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;
353 NAPI_STATE_SCHED, /* Poll is scheduled */
354 NAPI_STATE_MISSED, /* reschedule a napi */
355 NAPI_STATE_DISABLE, /* Disable pending */
356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
357 NAPI_STATE_LISTED, /* NAPI added to system lists */
358 NAPI_STATE_NO_BUSY_POLL, /* Do not add in napi_hash, no busy polling */
359 NAPI_STATE_IN_BUSY_POLL, /* sk_busy_loop() owns this NAPI */
360 NAPI_STATE_PREFER_BUSY_POLL, /* prefer busy-polling over softirq processing*/
364 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
365 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
366 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
367 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
368 NAPIF_STATE_LISTED = BIT(NAPI_STATE_LISTED),
369 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
370 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
371 NAPIF_STATE_PREFER_BUSY_POLL = BIT(NAPI_STATE_PREFER_BUSY_POLL),
381 typedef enum gro_result gro_result_t;
384 * enum rx_handler_result - Possible return values for rx_handlers.
385 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
387 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
388 * case skb->dev was changed by rx_handler.
389 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
390 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
392 * rx_handlers are functions called from inside __netif_receive_skb(), to do
393 * special processing of the skb, prior to delivery to protocol handlers.
395 * Currently, a net_device can only have a single rx_handler registered. Trying
396 * to register a second rx_handler will return -EBUSY.
398 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
399 * To unregister a rx_handler on a net_device, use
400 * netdev_rx_handler_unregister().
402 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
405 * If the rx_handler consumed the skb in some way, it should return
406 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
407 * the skb to be delivered in some other way.
409 * If the rx_handler changed skb->dev, to divert the skb to another
410 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
411 * new device will be called if it exists.
413 * If the rx_handler decides the skb should be ignored, it should return
414 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
415 * are registered on exact device (ptype->dev == skb->dev).
417 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
418 * delivered, it should return RX_HANDLER_PASS.
420 * A device without a registered rx_handler will behave as if rx_handler
421 * returned RX_HANDLER_PASS.
424 enum rx_handler_result {
430 typedef enum rx_handler_result rx_handler_result_t;
431 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
433 void __napi_schedule(struct napi_struct *n);
434 void __napi_schedule_irqoff(struct napi_struct *n);
436 static inline bool napi_disable_pending(struct napi_struct *n)
438 return test_bit(NAPI_STATE_DISABLE, &n->state);
441 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
443 return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
446 bool napi_schedule_prep(struct napi_struct *n);
449 * napi_schedule - schedule NAPI poll
452 * Schedule NAPI poll routine to be called if it is not already
455 static inline void napi_schedule(struct napi_struct *n)
457 if (napi_schedule_prep(n))
462 * napi_schedule_irqoff - schedule NAPI poll
465 * Variant of napi_schedule(), assuming hard irqs are masked.
467 static inline void napi_schedule_irqoff(struct napi_struct *n)
469 if (napi_schedule_prep(n))
470 __napi_schedule_irqoff(n);
473 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
474 static inline bool napi_reschedule(struct napi_struct *napi)
476 if (napi_schedule_prep(napi)) {
477 __napi_schedule(napi);
483 bool napi_complete_done(struct napi_struct *n, int work_done);
485 * napi_complete - NAPI processing complete
488 * Mark NAPI processing as complete.
489 * Consider using napi_complete_done() instead.
490 * Return false if device should avoid rearming interrupts.
492 static inline bool napi_complete(struct napi_struct *n)
494 return napi_complete_done(n, 0);
498 * napi_disable - prevent NAPI from scheduling
501 * Stop NAPI from being scheduled on this context.
502 * Waits till any outstanding processing completes.
504 void napi_disable(struct napi_struct *n);
507 * napi_enable - enable NAPI scheduling
510 * Resume NAPI from being scheduled on this context.
511 * Must be paired with napi_disable.
513 static inline void napi_enable(struct napi_struct *n)
515 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
516 smp_mb__before_atomic();
517 clear_bit(NAPI_STATE_SCHED, &n->state);
518 clear_bit(NAPI_STATE_NPSVC, &n->state);
522 * napi_synchronize - wait until NAPI is not running
525 * Wait until NAPI is done being scheduled on this context.
526 * Waits till any outstanding processing completes but
527 * does not disable future activations.
529 static inline void napi_synchronize(const struct napi_struct *n)
531 if (IS_ENABLED(CONFIG_SMP))
532 while (test_bit(NAPI_STATE_SCHED, &n->state))
539 * napi_if_scheduled_mark_missed - if napi is running, set the
543 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
548 unsigned long val, new;
551 val = READ_ONCE(n->state);
552 if (val & NAPIF_STATE_DISABLE)
555 if (!(val & NAPIF_STATE_SCHED))
558 new = val | NAPIF_STATE_MISSED;
559 } while (cmpxchg(&n->state, val, new) != val);
564 enum netdev_queue_state_t {
565 __QUEUE_STATE_DRV_XOFF,
566 __QUEUE_STATE_STACK_XOFF,
567 __QUEUE_STATE_FROZEN,
570 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
571 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
572 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
574 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
581 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
582 * netif_tx_* functions below are used to manipulate this flag. The
583 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
584 * queue independently. The netif_xmit_*stopped functions below are called
585 * to check if the queue has been stopped by the driver or stack (either
586 * of the XOFF bits are set in the state). Drivers should not need to call
587 * netif_xmit*stopped functions, they should only be using netif_tx_*.
590 struct netdev_queue {
594 struct net_device *dev;
595 struct Qdisc __rcu *qdisc;
596 struct Qdisc *qdisc_sleeping;
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
603 unsigned long tx_maxrate;
605 * Number of TX timeouts for this queue
606 * (/sys/class/net/DEV/Q/trans_timeout)
608 unsigned long trans_timeout;
610 /* Subordinate device that the queue has been assigned to */
611 struct net_device *sb_dev;
612 #ifdef CONFIG_XDP_SOCKETS
613 struct xsk_buff_pool *pool;
618 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
621 * Time (in jiffies) of last Tx
623 unsigned long trans_start;
630 } ____cacheline_aligned_in_smp;
632 extern int sysctl_fb_tunnels_only_for_init_net;
633 extern int sysctl_devconf_inherit_init_net;
636 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
637 * == 1 : For initns only
640 static inline bool net_has_fallback_tunnels(const struct net *net)
642 return !IS_ENABLED(CONFIG_SYSCTL) ||
643 !sysctl_fb_tunnels_only_for_init_net ||
644 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
647 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
649 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
656 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
658 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
665 * This structure holds an RPS map which can be of variable length. The
666 * map is an array of CPUs.
673 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
676 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
677 * tail pointer for that CPU's input queue at the time of last enqueue, and
678 * a hardware filter index.
680 struct rps_dev_flow {
683 unsigned int last_qtail;
685 #define RPS_NO_FILTER 0xffff
688 * The rps_dev_flow_table structure contains a table of flow mappings.
690 struct rps_dev_flow_table {
693 struct rps_dev_flow flows[];
695 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
696 ((_num) * sizeof(struct rps_dev_flow)))
699 * The rps_sock_flow_table contains mappings of flows to the last CPU
700 * on which they were processed by the application (set in recvmsg).
701 * Each entry is a 32bit value. Upper part is the high-order bits
702 * of flow hash, lower part is CPU number.
703 * rps_cpu_mask is used to partition the space, depending on number of
704 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
705 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
706 * meaning we use 32-6=26 bits for the hash.
708 struct rps_sock_flow_table {
711 u32 ents[] ____cacheline_aligned_in_smp;
713 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
715 #define RPS_NO_CPU 0xffff
717 extern u32 rps_cpu_mask;
718 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
720 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
724 unsigned int index = hash & table->mask;
725 u32 val = hash & ~rps_cpu_mask;
727 /* We only give a hint, preemption can change CPU under us */
728 val |= raw_smp_processor_id();
730 if (table->ents[index] != val)
731 table->ents[index] = val;
735 #ifdef CONFIG_RFS_ACCEL
736 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
739 #endif /* CONFIG_RPS */
741 /* This structure contains an instance of an RX queue. */
742 struct netdev_rx_queue {
744 struct rps_map __rcu *rps_map;
745 struct rps_dev_flow_table __rcu *rps_flow_table;
748 struct net_device *dev;
749 struct xdp_rxq_info xdp_rxq;
750 #ifdef CONFIG_XDP_SOCKETS
751 struct xsk_buff_pool *pool;
753 } ____cacheline_aligned_in_smp;
756 * RX queue sysfs structures and functions.
758 struct rx_queue_attribute {
759 struct attribute attr;
760 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
761 ssize_t (*store)(struct netdev_rx_queue *queue,
762 const char *buf, size_t len);
767 * This structure holds an XPS map which can be of variable length. The
768 * map is an array of queues.
772 unsigned int alloc_len;
776 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
777 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
778 - sizeof(struct xps_map)) / sizeof(u16))
781 * This structure holds all XPS maps for device. Maps are indexed by CPU.
783 struct xps_dev_maps {
785 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
788 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
789 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
791 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
792 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
794 #endif /* CONFIG_XPS */
796 #define TC_MAX_QUEUE 16
797 #define TC_BITMASK 15
798 /* HW offloaded queuing disciplines txq count and offset maps */
799 struct netdev_tc_txq {
804 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
806 * This structure is to hold information about the device
807 * configured to run FCoE protocol stack.
809 struct netdev_fcoe_hbainfo {
810 char manufacturer[64];
811 char serial_number[64];
812 char hardware_version[64];
813 char driver_version[64];
814 char optionrom_version[64];
815 char firmware_version[64];
817 char model_description[256];
821 #define MAX_PHYS_ITEM_ID_LEN 32
823 /* This structure holds a unique identifier to identify some
824 * physical item (port for example) used by a netdevice.
826 struct netdev_phys_item_id {
827 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
828 unsigned char id_len;
831 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
832 struct netdev_phys_item_id *b)
834 return a->id_len == b->id_len &&
835 memcmp(a->id, b->id, a->id_len) == 0;
838 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
840 struct net_device *sb_dev);
843 TC_SETUP_QDISC_MQPRIO,
846 TC_SETUP_CLSMATCHALL,
856 TC_SETUP_QDISC_TAPRIO,
863 /* These structures hold the attributes of bpf state that are being passed
864 * to the netdevice through the bpf op.
866 enum bpf_netdev_command {
867 /* Set or clear a bpf program used in the earliest stages of packet
868 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
869 * is responsible for calling bpf_prog_put on any old progs that are
870 * stored. In case of error, the callee need not release the new prog
871 * reference, but on success it takes ownership and must bpf_prog_put
872 * when it is no longer used.
876 /* BPF program for offload callbacks, invoked at program load time. */
877 BPF_OFFLOAD_MAP_ALLOC,
878 BPF_OFFLOAD_MAP_FREE,
882 struct bpf_prog_offload_ops;
883 struct netlink_ext_ack;
885 struct xdp_dev_bulk_queue;
895 struct bpf_xdp_entity {
896 struct bpf_prog *prog;
897 struct bpf_xdp_link *link;
901 enum bpf_netdev_command command;
906 struct bpf_prog *prog;
907 struct netlink_ext_ack *extack;
909 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
911 struct bpf_offloaded_map *offmap;
913 /* XDP_SETUP_XSK_POOL */
915 struct xsk_buff_pool *pool;
921 /* Flags for ndo_xsk_wakeup. */
922 #define XDP_WAKEUP_RX (1 << 0)
923 #define XDP_WAKEUP_TX (1 << 1)
925 #ifdef CONFIG_XFRM_OFFLOAD
927 int (*xdo_dev_state_add) (struct xfrm_state *x);
928 void (*xdo_dev_state_delete) (struct xfrm_state *x);
929 void (*xdo_dev_state_free) (struct xfrm_state *x);
930 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
931 struct xfrm_state *x);
932 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
937 struct rcu_head rcuhead;
944 struct netdev_name_node {
945 struct hlist_node hlist;
946 struct list_head list;
947 struct net_device *dev;
951 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
952 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
954 struct netdev_net_notifier {
955 struct list_head list;
956 struct notifier_block *nb;
960 * This structure defines the management hooks for network devices.
961 * The following hooks can be defined; unless noted otherwise, they are
962 * optional and can be filled with a null pointer.
964 * int (*ndo_init)(struct net_device *dev);
965 * This function is called once when a network device is registered.
966 * The network device can use this for any late stage initialization
967 * or semantic validation. It can fail with an error code which will
968 * be propagated back to register_netdev.
970 * void (*ndo_uninit)(struct net_device *dev);
971 * This function is called when device is unregistered or when registration
972 * fails. It is not called if init fails.
974 * int (*ndo_open)(struct net_device *dev);
975 * This function is called when a network device transitions to the up
978 * int (*ndo_stop)(struct net_device *dev);
979 * This function is called when a network device transitions to the down
982 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
983 * struct net_device *dev);
984 * Called when a packet needs to be transmitted.
985 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
986 * the queue before that can happen; it's for obsolete devices and weird
987 * corner cases, but the stack really does a non-trivial amount
988 * of useless work if you return NETDEV_TX_BUSY.
989 * Required; cannot be NULL.
991 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
992 * struct net_device *dev
993 * netdev_features_t features);
994 * Called by core transmit path to determine if device is capable of
995 * performing offload operations on a given packet. This is to give
996 * the device an opportunity to implement any restrictions that cannot
997 * be otherwise expressed by feature flags. The check is called with
998 * the set of features that the stack has calculated and it returns
999 * those the driver believes to be appropriate.
1001 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1002 * struct net_device *sb_dev);
1003 * Called to decide which queue to use when device supports multiple
1006 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1007 * This function is called to allow device receiver to make
1008 * changes to configuration when multicast or promiscuous is enabled.
1010 * void (*ndo_set_rx_mode)(struct net_device *dev);
1011 * This function is called device changes address list filtering.
1012 * If driver handles unicast address filtering, it should set
1013 * IFF_UNICAST_FLT in its priv_flags.
1015 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1016 * This function is called when the Media Access Control address
1017 * needs to be changed. If this interface is not defined, the
1018 * MAC address can not be changed.
1020 * int (*ndo_validate_addr)(struct net_device *dev);
1021 * Test if Media Access Control address is valid for the device.
1023 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1024 * Called when a user requests an ioctl which can't be handled by
1025 * the generic interface code. If not defined ioctls return
1026 * not supported error code.
1028 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1029 * Used to set network devices bus interface parameters. This interface
1030 * is retained for legacy reasons; new devices should use the bus
1031 * interface (PCI) for low level management.
1033 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1034 * Called when a user wants to change the Maximum Transfer Unit
1037 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1038 * Callback used when the transmitter has not made any progress
1039 * for dev->watchdog ticks.
1041 * void (*ndo_get_stats64)(struct net_device *dev,
1042 * struct rtnl_link_stats64 *storage);
1043 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1044 * Called when a user wants to get the network device usage
1045 * statistics. Drivers must do one of the following:
1046 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1047 * rtnl_link_stats64 structure passed by the caller.
1048 * 2. Define @ndo_get_stats to update a net_device_stats structure
1049 * (which should normally be dev->stats) and return a pointer to
1050 * it. The structure may be changed asynchronously only if each
1051 * field is written atomically.
1052 * 3. Update dev->stats asynchronously and atomically, and define
1053 * neither operation.
1055 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1056 * Return true if this device supports offload stats of this attr_id.
1058 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1060 * Get statistics for offload operations by attr_id. Write it into the
1061 * attr_data pointer.
1063 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1064 * If device supports VLAN filtering this function is called when a
1065 * VLAN id is registered.
1067 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1068 * If device supports VLAN filtering this function is called when a
1069 * VLAN id is unregistered.
1071 * void (*ndo_poll_controller)(struct net_device *dev);
1073 * SR-IOV management functions.
1074 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1075 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1076 * u8 qos, __be16 proto);
1077 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1079 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1080 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1081 * int (*ndo_get_vf_config)(struct net_device *dev,
1082 * int vf, struct ifla_vf_info *ivf);
1083 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1084 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1085 * struct nlattr *port[]);
1087 * Enable or disable the VF ability to query its RSS Redirection Table and
1088 * Hash Key. This is needed since on some devices VF share this information
1089 * with PF and querying it may introduce a theoretical security risk.
1090 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1091 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1092 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1094 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1095 * This is always called from the stack with the rtnl lock held and netif
1096 * tx queues stopped. This allows the netdevice to perform queue
1097 * management safely.
1099 * Fiber Channel over Ethernet (FCoE) offload functions.
1100 * int (*ndo_fcoe_enable)(struct net_device *dev);
1101 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1102 * so the underlying device can perform whatever needed configuration or
1103 * initialization to support acceleration of FCoE traffic.
1105 * int (*ndo_fcoe_disable)(struct net_device *dev);
1106 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1107 * so the underlying device can perform whatever needed clean-ups to
1108 * stop supporting acceleration of FCoE traffic.
1110 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1111 * struct scatterlist *sgl, unsigned int sgc);
1112 * Called when the FCoE Initiator wants to initialize an I/O that
1113 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1114 * perform necessary setup and returns 1 to indicate the device is set up
1115 * successfully to perform DDP on this I/O, otherwise this returns 0.
1117 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1118 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1119 * indicated by the FC exchange id 'xid', so the underlying device can
1120 * clean up and reuse resources for later DDP requests.
1122 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1123 * struct scatterlist *sgl, unsigned int sgc);
1124 * Called when the FCoE Target wants to initialize an I/O that
1125 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1126 * perform necessary setup and returns 1 to indicate the device is set up
1127 * successfully to perform DDP on this I/O, otherwise this returns 0.
1129 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1130 * struct netdev_fcoe_hbainfo *hbainfo);
1131 * Called when the FCoE Protocol stack wants information on the underlying
1132 * device. This information is utilized by the FCoE protocol stack to
1133 * register attributes with Fiber Channel management service as per the
1134 * FC-GS Fabric Device Management Information(FDMI) specification.
1136 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1137 * Called when the underlying device wants to override default World Wide
1138 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1139 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1140 * protocol stack to use.
1143 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1144 * u16 rxq_index, u32 flow_id);
1145 * Set hardware filter for RFS. rxq_index is the target queue index;
1146 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1147 * Return the filter ID on success, or a negative error code.
1149 * Slave management functions (for bridge, bonding, etc).
1150 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1151 * Called to make another netdev an underling.
1153 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1154 * Called to release previously enslaved netdev.
1156 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1157 * struct sk_buff *skb,
1159 * Get the xmit slave of master device. If all_slaves is true, function
1160 * assume all the slaves can transmit.
1162 * Feature/offload setting functions.
1163 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1164 * netdev_features_t features);
1165 * Adjusts the requested feature flags according to device-specific
1166 * constraints, and returns the resulting flags. Must not modify
1169 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1170 * Called to update device configuration to new features. Passed
1171 * feature set might be less than what was returned by ndo_fix_features()).
1172 * Must return >0 or -errno if it changed dev->features itself.
1174 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1175 * struct net_device *dev,
1176 * const unsigned char *addr, u16 vid, u16 flags,
1177 * struct netlink_ext_ack *extack);
1178 * Adds an FDB entry to dev for addr.
1179 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1180 * struct net_device *dev,
1181 * const unsigned char *addr, u16 vid)
1182 * Deletes the FDB entry from dev coresponding to addr.
1183 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1184 * struct net_device *dev, struct net_device *filter_dev,
1186 * Used to add FDB entries to dump requests. Implementers should add
1187 * entries to skb and update idx with the number of entries.
1189 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1190 * u16 flags, struct netlink_ext_ack *extack)
1191 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1192 * struct net_device *dev, u32 filter_mask,
1194 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1197 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1198 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1199 * which do not represent real hardware may define this to allow their
1200 * userspace components to manage their virtual carrier state. Devices
1201 * that determine carrier state from physical hardware properties (eg
1202 * network cables) or protocol-dependent mechanisms (eg
1203 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1205 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1206 * struct netdev_phys_item_id *ppid);
1207 * Called to get ID of physical port of this device. If driver does
1208 * not implement this, it is assumed that the hw is not able to have
1209 * multiple net devices on single physical port.
1211 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1212 * struct netdev_phys_item_id *ppid)
1213 * Called to get the parent ID of the physical port of this device.
1215 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1216 * struct net_device *dev)
1217 * Called by upper layer devices to accelerate switching or other
1218 * station functionality into hardware. 'pdev is the lowerdev
1219 * to use for the offload and 'dev' is the net device that will
1220 * back the offload. Returns a pointer to the private structure
1221 * the upper layer will maintain.
1222 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1223 * Called by upper layer device to delete the station created
1224 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1225 * the station and priv is the structure returned by the add
1227 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1228 * int queue_index, u32 maxrate);
1229 * Called when a user wants to set a max-rate limitation of specific
1231 * int (*ndo_get_iflink)(const struct net_device *dev);
1232 * Called to get the iflink value of this device.
1233 * void (*ndo_change_proto_down)(struct net_device *dev,
1235 * This function is used to pass protocol port error state information
1236 * to the switch driver. The switch driver can react to the proto_down
1237 * by doing a phys down on the associated switch port.
1238 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1239 * This function is used to get egress tunnel information for given skb.
1240 * This is useful for retrieving outer tunnel header parameters while
1242 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1243 * This function is used to specify the headroom that the skb must
1244 * consider when allocation skb during packet reception. Setting
1245 * appropriate rx headroom value allows avoiding skb head copy on
1246 * forward. Setting a negative value resets the rx headroom to the
1248 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1249 * This function is used to set or query state related to XDP on the
1250 * netdevice and manage BPF offload. See definition of
1251 * enum bpf_netdev_command for details.
1252 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1254 * This function is used to submit @n XDP packets for transmit on a
1255 * netdevice. Returns number of frames successfully transmitted, frames
1256 * that got dropped are freed/returned via xdp_return_frame().
1257 * Returns negative number, means general error invoking ndo, meaning
1258 * no frames were xmit'ed and core-caller will free all frames.
1259 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1260 * This function is used to wake up the softirq, ksoftirqd or kthread
1261 * responsible for sending and/or receiving packets on a specific
1262 * queue id bound to an AF_XDP socket. The flags field specifies if
1263 * only RX, only Tx, or both should be woken up using the flags
1264 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1265 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1266 * Get devlink port instance associated with a given netdev.
1267 * Called with a reference on the netdevice and devlink locks only,
1268 * rtnl_lock is not held.
1269 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1271 * Add, change, delete or get information on an IPv4 tunnel.
1272 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1273 * If a device is paired with a peer device, return the peer instance.
1274 * The caller must be under RCU read context.
1276 struct net_device_ops {
1277 int (*ndo_init)(struct net_device *dev);
1278 void (*ndo_uninit)(struct net_device *dev);
1279 int (*ndo_open)(struct net_device *dev);
1280 int (*ndo_stop)(struct net_device *dev);
1281 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1282 struct net_device *dev);
1283 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1284 struct net_device *dev,
1285 netdev_features_t features);
1286 u16 (*ndo_select_queue)(struct net_device *dev,
1287 struct sk_buff *skb,
1288 struct net_device *sb_dev);
1289 void (*ndo_change_rx_flags)(struct net_device *dev,
1291 void (*ndo_set_rx_mode)(struct net_device *dev);
1292 int (*ndo_set_mac_address)(struct net_device *dev,
1294 int (*ndo_validate_addr)(struct net_device *dev);
1295 int (*ndo_do_ioctl)(struct net_device *dev,
1296 struct ifreq *ifr, int cmd);
1297 int (*ndo_set_config)(struct net_device *dev,
1299 int (*ndo_change_mtu)(struct net_device *dev,
1301 int (*ndo_neigh_setup)(struct net_device *dev,
1302 struct neigh_parms *);
1303 void (*ndo_tx_timeout) (struct net_device *dev,
1304 unsigned int txqueue);
1306 void (*ndo_get_stats64)(struct net_device *dev,
1307 struct rtnl_link_stats64 *storage);
1308 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1309 int (*ndo_get_offload_stats)(int attr_id,
1310 const struct net_device *dev,
1312 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1314 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1315 __be16 proto, u16 vid);
1316 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1317 __be16 proto, u16 vid);
1318 #ifdef CONFIG_NET_POLL_CONTROLLER
1319 void (*ndo_poll_controller)(struct net_device *dev);
1320 int (*ndo_netpoll_setup)(struct net_device *dev,
1321 struct netpoll_info *info);
1322 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1324 int (*ndo_set_vf_mac)(struct net_device *dev,
1325 int queue, u8 *mac);
1326 int (*ndo_set_vf_vlan)(struct net_device *dev,
1327 int queue, u16 vlan,
1328 u8 qos, __be16 proto);
1329 int (*ndo_set_vf_rate)(struct net_device *dev,
1330 int vf, int min_tx_rate,
1332 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1333 int vf, bool setting);
1334 int (*ndo_set_vf_trust)(struct net_device *dev,
1335 int vf, bool setting);
1336 int (*ndo_get_vf_config)(struct net_device *dev,
1338 struct ifla_vf_info *ivf);
1339 int (*ndo_set_vf_link_state)(struct net_device *dev,
1340 int vf, int link_state);
1341 int (*ndo_get_vf_stats)(struct net_device *dev,
1343 struct ifla_vf_stats
1345 int (*ndo_set_vf_port)(struct net_device *dev,
1347 struct nlattr *port[]);
1348 int (*ndo_get_vf_port)(struct net_device *dev,
1349 int vf, struct sk_buff *skb);
1350 int (*ndo_get_vf_guid)(struct net_device *dev,
1352 struct ifla_vf_guid *node_guid,
1353 struct ifla_vf_guid *port_guid);
1354 int (*ndo_set_vf_guid)(struct net_device *dev,
1357 int (*ndo_set_vf_rss_query_en)(
1358 struct net_device *dev,
1359 int vf, bool setting);
1360 int (*ndo_setup_tc)(struct net_device *dev,
1361 enum tc_setup_type type,
1363 #if IS_ENABLED(CONFIG_FCOE)
1364 int (*ndo_fcoe_enable)(struct net_device *dev);
1365 int (*ndo_fcoe_disable)(struct net_device *dev);
1366 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1368 struct scatterlist *sgl,
1370 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1372 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1374 struct scatterlist *sgl,
1376 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1377 struct netdev_fcoe_hbainfo *hbainfo);
1380 #if IS_ENABLED(CONFIG_LIBFCOE)
1381 #define NETDEV_FCOE_WWNN 0
1382 #define NETDEV_FCOE_WWPN 1
1383 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1384 u64 *wwn, int type);
1387 #ifdef CONFIG_RFS_ACCEL
1388 int (*ndo_rx_flow_steer)(struct net_device *dev,
1389 const struct sk_buff *skb,
1393 int (*ndo_add_slave)(struct net_device *dev,
1394 struct net_device *slave_dev,
1395 struct netlink_ext_ack *extack);
1396 int (*ndo_del_slave)(struct net_device *dev,
1397 struct net_device *slave_dev);
1398 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1399 struct sk_buff *skb,
1401 struct net_device* (*ndo_sk_get_lower_dev)(struct net_device *dev,
1403 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1404 netdev_features_t features);
1405 int (*ndo_set_features)(struct net_device *dev,
1406 netdev_features_t features);
1407 int (*ndo_neigh_construct)(struct net_device *dev,
1408 struct neighbour *n);
1409 void (*ndo_neigh_destroy)(struct net_device *dev,
1410 struct neighbour *n);
1412 int (*ndo_fdb_add)(struct ndmsg *ndm,
1413 struct nlattr *tb[],
1414 struct net_device *dev,
1415 const unsigned char *addr,
1418 struct netlink_ext_ack *extack);
1419 int (*ndo_fdb_del)(struct ndmsg *ndm,
1420 struct nlattr *tb[],
1421 struct net_device *dev,
1422 const unsigned char *addr,
1424 int (*ndo_fdb_dump)(struct sk_buff *skb,
1425 struct netlink_callback *cb,
1426 struct net_device *dev,
1427 struct net_device *filter_dev,
1429 int (*ndo_fdb_get)(struct sk_buff *skb,
1430 struct nlattr *tb[],
1431 struct net_device *dev,
1432 const unsigned char *addr,
1433 u16 vid, u32 portid, u32 seq,
1434 struct netlink_ext_ack *extack);
1435 int (*ndo_bridge_setlink)(struct net_device *dev,
1436 struct nlmsghdr *nlh,
1438 struct netlink_ext_ack *extack);
1439 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1441 struct net_device *dev,
1444 int (*ndo_bridge_dellink)(struct net_device *dev,
1445 struct nlmsghdr *nlh,
1447 int (*ndo_change_carrier)(struct net_device *dev,
1449 int (*ndo_get_phys_port_id)(struct net_device *dev,
1450 struct netdev_phys_item_id *ppid);
1451 int (*ndo_get_port_parent_id)(struct net_device *dev,
1452 struct netdev_phys_item_id *ppid);
1453 int (*ndo_get_phys_port_name)(struct net_device *dev,
1454 char *name, size_t len);
1455 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1456 struct net_device *dev);
1457 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1460 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1463 int (*ndo_get_iflink)(const struct net_device *dev);
1464 int (*ndo_change_proto_down)(struct net_device *dev,
1466 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1467 struct sk_buff *skb);
1468 void (*ndo_set_rx_headroom)(struct net_device *dev,
1469 int needed_headroom);
1470 int (*ndo_bpf)(struct net_device *dev,
1471 struct netdev_bpf *bpf);
1472 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1473 struct xdp_frame **xdp,
1475 int (*ndo_xsk_wakeup)(struct net_device *dev,
1476 u32 queue_id, u32 flags);
1477 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1478 int (*ndo_tunnel_ctl)(struct net_device *dev,
1479 struct ip_tunnel_parm *p, int cmd);
1480 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1484 * enum netdev_priv_flags - &struct net_device priv_flags
1486 * These are the &struct net_device, they are only set internally
1487 * by drivers and used in the kernel. These flags are invisible to
1488 * userspace; this means that the order of these flags can change
1489 * during any kernel release.
1491 * You should have a pretty good reason to be extending these flags.
1493 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1494 * @IFF_EBRIDGE: Ethernet bridging device
1495 * @IFF_BONDING: bonding master or slave
1496 * @IFF_ISATAP: ISATAP interface (RFC4214)
1497 * @IFF_WAN_HDLC: WAN HDLC device
1498 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1500 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1501 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1502 * @IFF_MACVLAN_PORT: device used as macvlan port
1503 * @IFF_BRIDGE_PORT: device used as bridge port
1504 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1505 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1506 * @IFF_UNICAST_FLT: Supports unicast filtering
1507 * @IFF_TEAM_PORT: device used as team port
1508 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1509 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1510 * change when it's running
1511 * @IFF_MACVLAN: Macvlan device
1512 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1513 * underlying stacked devices
1514 * @IFF_L3MDEV_MASTER: device is an L3 master device
1515 * @IFF_NO_QUEUE: device can run without qdisc attached
1516 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1517 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1518 * @IFF_TEAM: device is a team device
1519 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1520 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1521 * entity (i.e. the master device for bridged veth)
1522 * @IFF_MACSEC: device is a MACsec device
1523 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1524 * @IFF_FAILOVER: device is a failover master device
1525 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1526 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1527 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1529 enum netdev_priv_flags {
1530 IFF_802_1Q_VLAN = 1<<0,
1534 IFF_WAN_HDLC = 1<<4,
1535 IFF_XMIT_DST_RELEASE = 1<<5,
1536 IFF_DONT_BRIDGE = 1<<6,
1537 IFF_DISABLE_NETPOLL = 1<<7,
1538 IFF_MACVLAN_PORT = 1<<8,
1539 IFF_BRIDGE_PORT = 1<<9,
1540 IFF_OVS_DATAPATH = 1<<10,
1541 IFF_TX_SKB_SHARING = 1<<11,
1542 IFF_UNICAST_FLT = 1<<12,
1543 IFF_TEAM_PORT = 1<<13,
1544 IFF_SUPP_NOFCS = 1<<14,
1545 IFF_LIVE_ADDR_CHANGE = 1<<15,
1546 IFF_MACVLAN = 1<<16,
1547 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1548 IFF_L3MDEV_MASTER = 1<<18,
1549 IFF_NO_QUEUE = 1<<19,
1550 IFF_OPENVSWITCH = 1<<20,
1551 IFF_L3MDEV_SLAVE = 1<<21,
1553 IFF_RXFH_CONFIGURED = 1<<23,
1554 IFF_PHONY_HEADROOM = 1<<24,
1556 IFF_NO_RX_HANDLER = 1<<26,
1557 IFF_FAILOVER = 1<<27,
1558 IFF_FAILOVER_SLAVE = 1<<28,
1559 IFF_L3MDEV_RX_HANDLER = 1<<29,
1560 IFF_LIVE_RENAME_OK = 1<<30,
1563 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1564 #define IFF_EBRIDGE IFF_EBRIDGE
1565 #define IFF_BONDING IFF_BONDING
1566 #define IFF_ISATAP IFF_ISATAP
1567 #define IFF_WAN_HDLC IFF_WAN_HDLC
1568 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1569 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1570 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1571 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1572 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1573 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1574 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1575 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1576 #define IFF_TEAM_PORT IFF_TEAM_PORT
1577 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1578 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1579 #define IFF_MACVLAN IFF_MACVLAN
1580 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1581 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1582 #define IFF_NO_QUEUE IFF_NO_QUEUE
1583 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1584 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1585 #define IFF_TEAM IFF_TEAM
1586 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1587 #define IFF_MACSEC IFF_MACSEC
1588 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1589 #define IFF_FAILOVER IFF_FAILOVER
1590 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1591 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1592 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1595 * struct net_device - The DEVICE structure.
1597 * Actually, this whole structure is a big mistake. It mixes I/O
1598 * data with strictly "high-level" data, and it has to know about
1599 * almost every data structure used in the INET module.
1601 * @name: This is the first field of the "visible" part of this structure
1602 * (i.e. as seen by users in the "Space.c" file). It is the name
1605 * @name_node: Name hashlist node
1606 * @ifalias: SNMP alias
1607 * @mem_end: Shared memory end
1608 * @mem_start: Shared memory start
1609 * @base_addr: Device I/O address
1610 * @irq: Device IRQ number
1612 * @state: Generic network queuing layer state, see netdev_state_t
1613 * @dev_list: The global list of network devices
1614 * @napi_list: List entry used for polling NAPI devices
1615 * @unreg_list: List entry when we are unregistering the
1616 * device; see the function unregister_netdev
1617 * @close_list: List entry used when we are closing the device
1618 * @ptype_all: Device-specific packet handlers for all protocols
1619 * @ptype_specific: Device-specific, protocol-specific packet handlers
1621 * @adj_list: Directly linked devices, like slaves for bonding
1622 * @features: Currently active device features
1623 * @hw_features: User-changeable features
1625 * @wanted_features: User-requested features
1626 * @vlan_features: Mask of features inheritable by VLAN devices
1628 * @hw_enc_features: Mask of features inherited by encapsulating devices
1629 * This field indicates what encapsulation
1630 * offloads the hardware is capable of doing,
1631 * and drivers will need to set them appropriately.
1633 * @mpls_features: Mask of features inheritable by MPLS
1634 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1636 * @ifindex: interface index
1637 * @group: The group the device belongs to
1639 * @stats: Statistics struct, which was left as a legacy, use
1640 * rtnl_link_stats64 instead
1642 * @rx_dropped: Dropped packets by core network,
1643 * do not use this in drivers
1644 * @tx_dropped: Dropped packets by core network,
1645 * do not use this in drivers
1646 * @rx_nohandler: nohandler dropped packets by core network on
1647 * inactive devices, do not use this in drivers
1648 * @carrier_up_count: Number of times the carrier has been up
1649 * @carrier_down_count: Number of times the carrier has been down
1651 * @wireless_handlers: List of functions to handle Wireless Extensions,
1653 * see <net/iw_handler.h> for details.
1654 * @wireless_data: Instance data managed by the core of wireless extensions
1656 * @netdev_ops: Includes several pointers to callbacks,
1657 * if one wants to override the ndo_*() functions
1658 * @ethtool_ops: Management operations
1659 * @l3mdev_ops: Layer 3 master device operations
1660 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1661 * discovery handling. Necessary for e.g. 6LoWPAN.
1662 * @xfrmdev_ops: Transformation offload operations
1663 * @tlsdev_ops: Transport Layer Security offload operations
1664 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1665 * of Layer 2 headers.
1667 * @flags: Interface flags (a la BSD)
1668 * @priv_flags: Like 'flags' but invisible to userspace,
1669 * see if.h for the definitions
1670 * @gflags: Global flags ( kept as legacy )
1671 * @padded: How much padding added by alloc_netdev()
1672 * @operstate: RFC2863 operstate
1673 * @link_mode: Mapping policy to operstate
1674 * @if_port: Selectable AUI, TP, ...
1676 * @mtu: Interface MTU value
1677 * @min_mtu: Interface Minimum MTU value
1678 * @max_mtu: Interface Maximum MTU value
1679 * @type: Interface hardware type
1680 * @hard_header_len: Maximum hardware header length.
1681 * @min_header_len: Minimum hardware header length
1683 * @needed_headroom: Extra headroom the hardware may need, but not in all
1684 * cases can this be guaranteed
1685 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1686 * cases can this be guaranteed. Some cases also use
1687 * LL_MAX_HEADER instead to allocate the skb
1689 * interface address info:
1691 * @perm_addr: Permanent hw address
1692 * @addr_assign_type: Hw address assignment type
1693 * @addr_len: Hardware address length
1694 * @upper_level: Maximum depth level of upper devices.
1695 * @lower_level: Maximum depth level of lower devices.
1696 * @neigh_priv_len: Used in neigh_alloc()
1697 * @dev_id: Used to differentiate devices that share
1698 * the same link layer address
1699 * @dev_port: Used to differentiate devices that share
1701 * @addr_list_lock: XXX: need comments on this one
1702 * @name_assign_type: network interface name assignment type
1703 * @uc_promisc: Counter that indicates promiscuous mode
1704 * has been enabled due to the need to listen to
1705 * additional unicast addresses in a device that
1706 * does not implement ndo_set_rx_mode()
1707 * @uc: unicast mac addresses
1708 * @mc: multicast mac addresses
1709 * @dev_addrs: list of device hw addresses
1710 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1711 * @promiscuity: Number of times the NIC is told to work in
1712 * promiscuous mode; if it becomes 0 the NIC will
1713 * exit promiscuous mode
1714 * @allmulti: Counter, enables or disables allmulticast mode
1716 * @vlan_info: VLAN info
1717 * @dsa_ptr: dsa specific data
1718 * @tipc_ptr: TIPC specific data
1719 * @atalk_ptr: AppleTalk link
1720 * @ip_ptr: IPv4 specific data
1721 * @dn_ptr: DECnet specific data
1722 * @ip6_ptr: IPv6 specific data
1723 * @ax25_ptr: AX.25 specific data
1724 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1725 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1727 * @mpls_ptr: mpls_dev struct pointer
1729 * @dev_addr: Hw address (before bcast,
1730 * because most packets are unicast)
1732 * @_rx: Array of RX queues
1733 * @num_rx_queues: Number of RX queues
1734 * allocated at register_netdev() time
1735 * @real_num_rx_queues: Number of RX queues currently active in device
1736 * @xdp_prog: XDP sockets filter program pointer
1737 * @gro_flush_timeout: timeout for GRO layer in NAPI
1738 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1739 * allow to avoid NIC hard IRQ, on busy queues.
1741 * @rx_handler: handler for received packets
1742 * @rx_handler_data: XXX: need comments on this one
1743 * @miniq_ingress: ingress/clsact qdisc specific data for
1744 * ingress processing
1745 * @ingress_queue: XXX: need comments on this one
1746 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1747 * @broadcast: hw bcast address
1749 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1750 * indexed by RX queue number. Assigned by driver.
1751 * This must only be set if the ndo_rx_flow_steer
1752 * operation is defined
1753 * @index_hlist: Device index hash chain
1755 * @_tx: Array of TX queues
1756 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1757 * @real_num_tx_queues: Number of TX queues currently active in device
1758 * @qdisc: Root qdisc from userspace point of view
1759 * @tx_queue_len: Max frames per queue allowed
1760 * @tx_global_lock: XXX: need comments on this one
1761 * @xdp_bulkq: XDP device bulk queue
1762 * @xps_cpus_map: all CPUs map for XPS device
1763 * @xps_rxqs_map: all RXQs map for XPS device
1765 * @xps_maps: XXX: need comments on this one
1766 * @miniq_egress: clsact qdisc specific data for
1768 * @qdisc_hash: qdisc hash table
1769 * @watchdog_timeo: Represents the timeout that is used by
1770 * the watchdog (see dev_watchdog())
1771 * @watchdog_timer: List of timers
1773 * @proto_down_reason: reason a netdev interface is held down
1774 * @pcpu_refcnt: Number of references to this device
1775 * @todo_list: Delayed register/unregister
1776 * @link_watch_list: XXX: need comments on this one
1778 * @reg_state: Register/unregister state machine
1779 * @dismantle: Device is going to be freed
1780 * @rtnl_link_state: This enum represents the phases of creating
1783 * @needs_free_netdev: Should unregister perform free_netdev?
1784 * @priv_destructor: Called from unregister
1785 * @npinfo: XXX: need comments on this one
1786 * @nd_net: Network namespace this network device is inside
1788 * @ml_priv: Mid-layer private
1789 * @lstats: Loopback statistics
1790 * @tstats: Tunnel statistics
1791 * @dstats: Dummy statistics
1792 * @vstats: Virtual ethernet statistics
1797 * @dev: Class/net/name entry
1798 * @sysfs_groups: Space for optional device, statistics and wireless
1801 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1802 * @rtnl_link_ops: Rtnl_link_ops
1804 * @gso_max_size: Maximum size of generic segmentation offload
1805 * @gso_max_segs: Maximum number of segments that can be passed to the
1808 * @dcbnl_ops: Data Center Bridging netlink ops
1809 * @num_tc: Number of traffic classes in the net device
1810 * @tc_to_txq: XXX: need comments on this one
1811 * @prio_tc_map: XXX: need comments on this one
1813 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1815 * @priomap: XXX: need comments on this one
1816 * @phydev: Physical device may attach itself
1817 * for hardware timestamping
1818 * @sfp_bus: attached &struct sfp_bus structure.
1820 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1821 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1823 * @proto_down: protocol port state information can be sent to the
1824 * switch driver and used to set the phys state of the
1827 * @wol_enabled: Wake-on-LAN is enabled
1829 * @net_notifier_list: List of per-net netdev notifier block
1830 * that follow this device when it is moved
1831 * to another network namespace.
1833 * @macsec_ops: MACsec offloading ops
1835 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1836 * offload capabilities of the device
1837 * @udp_tunnel_nic: UDP tunnel offload state
1838 * @xdp_state: stores info on attached XDP BPF programs
1840 * @nested_level: Used as as a parameter of spin_lock_nested() of
1841 * dev->addr_list_lock.
1842 * @unlink_list: As netif_addr_lock() can be called recursively,
1843 * keep a list of interfaces to be deleted.
1845 * FIXME: cleanup struct net_device such that network protocol info
1850 char name[IFNAMSIZ];
1851 struct netdev_name_node *name_node;
1852 struct dev_ifalias __rcu *ifalias;
1854 * I/O specific fields
1855 * FIXME: Merge these and struct ifmap into one
1857 unsigned long mem_end;
1858 unsigned long mem_start;
1859 unsigned long base_addr;
1863 * Some hardware also needs these fields (state,dev_list,
1864 * napi_list,unreg_list,close_list) but they are not
1865 * part of the usual set specified in Space.c.
1868 unsigned long state;
1870 struct list_head dev_list;
1871 struct list_head napi_list;
1872 struct list_head unreg_list;
1873 struct list_head close_list;
1874 struct list_head ptype_all;
1875 struct list_head ptype_specific;
1878 struct list_head upper;
1879 struct list_head lower;
1882 netdev_features_t features;
1883 netdev_features_t hw_features;
1884 netdev_features_t wanted_features;
1885 netdev_features_t vlan_features;
1886 netdev_features_t hw_enc_features;
1887 netdev_features_t mpls_features;
1888 netdev_features_t gso_partial_features;
1893 struct net_device_stats stats;
1895 atomic_long_t rx_dropped;
1896 atomic_long_t tx_dropped;
1897 atomic_long_t rx_nohandler;
1899 /* Stats to monitor link on/off, flapping */
1900 atomic_t carrier_up_count;
1901 atomic_t carrier_down_count;
1903 #ifdef CONFIG_WIRELESS_EXT
1904 const struct iw_handler_def *wireless_handlers;
1905 struct iw_public_data *wireless_data;
1907 const struct net_device_ops *netdev_ops;
1908 const struct ethtool_ops *ethtool_ops;
1909 #ifdef CONFIG_NET_L3_MASTER_DEV
1910 const struct l3mdev_ops *l3mdev_ops;
1912 #if IS_ENABLED(CONFIG_IPV6)
1913 const struct ndisc_ops *ndisc_ops;
1916 #ifdef CONFIG_XFRM_OFFLOAD
1917 const struct xfrmdev_ops *xfrmdev_ops;
1920 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1921 const struct tlsdev_ops *tlsdev_ops;
1924 const struct header_ops *header_ops;
1927 unsigned int priv_flags;
1929 unsigned short gflags;
1930 unsigned short padded;
1932 unsigned char operstate;
1933 unsigned char link_mode;
1935 unsigned char if_port;
1938 /* Note : dev->mtu is often read without holding a lock.
1939 * Writers usually hold RTNL.
1940 * It is recommended to use READ_ONCE() to annotate the reads,
1941 * and to use WRITE_ONCE() to annotate the writes.
1944 unsigned int min_mtu;
1945 unsigned int max_mtu;
1946 unsigned short type;
1947 unsigned short hard_header_len;
1948 unsigned char min_header_len;
1949 unsigned char name_assign_type;
1951 unsigned short needed_headroom;
1952 unsigned short needed_tailroom;
1954 /* Interface address info. */
1955 unsigned char perm_addr[MAX_ADDR_LEN];
1956 unsigned char addr_assign_type;
1957 unsigned char addr_len;
1958 unsigned char upper_level;
1959 unsigned char lower_level;
1961 unsigned short neigh_priv_len;
1962 unsigned short dev_id;
1963 unsigned short dev_port;
1964 spinlock_t addr_list_lock;
1966 struct netdev_hw_addr_list uc;
1967 struct netdev_hw_addr_list mc;
1968 struct netdev_hw_addr_list dev_addrs;
1971 struct kset *queues_kset;
1973 #ifdef CONFIG_LOCKDEP
1974 struct list_head unlink_list;
1976 unsigned int promiscuity;
1977 unsigned int allmulti;
1979 #ifdef CONFIG_LOCKDEP
1980 unsigned char nested_level;
1984 /* Protocol-specific pointers */
1986 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1987 struct vlan_info __rcu *vlan_info;
1989 #if IS_ENABLED(CONFIG_NET_DSA)
1990 struct dsa_port *dsa_ptr;
1992 #if IS_ENABLED(CONFIG_TIPC)
1993 struct tipc_bearer __rcu *tipc_ptr;
1995 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1998 struct in_device __rcu *ip_ptr;
1999 #if IS_ENABLED(CONFIG_DECNET)
2000 struct dn_dev __rcu *dn_ptr;
2002 struct inet6_dev __rcu *ip6_ptr;
2003 #if IS_ENABLED(CONFIG_AX25)
2006 struct wireless_dev *ieee80211_ptr;
2007 struct wpan_dev *ieee802154_ptr;
2008 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2009 struct mpls_dev __rcu *mpls_ptr;
2013 * Cache lines mostly used on receive path (including eth_type_trans())
2015 /* Interface address info used in eth_type_trans() */
2016 unsigned char *dev_addr;
2018 struct netdev_rx_queue *_rx;
2019 unsigned int num_rx_queues;
2020 unsigned int real_num_rx_queues;
2022 struct bpf_prog __rcu *xdp_prog;
2023 unsigned long gro_flush_timeout;
2024 int napi_defer_hard_irqs;
2025 rx_handler_func_t __rcu *rx_handler;
2026 void __rcu *rx_handler_data;
2028 #ifdef CONFIG_NET_CLS_ACT
2029 struct mini_Qdisc __rcu *miniq_ingress;
2031 struct netdev_queue __rcu *ingress_queue;
2032 #ifdef CONFIG_NETFILTER_INGRESS
2033 struct nf_hook_entries __rcu *nf_hooks_ingress;
2036 unsigned char broadcast[MAX_ADDR_LEN];
2037 #ifdef CONFIG_RFS_ACCEL
2038 struct cpu_rmap *rx_cpu_rmap;
2040 struct hlist_node index_hlist;
2043 * Cache lines mostly used on transmit path
2045 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2046 unsigned int num_tx_queues;
2047 unsigned int real_num_tx_queues;
2048 struct Qdisc *qdisc;
2049 unsigned int tx_queue_len;
2050 spinlock_t tx_global_lock;
2052 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2055 struct xps_dev_maps __rcu *xps_cpus_map;
2056 struct xps_dev_maps __rcu *xps_rxqs_map;
2058 #ifdef CONFIG_NET_CLS_ACT
2059 struct mini_Qdisc __rcu *miniq_egress;
2062 #ifdef CONFIG_NET_SCHED
2063 DECLARE_HASHTABLE (qdisc_hash, 4);
2065 /* These may be needed for future network-power-down code. */
2066 struct timer_list watchdog_timer;
2069 u32 proto_down_reason;
2071 struct list_head todo_list;
2072 int __percpu *pcpu_refcnt;
2074 struct list_head link_watch_list;
2076 enum { NETREG_UNINITIALIZED=0,
2077 NETREG_REGISTERED, /* completed register_netdevice */
2078 NETREG_UNREGISTERING, /* called unregister_netdevice */
2079 NETREG_UNREGISTERED, /* completed unregister todo */
2080 NETREG_RELEASED, /* called free_netdev */
2081 NETREG_DUMMY, /* dummy device for NAPI poll */
2087 RTNL_LINK_INITIALIZED,
2088 RTNL_LINK_INITIALIZING,
2089 } rtnl_link_state:16;
2091 bool needs_free_netdev;
2092 void (*priv_destructor)(struct net_device *dev);
2094 #ifdef CONFIG_NETPOLL
2095 struct netpoll_info __rcu *npinfo;
2098 possible_net_t nd_net;
2100 /* mid-layer private */
2103 struct pcpu_lstats __percpu *lstats;
2104 struct pcpu_sw_netstats __percpu *tstats;
2105 struct pcpu_dstats __percpu *dstats;
2108 #if IS_ENABLED(CONFIG_GARP)
2109 struct garp_port __rcu *garp_port;
2111 #if IS_ENABLED(CONFIG_MRP)
2112 struct mrp_port __rcu *mrp_port;
2116 const struct attribute_group *sysfs_groups[4];
2117 const struct attribute_group *sysfs_rx_queue_group;
2119 const struct rtnl_link_ops *rtnl_link_ops;
2121 /* for setting kernel sock attribute on TCP connection setup */
2122 #define GSO_MAX_SIZE 65536
2123 unsigned int gso_max_size;
2124 #define GSO_MAX_SEGS 65535
2128 const struct dcbnl_rtnl_ops *dcbnl_ops;
2131 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2132 u8 prio_tc_map[TC_BITMASK + 1];
2134 #if IS_ENABLED(CONFIG_FCOE)
2135 unsigned int fcoe_ddp_xid;
2137 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2138 struct netprio_map __rcu *priomap;
2140 struct phy_device *phydev;
2141 struct sfp_bus *sfp_bus;
2142 struct lock_class_key *qdisc_tx_busylock;
2143 struct lock_class_key *qdisc_running_key;
2145 unsigned wol_enabled:1;
2147 struct list_head net_notifier_list;
2149 #if IS_ENABLED(CONFIG_MACSEC)
2150 /* MACsec management functions */
2151 const struct macsec_ops *macsec_ops;
2153 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2154 struct udp_tunnel_nic *udp_tunnel_nic;
2156 /* protected by rtnl_lock */
2157 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2159 #define to_net_dev(d) container_of(d, struct net_device, dev)
2161 static inline bool netif_elide_gro(const struct net_device *dev)
2163 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2168 #define NETDEV_ALIGN 32
2171 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2173 return dev->prio_tc_map[prio & TC_BITMASK];
2177 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2179 if (tc >= dev->num_tc)
2182 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2186 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2187 void netdev_reset_tc(struct net_device *dev);
2188 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2189 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2192 int netdev_get_num_tc(struct net_device *dev)
2197 static inline void net_prefetch(void *p)
2200 #if L1_CACHE_BYTES < 128
2201 prefetch((u8 *)p + L1_CACHE_BYTES);
2205 static inline void net_prefetchw(void *p)
2208 #if L1_CACHE_BYTES < 128
2209 prefetchw((u8 *)p + L1_CACHE_BYTES);
2213 void netdev_unbind_sb_channel(struct net_device *dev,
2214 struct net_device *sb_dev);
2215 int netdev_bind_sb_channel_queue(struct net_device *dev,
2216 struct net_device *sb_dev,
2217 u8 tc, u16 count, u16 offset);
2218 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2219 static inline int netdev_get_sb_channel(struct net_device *dev)
2221 return max_t(int, -dev->num_tc, 0);
2225 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2228 return &dev->_tx[index];
2231 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2232 const struct sk_buff *skb)
2234 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2237 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2238 void (*f)(struct net_device *,
2239 struct netdev_queue *,
2245 for (i = 0; i < dev->num_tx_queues; i++)
2246 f(dev, &dev->_tx[i], arg);
2249 #define netdev_lockdep_set_classes(dev) \
2251 static struct lock_class_key qdisc_tx_busylock_key; \
2252 static struct lock_class_key qdisc_running_key; \
2253 static struct lock_class_key qdisc_xmit_lock_key; \
2254 static struct lock_class_key dev_addr_list_lock_key; \
2257 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2258 (dev)->qdisc_running_key = &qdisc_running_key; \
2259 lockdep_set_class(&(dev)->addr_list_lock, \
2260 &dev_addr_list_lock_key); \
2261 for (i = 0; i < (dev)->num_tx_queues; i++) \
2262 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2263 &qdisc_xmit_lock_key); \
2266 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2267 struct net_device *sb_dev);
2268 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2269 struct sk_buff *skb,
2270 struct net_device *sb_dev);
2272 /* returns the headroom that the master device needs to take in account
2273 * when forwarding to this dev
2275 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2277 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2280 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2282 if (dev->netdev_ops->ndo_set_rx_headroom)
2283 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2286 /* set the device rx headroom to the dev's default */
2287 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2289 netdev_set_rx_headroom(dev, -1);
2293 * Net namespace inlines
2296 struct net *dev_net(const struct net_device *dev)
2298 return read_pnet(&dev->nd_net);
2302 void dev_net_set(struct net_device *dev, struct net *net)
2304 write_pnet(&dev->nd_net, net);
2308 * netdev_priv - access network device private data
2309 * @dev: network device
2311 * Get network device private data
2313 static inline void *netdev_priv(const struct net_device *dev)
2315 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2318 /* Set the sysfs physical device reference for the network logical device
2319 * if set prior to registration will cause a symlink during initialization.
2321 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2323 /* Set the sysfs device type for the network logical device to allow
2324 * fine-grained identification of different network device types. For
2325 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2327 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2329 /* Default NAPI poll() weight
2330 * Device drivers are strongly advised to not use bigger value
2332 #define NAPI_POLL_WEIGHT 64
2335 * netif_napi_add - initialize a NAPI context
2336 * @dev: network device
2337 * @napi: NAPI context
2338 * @poll: polling function
2339 * @weight: default weight
2341 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2342 * *any* of the other NAPI-related functions.
2344 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2345 int (*poll)(struct napi_struct *, int), int weight);
2348 * netif_tx_napi_add - initialize a NAPI context
2349 * @dev: network device
2350 * @napi: NAPI context
2351 * @poll: polling function
2352 * @weight: default weight
2354 * This variant of netif_napi_add() should be used from drivers using NAPI
2355 * to exclusively poll a TX queue.
2356 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2358 static inline void netif_tx_napi_add(struct net_device *dev,
2359 struct napi_struct *napi,
2360 int (*poll)(struct napi_struct *, int),
2363 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2364 netif_napi_add(dev, napi, poll, weight);
2368 * __netif_napi_del - remove a NAPI context
2369 * @napi: NAPI context
2371 * Warning: caller must observe RCU grace period before freeing memory
2372 * containing @napi. Drivers might want to call this helper to combine
2373 * all the needed RCU grace periods into a single one.
2375 void __netif_napi_del(struct napi_struct *napi);
2378 * netif_napi_del - remove a NAPI context
2379 * @napi: NAPI context
2381 * netif_napi_del() removes a NAPI context from the network device NAPI list
2383 static inline void netif_napi_del(struct napi_struct *napi)
2385 __netif_napi_del(napi);
2389 struct napi_gro_cb {
2390 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2393 /* Length of frag0. */
2394 unsigned int frag0_len;
2396 /* This indicates where we are processing relative to skb->data. */
2399 /* This is non-zero if the packet cannot be merged with the new skb. */
2402 /* Save the IP ID here and check when we get to the transport layer */
2405 /* Number of segments aggregated. */
2408 /* Start offset for remote checksum offload */
2409 u16 gro_remcsum_start;
2411 /* jiffies when first packet was created/queued */
2414 /* Used in ipv6_gro_receive() and foo-over-udp */
2417 /* This is non-zero if the packet may be of the same flow. */
2420 /* Used in tunnel GRO receive */
2423 /* GRO checksum is valid */
2426 /* Number of checksums via CHECKSUM_UNNECESSARY */
2431 #define NAPI_GRO_FREE 1
2432 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2434 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2437 /* Used in GRE, set in fou/gue_gro_receive */
2440 /* Used to determine if flush_id can be ignored */
2443 /* Number of gro_receive callbacks this packet already went through */
2444 u8 recursion_counter:4;
2446 /* GRO is done by frag_list pointer chaining. */
2449 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2452 /* used in skb_gro_receive() slow path */
2453 struct sk_buff *last;
2456 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2458 #define GRO_RECURSION_LIMIT 15
2459 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2461 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2464 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2465 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2466 struct list_head *head,
2467 struct sk_buff *skb)
2469 if (unlikely(gro_recursion_inc_test(skb))) {
2470 NAPI_GRO_CB(skb)->flush |= 1;
2474 return cb(head, skb);
2477 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2479 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2481 struct list_head *head,
2482 struct sk_buff *skb)
2484 if (unlikely(gro_recursion_inc_test(skb))) {
2485 NAPI_GRO_CB(skb)->flush |= 1;
2489 return cb(sk, head, skb);
2492 struct packet_type {
2493 __be16 type; /* This is really htons(ether_type). */
2494 bool ignore_outgoing;
2495 struct net_device *dev; /* NULL is wildcarded here */
2496 int (*func) (struct sk_buff *,
2497 struct net_device *,
2498 struct packet_type *,
2499 struct net_device *);
2500 void (*list_func) (struct list_head *,
2501 struct packet_type *,
2502 struct net_device *);
2503 bool (*id_match)(struct packet_type *ptype,
2505 void *af_packet_priv;
2506 struct list_head list;
2509 struct offload_callbacks {
2510 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2511 netdev_features_t features);
2512 struct sk_buff *(*gro_receive)(struct list_head *head,
2513 struct sk_buff *skb);
2514 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2517 struct packet_offload {
2518 __be16 type; /* This is really htons(ether_type). */
2520 struct offload_callbacks callbacks;
2521 struct list_head list;
2524 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2525 struct pcpu_sw_netstats {
2530 struct u64_stats_sync syncp;
2531 } __aligned(4 * sizeof(u64));
2533 struct pcpu_lstats {
2534 u64_stats_t packets;
2536 struct u64_stats_sync syncp;
2537 } __aligned(2 * sizeof(u64));
2539 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2541 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2543 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2545 u64_stats_update_begin(&tstats->syncp);
2546 tstats->rx_bytes += len;
2547 tstats->rx_packets++;
2548 u64_stats_update_end(&tstats->syncp);
2551 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2552 unsigned int packets,
2555 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2557 u64_stats_update_begin(&tstats->syncp);
2558 tstats->tx_bytes += len;
2559 tstats->tx_packets += packets;
2560 u64_stats_update_end(&tstats->syncp);
2563 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2565 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2567 u64_stats_update_begin(&lstats->syncp);
2568 u64_stats_add(&lstats->bytes, len);
2569 u64_stats_inc(&lstats->packets);
2570 u64_stats_update_end(&lstats->syncp);
2573 #define __netdev_alloc_pcpu_stats(type, gfp) \
2575 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2578 for_each_possible_cpu(__cpu) { \
2579 typeof(type) *stat; \
2580 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2581 u64_stats_init(&stat->syncp); \
2587 #define netdev_alloc_pcpu_stats(type) \
2588 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2590 #define devm_netdev_alloc_pcpu_stats(dev, type) \
2592 typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2595 for_each_possible_cpu(__cpu) { \
2596 typeof(type) *stat; \
2597 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2598 u64_stats_init(&stat->syncp); \
2604 enum netdev_lag_tx_type {
2605 NETDEV_LAG_TX_TYPE_UNKNOWN,
2606 NETDEV_LAG_TX_TYPE_RANDOM,
2607 NETDEV_LAG_TX_TYPE_BROADCAST,
2608 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2609 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2610 NETDEV_LAG_TX_TYPE_HASH,
2613 enum netdev_lag_hash {
2614 NETDEV_LAG_HASH_NONE,
2616 NETDEV_LAG_HASH_L34,
2617 NETDEV_LAG_HASH_L23,
2618 NETDEV_LAG_HASH_E23,
2619 NETDEV_LAG_HASH_E34,
2620 NETDEV_LAG_HASH_VLAN_SRCMAC,
2621 NETDEV_LAG_HASH_UNKNOWN,
2624 struct netdev_lag_upper_info {
2625 enum netdev_lag_tx_type tx_type;
2626 enum netdev_lag_hash hash_type;
2629 struct netdev_lag_lower_state_info {
2634 #include <linux/notifier.h>
2636 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2637 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2641 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2643 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2644 detected a hardware crash and restarted
2645 - we can use this eg to kick tcp sessions
2647 NETDEV_CHANGE, /* Notify device state change */
2650 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2651 NETDEV_CHANGEADDR, /* notify after the address change */
2652 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2656 NETDEV_BONDING_FAILOVER,
2658 NETDEV_PRE_TYPE_CHANGE,
2659 NETDEV_POST_TYPE_CHANGE,
2662 NETDEV_NOTIFY_PEERS,
2666 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2667 NETDEV_CHANGEINFODATA,
2668 NETDEV_BONDING_INFO,
2669 NETDEV_PRECHANGEUPPER,
2670 NETDEV_CHANGELOWERSTATE,
2671 NETDEV_UDP_TUNNEL_PUSH_INFO,
2672 NETDEV_UDP_TUNNEL_DROP_INFO,
2673 NETDEV_CHANGE_TX_QUEUE_LEN,
2674 NETDEV_CVLAN_FILTER_PUSH_INFO,
2675 NETDEV_CVLAN_FILTER_DROP_INFO,
2676 NETDEV_SVLAN_FILTER_PUSH_INFO,
2677 NETDEV_SVLAN_FILTER_DROP_INFO,
2679 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2681 int register_netdevice_notifier(struct notifier_block *nb);
2682 int unregister_netdevice_notifier(struct notifier_block *nb);
2683 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2684 int unregister_netdevice_notifier_net(struct net *net,
2685 struct notifier_block *nb);
2686 int register_netdevice_notifier_dev_net(struct net_device *dev,
2687 struct notifier_block *nb,
2688 struct netdev_net_notifier *nn);
2689 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2690 struct notifier_block *nb,
2691 struct netdev_net_notifier *nn);
2693 struct netdev_notifier_info {
2694 struct net_device *dev;
2695 struct netlink_ext_ack *extack;
2698 struct netdev_notifier_info_ext {
2699 struct netdev_notifier_info info; /* must be first */
2705 struct netdev_notifier_change_info {
2706 struct netdev_notifier_info info; /* must be first */
2707 unsigned int flags_changed;
2710 struct netdev_notifier_changeupper_info {
2711 struct netdev_notifier_info info; /* must be first */
2712 struct net_device *upper_dev; /* new upper dev */
2713 bool master; /* is upper dev master */
2714 bool linking; /* is the notification for link or unlink */
2715 void *upper_info; /* upper dev info */
2718 struct netdev_notifier_changelowerstate_info {
2719 struct netdev_notifier_info info; /* must be first */
2720 void *lower_state_info; /* is lower dev state */
2723 struct netdev_notifier_pre_changeaddr_info {
2724 struct netdev_notifier_info info; /* must be first */
2725 const unsigned char *dev_addr;
2728 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2729 struct net_device *dev)
2732 info->extack = NULL;
2735 static inline struct net_device *
2736 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2741 static inline struct netlink_ext_ack *
2742 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2744 return info->extack;
2747 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2750 extern rwlock_t dev_base_lock; /* Device list lock */
2752 #define for_each_netdev(net, d) \
2753 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2754 #define for_each_netdev_reverse(net, d) \
2755 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2756 #define for_each_netdev_rcu(net, d) \
2757 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2758 #define for_each_netdev_safe(net, d, n) \
2759 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2760 #define for_each_netdev_continue(net, d) \
2761 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2762 #define for_each_netdev_continue_reverse(net, d) \
2763 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2765 #define for_each_netdev_continue_rcu(net, d) \
2766 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2767 #define for_each_netdev_in_bond_rcu(bond, slave) \
2768 for_each_netdev_rcu(&init_net, slave) \
2769 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2770 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2772 static inline struct net_device *next_net_device(struct net_device *dev)
2774 struct list_head *lh;
2778 lh = dev->dev_list.next;
2779 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2782 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2784 struct list_head *lh;
2788 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2789 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2792 static inline struct net_device *first_net_device(struct net *net)
2794 return list_empty(&net->dev_base_head) ? NULL :
2795 net_device_entry(net->dev_base_head.next);
2798 static inline struct net_device *first_net_device_rcu(struct net *net)
2800 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2802 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2805 int netdev_boot_setup_check(struct net_device *dev);
2806 unsigned long netdev_boot_base(const char *prefix, int unit);
2807 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2808 const char *hwaddr);
2809 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2810 void dev_add_pack(struct packet_type *pt);
2811 void dev_remove_pack(struct packet_type *pt);
2812 void __dev_remove_pack(struct packet_type *pt);
2813 void dev_add_offload(struct packet_offload *po);
2814 void dev_remove_offload(struct packet_offload *po);
2816 int dev_get_iflink(const struct net_device *dev);
2817 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2818 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2819 unsigned short mask);
2820 struct net_device *dev_get_by_name(struct net *net, const char *name);
2821 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2822 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2823 int dev_alloc_name(struct net_device *dev, const char *name);
2824 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2825 void dev_close(struct net_device *dev);
2826 void dev_close_many(struct list_head *head, bool unlink);
2827 void dev_disable_lro(struct net_device *dev);
2828 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2829 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2830 struct net_device *sb_dev);
2831 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2832 struct net_device *sb_dev);
2834 int dev_queue_xmit(struct sk_buff *skb);
2835 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2836 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2838 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
2842 ret = __dev_direct_xmit(skb, queue_id);
2843 if (!dev_xmit_complete(ret))
2848 int register_netdevice(struct net_device *dev);
2849 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2850 void unregister_netdevice_many(struct list_head *head);
2851 static inline void unregister_netdevice(struct net_device *dev)
2853 unregister_netdevice_queue(dev, NULL);
2856 int netdev_refcnt_read(const struct net_device *dev);
2857 void free_netdev(struct net_device *dev);
2858 void netdev_freemem(struct net_device *dev);
2859 int init_dummy_netdev(struct net_device *dev);
2861 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2862 struct sk_buff *skb,
2864 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
2866 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2867 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2868 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2869 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2870 int netdev_get_name(struct net *net, char *name, int ifindex);
2871 int dev_restart(struct net_device *dev);
2872 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2873 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2875 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2877 return NAPI_GRO_CB(skb)->data_offset;
2880 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2882 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2885 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2887 NAPI_GRO_CB(skb)->data_offset += len;
2890 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2891 unsigned int offset)
2893 return NAPI_GRO_CB(skb)->frag0 + offset;
2896 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2898 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2901 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2903 NAPI_GRO_CB(skb)->frag0 = NULL;
2904 NAPI_GRO_CB(skb)->frag0_len = 0;
2907 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2908 unsigned int offset)
2910 if (!pskb_may_pull(skb, hlen))
2913 skb_gro_frag0_invalidate(skb);
2914 return skb->data + offset;
2917 static inline void *skb_gro_network_header(struct sk_buff *skb)
2919 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2920 skb_network_offset(skb);
2923 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2924 const void *start, unsigned int len)
2926 if (NAPI_GRO_CB(skb)->csum_valid)
2927 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2928 csum_partial(start, len, 0));
2931 /* GRO checksum functions. These are logical equivalents of the normal
2932 * checksum functions (in skbuff.h) except that they operate on the GRO
2933 * offsets and fields in sk_buff.
2936 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2938 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2940 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2943 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2947 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2948 skb_checksum_start_offset(skb) <
2949 skb_gro_offset(skb)) &&
2950 !skb_at_gro_remcsum_start(skb) &&
2951 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2952 (!zero_okay || check));
2955 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2958 if (NAPI_GRO_CB(skb)->csum_valid &&
2959 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2962 NAPI_GRO_CB(skb)->csum = psum;
2964 return __skb_gro_checksum_complete(skb);
2967 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2969 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2970 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2971 NAPI_GRO_CB(skb)->csum_cnt--;
2973 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2974 * verified a new top level checksum or an encapsulated one
2975 * during GRO. This saves work if we fallback to normal path.
2977 __skb_incr_checksum_unnecessary(skb);
2981 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2984 __sum16 __ret = 0; \
2985 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2986 __ret = __skb_gro_checksum_validate_complete(skb, \
2987 compute_pseudo(skb, proto)); \
2989 skb_gro_incr_csum_unnecessary(skb); \
2993 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2994 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2996 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2998 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
3000 #define skb_gro_checksum_simple_validate(skb) \
3001 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
3003 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
3005 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
3006 !NAPI_GRO_CB(skb)->csum_valid);
3009 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
3012 NAPI_GRO_CB(skb)->csum = ~pseudo;
3013 NAPI_GRO_CB(skb)->csum_valid = 1;
3016 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
3018 if (__skb_gro_checksum_convert_check(skb)) \
3019 __skb_gro_checksum_convert(skb, \
3020 compute_pseudo(skb, proto)); \
3023 struct gro_remcsum {
3028 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3034 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3035 unsigned int off, size_t hdrlen,
3036 int start, int offset,
3037 struct gro_remcsum *grc,
3041 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3043 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3046 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3050 ptr = skb_gro_header_fast(skb, off);
3051 if (skb_gro_header_hard(skb, off + plen)) {
3052 ptr = skb_gro_header_slow(skb, off + plen, off);
3057 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3060 /* Adjust skb->csum since we changed the packet */
3061 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3063 grc->offset = off + hdrlen + offset;
3069 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3070 struct gro_remcsum *grc)
3073 size_t plen = grc->offset + sizeof(u16);
3078 ptr = skb_gro_header_fast(skb, grc->offset);
3079 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3080 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3085 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3088 #ifdef CONFIG_XFRM_OFFLOAD
3089 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3091 if (PTR_ERR(pp) != -EINPROGRESS)
3092 NAPI_GRO_CB(skb)->flush |= flush;
3094 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3097 struct gro_remcsum *grc)
3099 if (PTR_ERR(pp) != -EINPROGRESS) {
3100 NAPI_GRO_CB(skb)->flush |= flush;
3101 skb_gro_remcsum_cleanup(skb, grc);
3102 skb->remcsum_offload = 0;
3106 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3108 NAPI_GRO_CB(skb)->flush |= flush;
3110 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3113 struct gro_remcsum *grc)
3115 NAPI_GRO_CB(skb)->flush |= flush;
3116 skb_gro_remcsum_cleanup(skb, grc);
3117 skb->remcsum_offload = 0;
3121 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3122 unsigned short type,
3123 const void *daddr, const void *saddr,
3126 if (!dev->header_ops || !dev->header_ops->create)
3129 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3132 static inline int dev_parse_header(const struct sk_buff *skb,
3133 unsigned char *haddr)
3135 const struct net_device *dev = skb->dev;
3137 if (!dev->header_ops || !dev->header_ops->parse)
3139 return dev->header_ops->parse(skb, haddr);
3142 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3144 const struct net_device *dev = skb->dev;
3146 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3148 return dev->header_ops->parse_protocol(skb);
3151 /* ll_header must have at least hard_header_len allocated */
3152 static inline bool dev_validate_header(const struct net_device *dev,
3153 char *ll_header, int len)
3155 if (likely(len >= dev->hard_header_len))
3157 if (len < dev->min_header_len)
3160 if (capable(CAP_SYS_RAWIO)) {
3161 memset(ll_header + len, 0, dev->hard_header_len - len);
3165 if (dev->header_ops && dev->header_ops->validate)
3166 return dev->header_ops->validate(ll_header, len);
3171 static inline bool dev_has_header(const struct net_device *dev)
3173 return dev->header_ops && dev->header_ops->create;
3176 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3178 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3179 static inline int unregister_gifconf(unsigned int family)
3181 return register_gifconf(family, NULL);
3184 #ifdef CONFIG_NET_FLOW_LIMIT
3185 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3186 struct sd_flow_limit {
3188 unsigned int num_buckets;
3189 unsigned int history_head;
3190 u16 history[FLOW_LIMIT_HISTORY];
3194 extern int netdev_flow_limit_table_len;
3195 #endif /* CONFIG_NET_FLOW_LIMIT */
3198 * Incoming packets are placed on per-CPU queues
3200 struct softnet_data {
3201 struct list_head poll_list;
3202 struct sk_buff_head process_queue;
3205 unsigned int processed;
3206 unsigned int time_squeeze;
3207 unsigned int received_rps;
3209 struct softnet_data *rps_ipi_list;
3211 #ifdef CONFIG_NET_FLOW_LIMIT
3212 struct sd_flow_limit __rcu *flow_limit;
3214 struct Qdisc *output_queue;
3215 struct Qdisc **output_queue_tailp;
3216 struct sk_buff *completion_queue;
3217 #ifdef CONFIG_XFRM_OFFLOAD
3218 struct sk_buff_head xfrm_backlog;
3220 /* written and read only by owning cpu: */
3226 /* input_queue_head should be written by cpu owning this struct,
3227 * and only read by other cpus. Worth using a cache line.
3229 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3231 /* Elements below can be accessed between CPUs for RPS/RFS */
3232 call_single_data_t csd ____cacheline_aligned_in_smp;
3233 struct softnet_data *rps_ipi_next;
3235 unsigned int input_queue_tail;
3237 unsigned int dropped;
3238 struct sk_buff_head input_pkt_queue;
3239 struct napi_struct backlog;
3243 static inline void input_queue_head_incr(struct softnet_data *sd)
3246 sd->input_queue_head++;
3250 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3251 unsigned int *qtail)
3254 *qtail = ++sd->input_queue_tail;
3258 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3260 static inline int dev_recursion_level(void)
3262 return this_cpu_read(softnet_data.xmit.recursion);
3265 #define XMIT_RECURSION_LIMIT 8
3266 static inline bool dev_xmit_recursion(void)
3268 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3269 XMIT_RECURSION_LIMIT);
3272 static inline void dev_xmit_recursion_inc(void)
3274 __this_cpu_inc(softnet_data.xmit.recursion);
3277 static inline void dev_xmit_recursion_dec(void)
3279 __this_cpu_dec(softnet_data.xmit.recursion);
3282 void __netif_schedule(struct Qdisc *q);
3283 void netif_schedule_queue(struct netdev_queue *txq);
3285 static inline void netif_tx_schedule_all(struct net_device *dev)
3289 for (i = 0; i < dev->num_tx_queues; i++)
3290 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3293 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3295 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3299 * netif_start_queue - allow transmit
3300 * @dev: network device
3302 * Allow upper layers to call the device hard_start_xmit routine.
3304 static inline void netif_start_queue(struct net_device *dev)
3306 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3309 static inline void netif_tx_start_all_queues(struct net_device *dev)
3313 for (i = 0; i < dev->num_tx_queues; i++) {
3314 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3315 netif_tx_start_queue(txq);
3319 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3322 * netif_wake_queue - restart transmit
3323 * @dev: network device
3325 * Allow upper layers to call the device hard_start_xmit routine.
3326 * Used for flow control when transmit resources are available.
3328 static inline void netif_wake_queue(struct net_device *dev)
3330 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3333 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3337 for (i = 0; i < dev->num_tx_queues; i++) {
3338 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3339 netif_tx_wake_queue(txq);
3343 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3345 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3349 * netif_stop_queue - stop transmitted packets
3350 * @dev: network device
3352 * Stop upper layers calling the device hard_start_xmit routine.
3353 * Used for flow control when transmit resources are unavailable.
3355 static inline void netif_stop_queue(struct net_device *dev)
3357 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3360 void netif_tx_stop_all_queues(struct net_device *dev);
3362 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3364 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3368 * netif_queue_stopped - test if transmit queue is flowblocked
3369 * @dev: network device
3371 * Test if transmit queue on device is currently unable to send.
3373 static inline bool netif_queue_stopped(const struct net_device *dev)
3375 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3378 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3380 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3384 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3386 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3390 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3392 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3396 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3397 * @dev_queue: pointer to transmit queue
3399 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3400 * to give appropriate hint to the CPU.
3402 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3405 prefetchw(&dev_queue->dql.num_queued);
3410 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3411 * @dev_queue: pointer to transmit queue
3413 * BQL enabled drivers might use this helper in their TX completion path,
3414 * to give appropriate hint to the CPU.
3416 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3419 prefetchw(&dev_queue->dql.limit);
3423 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3427 dql_queued(&dev_queue->dql, bytes);
3429 if (likely(dql_avail(&dev_queue->dql) >= 0))
3432 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3435 * The XOFF flag must be set before checking the dql_avail below,
3436 * because in netdev_tx_completed_queue we update the dql_completed
3437 * before checking the XOFF flag.
3441 /* check again in case another CPU has just made room avail */
3442 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3443 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3447 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3448 * that they should not test BQL status themselves.
3449 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3451 * Returns true if the doorbell must be used to kick the NIC.
3453 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3459 dql_queued(&dev_queue->dql, bytes);
3461 return netif_tx_queue_stopped(dev_queue);
3463 netdev_tx_sent_queue(dev_queue, bytes);
3468 * netdev_sent_queue - report the number of bytes queued to hardware
3469 * @dev: network device
3470 * @bytes: number of bytes queued to the hardware device queue
3472 * Report the number of bytes queued for sending/completion to the network
3473 * device hardware queue. @bytes should be a good approximation and should
3474 * exactly match netdev_completed_queue() @bytes
3476 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3478 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3481 static inline bool __netdev_sent_queue(struct net_device *dev,
3485 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3489 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3490 unsigned int pkts, unsigned int bytes)
3493 if (unlikely(!bytes))
3496 dql_completed(&dev_queue->dql, bytes);
3499 * Without the memory barrier there is a small possiblity that
3500 * netdev_tx_sent_queue will miss the update and cause the queue to
3501 * be stopped forever
3505 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3508 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3509 netif_schedule_queue(dev_queue);
3514 * netdev_completed_queue - report bytes and packets completed by device
3515 * @dev: network device
3516 * @pkts: actual number of packets sent over the medium
3517 * @bytes: actual number of bytes sent over the medium
3519 * Report the number of bytes and packets transmitted by the network device
3520 * hardware queue over the physical medium, @bytes must exactly match the
3521 * @bytes amount passed to netdev_sent_queue()
3523 static inline void netdev_completed_queue(struct net_device *dev,
3524 unsigned int pkts, unsigned int bytes)
3526 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3529 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3532 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3538 * netdev_reset_queue - reset the packets and bytes count of a network device
3539 * @dev_queue: network device
3541 * Reset the bytes and packet count of a network device and clear the
3542 * software flow control OFF bit for this network device
3544 static inline void netdev_reset_queue(struct net_device *dev_queue)
3546 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3550 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3551 * @dev: network device
3552 * @queue_index: given tx queue index
3554 * Returns 0 if given tx queue index >= number of device tx queues,
3555 * otherwise returns the originally passed tx queue index.
3557 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3559 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3560 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3561 dev->name, queue_index,
3562 dev->real_num_tx_queues);
3570 * netif_running - test if up
3571 * @dev: network device
3573 * Test if the device has been brought up.
3575 static inline bool netif_running(const struct net_device *dev)
3577 return test_bit(__LINK_STATE_START, &dev->state);
3581 * Routines to manage the subqueues on a device. We only need start,
3582 * stop, and a check if it's stopped. All other device management is
3583 * done at the overall netdevice level.
3584 * Also test the device if we're multiqueue.
3588 * netif_start_subqueue - allow sending packets on subqueue
3589 * @dev: network device
3590 * @queue_index: sub queue index
3592 * Start individual transmit queue of a device with multiple transmit queues.
3594 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3596 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3598 netif_tx_start_queue(txq);
3602 * netif_stop_subqueue - stop sending packets on subqueue
3603 * @dev: network device
3604 * @queue_index: sub queue index
3606 * Stop individual transmit queue of a device with multiple transmit queues.
3608 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3610 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3611 netif_tx_stop_queue(txq);
3615 * __netif_subqueue_stopped - test status of subqueue
3616 * @dev: network device
3617 * @queue_index: sub queue index
3619 * Check individual transmit queue of a device with multiple transmit queues.
3621 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3624 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3626 return netif_tx_queue_stopped(txq);
3630 * netif_subqueue_stopped - test status of subqueue
3631 * @dev: network device
3632 * @skb: sub queue buffer pointer
3634 * Check individual transmit queue of a device with multiple transmit queues.
3636 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3637 struct sk_buff *skb)
3639 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3643 * netif_wake_subqueue - allow sending packets on subqueue
3644 * @dev: network device
3645 * @queue_index: sub queue index
3647 * Resume individual transmit queue of a device with multiple transmit queues.
3649 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3651 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3653 netif_tx_wake_queue(txq);
3657 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3659 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3660 u16 index, bool is_rxqs_map);
3663 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3664 * @j: CPU/Rx queue index
3665 * @mask: bitmask of all cpus/rx queues
3666 * @nr_bits: number of bits in the bitmask
3668 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3670 static inline bool netif_attr_test_mask(unsigned long j,
3671 const unsigned long *mask,
3672 unsigned int nr_bits)
3674 cpu_max_bits_warn(j, nr_bits);
3675 return test_bit(j, mask);
3679 * netif_attr_test_online - Test for online CPU/Rx queue
3680 * @j: CPU/Rx queue index
3681 * @online_mask: bitmask for CPUs/Rx queues that are online
3682 * @nr_bits: number of bits in the bitmask
3684 * Returns true if a CPU/Rx queue is online.
3686 static inline bool netif_attr_test_online(unsigned long j,
3687 const unsigned long *online_mask,
3688 unsigned int nr_bits)
3690 cpu_max_bits_warn(j, nr_bits);
3693 return test_bit(j, online_mask);
3695 return (j < nr_bits);
3699 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3700 * @n: CPU/Rx queue index
3701 * @srcp: the cpumask/Rx queue mask pointer
3702 * @nr_bits: number of bits in the bitmask
3704 * Returns >= nr_bits if no further CPUs/Rx queues set.
3706 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3707 unsigned int nr_bits)
3709 /* -1 is a legal arg here. */
3711 cpu_max_bits_warn(n, nr_bits);
3714 return find_next_bit(srcp, nr_bits, n + 1);
3720 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3721 * @n: CPU/Rx queue index
3722 * @src1p: the first CPUs/Rx queues mask pointer
3723 * @src2p: the second CPUs/Rx queues mask pointer
3724 * @nr_bits: number of bits in the bitmask
3726 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3728 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3729 const unsigned long *src2p,
3730 unsigned int nr_bits)
3732 /* -1 is a legal arg here. */
3734 cpu_max_bits_warn(n, nr_bits);
3737 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3739 return find_next_bit(src1p, nr_bits, n + 1);
3741 return find_next_bit(src2p, nr_bits, n + 1);
3746 static inline int netif_set_xps_queue(struct net_device *dev,
3747 const struct cpumask *mask,
3753 static inline int __netif_set_xps_queue(struct net_device *dev,
3754 const unsigned long *mask,
3755 u16 index, bool is_rxqs_map)
3762 * netif_is_multiqueue - test if device has multiple transmit queues
3763 * @dev: network device
3765 * Check if device has multiple transmit queues
3767 static inline bool netif_is_multiqueue(const struct net_device *dev)
3769 return dev->num_tx_queues > 1;
3772 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3775 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3777 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3780 dev->real_num_rx_queues = rxqs;
3785 static inline struct netdev_rx_queue *
3786 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3788 return dev->_rx + rxq;
3792 static inline unsigned int get_netdev_rx_queue_index(
3793 struct netdev_rx_queue *queue)
3795 struct net_device *dev = queue->dev;
3796 int index = queue - dev->_rx;
3798 BUG_ON(index >= dev->num_rx_queues);
3803 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3804 int netif_get_num_default_rss_queues(void);
3806 enum skb_free_reason {
3807 SKB_REASON_CONSUMED,
3811 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3812 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3815 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3816 * interrupt context or with hardware interrupts being disabled.
3817 * (in_irq() || irqs_disabled())
3819 * We provide four helpers that can be used in following contexts :
3821 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3822 * replacing kfree_skb(skb)
3824 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3825 * Typically used in place of consume_skb(skb) in TX completion path
3827 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3828 * replacing kfree_skb(skb)
3830 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3831 * and consumed a packet. Used in place of consume_skb(skb)
3833 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3835 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3838 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3840 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3843 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3845 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3848 static inline void dev_consume_skb_any(struct sk_buff *skb)
3850 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3853 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3854 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3855 int netif_rx(struct sk_buff *skb);
3856 int netif_rx_ni(struct sk_buff *skb);
3857 int netif_rx_any_context(struct sk_buff *skb);
3858 int netif_receive_skb(struct sk_buff *skb);
3859 int netif_receive_skb_core(struct sk_buff *skb);
3860 void netif_receive_skb_list(struct list_head *head);
3861 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3862 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3863 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3864 gro_result_t napi_gro_frags(struct napi_struct *napi);
3865 struct packet_offload *gro_find_receive_by_type(__be16 type);
3866 struct packet_offload *gro_find_complete_by_type(__be16 type);
3868 static inline void napi_free_frags(struct napi_struct *napi)
3870 kfree_skb(napi->skb);
3874 bool netdev_is_rx_handler_busy(struct net_device *dev);
3875 int netdev_rx_handler_register(struct net_device *dev,
3876 rx_handler_func_t *rx_handler,
3877 void *rx_handler_data);
3878 void netdev_rx_handler_unregister(struct net_device *dev);
3880 bool dev_valid_name(const char *name);
3881 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3882 bool *need_copyout);
3883 int dev_ifconf(struct net *net, struct ifconf *, int);
3884 int dev_ethtool(struct net *net, struct ifreq *);
3885 unsigned int dev_get_flags(const struct net_device *);
3886 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3887 struct netlink_ext_ack *extack);
3888 int dev_change_flags(struct net_device *dev, unsigned int flags,
3889 struct netlink_ext_ack *extack);
3890 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3891 unsigned int gchanges);
3892 int dev_change_name(struct net_device *, const char *);
3893 int dev_set_alias(struct net_device *, const char *, size_t);
3894 int dev_get_alias(const struct net_device *, char *, size_t);
3895 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3896 int __dev_set_mtu(struct net_device *, int);
3897 int dev_validate_mtu(struct net_device *dev, int mtu,
3898 struct netlink_ext_ack *extack);
3899 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3900 struct netlink_ext_ack *extack);
3901 int dev_set_mtu(struct net_device *, int);
3902 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3903 void dev_set_group(struct net_device *, int);
3904 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3905 struct netlink_ext_ack *extack);
3906 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3907 struct netlink_ext_ack *extack);
3908 int dev_change_carrier(struct net_device *, bool new_carrier);
3909 int dev_get_phys_port_id(struct net_device *dev,
3910 struct netdev_phys_item_id *ppid);
3911 int dev_get_phys_port_name(struct net_device *dev,
3912 char *name, size_t len);
3913 int dev_get_port_parent_id(struct net_device *dev,
3914 struct netdev_phys_item_id *ppid, bool recurse);
3915 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3916 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3917 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3918 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3920 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3921 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3922 struct netdev_queue *txq, int *ret);
3924 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3925 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3926 int fd, int expected_fd, u32 flags);
3927 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3928 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3930 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3932 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3933 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3934 bool is_skb_forwardable(const struct net_device *dev,
3935 const struct sk_buff *skb);
3937 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3938 struct sk_buff *skb)
3940 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3941 unlikely(!is_skb_forwardable(dev, skb))) {
3942 atomic_long_inc(&dev->rx_dropped);
3947 skb_scrub_packet(skb, true);
3952 bool dev_nit_active(struct net_device *dev);
3953 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3955 extern int netdev_budget;
3956 extern unsigned int netdev_budget_usecs;
3958 /* Called by rtnetlink.c:rtnl_unlock() */
3959 void netdev_run_todo(void);
3962 * dev_put - release reference to device
3963 * @dev: network device
3965 * Release reference to device to allow it to be freed.
3967 static inline void dev_put(struct net_device *dev)
3969 this_cpu_dec(*dev->pcpu_refcnt);
3973 * dev_hold - get reference to device
3974 * @dev: network device
3976 * Hold reference to device to keep it from being freed.
3978 static inline void dev_hold(struct net_device *dev)
3980 this_cpu_inc(*dev->pcpu_refcnt);
3983 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3984 * and _off may be called from IRQ context, but it is caller
3985 * who is responsible for serialization of these calls.
3987 * The name carrier is inappropriate, these functions should really be
3988 * called netif_lowerlayer_*() because they represent the state of any
3989 * kind of lower layer not just hardware media.
3992 void linkwatch_init_dev(struct net_device *dev);
3993 void linkwatch_fire_event(struct net_device *dev);
3994 void linkwatch_forget_dev(struct net_device *dev);
3997 * netif_carrier_ok - test if carrier present
3998 * @dev: network device
4000 * Check if carrier is present on device
4002 static inline bool netif_carrier_ok(const struct net_device *dev)
4004 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4007 unsigned long dev_trans_start(struct net_device *dev);
4009 void __netdev_watchdog_up(struct net_device *dev);
4011 void netif_carrier_on(struct net_device *dev);
4013 void netif_carrier_off(struct net_device *dev);
4016 * netif_dormant_on - mark device as dormant.
4017 * @dev: network device
4019 * Mark device as dormant (as per RFC2863).
4021 * The dormant state indicates that the relevant interface is not
4022 * actually in a condition to pass packets (i.e., it is not 'up') but is
4023 * in a "pending" state, waiting for some external event. For "on-
4024 * demand" interfaces, this new state identifies the situation where the
4025 * interface is waiting for events to place it in the up state.
4027 static inline void netif_dormant_on(struct net_device *dev)
4029 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4030 linkwatch_fire_event(dev);
4034 * netif_dormant_off - set device as not dormant.
4035 * @dev: network device
4037 * Device is not in dormant state.
4039 static inline void netif_dormant_off(struct net_device *dev)
4041 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4042 linkwatch_fire_event(dev);
4046 * netif_dormant - test if device is dormant
4047 * @dev: network device
4049 * Check if device is dormant.
4051 static inline bool netif_dormant(const struct net_device *dev)
4053 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4058 * netif_testing_on - mark device as under test.
4059 * @dev: network device
4061 * Mark device as under test (as per RFC2863).
4063 * The testing state indicates that some test(s) must be performed on
4064 * the interface. After completion, of the test, the interface state
4065 * will change to up, dormant, or down, as appropriate.
4067 static inline void netif_testing_on(struct net_device *dev)
4069 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4070 linkwatch_fire_event(dev);
4074 * netif_testing_off - set device as not under test.
4075 * @dev: network device
4077 * Device is not in testing state.
4079 static inline void netif_testing_off(struct net_device *dev)
4081 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4082 linkwatch_fire_event(dev);
4086 * netif_testing - test if device is under test
4087 * @dev: network device
4089 * Check if device is under test
4091 static inline bool netif_testing(const struct net_device *dev)
4093 return test_bit(__LINK_STATE_TESTING, &dev->state);
4098 * netif_oper_up - test if device is operational
4099 * @dev: network device
4101 * Check if carrier is operational
4103 static inline bool netif_oper_up(const struct net_device *dev)
4105 return (dev->operstate == IF_OPER_UP ||
4106 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4110 * netif_device_present - is device available or removed
4111 * @dev: network device
4113 * Check if device has not been removed from system.
4115 static inline bool netif_device_present(struct net_device *dev)
4117 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4120 void netif_device_detach(struct net_device *dev);
4122 void netif_device_attach(struct net_device *dev);
4125 * Network interface message level settings
4130 NETIF_MSG_PROBE_BIT,
4132 NETIF_MSG_TIMER_BIT,
4133 NETIF_MSG_IFDOWN_BIT,
4135 NETIF_MSG_RX_ERR_BIT,
4136 NETIF_MSG_TX_ERR_BIT,
4137 NETIF_MSG_TX_QUEUED_BIT,
4139 NETIF_MSG_TX_DONE_BIT,
4140 NETIF_MSG_RX_STATUS_BIT,
4141 NETIF_MSG_PKTDATA_BIT,
4145 /* When you add a new bit above, update netif_msg_class_names array
4146 * in net/ethtool/common.c
4148 NETIF_MSG_CLASS_COUNT,
4150 /* Both ethtool_ops interface and internal driver implementation use u32 */
4151 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4153 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4154 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4156 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4157 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4158 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4159 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4160 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4161 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4162 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4163 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4164 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4165 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4166 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4167 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4168 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4169 #define NETIF_MSG_HW __NETIF_MSG(HW)
4170 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4172 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4173 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4174 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4175 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4176 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4177 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4178 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4179 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4180 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4181 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4182 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4183 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4184 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4185 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4186 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4188 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4191 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4192 return default_msg_enable_bits;
4193 if (debug_value == 0) /* no output */
4195 /* set low N bits */
4196 return (1U << debug_value) - 1;
4199 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4201 spin_lock(&txq->_xmit_lock);
4202 txq->xmit_lock_owner = cpu;
4205 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4207 __acquire(&txq->_xmit_lock);
4211 static inline void __netif_tx_release(struct netdev_queue *txq)
4213 __release(&txq->_xmit_lock);
4216 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4218 spin_lock_bh(&txq->_xmit_lock);
4219 txq->xmit_lock_owner = smp_processor_id();
4222 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4224 bool ok = spin_trylock(&txq->_xmit_lock);
4226 txq->xmit_lock_owner = smp_processor_id();
4230 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4232 txq->xmit_lock_owner = -1;
4233 spin_unlock(&txq->_xmit_lock);
4236 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4238 txq->xmit_lock_owner = -1;
4239 spin_unlock_bh(&txq->_xmit_lock);
4242 static inline void txq_trans_update(struct netdev_queue *txq)
4244 if (txq->xmit_lock_owner != -1)
4245 txq->trans_start = jiffies;
4248 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4249 static inline void netif_trans_update(struct net_device *dev)
4251 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4253 if (txq->trans_start != jiffies)
4254 txq->trans_start = jiffies;
4258 * netif_tx_lock - grab network device transmit lock
4259 * @dev: network device
4261 * Get network device transmit lock
4263 static inline void netif_tx_lock(struct net_device *dev)
4268 spin_lock(&dev->tx_global_lock);
4269 cpu = smp_processor_id();
4270 for (i = 0; i < dev->num_tx_queues; i++) {
4271 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4273 /* We are the only thread of execution doing a
4274 * freeze, but we have to grab the _xmit_lock in
4275 * order to synchronize with threads which are in
4276 * the ->hard_start_xmit() handler and already
4277 * checked the frozen bit.
4279 __netif_tx_lock(txq, cpu);
4280 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4281 __netif_tx_unlock(txq);
4285 static inline void netif_tx_lock_bh(struct net_device *dev)
4291 static inline void netif_tx_unlock(struct net_device *dev)
4295 for (i = 0; i < dev->num_tx_queues; i++) {
4296 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4298 /* No need to grab the _xmit_lock here. If the
4299 * queue is not stopped for another reason, we
4302 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4303 netif_schedule_queue(txq);
4305 spin_unlock(&dev->tx_global_lock);
4308 static inline void netif_tx_unlock_bh(struct net_device *dev)
4310 netif_tx_unlock(dev);
4314 #define HARD_TX_LOCK(dev, txq, cpu) { \
4315 if ((dev->features & NETIF_F_LLTX) == 0) { \
4316 __netif_tx_lock(txq, cpu); \
4318 __netif_tx_acquire(txq); \
4322 #define HARD_TX_TRYLOCK(dev, txq) \
4323 (((dev->features & NETIF_F_LLTX) == 0) ? \
4324 __netif_tx_trylock(txq) : \
4325 __netif_tx_acquire(txq))
4327 #define HARD_TX_UNLOCK(dev, txq) { \
4328 if ((dev->features & NETIF_F_LLTX) == 0) { \
4329 __netif_tx_unlock(txq); \
4331 __netif_tx_release(txq); \
4335 static inline void netif_tx_disable(struct net_device *dev)
4341 cpu = smp_processor_id();
4342 for (i = 0; i < dev->num_tx_queues; i++) {
4343 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4345 __netif_tx_lock(txq, cpu);
4346 netif_tx_stop_queue(txq);
4347 __netif_tx_unlock(txq);
4352 static inline void netif_addr_lock(struct net_device *dev)
4354 unsigned char nest_level = 0;
4356 #ifdef CONFIG_LOCKDEP
4357 nest_level = dev->nested_level;
4359 spin_lock_nested(&dev->addr_list_lock, nest_level);
4362 static inline void netif_addr_lock_bh(struct net_device *dev)
4364 unsigned char nest_level = 0;
4366 #ifdef CONFIG_LOCKDEP
4367 nest_level = dev->nested_level;
4370 spin_lock_nested(&dev->addr_list_lock, nest_level);
4373 static inline void netif_addr_unlock(struct net_device *dev)
4375 spin_unlock(&dev->addr_list_lock);
4378 static inline void netif_addr_unlock_bh(struct net_device *dev)
4380 spin_unlock_bh(&dev->addr_list_lock);
4384 * dev_addrs walker. Should be used only for read access. Call with
4385 * rcu_read_lock held.
4387 #define for_each_dev_addr(dev, ha) \
4388 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4390 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4392 void ether_setup(struct net_device *dev);
4394 /* Support for loadable net-drivers */
4395 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4396 unsigned char name_assign_type,
4397 void (*setup)(struct net_device *),
4398 unsigned int txqs, unsigned int rxqs);
4399 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4400 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4402 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4403 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4406 int register_netdev(struct net_device *dev);
4407 void unregister_netdev(struct net_device *dev);
4409 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4411 /* General hardware address lists handling functions */
4412 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4413 struct netdev_hw_addr_list *from_list, int addr_len);
4414 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4415 struct netdev_hw_addr_list *from_list, int addr_len);
4416 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4417 struct net_device *dev,
4418 int (*sync)(struct net_device *, const unsigned char *),
4419 int (*unsync)(struct net_device *,
4420 const unsigned char *));
4421 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4422 struct net_device *dev,
4423 int (*sync)(struct net_device *,
4424 const unsigned char *, int),
4425 int (*unsync)(struct net_device *,
4426 const unsigned char *, int));
4427 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4428 struct net_device *dev,
4429 int (*unsync)(struct net_device *,
4430 const unsigned char *, int));
4431 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4432 struct net_device *dev,
4433 int (*unsync)(struct net_device *,
4434 const unsigned char *));
4435 void __hw_addr_init(struct netdev_hw_addr_list *list);
4437 /* Functions used for device addresses handling */
4438 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4439 unsigned char addr_type);
4440 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4441 unsigned char addr_type);
4442 void dev_addr_flush(struct net_device *dev);
4443 int dev_addr_init(struct net_device *dev);
4445 /* Functions used for unicast addresses handling */
4446 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4447 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4448 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4449 int dev_uc_sync(struct net_device *to, struct net_device *from);
4450 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4451 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4452 void dev_uc_flush(struct net_device *dev);
4453 void dev_uc_init(struct net_device *dev);
4456 * __dev_uc_sync - Synchonize device's unicast list
4457 * @dev: device to sync
4458 * @sync: function to call if address should be added
4459 * @unsync: function to call if address should be removed
4461 * Add newly added addresses to the interface, and release
4462 * addresses that have been deleted.
4464 static inline int __dev_uc_sync(struct net_device *dev,
4465 int (*sync)(struct net_device *,
4466 const unsigned char *),
4467 int (*unsync)(struct net_device *,
4468 const unsigned char *))
4470 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4474 * __dev_uc_unsync - Remove synchronized addresses from device
4475 * @dev: device to sync
4476 * @unsync: function to call if address should be removed
4478 * Remove all addresses that were added to the device by dev_uc_sync().
4480 static inline void __dev_uc_unsync(struct net_device *dev,
4481 int (*unsync)(struct net_device *,
4482 const unsigned char *))
4484 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4487 /* Functions used for multicast addresses handling */
4488 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4489 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4490 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4491 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4492 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4493 int dev_mc_sync(struct net_device *to, struct net_device *from);
4494 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4495 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4496 void dev_mc_flush(struct net_device *dev);
4497 void dev_mc_init(struct net_device *dev);
4500 * __dev_mc_sync - Synchonize device's multicast list
4501 * @dev: device to sync
4502 * @sync: function to call if address should be added
4503 * @unsync: function to call if address should be removed
4505 * Add newly added addresses to the interface, and release
4506 * addresses that have been deleted.
4508 static inline int __dev_mc_sync(struct net_device *dev,
4509 int (*sync)(struct net_device *,
4510 const unsigned char *),
4511 int (*unsync)(struct net_device *,
4512 const unsigned char *))
4514 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4518 * __dev_mc_unsync - Remove synchronized addresses from device
4519 * @dev: device to sync
4520 * @unsync: function to call if address should be removed
4522 * Remove all addresses that were added to the device by dev_mc_sync().
4524 static inline void __dev_mc_unsync(struct net_device *dev,
4525 int (*unsync)(struct net_device *,
4526 const unsigned char *))
4528 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4531 /* Functions used for secondary unicast and multicast support */
4532 void dev_set_rx_mode(struct net_device *dev);
4533 void __dev_set_rx_mode(struct net_device *dev);
4534 int dev_set_promiscuity(struct net_device *dev, int inc);
4535 int dev_set_allmulti(struct net_device *dev, int inc);
4536 void netdev_state_change(struct net_device *dev);
4537 void __netdev_notify_peers(struct net_device *dev);
4538 void netdev_notify_peers(struct net_device *dev);
4539 void netdev_features_change(struct net_device *dev);
4540 /* Load a device via the kmod */
4541 void dev_load(struct net *net, const char *name);
4542 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4543 struct rtnl_link_stats64 *storage);
4544 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4545 const struct net_device_stats *netdev_stats);
4546 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4547 const struct pcpu_sw_netstats __percpu *netstats);
4548 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4550 extern int netdev_max_backlog;
4551 extern int netdev_tstamp_prequeue;
4552 extern int weight_p;
4553 extern int dev_weight_rx_bias;
4554 extern int dev_weight_tx_bias;
4555 extern int dev_rx_weight;
4556 extern int dev_tx_weight;
4557 extern int gro_normal_batch;
4560 NESTED_SYNC_IMM_BIT,
4561 NESTED_SYNC_TODO_BIT,
4564 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4565 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4567 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4568 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4570 struct netdev_nested_priv {
4571 unsigned char flags;
4575 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4576 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4577 struct list_head **iter);
4578 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4579 struct list_head **iter);
4581 #ifdef CONFIG_LOCKDEP
4582 static LIST_HEAD(net_unlink_list);
4584 static inline void net_unlink_todo(struct net_device *dev)
4586 if (list_empty(&dev->unlink_list))
4587 list_add_tail(&dev->unlink_list, &net_unlink_list);
4591 /* iterate through upper list, must be called under RCU read lock */
4592 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4593 for (iter = &(dev)->adj_list.upper, \
4594 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4596 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4598 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4599 int (*fn)(struct net_device *upper_dev,
4600 struct netdev_nested_priv *priv),
4601 struct netdev_nested_priv *priv);
4603 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4604 struct net_device *upper_dev);
4606 bool netdev_has_any_upper_dev(struct net_device *dev);
4608 void *netdev_lower_get_next_private(struct net_device *dev,
4609 struct list_head **iter);
4610 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4611 struct list_head **iter);
4613 #define netdev_for_each_lower_private(dev, priv, iter) \
4614 for (iter = (dev)->adj_list.lower.next, \
4615 priv = netdev_lower_get_next_private(dev, &(iter)); \
4617 priv = netdev_lower_get_next_private(dev, &(iter)))
4619 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4620 for (iter = &(dev)->adj_list.lower, \
4621 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4623 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4625 void *netdev_lower_get_next(struct net_device *dev,
4626 struct list_head **iter);
4628 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4629 for (iter = (dev)->adj_list.lower.next, \
4630 ldev = netdev_lower_get_next(dev, &(iter)); \
4632 ldev = netdev_lower_get_next(dev, &(iter)))
4634 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4635 struct list_head **iter);
4636 int netdev_walk_all_lower_dev(struct net_device *dev,
4637 int (*fn)(struct net_device *lower_dev,
4638 struct netdev_nested_priv *priv),
4639 struct netdev_nested_priv *priv);
4640 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4641 int (*fn)(struct net_device *lower_dev,
4642 struct netdev_nested_priv *priv),
4643 struct netdev_nested_priv *priv);
4645 void *netdev_adjacent_get_private(struct list_head *adj_list);
4646 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4647 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4648 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4649 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4650 struct netlink_ext_ack *extack);
4651 int netdev_master_upper_dev_link(struct net_device *dev,
4652 struct net_device *upper_dev,
4653 void *upper_priv, void *upper_info,
4654 struct netlink_ext_ack *extack);
4655 void netdev_upper_dev_unlink(struct net_device *dev,
4656 struct net_device *upper_dev);
4657 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4658 struct net_device *new_dev,
4659 struct net_device *dev,
4660 struct netlink_ext_ack *extack);
4661 void netdev_adjacent_change_commit(struct net_device *old_dev,
4662 struct net_device *new_dev,
4663 struct net_device *dev);
4664 void netdev_adjacent_change_abort(struct net_device *old_dev,
4665 struct net_device *new_dev,
4666 struct net_device *dev);
4667 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4668 void *netdev_lower_dev_get_private(struct net_device *dev,
4669 struct net_device *lower_dev);
4670 void netdev_lower_state_changed(struct net_device *lower_dev,
4671 void *lower_state_info);
4673 /* RSS keys are 40 or 52 bytes long */
4674 #define NETDEV_RSS_KEY_LEN 52
4675 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4676 void netdev_rss_key_fill(void *buffer, size_t len);
4678 int skb_checksum_help(struct sk_buff *skb);
4679 int skb_crc32c_csum_help(struct sk_buff *skb);
4680 int skb_csum_hwoffload_help(struct sk_buff *skb,
4681 const netdev_features_t features);
4683 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4684 netdev_features_t features, bool tx_path);
4685 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4686 netdev_features_t features);
4688 struct netdev_bonding_info {
4693 struct netdev_notifier_bonding_info {
4694 struct netdev_notifier_info info; /* must be first */
4695 struct netdev_bonding_info bonding_info;
4698 void netdev_bonding_info_change(struct net_device *dev,
4699 struct netdev_bonding_info *bonding_info);
4701 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4702 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4704 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4711 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4713 return __skb_gso_segment(skb, features, true);
4715 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4717 static inline bool can_checksum_protocol(netdev_features_t features,
4720 if (protocol == htons(ETH_P_FCOE))
4721 return !!(features & NETIF_F_FCOE_CRC);
4723 /* Assume this is an IP checksum (not SCTP CRC) */
4725 if (features & NETIF_F_HW_CSUM) {
4726 /* Can checksum everything */
4731 case htons(ETH_P_IP):
4732 return !!(features & NETIF_F_IP_CSUM);
4733 case htons(ETH_P_IPV6):
4734 return !!(features & NETIF_F_IPV6_CSUM);
4741 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4743 static inline void netdev_rx_csum_fault(struct net_device *dev,
4744 struct sk_buff *skb)
4748 /* rx skb timestamps */
4749 void net_enable_timestamp(void);
4750 void net_disable_timestamp(void);
4752 #ifdef CONFIG_PROC_FS
4753 int __init dev_proc_init(void);
4755 #define dev_proc_init() 0
4758 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4759 struct sk_buff *skb, struct net_device *dev,
4762 __this_cpu_write(softnet_data.xmit.more, more);
4763 return ops->ndo_start_xmit(skb, dev);
4766 static inline bool netdev_xmit_more(void)
4768 return __this_cpu_read(softnet_data.xmit.more);
4771 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4772 struct netdev_queue *txq, bool more)
4774 const struct net_device_ops *ops = dev->netdev_ops;
4777 rc = __netdev_start_xmit(ops, skb, dev, more);
4778 if (rc == NETDEV_TX_OK)
4779 txq_trans_update(txq);
4784 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4786 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4789 extern const struct kobj_ns_type_operations net_ns_type_operations;
4791 const char *netdev_drivername(const struct net_device *dev);
4793 void linkwatch_run_queue(void);
4795 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4796 netdev_features_t f2)
4798 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4799 if (f1 & NETIF_F_HW_CSUM)
4800 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4802 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4808 static inline netdev_features_t netdev_get_wanted_features(
4809 struct net_device *dev)
4811 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4813 netdev_features_t netdev_increment_features(netdev_features_t all,
4814 netdev_features_t one, netdev_features_t mask);
4816 /* Allow TSO being used on stacked device :
4817 * Performing the GSO segmentation before last device
4818 * is a performance improvement.
4820 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4821 netdev_features_t mask)
4823 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4826 int __netdev_update_features(struct net_device *dev);
4827 void netdev_update_features(struct net_device *dev);
4828 void netdev_change_features(struct net_device *dev);
4830 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4831 struct net_device *dev);
4833 netdev_features_t passthru_features_check(struct sk_buff *skb,
4834 struct net_device *dev,
4835 netdev_features_t features);
4836 netdev_features_t netif_skb_features(struct sk_buff *skb);
4838 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4840 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4842 /* check flags correspondence */
4843 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4844 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4845 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4846 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4847 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4848 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4849 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4850 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4851 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4852 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4853 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4854 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4855 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4856 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4857 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4858 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4859 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4860 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4861 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4863 return (features & feature) == feature;
4866 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4868 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4869 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4872 static inline bool netif_needs_gso(struct sk_buff *skb,
4873 netdev_features_t features)
4875 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4876 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4877 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4880 static inline void netif_set_gso_max_size(struct net_device *dev,
4883 dev->gso_max_size = size;
4886 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4887 int pulled_hlen, u16 mac_offset,
4890 skb->protocol = protocol;
4891 skb->encapsulation = 1;
4892 skb_push(skb, pulled_hlen);
4893 skb_reset_transport_header(skb);
4894 skb->mac_header = mac_offset;
4895 skb->network_header = skb->mac_header + mac_len;
4896 skb->mac_len = mac_len;
4899 static inline bool netif_is_macsec(const struct net_device *dev)
4901 return dev->priv_flags & IFF_MACSEC;
4904 static inline bool netif_is_macvlan(const struct net_device *dev)
4906 return dev->priv_flags & IFF_MACVLAN;
4909 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4911 return dev->priv_flags & IFF_MACVLAN_PORT;
4914 static inline bool netif_is_bond_master(const struct net_device *dev)
4916 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4919 static inline bool netif_is_bond_slave(const struct net_device *dev)
4921 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4924 static inline bool netif_supports_nofcs(struct net_device *dev)
4926 return dev->priv_flags & IFF_SUPP_NOFCS;
4929 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4931 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4934 static inline bool netif_is_l3_master(const struct net_device *dev)
4936 return dev->priv_flags & IFF_L3MDEV_MASTER;
4939 static inline bool netif_is_l3_slave(const struct net_device *dev)
4941 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4944 static inline bool netif_is_bridge_master(const struct net_device *dev)
4946 return dev->priv_flags & IFF_EBRIDGE;
4949 static inline bool netif_is_bridge_port(const struct net_device *dev)
4951 return dev->priv_flags & IFF_BRIDGE_PORT;
4954 static inline bool netif_is_ovs_master(const struct net_device *dev)
4956 return dev->priv_flags & IFF_OPENVSWITCH;
4959 static inline bool netif_is_ovs_port(const struct net_device *dev)
4961 return dev->priv_flags & IFF_OVS_DATAPATH;
4964 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4966 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4969 static inline bool netif_is_team_master(const struct net_device *dev)
4971 return dev->priv_flags & IFF_TEAM;
4974 static inline bool netif_is_team_port(const struct net_device *dev)
4976 return dev->priv_flags & IFF_TEAM_PORT;
4979 static inline bool netif_is_lag_master(const struct net_device *dev)
4981 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4984 static inline bool netif_is_lag_port(const struct net_device *dev)
4986 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4989 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4991 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4994 static inline bool netif_is_failover(const struct net_device *dev)
4996 return dev->priv_flags & IFF_FAILOVER;
4999 static inline bool netif_is_failover_slave(const struct net_device *dev)
5001 return dev->priv_flags & IFF_FAILOVER_SLAVE;
5004 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5005 static inline void netif_keep_dst(struct net_device *dev)
5007 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5010 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5011 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5013 /* TODO: reserve and use an additional IFF bit, if we get more users */
5014 return dev->priv_flags & IFF_MACSEC;
5017 extern struct pernet_operations __net_initdata loopback_net_ops;
5019 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5021 /* netdev_printk helpers, similar to dev_printk */
5023 static inline const char *netdev_name(const struct net_device *dev)
5025 if (!dev->name[0] || strchr(dev->name, '%'))
5026 return "(unnamed net_device)";
5030 static inline bool netdev_unregistering(const struct net_device *dev)
5032 return dev->reg_state == NETREG_UNREGISTERING;
5035 static inline const char *netdev_reg_state(const struct net_device *dev)
5037 switch (dev->reg_state) {
5038 case NETREG_UNINITIALIZED: return " (uninitialized)";
5039 case NETREG_REGISTERED: return "";
5040 case NETREG_UNREGISTERING: return " (unregistering)";
5041 case NETREG_UNREGISTERED: return " (unregistered)";
5042 case NETREG_RELEASED: return " (released)";
5043 case NETREG_DUMMY: return " (dummy)";
5046 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5047 return " (unknown)";
5050 __printf(3, 4) __cold
5051 void netdev_printk(const char *level, const struct net_device *dev,
5052 const char *format, ...);
5053 __printf(2, 3) __cold
5054 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5055 __printf(2, 3) __cold
5056 void netdev_alert(const struct net_device *dev, const char *format, ...);
5057 __printf(2, 3) __cold
5058 void netdev_crit(const struct net_device *dev, const char *format, ...);
5059 __printf(2, 3) __cold
5060 void netdev_err(const struct net_device *dev, const char *format, ...);
5061 __printf(2, 3) __cold
5062 void netdev_warn(const struct net_device *dev, const char *format, ...);
5063 __printf(2, 3) __cold
5064 void netdev_notice(const struct net_device *dev, const char *format, ...);
5065 __printf(2, 3) __cold
5066 void netdev_info(const struct net_device *dev, const char *format, ...);
5068 #define netdev_level_once(level, dev, fmt, ...) \
5070 static bool __print_once __read_mostly; \
5072 if (!__print_once) { \
5073 __print_once = true; \
5074 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5078 #define netdev_emerg_once(dev, fmt, ...) \
5079 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5080 #define netdev_alert_once(dev, fmt, ...) \
5081 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5082 #define netdev_crit_once(dev, fmt, ...) \
5083 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5084 #define netdev_err_once(dev, fmt, ...) \
5085 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5086 #define netdev_warn_once(dev, fmt, ...) \
5087 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5088 #define netdev_notice_once(dev, fmt, ...) \
5089 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5090 #define netdev_info_once(dev, fmt, ...) \
5091 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5093 #define MODULE_ALIAS_NETDEV(device) \
5094 MODULE_ALIAS("netdev-" device)
5096 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5097 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5098 #define netdev_dbg(__dev, format, args...) \
5100 dynamic_netdev_dbg(__dev, format, ##args); \
5102 #elif defined(DEBUG)
5103 #define netdev_dbg(__dev, format, args...) \
5104 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5106 #define netdev_dbg(__dev, format, args...) \
5109 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5113 #if defined(VERBOSE_DEBUG)
5114 #define netdev_vdbg netdev_dbg
5117 #define netdev_vdbg(dev, format, args...) \
5120 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5126 * netdev_WARN() acts like dev_printk(), but with the key difference
5127 * of using a WARN/WARN_ON to get the message out, including the
5128 * file/line information and a backtrace.
5130 #define netdev_WARN(dev, format, args...) \
5131 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5132 netdev_reg_state(dev), ##args)
5134 #define netdev_WARN_ONCE(dev, format, args...) \
5135 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5136 netdev_reg_state(dev), ##args)
5138 /* netif printk helpers, similar to netdev_printk */
5140 #define netif_printk(priv, type, level, dev, fmt, args...) \
5142 if (netif_msg_##type(priv)) \
5143 netdev_printk(level, (dev), fmt, ##args); \
5146 #define netif_level(level, priv, type, dev, fmt, args...) \
5148 if (netif_msg_##type(priv)) \
5149 netdev_##level(dev, fmt, ##args); \
5152 #define netif_emerg(priv, type, dev, fmt, args...) \
5153 netif_level(emerg, priv, type, dev, fmt, ##args)
5154 #define netif_alert(priv, type, dev, fmt, args...) \
5155 netif_level(alert, priv, type, dev, fmt, ##args)
5156 #define netif_crit(priv, type, dev, fmt, args...) \
5157 netif_level(crit, priv, type, dev, fmt, ##args)
5158 #define netif_err(priv, type, dev, fmt, args...) \
5159 netif_level(err, priv, type, dev, fmt, ##args)
5160 #define netif_warn(priv, type, dev, fmt, args...) \
5161 netif_level(warn, priv, type, dev, fmt, ##args)
5162 #define netif_notice(priv, type, dev, fmt, args...) \
5163 netif_level(notice, priv, type, dev, fmt, ##args)
5164 #define netif_info(priv, type, dev, fmt, args...) \
5165 netif_level(info, priv, type, dev, fmt, ##args)
5167 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5168 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5169 #define netif_dbg(priv, type, netdev, format, args...) \
5171 if (netif_msg_##type(priv)) \
5172 dynamic_netdev_dbg(netdev, format, ##args); \
5174 #elif defined(DEBUG)
5175 #define netif_dbg(priv, type, dev, format, args...) \
5176 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5178 #define netif_dbg(priv, type, dev, format, args...) \
5181 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5186 /* if @cond then downgrade to debug, else print at @level */
5187 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5190 netif_dbg(priv, type, netdev, fmt, ##args); \
5192 netif_ ## level(priv, type, netdev, fmt, ##args); \
5195 #if defined(VERBOSE_DEBUG)
5196 #define netif_vdbg netif_dbg
5198 #define netif_vdbg(priv, type, dev, format, args...) \
5201 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5207 * The list of packet types we will receive (as opposed to discard)
5208 * and the routines to invoke.
5210 * Why 16. Because with 16 the only overlap we get on a hash of the
5211 * low nibble of the protocol value is RARP/SNAP/X.25.
5225 #define PTYPE_HASH_SIZE (16)
5226 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5228 extern struct net_device *blackhole_netdev;
5230 #endif /* _LINUX_NETDEVICE_H */