_dev __rcu *ip6_ptr;
2027 #if IS_ENABLED(CONFIG_AX25)
2030 struct wireless_dev *ieee80211_ptr;
2031 struct wpan_dev *ieee802154_ptr;
2032 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2033 struct mpls_dev __rcu *mpls_ptr;
2037 * Cache lines mostly used on receive path (including eth_type_trans())
2039 /* Interface address info used in eth_type_trans() */
2040 unsigned char *dev_addr;
2042 struct netdev_rx_queue *_rx;
2043 unsigned int num_rx_queues;
2044 unsigned int real_num_rx_queues;
2046 struct bpf_prog __rcu *xdp_prog;
2047 unsigned long gro_flush_timeout;
2048 int napi_defer_hard_irqs;
2049 rx_handler_func_t __rcu *rx_handler;
2050 void __rcu *rx_handler_data;
2052 #ifdef CONFIG_NET_CLS_ACT
2053 struct mini_Qdisc __rcu *miniq_ingress;
2055 struct netdev_queue __rcu *ingress_queue;
2056 #ifdef CONFIG_NETFILTER_INGRESS
2057 struct nf_hook_entries __rcu *nf_hooks_ingress;
2060 unsigned char broadcast[MAX_ADDR_LEN];
2061 #ifdef CONFIG_RFS_ACCEL
2062 struct cpu_rmap *rx_cpu_rmap;
2064 struct hlist_node index_hlist;
2067 * Cache lines mostly used on transmit path
2069 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2070 unsigned int num_tx_queues;
2071 unsigned int real_num_tx_queues;
2072 struct Qdisc *qdisc;
2073 unsigned int tx_queue_len;
2074 spinlock_t tx_global_lock;
2076 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2079 struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2081 #ifdef CONFIG_NET_CLS_ACT
2082 struct mini_Qdisc __rcu *miniq_egress;
2085 #ifdef CONFIG_NET_SCHED
2086 DECLARE_HASHTABLE (qdisc_hash, 4);
2088 /* These may be needed for future network-power-down code. */
2089 struct timer_list watchdog_timer;
2092 u32 proto_down_reason;
2094 struct list_head todo_list;
2096 #ifdef CONFIG_PCPU_DEV_REFCNT
2097 int __percpu *pcpu_refcnt;
2099 refcount_t dev_refcnt;
2102 struct list_head link_watch_list;
2104 enum { NETREG_UNINITIALIZED=0,
2105 NETREG_REGISTERED, /* completed register_netdevice */
2106 NETREG_UNREGISTERING, /* called unregister_netdevice */
2107 NETREG_UNREGISTERED, /* completed unregister todo */
2108 NETREG_RELEASED, /* called free_netdev */
2109 NETREG_DUMMY, /* dummy device for NAPI poll */
2115 RTNL_LINK_INITIALIZED,
2116 RTNL_LINK_INITIALIZING,
2117 } rtnl_link_state:16;
2119 bool needs_free_netdev;
2120 void (*priv_destructor)(struct net_device *dev);
2122 #ifdef CONFIG_NETPOLL
2123 struct netpoll_info __rcu *npinfo;
2126 possible_net_t nd_net;
2128 /* mid-layer private */
2130 enum netdev_ml_priv_type ml_priv_type;
2133 struct pcpu_lstats __percpu *lstats;
2134 struct pcpu_sw_netstats __percpu *tstats;
2135 struct pcpu_dstats __percpu *dstats;
2138 #if IS_ENABLED(CONFIG_GARP)
2139 struct garp_port __rcu *garp_port;
2141 #if IS_ENABLED(CONFIG_MRP)
2142 struct mrp_port __rcu *mrp_port;
2146 const struct attribute_group *sysfs_groups[4];
2147 const struct attribute_group *sysfs_rx_queue_group;
2149 const struct rtnl_link_ops *rtnl_link_ops;
2151 /* for setting kernel sock attribute on TCP connection setup */
2152 #define GSO_MAX_SIZE 65536
2153 unsigned int gso_max_size;
2154 #define GSO_MAX_SEGS 65535
2158 const struct dcbnl_rtnl_ops *dcbnl_ops;
2161 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2162 u8 prio_tc_map[TC_BITMASK + 1];
2164 #if IS_ENABLED(CONFIG_FCOE)
2165 unsigned int fcoe_ddp_xid;
2167 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2168 struct netprio_map __rcu *priomap;
2170 struct phy_device *phydev;
2171 struct sfp_bus *sfp_bus;
2172 struct lock_class_key *qdisc_tx_busylock;
2173 struct lock_class_key *qdisc_running_key;
2175 unsigned wol_enabled:1;
2176 unsigned threaded:1;
2178 struct list_head net_notifier_list;
2180 #if IS_ENABLED(CONFIG_MACSEC)
2181 /* MACsec management functions */
2182 const struct macsec_ops *macsec_ops;
2184 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2185 struct udp_tunnel_nic *udp_tunnel_nic;
2187 /* protected by rtnl_lock */
2188 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2190 #define to_net_dev(d) container_of(d, struct net_device, dev)
2192 static inline bool netif_elide_gro(const struct net_device *dev)
2194 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2199 #define NETDEV_ALIGN 32
2202 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2204 return dev->prio_tc_map[prio & TC_BITMASK];
2208 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2210 if (tc >= dev->num_tc)
2213 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2217 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2218 void netdev_reset_tc(struct net_device *dev);
2219 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2220 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2223 int netdev_get_num_tc(struct net_device *dev)
2228 static inline void net_prefetch(void *p)
2231 #if L1_CACHE_BYTES < 128
2232 prefetch((u8 *)p + L1_CACHE_BYTES);
2236 static inline void net_prefetchw(void *p)
2239 #if L1_CACHE_BYTES < 128
2240 prefetchw((u8 *)p + L1_CACHE_BYTES);
2244 void netdev_unbind_sb_channel(struct net_device *dev,
2245 struct net_device *sb_dev);
2246 int netdev_bind_sb_channel_queue(struct net_device *dev,
2247 struct net_device *sb_dev,
2248 u8 tc, u16 count, u16 offset);
2249 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2250 static inline int netdev_get_sb_channel(struct net_device *dev)
2252 return max_t(int, -dev->num_tc, 0);
2256 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2259 return &dev->_tx[index];
2262 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2263 const struct sk_buff *skb)
2265 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2268 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2269 void (*f)(struct net_device *,
2270 struct netdev_queue *,
2276 for (i = 0; i < dev->num_tx_queues; i++)
2277 f(dev, &dev->_tx[i], arg);
2280 #define netdev_lockdep_set_classes(dev) \
2282 static struct lock_class_key qdisc_tx_busylock_key; \
2283 static struct lock_class_key qdisc_running_key; \
2284 static struct lock_class_key qdisc_xmit_lock_key; \
2285 static struct lock_class_key dev_addr_list_lock_key; \
2288 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2289 (dev)->qdisc_running_key = &qdisc_running_key; \
2290 lockdep_set_class(&(dev)->addr_list_lock, \
2291 &dev_addr_list_lock_key); \
2292 for (i = 0; i < (dev)->num_tx_queues; i++) \
2293 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2294 &qdisc_xmit_lock_key); \
2297 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2298 struct net_device *sb_dev);
2299 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2300 struct sk_buff *skb,
2301 struct net_device *sb_dev);
2303 /* returns the headroom that the master device needs to take in account
2304 * when forwarding to this dev
2306 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2308 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2311 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2313 if (dev->netdev_ops->ndo_set_rx_headroom)
2314 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2317 /* set the device rx headroom to the dev's default */
2318 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2320 netdev_set_rx_headroom(dev, -1);
2323 static inline void *netdev_get_ml_priv(struct net_device *dev,
2324 enum netdev_ml_priv_type type)
2326 if (dev->ml_priv_type != type)
2329 return dev->ml_priv;
2332 static inline void netdev_set_ml_priv(struct net_device *dev,
2334 enum netdev_ml_priv_type type)
2336 WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2337 "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2338 dev->ml_priv_type, type);
2339 WARN(!dev->ml_priv_type && dev->ml_priv,
2340 "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2342 dev->ml_priv = ml_priv;
2343 dev->ml_priv_type = type;
2347 * Net namespace inlines
2350 struct net *dev_net(const struct net_device *dev)
2352 return read_pnet(&dev->nd_net);
2356 void dev_net_set(struct net_device *dev, struct net *net)
2358 write_pnet(&dev->nd_net, net);
2362 * netdev_priv - access network device private data
2363 * @dev: network device
2365 * Get network device private data
2367 static inline void *netdev_priv(const struct net_device *dev)
2369 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2372 /* Set the sysfs physical device reference for the network logical device
2373 * if set prior to registration will cause a symlink during initialization.
2375 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2377 /* Set the sysfs device type for the network logical device to allow
2378 * fine-grained identification of different network device types. For
2379 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2381 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2383 /* Default NAPI poll() weight
2384 * Device drivers are strongly advised to not use bigger value
2386 #define NAPI_POLL_WEIGHT 64
2389 * netif_napi_add - initialize a NAPI context
2390 * @dev: network device
2391 * @napi: NAPI context
2392 * @poll: polling function
2393 * @weight: default weight
2395 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2396 * *any* of the other NAPI-related functions.
2398 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2399 int (*poll)(struct napi_struct *, int), int weight);
2402 * netif_tx_napi_add - initialize a NAPI context
2403 * @dev: network device
2404 * @napi: NAPI context
2405 * @poll: polling function
2406 * @weight: default weight
2408 * This variant of netif_napi_add() should be used from drivers using NAPI
2409 * to exclusively poll a TX queue.
2410 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2412 static inline void netif_tx_napi_add(struct net_device *dev,
2413 struct napi_struct *napi,
2414 int (*poll)(struct napi_struct *, int),
2417 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2418 netif_napi_add(dev, napi, poll, weight);
2422 * __netif_napi_del - remove a NAPI context
2423 * @napi: NAPI context
2425 * Warning: caller must observe RCU grace period before freeing memory
2426 * containing @napi. Drivers might want to call this helper to combine
2427 * all the needed RCU grace periods into a single one.
2429 void __netif_napi_del(struct napi_struct *napi);
2432 * netif_napi_del - remove a NAPI context
2433 * @napi: NAPI context
2435 * netif_napi_del() removes a NAPI context from the network device NAPI list
2437 static inline void netif_napi_del(struct napi_struct *napi)
2439 __netif_napi_del(napi);
2443 struct napi_gro_cb {
2444 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2447 /* Length of frag0. */
2448 unsigned int frag0_len;
2450 /* This indicates where we are processing relative to skb->data. */
2453 /* This is non-zero if the packet cannot be merged with the new skb. */
2456 /* Save the IP ID here and check when we get to the transport layer */
2459 /* Number of segments aggregated. */
2462 /* Start offset for remote checksum offload */
2463 u16 gro_remcsum_start;
2465 /* jiffies when first packet was created/queued */
2468 /* Used in ipv6_gro_receive() and foo-over-udp */
2471 /* This is non-zero if the packet may be of the same flow. */
2474 /* Used in tunnel GRO receive */
2477 /* GRO checksum is valid */
2480 /* Number of checksums via CHECKSUM_UNNECESSARY */
2485 #define NAPI_GRO_FREE 1
2486 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2488 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2491 /* Used in GRE, set in fou/gue_gro_receive */
2494 /* Used to determine if flush_id can be ignored */
2497 /* Number of gro_receive callbacks this packet already went through */
2498 u8 recursion_counter:4;
2500 /* GRO is done by frag_list pointer chaining. */
2503 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2506 /* used in skb_gro_receive() slow path */
2507 struct sk_buff *last;
2510 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2512 #define GRO_RECURSION_LIMIT 15
2513 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2515 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2518 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2519 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2520 struct list_head *head,
2521 struct sk_buff *skb)
2523 if (unlikely(gro_recursion_inc_test(skb))) {
2524 NAPI_GRO_CB(skb)->flush |= 1;
2528 return cb(head, skb);
2531 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2533 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2535 struct list_head *head,
2536 struct sk_buff *skb)
2538 if (unlikely(gro_recursion_inc_test(skb))) {
2539 NAPI_GRO_CB(skb)->flush |= 1;
2543 return cb(sk, head, skb);
2546 struct packet_type {
2547 __be16 type; /* This is really htons(ether_type). */
2548 bool ignore_outgoing;
2549 struct net_device *dev; /* NULL is wildcarded here */
2550 int (*func) (struct sk_buff *,
2551 struct net_device *,
2552 struct packet_type *,
2553 struct net_device *);
2554 void (*list_func) (struct list_head *,
2555 struct packet_type *,
2556 struct net_device *);
2557 bool (*id_match)(struct packet_type *ptype,
2559 void *af_packet_priv;
2560 struct list_head list;
2563 struct offload_callbacks {
2564 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2565 netdev_features_t features);
2566 struct sk_buff *(*gro_receive)(struct list_head *head,
2567 struct sk_buff *skb);
2568 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2571 struct packet_offload {
2572 __be16 type; /* This is really htons(ether_type). */
2574 struct offload_callbacks callbacks;
2575 struct list_head list;
2578 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2579 struct pcpu_sw_netstats {
2584 struct u64_stats_sync syncp;
2585 } __aligned(4 * sizeof(u64));
2587 struct pcpu_lstats {
2588 u64_stats_t packets;
2590 struct u64_stats_sync syncp;
2591 } __aligned(2 * sizeof(u64));
2593 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2595 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2597 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2599 u64_stats_update_begin(&tstats->syncp);
2600 tstats->rx_bytes += len;
2601 tstats->rx_packets++;
2602 u64_stats_update_end(&tstats->syncp);
2605 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2606 unsigned int packets,
2609 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2611 u64_stats_update_begin(&tstats->syncp);
2612 tstats->tx_bytes += len;
2613 tstats->tx_packets += packets;
2614 u64_stats_update_end(&tstats->syncp);
2617 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2619 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2621 u64_stats_update_begin(&lstats->syncp);
2622 u64_stats_add(&lstats->bytes, len);
2623 u64_stats_inc(&lstats->packets);
2624 u64_stats_update_end(&lstats->syncp);
2627 #define __netdev_alloc_pcpu_stats(type, gfp) \
2629 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2632 for_each_possible_cpu(__cpu) { \
2633 typeof(type) *stat; \
2634 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2635 u64_stats_init(&stat->syncp); \
2641 #define netdev_alloc_pcpu_stats(type) \
2642 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2644 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2646 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2649 for_each_possible_cpu(__cpu) { \
2650 typeof(type) *stat; \
2651 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2652 u64_stats_init(&stat->syncp); \
2658 enum netdev_lag_tx_type {
2659 NETDEV_LAG_TX_TYPE_UNKNOWN,
2660 NETDEV_LAG_TX_TYPE_RANDOM,
2661 NETDEV_LAG_TX_TYPE_BROADCAST,
2662 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2663 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2664 NETDEV_LAG_TX_TYPE_HASH,
2667 enum netdev_lag_hash {
2668 NETDEV_LAG_HASH_NONE,
2670 NETDEV_LAG_HASH_L34,
2671 NETDEV_LAG_HASH_L23,
2672 NETDEV_LAG_HASH_E23,
2673 NETDEV_LAG_HASH_E34,
2674 NETDEV_LAG_HASH_VLAN_SRCMAC,
2675 NETDEV_LAG_HASH_UNKNOWN,
2678 struct netdev_lag_upper_info {
2679 enum netdev_lag_tx_type tx_type;
2680 enum netdev_lag_hash hash_type;
2683 struct netdev_lag_lower_state_info {
2688 #include <linux/notifier.h>
2690 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2691 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2695 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2697 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2698 detected a hardware crash and restarted
2699 - we can use this eg to kick tcp sessions
2701 NETDEV_CHANGE, /* Notify device state change */
2704 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2705 NETDEV_CHANGEADDR, /* notify after the address change */
2706 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2710 NETDEV_BONDING_FAILOVER,
2712 NETDEV_PRE_TYPE_CHANGE,
2713 NETDEV_POST_TYPE_CHANGE,
2716 NETDEV_NOTIFY_PEERS,
2720 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2721 NETDEV_CHANGEINFODATA,
2722 NETDEV_BONDING_INFO,
2723 NETDEV_PRECHANGEUPPER,
2724 NETDEV_CHANGELOWERSTATE,
2725 NETDEV_UDP_TUNNEL_PUSH_INFO,
2726 NETDEV_UDP_TUNNEL_DROP_INFO,
2727 NETDEV_CHANGE_TX_QUEUE_LEN,
2728 NETDEV_CVLAN_FILTER_PUSH_INFO,
2729 NETDEV_CVLAN_FILTER_DROP_INFO,
2730 NETDEV_SVLAN_FILTER_PUSH_INFO,
2731 NETDEV_SVLAN_FILTER_DROP_INFO,
2733 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2735 int register_netdevice_notifier(struct notifier_block *nb);
2736 int unregister_netdevice_notifier(struct notifier_block *nb);
2737 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2738 int unregister_netdevice_notifier_net(struct net *net,
2739 struct notifier_block *nb);
2740 int register_netdevice_notifier_dev_net(struct net_device *dev,
2741 struct notifier_block *nb,
2742 struct netdev_net_notifier *nn);
2743 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2744 struct notifier_block *nb,
2745 struct netdev_net_notifier *nn);
2747 struct netdev_notifier_info {
2748 struct net_device *dev;
2749 struct netlink_ext_ack *extack;
2752 struct netdev_notifier_info_ext {
2753 struct netdev_notifier_info info; /* must be first */
2759 struct netdev_notifier_change_info {
2760 struct netdev_notifier_info info; /* must be first */
2761 unsigned int flags_changed;
2764 struct netdev_notifier_changeupper_info {
2765 struct netdev_notifier_info info; /* must be first */
2766 struct net_device *upper_dev; /* new upper dev */
2767 bool master; /* is upper dev master */
2768 bool linking; /* is the notification for link or unlink */
2769 void *upper_info; /* upper dev info */
2772 struct netdev_notifier_changelowerstate_info {
2773 struct netdev_notifier_info info; /* must be first */
2774 void *lower_state_info; /* is lower dev state */
2777 struct netdev_notifier_pre_changeaddr_info {
2778 struct netdev_notifier_info info; /* must be first */
2779 const unsigned char *dev_addr;
2782 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2783 struct net_device *dev)
2786 info->extack = NULL;
2789 static inline struct net_device *
2790 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2795 static inline struct netlink_ext_ack *
2796 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2798 return info->extack;
2801 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2804 extern rwlock_t dev_base_lock; /* Device list lock */
2806 #define for_each_netdev(net, d) \
2807 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2808 #define for_each_netdev_reverse(net, d) \
2809 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2810 #define for_each_netdev_rcu(net, d) \
2811 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2812 #define for_each_netdev_safe(net, d, n) \
2813 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2814 #define for_each_netdev_continue(net, d) \
2815 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2816 #define for_each_netdev_continue_reverse(net, d) \
2817 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2819 #define for_each_netdev_continue_rcu(net, d) \
2820 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2821 #define for_each_netdev_in_bond_rcu(bond, slave) \
2822 for_each_netdev_rcu(&init_net, slave) \
2823 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2824 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2826 static inline struct net_device *next_net_device(struct net_device *dev)
2828 struct list_head *lh;
2832 lh = dev->dev_list.next;
2833 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2836 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2838 struct list_head *lh;
2842 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2843 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2846 static inline struct net_device *first_net_device(struct net *net)
2848 return list_empty(&net->dev_base_head) ? NULL :
2849 net_device_entry(net->dev_base_head.next);
2852 static inline struct net_device *first_net_device_rcu(struct net *net)
2854 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2856 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2859 int netdev_boot_setup_check(struct net_device *dev);
2860 unsigned long netdev_boot_base(const char *prefix, int unit);
2861 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2862 const char *hwaddr);
2863 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2864 void dev_add_pack(struct packet_type *pt);
2865 void dev_remove_pack(struct packet_type *pt);
2866 void __dev_remove_pack(struct packet_type *pt);
2867 void dev_add_offload(struct packet_offload *po);
2868 void dev_remove_offload(struct packet_offload *po);
2870 int dev_get_iflink(const struct net_device *dev);
2871 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2872 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2873 unsigned short mask);
2874 struct net_device *dev_get_by_name(struct net *net, const char *name);
2875 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2876 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2877 int dev_alloc_name(struct net_device *dev, const char *name);
2878 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2879 void dev_close(struct net_device *dev);
2880 void dev_close_many(struct list_head *head, bool unlink);
2881 void dev_disable_lro(struct net_device *dev);
2882 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2883 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2884 struct net_device *sb_dev);
2885 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2886 struct net_device *sb_dev);
2888 int dev_queue_xmit(struct sk_buff *skb);
2889 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2890 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2892 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2896 ret = __dev_direct_xmit(skb, queue_id);
2897 if (!dev_xmit_complete(ret))
2902 int register_netdevice(struct net_device *dev);
2903 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2904 void unregister_netdevice_many(struct list_head *head);
2905 static inline void unregister_netdevice(struct net_device *dev)
2907 unregister_netdevice_queue(dev, NULL);
2910 int netdev_refcnt_read(const struct net_device *dev);
2911 void free_netdev(struct net_device *dev);
2912 void netdev_freemem(struct net_device *dev);
2913 int init_dummy_netdev(struct net_device *dev);
2915 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2916 struct sk_buff *skb,
2918 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2920 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2921 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2922 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2923 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2924 int netdev_get_name(struct net *net, char *name, int ifindex);
2925 int dev_restart(struct net_device *dev);
2926 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2927 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2929 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2931 return NAPI_GRO_CB(skb)->data_offset;
2934 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2936 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2939 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2941 NAPI_GRO_CB(skb)->data_offset += len;
2944 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2945 unsigned int offset)
2947 return NAPI_GRO_CB(skb)->frag0 + offset;
2950 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2952 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2955 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2957 NAPI_GRO_CB(skb)->frag0 = NULL;
2958 NAPI_GRO_CB(skb)->frag0_len = 0;
2961 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2962 unsigned int offset)
2964 if (!pskb_may_pull(skb, hlen))
2967 skb_gro_frag0_invalidate(skb);
2968 return skb->data + offset;
2971 static inline void *skb_gro_network_header(struct sk_buff *skb)
2973 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2974 skb_network_offset(skb);
2977 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2978 const void *start, unsigned int len)
2980 if (NAPI_GRO_CB(skb)->csum_valid)
2981 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2982 csum_partial(start, len, 0));
2985 /* GRO checksum functions. These are logical equivalents of the normal
2986 * checksum functions (in skbuff.h) except that they operate on the GRO
2987 * offsets and fields in sk_buff.
2990 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2992 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2994 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2997 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
3001 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
3002 skb_checksum_start_offset(skb) <
3003 skb_gro_offset(skb)) &&
3004 !skb_at_gro_remcsum_start(skb) &&
3005 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3006 (!zero_okay || check));
3009 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
3012 if (NAPI_GRO_CB(skb)->csum_valid &&
3013 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
3016 NAPI_GRO_CB(skb)->csum = psum;
3018 return __skb_gro_checksum_complete(skb);
3021 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
3023 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
3024 /* Consume a checksum from CHECKSUM_UNNECESSARY */
3025 NAPI_GRO_CB(skb)->csum_cnt--;
3027 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
3028 * verified a new top level checksum or an encapsulated one
3029 * during GRO. This saves work if we fallback to normal path.
3031 __skb_incr_checksum_unnecessary(skb);
3035 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
3038 __sum16 __ret = 0; \
3039 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
3040 __ret = __skb_gro_checksum_validate_complete(skb, \
3041 compute_pseudo(skb, proto)); \
3043 skb_gro_incr_csum_unnecessary(skb); \
3047 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
3048 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
3050 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
3052 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3054 #define skb_gro_checksum_simple_validate(skb) \
3055 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3057 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3059 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3060 !NAPI_GRO_CB(skb)->csum_valid);
3063 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3066 NAPI_GRO_CB(skb)->csum = ~pseudo;
3067 NAPI_GRO_CB(skb)->csum_valid = 1;
3070 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3072 if (__skb_gro_checksum_convert_check(skb)) \
3073 __skb_gro_checksum_convert(skb, \
3074 compute_pseudo(skb, proto)); \
3077 struct gro_remcsum {
3082 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3088 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3089 unsigned int off, size_t hdrlen,
3090 int start, int offset,
3091 struct gro_remcsum *grc,
3095 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3097 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3100 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3104 ptr = skb_gro_header_fast(skb, off);
3105 if (skb_gro_header_hard(skb, off + plen)) {
3106 ptr = skb_gro_header_slow(skb, off + plen, off);
3111 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3114 /* Adjust skb->csum since we changed the packet */
3115 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3117 grc->offset = off + hdrlen + offset;
3123 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3124 struct gro_remcsum *grc)
3127 size_t plen = grc->offset + sizeof(u16);
3132 ptr = skb_gro_header_fast(skb, grc->offset);
3133 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3134 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3139 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3142 #ifdef CONFIG_XFRM_OFFLOAD
3143 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3145 if (PTR_ERR(pp) != -EINPROGRESS)
3146 NAPI_GRO_CB(skb)->flush |= flush;
3148 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3151 struct gro_remcsum *grc)
3153 if (PTR_ERR(pp) != -EINPROGRESS) {
3154 NAPI_GRO_CB(skb)->flush |= flush;
3155 skb_gro_remcsum_cleanup(skb, grc);
3156 skb->remcsum_offload = 0;
3160 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3162 NAPI_GRO_CB(skb)->flush |= flush;
3164 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3167 struct gro_remcsum *grc)
3169 NAPI_GRO_CB(skb)->flush |= flush;
3170 skb_gro_remcsum_cleanup(skb, grc);
3171 skb->remcsum_offload = 0;
3175 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3176 unsigned short type,
3177 const void *daddr, const void *saddr,
3180 if (!dev->header_ops || !dev->header_ops->create)
3183 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3186 static inline int dev_parse_header(const struct sk_buff *skb,
3187 unsigned char *haddr)
3189 const struct net_device *dev = skb->dev;
3191 if (!dev->header_ops || !dev->header_ops->parse)
3193 return dev->header_ops->parse(skb, haddr);
3196 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3198 const struct net_device *dev = skb->dev;
3200 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3202 return dev->header_ops->parse_protocol(skb);
3205 /* ll_header must have at least hard_header_len allocated */
3206 static inline bool dev_validate_header(const struct net_device *dev,
3207 char *ll_header, int len)
3209 if (likely(len >= dev->hard_header_len))
3211 if (len < dev->min_header_len)
3214 if (capable(CAP_SYS_RAWIO)) {
3215 memset(ll_header + len, 0, dev->hard_header_len - len);
3219 if (dev->header_ops && dev->header_ops->validate)
3220 return dev->header_ops->validate(ll_header, len);
3225 static inline bool dev_has_header(const struct net_device *dev)
3227 return dev->header_ops && dev->header_ops->create;
3230 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3232 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3233 static inline int unregister_gifconf(unsigned int family)
3235 return register_gifconf(family, NULL);
3238 #ifdef CONFIG_NET_FLOW_LIMIT
3239 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3240 struct sd_flow_limit {
3242 unsigned int num_buckets;
3243 unsigned int history_head;
3244 u16 history[FLOW_LIMIT_HISTORY];
3248 extern int netdev_flow_limit_table_len;
3249 #endif /* CONFIG_NET_FLOW_LIMIT */
3252 * Incoming packets are placed on per-CPU queues
3254 struct softnet_data {
3255 struct list_head poll_list;
3256 struct sk_buff_head process_queue;
3259 unsigned int processed;
3260 unsigned int time_squeeze;
3261 unsigned int received_rps;
3263 struct softnet_data *rps_ipi_list;
3265 #ifdef CONFIG_NET_FLOW_LIMIT
3266 struct sd_flow_limit __rcu *flow_limit;
3268 struct Qdisc *output_queue;
3269 struct Qdisc **output_queue_tailp;
3270 struct sk_buff *completion_queue;
3271 #ifdef CONFIG_XFRM_OFFLOAD
3272 struct sk_buff_head xfrm_backlog;
3274 /* written and read only by owning cpu: */
3280 /* input_queue_head should be written by cpu owning this struct,
3281 * and only read by other cpus. Worth using a cache line.
3283 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3285 /* Elements below can be accessed between CPUs for RPS/RFS */
3286 call_single_data_t csd ____cacheline_aligned_in_smp;
3287 struct softnet_data *rps_ipi_next;
3289 unsigned int input_queue_tail;
3291 unsigned int dropped;
3292 struct sk_buff_head input_pkt_queue;
3293 struct napi_struct backlog;
3297 static inline void input_queue_head_incr(struct softnet_data *sd)
3300 sd->input_queue_head++;
3304 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3305 unsigned int *qtail)
3308 *qtail = ++sd->input_queue_tail;
3312 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3314 static inline int dev_recursion_level(void)
3316 return this_cpu_read(softnet_data.xmit.recursion);
3319 #define XMIT_RECURSION_LIMIT 8
3320 static inline bool dev_xmit_recursion(void)
3322 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3323 XMIT_RECURSION_LIMIT);
3326 static inline void dev_xmit_recursion_inc(void)
3328 __this_cpu_inc(softnet_data.xmit.recursion);
3331 static inline void dev_xmit_recursion_dec(void)
3333 __this_cpu_dec(softnet_data.xmit.recursion);
3336 void __netif_schedule(struct Qdisc *q);
3337 void netif_schedule_queue(struct netdev_queue *txq);
3339 static inline void netif_tx_schedule_all(struct net_device *dev)
3343 for (i = 0; i < dev->num_tx_queues; i++)
3344 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3347 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3349 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3353 * netif_start_queue - allow transmit
3354 * @dev: network device
3356 * Allow upper layers to call the device hard_start_xmit routine.
3358 static inline void netif_start_queue(struct net_device *dev)
3360 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3363 static inline void netif_tx_start_all_queues(struct net_device *dev)
3367 for (i = 0; i < dev->num_tx_queues; i++) {
3368 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3369 netif_tx_start_queue(txq);
3373 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3376 * netif_wake_queue - restart transmit
3377 * @dev: network device
3379 * Allow upper layers to call the device hard_start_xmit routine.
3380 * Used for flow control when transmit resources are available.
3382 static inline void netif_wake_queue(struct net_device *dev)
3384 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3387 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3391 for (i = 0; i < dev->num_tx_queues; i++) {
3392 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3393 netif_tx_wake_queue(txq);
3397 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3399 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3403 * netif_stop_queue - stop transmitted packets
3404 * @dev: network device
3406 * Stop upper layers calling the device hard_start_xmit routine.
3407 * Used for flow control when transmit resources are unavailable.
3409 static inline void netif_stop_queue(struct net_device *dev)
3411 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3414 void netif_tx_stop_all_queues(struct net_device *dev);
3416 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3418 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3422 * netif_queue_stopped - test if transmit queue is flowblocked
3423 * @dev: network device
3425 * Test if transmit queue on device is currently unable to send.
3427 static inline bool netif_queue_stopped(const struct net_device *dev)
3429 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3432 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3434 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3438 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3440 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3444 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3446 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3450 * netdev_queue_set_dql_min_limit - set dql minimum limit
3451 * @dev_queue: pointer to transmit queue
3452 * @min_limit: dql minimum limit
3454 * Forces xmit_more() to return true until the minimum threshold
3455 * defined by @min_limit is reached (or until the tx queue is
3456 * empty). Warning: to be use with care, misuse will impact the
3459 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3460 unsigned int min_limit)
3463 dev_queue->dql.min_limit = min_limit;
3468 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3469 * @dev_queue: pointer to transmit queue
3471 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3472 * to give appropriate hint to the CPU.
3474 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3477 prefetchw(&dev_queue->dql.num_queued);
3482 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3483 * @dev_queue: pointer to transmit queue
3485 * BQL enabled drivers might use this helper in their TX completion path,
3486 * to give appropriate hint to the CPU.
3488 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3491 prefetchw(&dev_queue->dql.limit);
3495 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3499 dql_queued(&dev_queue->dql, bytes);
3501 if (likely(dql_avail(&dev_queue->dql) >= 0))
3504 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3507 * The XOFF flag must be set before checking the dql_avail below,
3508 * because in netdev_tx_completed_queue we update the dql_completed
3509 * before checking the XOFF flag.
3513 /* check again in case another CPU has just made room avail */
3514 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3515 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3519 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3520 * that they should not test BQL status themselves.
3521 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3523 * Returns true if the doorbell must be used to kick the NIC.
3525 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3531 dql_queued(&dev_queue->dql, bytes);
3533 return netif_tx_queue_stopped(dev_queue);
3535 netdev_tx_sent_queue(dev_queue, bytes);
3540 * netdev_sent_queue - report the number of bytes queued to hardware
3541 * @dev: network device
3542 * @bytes: number of bytes queued to the hardware device queue
3544 * Report the number of bytes queued for sending/completion to the network
3545 * device hardware queue. @bytes should be a good approximation and should
3546 * exactly match netdev_completed_queue() @bytes
3548 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3550 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3553 static inline bool __netdev_sent_queue(struct net_device *dev,
3557 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3561 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3562 unsigned int pkts, unsigned int bytes)
3565 if (unlikely(!bytes))
3568 dql_completed(&dev_queue->dql, bytes);
3571 * Without the memory barrier there is a small possiblity that
3572 * netdev_tx_sent_queue will miss the update and cause the queue to
3573 * be stopped forever
3577 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3580 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3581 netif_schedule_queue(dev_queue);
3586 * netdev_completed_queue - report bytes and packets completed by device
3587 * @dev: network device
3588 * @pkts: actual number of packets sent over the medium
3589 * @bytes: actual number of bytes sent over the medium
3591 * Report the number of bytes and packets transmitted by the network device
3592 * hardware queue over the physical medium, @bytes must exactly match the
3593 * @bytes amount passed to netdev_sent_queue()
3595 static inline void netdev_completed_queue(struct net_device *dev,
3596 unsigned int pkts, unsigned int bytes)
3598 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3601 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3604 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3610 * netdev_reset_queue - reset the packets and bytes count of a network device
3611 * @dev_queue: network device
3613 * Reset the bytes and packet count of a network device and clear the
3614 * software flow control OFF bit for this network device
3616 static inline void netdev_reset_queue(struct net_device *dev_queue)
3618 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3622 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3623 * @dev: network device
3624 * @queue_index: given tx queue index
3626 * Returns 0 if given tx queue index >= number of device tx queues,
3627 * otherwise returns the originally passed tx queue index.
3629 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3631 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3632 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3633 dev->name, queue_index,
3634 dev->real_num_tx_queues);
3642 * netif_running - test if up
3643 * @dev: network device
3645 * Test if the device has been brought up.
3647 static inline bool netif_running(const struct net_device *dev)
3649 return test_bit(__LINK_STATE_START, &dev->state);
3653 * Routines to manage the subqueues on a device. We only need start,
3654 * stop, and a check if it's stopped. All other device management is
3655 * done at the overall netdevice level.
3656 * Also test the device if we're multiqueue.
3660 * netif_start_subqueue - allow sending packets on subqueue
3661 * @dev: network device
3662 * @queue_index: sub queue index
3664 * Start individual transmit queue of a device with multiple transmit queues.
3666 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3668 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3670 netif_tx_start_queue(txq);
3674 * netif_stop_subqueue - stop sending packets on subqueue
3675 * @dev: network device
3676 * @queue_index: sub queue index
3678 * Stop individual transmit queue of a device with multiple transmit queues.
3680 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3682 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3683 netif_tx_stop_queue(txq);
3687 * __netif_subqueue_stopped - test status of subqueue
3688 * @dev: network device
3689 * @queue_index: sub queue index
3691 * Check individual transmit queue of a device with multiple transmit queues.
3693 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3696 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3698 return netif_tx_queue_stopped(txq);
3702 * netif_subqueue_stopped - test status of subqueue
3703 * @dev: network device
3704 * @skb: sub queue buffer pointer
3706 * Check individual transmit queue of a device with multiple transmit queues.
3708 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3709 struct sk_buff *skb)
3711 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3715 * netif_wake_subqueue - allow sending packets on subqueue
3716 * @dev: network device
3717 * @queue_index: sub queue index
3719 * Resume individual transmit queue of a device with multiple transmit queues.
3721 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3723 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3725 netif_tx_wake_queue(txq);
3729 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3731 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3732 u16 index, enum xps_map_type type);
3735 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3736 * @j: CPU/Rx queue index
3737 * @mask: bitmask of all cpus/rx queues
3738 * @nr_bits: number of bits in the bitmask
3740 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3742 static inline bool netif_attr_test_mask(unsigned long j,
3743 const unsigned long *mask,
3744 unsigned int nr_bits)
3746 cpu_max_bits_warn(j, nr_bits);
3747 return test_bit(j, mask);
3751 * netif_attr_test_online - Test for online CPU/Rx queue
3752 * @j: CPU/Rx queue index
3753 * @online_mask: bitmask for CPUs/Rx queues that are online
3754 * @nr_bits: number of bits in the bitmask
3756 * Returns true if a CPU/Rx queue is online.
3758 static inline bool netif_attr_test_online(unsigned long j,
3759 const unsigned long *online_mask,
3760 unsigned int nr_bits)
3762 cpu_max_bits_warn(j, nr_bits);
3765 return test_bit(j, online_mask);
3767 return (j < nr_bits);
3771 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3772 * @n: CPU/Rx queue index
3773 * @srcp: the cpumask/Rx queue mask pointer
3774 * @nr_bits: number of bits in the bitmask
3776 * Returns >= nr_bits if no further CPUs/Rx queues set.
3778 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3779 unsigned int nr_bits)
3781 /* -1 is a legal arg here. */
3783 cpu_max_bits_warn(n, nr_bits);
3786 return find_next_bit(srcp, nr_bits, n + 1);
3792 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3793 * @n: CPU/Rx queue index
3794 * @src1p: the first CPUs/Rx queues mask pointer
3795 * @src2p: the second CPUs/Rx queues mask pointer
3796 * @nr_bits: number of bits in the bitmask
3798 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3800 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3801 const unsigned long *src2p,
3802 unsigned int nr_bits)
3804 /* -1 is a legal arg here. */
3806 cpu_max_bits_warn(n, nr_bits);
3809 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3811 return find_next_bit(src1p, nr_bits, n + 1);
3813 return find_next_bit(src2p, nr_bits, n + 1);
3818 static inline int netif_set_xps_queue(struct net_device *dev,
3819 const struct cpumask *mask,
3825 static inline int __netif_set_xps_queue(struct net_device *dev,
3826 const unsigned long *mask,
3827 u16 index, enum xps_map_type type)
3834 * netif_is_multiqueue - test if device has multiple transmit queues
3835 * @dev: network device
3837 * Check if device has multiple transmit queues
3839 static inline bool netif_is_multiqueue(const struct net_device *dev)
3841 return dev->num_tx_queues > 1;
3844 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3847 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3849 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3852 dev->real_num_rx_queues = rxqs;
3857 static inline struct netdev_rx_queue *
3858 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3860 return dev->_rx + rxq;
3864 static inline unsigned int get_netdev_rx_queue_index(
3865 struct netdev_rx_queue *queue)
3867 struct net_device *dev = queue->dev;
3868 int index = queue - dev->_rx;
3870 BUG_ON(index >= dev->num_rx_queues);
3875 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3876 int netif_get_num_default_rss_queues(void);
3878 enum skb_free_reason {
3879 SKB_REASON_CONSUMED,
3883 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3884 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3887 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3888 * interrupt context or with hardware interrupts being disabled.
3889 * (in_irq() || irqs_disabled())
3891 * We provide four helpers that can be used in following contexts :
3893 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3894 * replacing kfree_skb(skb)
3896 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3897 * Typically used in place of consume_skb(skb) in TX completion path
3899 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3900 * replacing kfree_skb(skb)
3902 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3903 * and consumed a packet. Used in place of consume_skb(skb)
3905 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3907 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3910 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3912 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3915 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3917 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3920 static inline void dev_consume_skb_any(struct sk_buff *skb)
3922 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3925 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3926 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3927 int netif_rx(struct sk_buff *skb);
3928 int netif_rx_ni(struct sk_buff *skb);
3929 int netif_rx_any_context(struct sk_buff *skb);
3930 int netif_receive_skb(struct sk_buff *skb);
3931 int netif_receive_skb_core(struct sk_buff *skb);
3932 void netif_receive_skb_list(struct list_head *head);
3933 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3934 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3935 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3936 gro_result_t napi_gro_frags(struct napi_struct *napi);
3937 struct packet_offload *gro_find_receive_by_type(__be16 type);
3938 struct packet_offload *gro_find_complete_by_type(__be16 type);
3940 static inline void napi_free_frags(struct napi_struct *napi)
3942 kfree_skb(napi->skb);
3946 bool netdev_is_rx_handler_busy(struct net_device *dev);
3947 int netdev_rx_handler_register(struct net_device *dev,
3948 rx_handler_func_t *rx_handler,
3949 void *rx_handler_data);
3950 void netdev_rx_handler_unregister(struct net_device *dev);
3952 bool dev_valid_name(const char *name);
3953 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3954 bool *need_copyout);
3955 int dev_ifconf(struct net *net, struct ifconf *, int);
3956 int dev_ethtool(struct net *net, struct ifreq *);
3957 unsigned int dev_get_flags(const struct net_device *);
3958 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3959 struct netlink_ext_ack *extack);
3960 int dev_change_flags(struct net_device *dev, unsigned int flags,
3961 struct netlink_ext_ack *extack);
3962 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3963 unsigned int gchanges);
3964 int dev_change_name(struct net_device *, const char *);
3965 int dev_set_alias(struct net_device *, const char *, size_t);
3966 int dev_get_alias(const struct net_device *, char *, size_t);
3967 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3968 int __dev_set_mtu(struct net_device *, int);
3969 int dev_validate_mtu(struct net_device *dev, int mtu,
3970 struct netlink_ext_ack *extack);
3971 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3972 struct netlink_ext_ack *extack);
3973 int dev_set_mtu(struct net_device *, int);
3974 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3975 void dev_set_group(struct net_device *, int);
3976 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3977 struct netlink_ext_ack *extack);
3978 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3979 struct netlink_ext_ack *extack);
3980 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3981 struct netlink_ext_ack *extack);
3982 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3983 int dev_change_carrier(struct net_device *, bool new_carrier);
3984 int dev_get_phys_port_id(struct net_device *dev,
3985 struct netdev_phys_item_id *ppid);
3986 int dev_get_phys_port_name(struct net_device *dev,
3987 char *name, size_t len);
3988 int dev_get_port_parent_id(struct net_device *dev,
3989 struct netdev_phys_item_id *ppid, bool recurse);
3990 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3991 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3992 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3993 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3995 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3996 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3997 struct netdev_queue *txq, int *ret);
3999 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
4000 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
4001 int fd, int expected_fd, u32 flags);
4002 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
4003 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
4005 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4006 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
4007 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
4008 bool is_skb_forwardable(const struct net_device *dev,
4009 const struct sk_buff *skb);
4011 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
4012 const struct sk_buff *skb,
4013 const bool check_mtu)
4015 const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
4018 if (!(dev->flags & IFF_UP))
4024 len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
4025 if (skb->len <= len)
4028 /* if TSO is enabled, we don't care about the length as the packet
4029 * could be forwarded without being segmented before
4031 if (skb_is_gso(skb))
4037 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4038 struct sk_buff *skb,
4039 const bool check_mtu)
4041 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4042 unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4043 atomic_long_inc(&dev->rx_dropped);
4048 skb_scrub_packet(skb, true);
4053 bool dev_nit_active(struct net_device *dev);
4054 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4056 extern int netdev_budget;
4057 extern unsigned int netdev_budget_usecs;
4059 /* Called by rtnetlink.c:rtnl_unlock() */
4060 void netdev_run_todo(void);
4063 * dev_put - release reference to device
4064 * @dev: network device
4066 * Release reference to device to allow it to be freed.
4068 static inline void dev_put(struct net_device *dev)
4070 #ifdef CONFIG_PCPU_DEV_REFCNT
4071 this_cpu_dec(*dev->pcpu_refcnt);
4073 refcount_dec(&dev->dev_refcnt);
4078 * dev_hold - get reference to device
4079 * @dev: network device
4081 * Hold reference to device to keep it from being freed.
4083 static inline void dev_hold(struct net_device *dev)
4085 #ifdef CONFIG_PCPU_DEV_REFCNT
4086 this_cpu_inc(*dev->pcpu_refcnt);
4088 refcount_inc(&dev->dev_refcnt);
4092 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4093 * and _off may be called from IRQ context, but it is caller
4094 * who is responsible for serialization of these calls.
4096 * The name carrier is inappropriate, these functions should really be
4097 * called netif_lowerlayer_*() because they represent the state of any
4098 * kind of lower layer not just hardware media.
4101 void linkwatch_init_dev(struct net_device *dev);
4102 void linkwatch_fire_event(struct net_device *dev);
4103 void linkwatch_forget_dev(struct net_device *dev);
4106 * netif_carrier_ok - test if carrier present
4107 * @dev: network device
4109 * Check if carrier is present on device
4111 static inline bool netif_carrier_ok(const struct net_device *dev)
4113 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4116 unsigned long dev_trans_start(struct net_device *dev);
4118 void __netdev_watchdog_up(struct net_device *dev);
4120 void netif_carrier_on(struct net_device *dev);
4122 void netif_carrier_off(struct net_device *dev);
4125 * netif_dormant_on - mark device as dormant.
4126 * @dev: network device
4128 * Mark device as dormant (as per RFC2863).
4130 * The dormant state indicates that the relevant interface is not
4131 * actually in a condition to pass packets (i.e., it is not 'up') but is
4132 * in a "pending" state, waiting for some external event. For "on-
4133 * demand" interfaces, this new state identifies the situation where the
4134 * interface is waiting for events to place it in the up state.
4136 static inline void netif_dormant_on(struct net_device *dev)
4138 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4139 linkwatch_fire_event(dev);
4143 * netif_dormant_off - set device as not dormant.
4144 * @dev: network device
4146 * Device is not in dormant state.
4148 static inline void netif_dormant_off(struct net_device *dev)
4150 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4151 linkwatch_fire_event(dev);
4155 * netif_dormant - test if device is dormant
4156 * @dev: network device
4158 * Check if device is dormant.
4160 static inline bool netif_dormant(const struct net_device *dev)
4162 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4167 * netif_testing_on - mark device as under test.
4168 * @dev: network device
4170 * Mark device as under test (as per RFC2863).
4172 * The testing state indicates that some test(s) must be performed on
4173 * the interface. After completion, of the test, the interface state
4174 * will change to up, dormant, or down, as appropriate.
4176 static inline void netif_testing_on(struct net_device *dev)
4178 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4179 linkwatch_fire_event(dev);
4183 * netif_testing_off - set device as not under test.
4184 * @dev: network device
4186 * Device is not in testing state.
4188 static inline void netif_testing_off(struct net_device *dev)
4190 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4191 linkwatch_fire_event(dev);
4195 * netif_testing - test if device is under test
4196 * @dev: network device
4198 * Check if device is under test
4200 static inline bool netif_testing(const struct net_device *dev)
4202 return test_bit(__LINK_STATE_TESTING, &dev->state);
4207 * netif_oper_up - test if device is operational
4208 * @dev: network device
4210 * Check if carrier is operational
4212 static inline bool netif_oper_up(const struct net_device *dev)
4214 return (dev->operstate == IF_OPER_UP ||
4215 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4219 * netif_device_present - is device available or removed
4220 * @dev: network device
4222 * Check if device has not been removed from system.
4224 static inline bool netif_device_present(const struct net_device *dev)
4226 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4229 void netif_device_detach(struct net_device *dev);
4231 void netif_device_attach(struct net_device *dev);
4234 * Network interface message level settings
4239 NETIF_MSG_PROBE_BIT,
4241 NETIF_MSG_TIMER_BIT,
4242 NETIF_MSG_IFDOWN_BIT,
4244 NETIF_MSG_RX_ERR_BIT,
4245 NETIF_MSG_TX_ERR_BIT,
4246 NETIF_MSG_TX_QUEUED_BIT,
4248 NETIF_MSG_TX_DONE_BIT,
4249 NETIF_MSG_RX_STATUS_BIT,
4250 NETIF_MSG_PKTDATA_BIT,
4254 /* When you add a new bit above, update netif_msg_class_names array
4255 * in net/ethtool/common.c
4257 NETIF_MSG_CLASS_COUNT,
4259 /* Both ethtool_ops interface and internal driver implementation use u32 */
4260 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4262 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4263 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4265 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4266 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4267 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4268 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4269 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4270 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4271 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4272 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4273 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4274 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4275 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4276 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4277 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4278 #define NETIF_MSG_HW __NETIF_MSG(HW)
4279 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4281 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4282 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4283 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4284 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4285 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4286 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4287 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4288 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4289 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4290 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4291 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4292 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4293 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4294 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4295 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4297 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4300 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4301 return default_msg_enable_bits;
4302 if (debug_value == 0) /* no output */
4304 /* set low N bits */
4305 return (1U << debug_value) - 1;
4308 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4310 spin_lock(&txq->_xmit_lock);
4311 txq->xmit_lock_owner = cpu;
4314 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4316 __acquire(&txq->_xmit_lock);
4320 static inline void __netif_tx_release(struct netdev_queue *txq)
4322 __release(&txq->_xmit_lock);
4325 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4327 spin_lock_bh(&txq->_xmit_lock);
4328 txq->xmit_lock_owner = smp_processor_id();
4331 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4333 bool ok = spin_trylock(&txq->_xmit_lock);
4335 txq->xmit_lock_owner = smp_processor_id();
4339 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4341 txq->xmit_lock_owner = -1;
4342 spin_unlock(&txq->_xmit_lock);
4345 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4347 txq->xmit_lock_owner = -1;
4348 spin_unlock_bh(&txq->_xmit_lock);
4351 static inline void txq_trans_update(struct netdev_queue *txq)
4353 if (txq->xmit_lock_owner != -1)
4354 txq->trans_start = jiffies;
4357 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4358 static inline void netif_trans_update(struct net_device *dev)
4360 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4362 if (txq->trans_start != jiffies)
4363 txq->trans_start = jiffies;
4367 * netif_tx_lock - grab network device transmit lock
4368 * @dev: network device
4370 * Get network device transmit lock
4372 static inline void netif_tx_lock(struct net_device *dev)
4377 spin_lock(&dev->tx_global_lock);
4378 cpu = smp_processor_id();
4379 for (i = 0; i < dev->num_tx_queues; i++) {
4380 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4382 /* We are the only thread of execution doing a
4383 * freeze, but we have to grab the _xmit_lock in
4384 * order to synchronize with threads which are in
4385 * the ->hard_start_xmit() handler and already
4386 * checked the frozen bit.
4388 __netif_tx_lock(txq, cpu);
4389 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4390 __netif_tx_unlock(txq);
4394 static inline void netif_tx_lock_bh(struct net_device *dev)
4400 static inline void netif_tx_unlock(struct net_device *dev)
4404 for (i = 0; i < dev->num_tx_queues; i++) {
4405 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4407 /* No need to grab the _xmit_lock here. If the
4408 * queue is not stopped for another reason, we
4411 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4412 netif_schedule_queue(txq);
4414 spin_unlock(&dev->tx_global_lock);
4417 static inline void netif_tx_unlock_bh(struct net_device *dev)
4419 netif_tx_unlock(dev);
4423 #define HARD_TX_LOCK(dev, txq, cpu) { \
4424 if ((dev->features & NETIF_F_LLTX) == 0) { \
4425 __netif_tx_lock(txq, cpu); \
4427 __netif_tx_acquire(txq); \
4431 #define HARD_TX_TRYLOCK(dev, txq) \
4432 (((dev->features & NETIF_F_LLTX) == 0) ? \
4433 __netif_tx_trylock(txq) : \
4434 __netif_tx_acquire(txq))
4436 #define HARD_TX_UNLOCK(dev, txq) { \
4437 if ((dev->features & NETIF_F_LLTX) == 0) { \
4438 __netif_tx_unlock(txq); \
4440 __netif_tx_release(txq); \
4444 static inline void netif_tx_disable(struct net_device *dev)
4450 cpu = smp_processor_id();
4451 spin_lock(&dev->tx_global_lock);
4452 for (i = 0; i < dev->num_tx_queues; i++) {
4453 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4455 __netif_tx_lock(txq, cpu);
4456 netif_tx_stop_queue(txq);
4457 __netif_tx_unlock(txq);
4459 spin_unlock(&dev->tx_global_lock);
4463 static inline void netif_addr_lock(struct net_device *dev)
4465 unsigned char nest_level = 0;
4467 #ifdef CONFIG_LOCKDEP
4468 nest_level = dev->nested_level;
4470 spin_lock_nested(&dev->addr_list_lock, nest_level);
4473 static inline void netif_addr_lock_bh(struct net_device *dev)
4475 unsigned char nest_level = 0;
4477 #ifdef CONFIG_LOCKDEP
4478 nest_level = dev->nested_level;
4481 spin_lock_nested(&dev->addr_list_lock, nest_level);
4484 static inline void netif_addr_unlock(struct net_device *dev)
4486 spin_unlock(&dev->addr_list_lock);
4489 static inline void netif_addr_unlock_bh(struct net_device *dev)
4491 spin_unlock_bh(&dev->addr_list_lock);
4495 * dev_addrs walker. Should be used only for read access. Call with
4496 * rcu_read_lock held.
4498 #define for_each_dev_addr(dev, ha) \
4499 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4501 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4503 void ether_setup(struct net_device *dev);
4505 /* Support for loadable net-drivers */
4506 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4507 unsigned char name_assign_type,
4508 void (*setup)(struct net_device *),
4509 unsigned int txqs, unsigned int rxqs);
4510 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4511 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4513 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4514 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4517 int register_netdev(struct net_device *dev);
4518 void unregister_netdev(struct net_device *dev);
4520 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4522 /* General hardware address lists handling functions */
4523 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4524 struct netdev_hw_addr_list *from_list, int addr_len);
4525 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4526 struct netdev_hw_addr_list *from_list, int addr_len);
4527 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4528 struct net_device *dev,
4529 int (*sync)(struct net_device *, const unsigned char *),
4530 int (*unsync)(struct net_device *,
4531 const unsigned char *));
4532 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4533 struct net_device *dev,
4534 int (*sync)(struct net_device *,
4535 const unsigned char *, int),
4536 int (*unsync)(struct net_device *,
4537 const unsigned char *, int));
4538 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4539 struct net_device *dev,
4540 int (*unsync)(struct net_device *,
4541 const unsigned char *, int));
4542 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4543 struct net_device *dev,
4544 int (*unsync)(struct net_device *,
4545 const unsigned char *));
4546 void __hw_addr_init(struct netdev_hw_addr_list *list);
4548 /* Functions used for device addresses handling */
4549 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4550 unsigned char addr_type);
4551 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4552 unsigned char addr_type);
4553 void dev_addr_flush(struct net_device *dev);
4554 int dev_addr_init(struct net_device *dev);
4556 /* Functions used for unicast addresses handling */
4557 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4558 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4559 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4560 int dev_uc_sync(struct net_device *to, struct net_device *from);
4561 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4562 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4563 void dev_uc_flush(struct net_device *dev);
4564 void dev_uc_init(struct net_device *dev);
4567 * __dev_uc_sync - Synchonize device's unicast list
4568 * @dev: device to sync
4569 * @sync: function to call if address should be added
4570 * @unsync: function to call if address should be removed
4572 * Add newly added addresses to the interface, and release
4573 * addresses that have been deleted.
4575 static inline int __dev_uc_sync(struct net_device *dev,
4576 int (*sync)(struct net_device *,
4577 const unsigned char *),
4578 int (*unsync)(struct net_device *,
4579 const unsigned char *))
4581 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4585 * __dev_uc_unsync - Remove synchronized addresses from device
4586 * @dev: device to sync
4587 * @unsync: function to call if address should be removed
4589 * Remove all addresses that were added to the device by dev_uc_sync().
4591 static inline void __dev_uc_unsync(struct net_device *dev,
4592 int (*unsync)(struct net_device *,
4593 const unsigned char *))
4595 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4598 /* Functions used for multicast addresses handling */
4599 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4600 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4601 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4602 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4603 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4604 int dev_mc_sync(struct net_device *to, struct net_device *from);
4605 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4606 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4607 void dev_mc_flush(struct net_device *dev);
4608 void dev_mc_init(struct net_device *dev);
4611 * __dev_mc_sync - Synchonize device's multicast list
4612 * @dev: device to sync
4613 * @sync: function to call if address should be added
4614 * @unsync: function to call if address should be removed
4616 * Add newly added addresses to the interface, and release
4617 * addresses that have been deleted.
4619 static inline int __dev_mc_sync(struct net_device *dev,
4620 int (*sync)(struct net_device *,
4621 const unsigned char *),
4622 int (*unsync)(struct net_device *,
4623 const unsigned char *))
4625 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4629 * __dev_mc_unsync - Remove synchronized addresses from device
4630 * @dev: device to sync
4631 * @unsync: function to call if address should be removed
4633 * Remove all addresses that were added to the device by dev_mc_sync().
4635 static inline void __dev_mc_unsync(struct net_device *dev,
4636 int (*unsync)(struct net_device *,
4637 const unsigned char *))
4639 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4642 /* Functions used for secondary unicast and multicast support */
4643 void dev_set_rx_mode(struct net_device *dev);
4644 void __dev_set_rx_mode(struct net_device *dev);
4645 int dev_set_promiscuity(struct net_device *dev, int inc);
4646 int dev_set_allmulti(struct net_device *dev, int inc);
4647 void netdev_state_change(struct net_device *dev);
4648 void __netdev_notify_peers(struct net_device *dev);
4649 void netdev_notify_peers(struct net_device *dev);
4650 void netdev_features_change(struct net_device *dev);
4651 /* Load a device via the kmod */
4652 void dev_load(struct net *net, const char *name);
4653 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4654 struct rtnl_link_stats64 *storage);
4655 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4656 const struct net_device_stats *netdev_stats);
4657 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4658 const struct pcpu_sw_netstats __percpu *netstats);
4659 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4661 extern int netdev_max_backlog;
4662 extern int netdev_tstamp_prequeue;
4663 extern int weight_p;
4664 extern int dev_weight_rx_bias;
4665 extern int dev_weight_tx_bias;
4666 extern int dev_rx_weight;
4667 extern int dev_tx_weight;
4668 extern int gro_normal_batch;
4671 NESTED_SYNC_IMM_BIT,
4672 NESTED_SYNC_TODO_BIT,
4675 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4676 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4678 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4679 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4681 struct netdev_nested_priv {
4682 unsigned char flags;
4686 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4687 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4688 struct list_head **iter);
4689 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4690 struct list_head **iter);
4692 #ifdef CONFIG_LOCKDEP
4693 static LIST_HEAD(net_unlink_list);
4695 static inline void net_unlink_todo(struct net_device *dev)
4697 if (list_empty(&dev->unlink_list))
4698 list_add_tail(&dev->unlink_list, &net_unlink_list);
4702 /* iterate through upper list, must be called under RCU read lock */
4703 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4704 for (iter = &(dev)->adj_list.upper, \
4705 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4707 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4709 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4710 int (*fn)(struct net_device *upper_dev,
4711 struct netdev_nested_priv *priv),
4712 struct netdev_nested_priv *priv);
4714 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4715 struct net_device *upper_dev);
4717 bool netdev_has_any_upper_dev(struct net_device *dev);
4719 void *netdev_lower_get_next_private(struct net_device *dev,
4720 struct list_head **iter);
4721 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4722 struct list_head **iter);
4724 #define netdev_for_each_lower_private(dev, priv, iter) \
4725 for (iter = (dev)->adj_list.lower.next, \
4726 priv = netdev_lower_get_next_private(dev, &(iter)); \
4728 priv = netdev_lower_get_next_private(dev, &(iter)))
4730 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4731 for (iter = &(dev)->adj_list.lower, \
4732 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4734 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4736 void *netdev_lower_get_next(struct net_device *dev,
4737 struct list_head **iter);
4739 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4740 for (iter = (dev)->adj_list.lower.next, \
4741 ldev = netdev_lower_get_next(dev, &(iter)); \
4743 ldev = netdev_lower_get_next(dev, &(iter)))
4745 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4746 struct list_head **iter);
4747 int netdev_walk_all_lower_dev(struct net_device *dev,
4748 int (*fn)(struct net_device *lower_dev,
4749 struct netdev_nested_priv *priv),
4750 struct netdev_nested_priv *priv);
4751 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4752 int (*fn)(struct net_device *lower_dev,
4753 struct netdev_nested_priv *priv),
4754 struct netdev_nested_priv *priv);
4756 void *netdev_adjacent_get_private(struct list_head *adj_list);
4757 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4758 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4759 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4760 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4761 struct netlink_ext_ack *extack);
4762 int netdev_master_upper_dev_link(struct net_device *dev,
4763 struct net_device *upper_dev,
4764 void *upper_priv, void *upper_info,
4765 struct netlink_ext_ack *extack);
4766 void netdev_upper_dev_unlink(struct net_device *dev,
4767 struct net_device *upper_dev);
4768 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4769 struct net_device *new_dev,
4770 struct net_device *dev,
4771 struct netlink_ext_ack *extack);
4772 void netdev_adjacent_change_commit(struct net_device *old_dev,
4773 struct net_device *new_dev,
4774 struct net_device *dev);
4775 void netdev_adjacent_change_abort(struct net_device *old_dev,
4776 struct net_device *new_dev,
4777 struct net_device *dev);
4778 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4779 void *netdev_lower_dev_get_private(struct net_device *dev,
4780 struct net_device *lower_dev);
4781 void netdev_lower_state_changed(struct net_device *lower_dev,
4782 void *lower_state_info);
4784 /* RSS keys are 40 or 52 bytes long */
4785 #define NETDEV_RSS_KEY_LEN 52
4786 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4787 void netdev_rss_key_fill(void *buffer, size_t len);
4789 int skb_checksum_help(struct sk_buff *skb);
4790 int skb_crc32c_csum_help(struct sk_buff *skb);
4791 int skb_csum_hwoffload_help(struct sk_buff *skb,
4792 const netdev_features_t features);
4794 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4795 netdev_features_t features, bool tx_path);
4796 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4797 netdev_features_t features);
4799 struct netdev_bonding_info {
4804 struct netdev_notifier_bonding_info {
4805 struct netdev_notifier_info info; /* must be first */
4806 struct netdev_bonding_info bonding_info;
4809 void netdev_bonding_info_change(struct net_device *dev,
4810 struct netdev_bonding_info *bonding_info);
4812 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4813 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4815 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4822 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4824 return __skb_gso_segment(skb, features, true);
4826 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4828 static inline bool can_checksum_protocol(netdev_features_t features,
4831 if (protocol == htons(ETH_P_FCOE))
4832 return !!(features & NETIF_F_FCOE_CRC);
4834 /* Assume this is an IP checksum (not SCTP CRC) */
4836 if (features & NETIF_F_HW_CSUM) {
4837 /* Can checksum everything */
4842 case htons(ETH_P_IP):
4843 return !!(features & NETIF_F_IP_CSUM);
4844 case htons(ETH_P_IPV6):
4845 return !!(features & NETIF_F_IPV6_CSUM);
4852 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4854 static inline void netdev_rx_csum_fault(struct net_device *dev,
4855 struct sk_buff *skb)
4859 /* rx skb timestamps */
4860 void net_enable_timestamp(void);
4861 void net_disable_timestamp(void);
4863 #ifdef CONFIG_PROC_FS
4864 int __init dev_proc_init(void);
4866 #define dev_proc_init() 0
4869 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4870 struct sk_buff *skb, struct net_device *dev,
4873 __this_cpu_write(softnet_data.xmit.more, more);
4874 return ops->ndo_start_xmit(skb, dev);
4877 static inline bool netdev_xmit_more(void)
4879 return __this_cpu_read(softnet_data.xmit.more);
4882 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4883 struct netdev_queue *txq, bool more)
4885 const struct net_device_ops *ops = dev->netdev_ops;
4888 rc = __netdev_start_xmit(ops, skb, dev, more);
4889 if (rc == NETDEV_TX_OK)
4890 txq_trans_update(txq);
4895 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4897 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4900 extern const struct kobj_ns_type_operations net_ns_type_operations;
4902 const char *netdev_drivername(const struct net_device *dev);
4904 void linkwatch_run_queue(void);
4906 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4907 netdev_features_t f2)
4909 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4910 if (f1 & NETIF_F_HW_CSUM)
4911 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4913 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4919 static inline netdev_features_t netdev_get_wanted_features(
4920 struct net_device *dev)
4922 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4924 netdev_features_t netdev_increment_features(netdev_features_t all,
4925 netdev_features_t one, netdev_features_t mask);
4927 /* Allow TSO being used on stacked device :
4928 * Performing the GSO segmentation before last device
4929 * is a performance improvement.
4931 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4932 netdev_features_t mask)
4934 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4937 int __netdev_update_features(struct net_device *dev);
4938 void netdev_update_features(struct net_device *dev);
4939 void netdev_change_features(struct net_device *dev);
4941 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4942 struct net_device *dev);
4944 netdev_features_t passthru_features_check(struct sk_buff *skb,
4945 struct net_device *dev,
4946 netdev_features_t features);
4947 netdev_features_t netif_skb_features(struct sk_buff *skb);
4949 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4951 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4953 /* check flags correspondence */
4954 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4955 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4956 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4957 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4958 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4959 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4960 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4961 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4962 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4963 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4964 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4965 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4966 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4967 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4968 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4969 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4970 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4971 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4972 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4974 return (features & feature) == feature;
4977 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4979 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4980 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4983 static inline bool netif_needs_gso(struct sk_buff *skb,
4984 netdev_features_t features)
4986 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4987 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4988 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4991 static inline void netif_set_gso_max_size(struct net_device *dev,
4994 dev->gso_max_size = size;
4997 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4998 int pulled_hlen, u16 mac_offset,
5001 skb->protocol = protocol;
5002 skb->encapsulation = 1;
5003 skb_push(skb, pulled_hlen);
5004 skb_reset_transport_header(skb);
5005 skb->mac_header = mac_offset;
5006 skb->network_header = skb->mac_header + mac_len;
5007 skb->mac_len = mac_len;
5010 static inline bool netif_is_macsec(const struct net_device *dev)
5012 return dev->priv_flags & IFF_MACSEC;
5015 static inline bool netif_is_macvlan(const struct net_device *dev)
5017 return dev->priv_flags & IFF_MACVLAN;
5020 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5022 return dev->priv_flags & IFF_MACVLAN_PORT;
5025 static inline bool netif_is_bond_master(const struct net_device *dev)
5027 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5030 static inline bool netif_is_bond_slave(const struct net_device *dev)
5032 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5035 static inline bool netif_supports_nofcs(struct net_device *dev)
5037 return dev->priv_flags & IFF_SUPP_NOFCS;
5040 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5042 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5045 static inline bool netif_is_l3_master(const struct net_device *dev)
5047 return dev->priv_flags & IFF_L3MDEV_MASTER;
5050 static inline bool netif_is_l3_slave(const struct net_device *dev)
5052 return dev->priv_flags & IFF_L3MDEV_SLAVE;
5055 static inline bool netif_is_bridge_master(const struct net_device *dev)
5057 return dev->priv_flags & IFF_EBRIDGE;
5060 static inline bool netif_is_bridge_port(const struct net_device *dev)
5062 return dev->priv_flags & IFF_BRIDGE_PORT;
5065 static inline bool netif_is_ovs_master(const struct net_device *dev)
5067 return dev->priv_flags & IFF_OPENVSWITCH;
5070 static inline bool netif_is_ovs_port(const struct net_device *dev)
5072 return dev->priv_flags & IFF_OVS_DATAPATH;
5075 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5077 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5080 static inline bool netif_is_team_master(const struct net_device *dev)
5082 return dev->priv_flags & IFF_TEAM;
5085 static inline bool netif_is_team_port(const struct net_device *dev)
5087 return dev->priv_flags & IFF_TEAM_PORT;
5090 static inline bool netif_is_lag_master(const struct net_device *dev)
5092 return netif_is_bond_master(dev) || netif_is_team_master(dev);
5095 static inline bool netif_is_lag_port(const struct net_device *dev)
5097 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5100 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5102 return dev->priv_flags & IFF_RXFH_CONFIGURED;
5105 static inline bool netif_is_failover(const struct net_device *dev)
5107 return dev->priv_flags & IFF_FAILOVER;
5110 static inline bool netif_is_failover_slave(const struct net_device *dev)
5112 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5115 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5116 static inline void netif_keep_dst(struct net_device *dev)
5118 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5121 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5122 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5124 /* TODO: reserve and use an additional IFF bit, if we get more users */
5125 return dev->priv_flags & IFF_MACSEC;
5128 extern struct pernet_operations __net_initdata loopback_net_ops;
5130 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5132 /* netdev_printk helpers, similar to dev_printk */
5134 static inline const char *netdev_name(const struct net_device *dev)
5136 if (!dev->name[0] || strchr(dev->name, '%'))
5137 return "(unnamed net_device)";
5141 static inline bool netdev_unregistering(const struct net_device *dev)
5143 return dev->reg_state == NETREG_UNREGISTERING;
5146 static inline const char *netdev_reg_state(const struct net_device *dev)
5148 switch (dev->reg_state) {
5149 case NETREG_UNINITIALIZED: return " (uninitialized)";
5150 case NETREG_REGISTERED: return "";
5151 case NETREG_UNREGISTERING: return " (unregistering)";
5152 case NETREG_UNREGISTERED: return " (unregistered)";
5153 case NETREG_RELEASED: return " (released)";
5154 case NETREG_DUMMY: return " (dummy)";
5157 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5158 return " (unknown)";
5161 __printf(3, 4) __cold
5162 void netdev_printk(const char *level, const struct net_device *dev,
5163 const char *format, ...);
5164 __printf(2, 3) __cold
5165 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5166 __printf(2, 3) __cold
5167 void netdev_alert(const struct net_device *dev, const char *format, ...);
5168 __printf(2, 3) __cold
5169 void netdev_crit(const struct net_device *dev, const char *format, ...);
5170 __printf(2, 3) __cold
5171 void netdev_err(const struct net_device *dev, const char *format, ...);
5172 __printf(2, 3) __cold
5173 void netdev_warn(const struct net_device *dev, const char *format, ...);
5174 __printf(2, 3) __cold
5175 void netdev_notice(const struct net_device *dev, const char *format, ...);
5176 __printf(2, 3) __cold
5177 void netdev_info(const struct net_device *dev, const char *format, ...);
5179 #define netdev_level_once(level, dev, fmt, ...) \
5181 static bool __print_once __read_mostly; \
5183 if (!__print_once) { \
5184 __print_once = true; \
5185 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5189 #define netdev_emerg_once(dev, fmt, ...) \
5190 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5191 #define netdev_alert_once(dev, fmt, ...) \
5192 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5193 #define netdev_crit_once(dev, fmt, ...) \
5194 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5195 #define netdev_err_once(dev, fmt, ...) \
5196 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5197 #define netdev_warn_once(dev, fmt, ...) \
5198 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5199 #define netdev_notice_once(dev, fmt, ...) \
5200 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5201 #define netdev_info_once(dev, fmt, ...) \
5202 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5204 #define MODULE_ALIAS_NETDEV(device) \
5205 MODULE_ALIAS("netdev-" device)
5207 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5208 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5209 #define netdev_dbg(__dev, format, args...) \
5211 dynamic_netdev_dbg(__dev, format, ##args); \
5213 #elif defined(DEBUG)
5214 #define netdev_dbg(__dev, format, args...) \
5215 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5217 #define netdev_dbg(__dev, format, args...) \
5220 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5224 #if defined(VERBOSE_DEBUG)
5225 #define netdev_vdbg netdev_dbg
5228 #define netdev_vdbg(dev, format, args...) \
5231 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5237 * netdev_WARN() acts like dev_printk(), but with the key difference
5238 * of using a WARN/WARN_ON to get the message out, including the
5239 * file/line information and a backtrace.
5241 #define netdev_WARN(dev, format, args...) \
5242 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5243 netdev_reg_state(dev), ##args)
5245 #define netdev_WARN_ONCE(dev, format, args...) \
5246 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5247 netdev_reg_state(dev), ##args)
5249 /* netif printk helpers, similar to netdev_printk */
5251 #define netif_printk(priv, type, level, dev, fmt, args...) \
5253 if (netif_msg_##type(priv)) \
5254 netdev_printk(level, (dev), fmt, ##args); \
5257 #define netif_level(level, priv, type, dev, fmt, args...) \
5259 if (netif_msg_##type(priv)) \
5260 netdev_##level(dev, fmt, ##args); \
5263 #define netif_emerg(priv, type, dev, fmt, args...) \
5264 netif_level(emerg, priv, type, dev, fmt, ##args)
5265 #define netif_alert(priv, type, dev, fmt, args...) \
5266 netif_level(alert, priv, type, dev, fmt, ##args)
5267 #define netif_crit(priv, type, dev, fmt, args...) \
5268 netif_level(crit, priv, type, dev, fmt, ##args)
5269 #define netif_err(priv, type, dev, fmt, args...) \
5270 netif_level(err, priv, type, dev, fmt, ##args)
5271 #define netif_warn(priv, type, dev, fmt, args...) \
5272 netif_level(warn, priv, type, dev, fmt, ##args)
5273 #define netif_notice(priv, type, dev, fmt, args...) \
5274 netif_level(notice, priv, type, dev, fmt, ##args)
5275 #define netif_info(priv, type, dev, fmt, args...) \
5276 netif_level(info, priv, type, dev, fmt, ##args)
5278 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5279 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5280 #define netif_dbg(priv, type, netdev, format, args...) \
5282 if (netif_msg_##type(priv)) \
5283 dynamic_netdev_dbg(netdev, format, ##args); \
5285 #elif defined(DEBUG)
5286 #define netif_dbg(priv, type, dev, format, args...) \
5287 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5289 #define netif_dbg(priv, type, dev, format, args...) \
5292 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5297 /* if @cond then downgrade to debug, else print at @level */
5298 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5301 netif_dbg(priv, type, netdev, fmt, ##args); \
5303 netif_ ## level(priv, type, netdev, fmt, ##args); \
5306 #if defined(VERBOSE_DEBUG)
5307 #define netif_vdbg netif_dbg
5309 #define netif_vdbg(priv, type, dev, format, args...) \
5312 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5318 * The list of packet types we will receive (as opposed to discard)
5319 * and the routines to invoke.
5321 * Why 16. Because with 16 the only overlap we get on a hash of the
5322 * low nibble of the protocol value is RARP/SNAP/X.25.
5336 #define PTYPE_HASH_SIZE (16)
5337 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5339 extern struct net_device *blackhole_netdev;
5341 #endif /* _LINUX_NETDEVICE_H */