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
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 <linux/ethtool.h>
38 #include <net/net_namespace.h>
40 #include <net/dcbnl.h>
42 #include <net/netprio_cgroup.h>
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #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 netdev_set_default_ethtool_ops(struct net_device *dev,
74 const struct ethtool_ops *ops);
76 /* Backlog congestion levels */
77 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
78 #define NET_RX_DROP 1 /* packet dropped */
80 #define MAX_NEST_DEV 8
83 * Transmit return codes: transmit return codes originate from three different
86 * - qdisc return codes
87 * - driver transmit return codes
90 * Drivers are allowed to return any one of those in their hard_start_xmit()
91 * function. Real network devices commonly used with qdiscs should only return
92 * the driver transmit return codes though - when qdiscs are used, the actual
93 * transmission happens asynchronously, so the value is not propagated to
94 * higher layers. Virtual network devices transmit synchronously; in this case
95 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
96 * others are propagated to higher layers.
99 /* qdisc ->enqueue() return codes. */
100 #define NET_XMIT_SUCCESS 0x00
101 #define NET_XMIT_DROP 0x01 /* skb dropped */
102 #define NET_XMIT_CN 0x02 /* congestion notification */
103 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
105 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
106 * indicates that the device will soon be dropping packets, or already drops
107 * some packets of the same priority; prompting us to send less aggressively. */
108 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
109 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
111 /* Driver transmit return codes */
112 #define NETDEV_TX_MASK 0xf0
115 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
116 NETDEV_TX_OK = 0x00, /* driver took care of packet */
117 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
119 typedef enum netdev_tx netdev_tx_t;
122 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
123 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
125 static inline bool dev_xmit_complete(int rc)
128 * Positive cases with an skb consumed by a driver:
129 * - successful transmission (rc == NETDEV_TX_OK)
130 * - error while transmitting (rc < 0)
131 * - error while queueing to a different device (rc & NET_XMIT_MASK)
133 if (likely(rc < NET_XMIT_MASK))
140 * Compute the worst-case header length according to the protocols
144 #if defined(CONFIG_HYPERV_NET)
145 # define LL_MAX_HEADER 128
146 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
147 # if defined(CONFIG_MAC80211_MESH)
148 # define LL_MAX_HEADER 128
150 # define LL_MAX_HEADER 96
153 # define LL_MAX_HEADER 32
156 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
157 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
158 #define MAX_HEADER LL_MAX_HEADER
160 #define MAX_HEADER (LL_MAX_HEADER + 48)
164 * Old network device statistics. Fields are native words
165 * (unsigned long) so they can be read and written atomically.
168 struct net_device_stats {
169 unsigned long rx_packets;
170 unsigned long tx_packets;
171 unsigned long rx_bytes;
172 unsigned long tx_bytes;
173 unsigned long rx_errors;
174 unsigned long tx_errors;
175 unsigned long rx_dropped;
176 unsigned long tx_dropped;
177 unsigned long multicast;
178 unsigned long collisions;
179 unsigned long rx_length_errors;
180 unsigned long rx_over_errors;
181 unsigned long rx_crc_errors;
182 unsigned long rx_frame_errors;
183 unsigned long rx_fifo_errors;
184 unsigned long rx_missed_errors;
185 unsigned long tx_aborted_errors;
186 unsigned long tx_carrier_errors;
187 unsigned long tx_fifo_errors;
188 unsigned long tx_heartbeat_errors;
189 unsigned long tx_window_errors;
190 unsigned long rx_compressed;
191 unsigned long tx_compressed;
195 #include <linux/cache.h>
196 #include <linux/skbuff.h>
199 #include <linux/static_key.h>
200 extern struct static_key_false rps_needed;
201 extern struct static_key_false rfs_needed;
208 struct netdev_hw_addr {
209 struct list_head list;
210 unsigned char addr[MAX_ADDR_LEN];
212 #define NETDEV_HW_ADDR_T_LAN 1
213 #define NETDEV_HW_ADDR_T_SAN 2
214 #define NETDEV_HW_ADDR_T_SLAVE 3
215 #define NETDEV_HW_ADDR_T_UNICAST 4
216 #define NETDEV_HW_ADDR_T_MULTICAST 5
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_HASHED, /* In NAPI hash (busy polling possible) */
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 */
363 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
364 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
365 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
366 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
367 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
368 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
369 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
380 typedef enum gro_result gro_result_t;
383 * enum rx_handler_result - Possible return values for rx_handlers.
384 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
386 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
387 * case skb->dev was changed by rx_handler.
388 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
389 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
391 * rx_handlers are functions called from inside __netif_receive_skb(), to do
392 * special processing of the skb, prior to delivery to protocol handlers.
394 * Currently, a net_device can only have a single rx_handler registered. Trying
395 * to register a second rx_handler will return -EBUSY.
397 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
398 * To unregister a rx_handler on a net_device, use
399 * netdev_rx_handler_unregister().
401 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
404 * If the rx_handler consumed the skb in some way, it should return
405 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
406 * the skb to be delivered in some other way.
408 * If the rx_handler changed skb->dev, to divert the skb to another
409 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
410 * new device will be called if it exists.
412 * If the rx_handler decides the skb should be ignored, it should return
413 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
414 * are registered on exact device (ptype->dev == skb->dev).
416 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
417 * delivered, it should return RX_HANDLER_PASS.
419 * A device without a registered rx_handler will behave as if rx_handler
420 * returned RX_HANDLER_PASS.
423 enum rx_handler_result {
429 typedef enum rx_handler_result rx_handler_result_t;
430 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
432 void __napi_schedule(struct napi_struct *n);
433 void __napi_schedule_irqoff(struct napi_struct *n);
435 static inline bool napi_disable_pending(struct napi_struct *n)
437 return test_bit(NAPI_STATE_DISABLE, &n->state);
440 bool napi_schedule_prep(struct napi_struct *n);
443 * napi_schedule - schedule NAPI poll
446 * Schedule NAPI poll routine to be called if it is not already
449 static inline void napi_schedule(struct napi_struct *n)
451 if (napi_schedule_prep(n))
456 * napi_schedule_irqoff - schedule NAPI poll
459 * Variant of napi_schedule(), assuming hard irqs are masked.
461 static inline void napi_schedule_irqoff(struct napi_struct *n)
463 if (napi_schedule_prep(n))
464 __napi_schedule_irqoff(n);
467 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
468 static inline bool napi_reschedule(struct napi_struct *napi)
470 if (napi_schedule_prep(napi)) {
471 __napi_schedule(napi);
477 bool napi_complete_done(struct napi_struct *n, int work_done);
479 * napi_complete - NAPI processing complete
482 * Mark NAPI processing as complete.
483 * Consider using napi_complete_done() instead.
484 * Return false if device should avoid rearming interrupts.
486 static inline bool napi_complete(struct napi_struct *n)
488 return napi_complete_done(n, 0);
492 * napi_hash_del - remove a NAPI from global table
493 * @napi: NAPI context
495 * Warning: caller must observe RCU grace period
496 * before freeing memory containing @napi, if
497 * this function returns true.
498 * Note: core networking stack automatically calls it
499 * from netif_napi_del().
500 * Drivers might want to call this helper to combine all
501 * the needed RCU grace periods into a single one.
503 bool napi_hash_del(struct napi_struct *napi);
506 * napi_disable - prevent NAPI from scheduling
509 * Stop NAPI from being scheduled on this context.
510 * Waits till any outstanding processing completes.
512 void napi_disable(struct napi_struct *n);
515 * napi_enable - enable NAPI scheduling
518 * Resume NAPI from being scheduled on this context.
519 * Must be paired with napi_disable.
521 static inline void napi_enable(struct napi_struct *n)
523 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
524 smp_mb__before_atomic();
525 clear_bit(NAPI_STATE_SCHED, &n->state);
526 clear_bit(NAPI_STATE_NPSVC, &n->state);
530 * napi_synchronize - wait until NAPI is not running
533 * Wait until NAPI is done being scheduled on this context.
534 * Waits till any outstanding processing completes but
535 * does not disable future activations.
537 static inline void napi_synchronize(const struct napi_struct *n)
539 if (IS_ENABLED(CONFIG_SMP))
540 while (test_bit(NAPI_STATE_SCHED, &n->state))
547 * napi_if_scheduled_mark_missed - if napi is running, set the
551 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
554 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
556 unsigned long val, new;
559 val = READ_ONCE(n->state);
560 if (val & NAPIF_STATE_DISABLE)
563 if (!(val & NAPIF_STATE_SCHED))
566 new = val | NAPIF_STATE_MISSED;
567 } while (cmpxchg(&n->state, val, new) != val);
572 enum netdev_queue_state_t {
573 __QUEUE_STATE_DRV_XOFF,
574 __QUEUE_STATE_STACK_XOFF,
575 __QUEUE_STATE_FROZEN,
578 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
579 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
580 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
582 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
583 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
585 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
589 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
590 * netif_tx_* functions below are used to manipulate this flag. The
591 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
592 * queue independently. The netif_xmit_*stopped functions below are called
593 * to check if the queue has been stopped by the driver or stack (either
594 * of the XOFF bits are set in the state). Drivers should not need to call
595 * netif_xmit*stopped functions, they should only be using netif_tx_*.
598 struct netdev_queue {
602 struct net_device *dev;
603 struct Qdisc __rcu *qdisc;
604 struct Qdisc *qdisc_sleeping;
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
611 unsigned long tx_maxrate;
613 * Number of TX timeouts for this queue
614 * (/sys/class/net/DEV/Q/trans_timeout)
616 unsigned long trans_timeout;
618 /* Subordinate device that the queue has been assigned to */
619 struct net_device *sb_dev;
620 #ifdef CONFIG_XDP_SOCKETS
621 struct xdp_umem *umem;
626 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
629 * Time (in jiffies) of last Tx
631 unsigned long trans_start;
638 } ____cacheline_aligned_in_smp;
640 extern int sysctl_fb_tunnels_only_for_init_net;
641 extern int sysctl_devconf_inherit_init_net;
643 static inline bool net_has_fallback_tunnels(const struct net *net)
645 return net == &init_net ||
646 !IS_ENABLED(CONFIG_SYSCTL) ||
647 !sysctl_fb_tunnels_only_for_init_net;
650 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
661 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
668 * This structure holds an RPS map which can be of variable length. The
669 * map is an array of CPUs.
676 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
679 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
680 * tail pointer for that CPU's input queue at the time of last enqueue, and
681 * a hardware filter index.
683 struct rps_dev_flow {
686 unsigned int last_qtail;
688 #define RPS_NO_FILTER 0xffff
691 * The rps_dev_flow_table structure contains a table of flow mappings.
693 struct rps_dev_flow_table {
696 struct rps_dev_flow flows[];
698 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
699 ((_num) * sizeof(struct rps_dev_flow)))
702 * The rps_sock_flow_table contains mappings of flows to the last CPU
703 * on which they were processed by the application (set in recvmsg).
704 * Each entry is a 32bit value. Upper part is the high-order bits
705 * of flow hash, lower part is CPU number.
706 * rps_cpu_mask is used to partition the space, depending on number of
707 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
708 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
709 * meaning we use 32-6=26 bits for the hash.
711 struct rps_sock_flow_table {
714 u32 ents[] ____cacheline_aligned_in_smp;
716 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
718 #define RPS_NO_CPU 0xffff
720 extern u32 rps_cpu_mask;
721 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
723 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
727 unsigned int index = hash & table->mask;
728 u32 val = hash & ~rps_cpu_mask;
730 /* We only give a hint, preemption can change CPU under us */
731 val |= raw_smp_processor_id();
733 if (table->ents[index] != val)
734 table->ents[index] = val;
738 #ifdef CONFIG_RFS_ACCEL
739 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
742 #endif /* CONFIG_RPS */
744 /* This structure contains an instance of an RX queue. */
745 struct netdev_rx_queue {
747 struct rps_map __rcu *rps_map;
748 struct rps_dev_flow_table __rcu *rps_flow_table;
751 struct net_device *dev;
752 struct xdp_rxq_info xdp_rxq;
753 #ifdef CONFIG_XDP_SOCKETS
754 struct xdp_umem *umem;
756 } ____cacheline_aligned_in_smp;
759 * RX queue sysfs structures and functions.
761 struct rx_queue_attribute {
762 struct attribute attr;
763 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
764 ssize_t (*store)(struct netdev_rx_queue *queue,
765 const char *buf, size_t len);
770 * This structure holds an XPS map which can be of variable length. The
771 * map is an array of queues.
775 unsigned int alloc_len;
779 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
780 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
781 - sizeof(struct xps_map)) / sizeof(u16))
784 * This structure holds all XPS maps for device. Maps are indexed by CPU.
786 struct xps_dev_maps {
788 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
791 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
792 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
794 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
795 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
797 #endif /* CONFIG_XPS */
799 #define TC_MAX_QUEUE 16
800 #define TC_BITMASK 15
801 /* HW offloaded queuing disciplines txq count and offset maps */
802 struct netdev_tc_txq {
807 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
809 * This structure is to hold information about the device
810 * configured to run FCoE protocol stack.
812 struct netdev_fcoe_hbainfo {
813 char manufacturer[64];
814 char serial_number[64];
815 char hardware_version[64];
816 char driver_version[64];
817 char optionrom_version[64];
818 char firmware_version[64];
820 char model_description[256];
824 #define MAX_PHYS_ITEM_ID_LEN 32
826 /* This structure holds a unique identifier to identify some
827 * physical item (port for example) used by a netdevice.
829 struct netdev_phys_item_id {
830 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
831 unsigned char id_len;
834 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
835 struct netdev_phys_item_id *b)
837 return a->id_len == b->id_len &&
838 memcmp(a->id, b->id, a->id_len) == 0;
841 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
843 struct net_device *sb_dev);
846 TC_SETUP_QDISC_MQPRIO,
849 TC_SETUP_CLSMATCHALL,
859 TC_SETUP_QDISC_TAPRIO,
866 /* These structures hold the attributes of bpf state that are being passed
867 * to the netdevice through the bpf op.
869 enum bpf_netdev_command {
870 /* Set or clear a bpf program used in the earliest stages of packet
871 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
872 * is responsible for calling bpf_prog_put on any old progs that are
873 * stored. In case of error, the callee need not release the new prog
874 * reference, but on success it takes ownership and must bpf_prog_put
875 * when it is no longer used.
879 /* BPF program for offload callbacks, invoked at program load time. */
880 BPF_OFFLOAD_MAP_ALLOC,
881 BPF_OFFLOAD_MAP_FREE,
885 struct bpf_prog_offload_ops;
886 struct netlink_ext_ack;
888 struct xdp_dev_bulk_queue;
898 struct bpf_xdp_entity {
899 struct bpf_prog *prog;
900 struct bpf_xdp_link *link;
904 enum bpf_netdev_command command;
909 struct bpf_prog *prog;
910 struct netlink_ext_ack *extack;
912 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
914 struct bpf_offloaded_map *offmap;
916 /* XDP_SETUP_XSK_UMEM */
918 struct xdp_umem *umem;
924 /* Flags for ndo_xsk_wakeup. */
925 #define XDP_WAKEUP_RX (1 << 0)
926 #define XDP_WAKEUP_TX (1 << 1)
928 #ifdef CONFIG_XFRM_OFFLOAD
930 int (*xdo_dev_state_add) (struct xfrm_state *x);
931 void (*xdo_dev_state_delete) (struct xfrm_state *x);
932 void (*xdo_dev_state_free) (struct xfrm_state *x);
933 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
934 struct xfrm_state *x);
935 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
940 struct rcu_head rcuhead;
947 struct netdev_name_node {
948 struct hlist_node hlist;
949 struct list_head list;
950 struct net_device *dev;
954 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
955 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
957 struct netdev_net_notifier {
958 struct list_head list;
959 struct notifier_block *nb;
963 * This structure defines the management hooks for network devices.
964 * The following hooks can be defined; unless noted otherwise, they are
965 * optional and can be filled with a null pointer.
967 * int (*ndo_init)(struct net_device *dev);
968 * This function is called once when a network device is registered.
969 * The network device can use this for any late stage initialization
970 * or semantic validation. It can fail with an error code which will
971 * be propagated back to register_netdev.
973 * void (*ndo_uninit)(struct net_device *dev);
974 * This function is called when device is unregistered or when registration
975 * fails. It is not called if init fails.
977 * int (*ndo_open)(struct net_device *dev);
978 * This function is called when a network device transitions to the up
981 * int (*ndo_stop)(struct net_device *dev);
982 * This function is called when a network device transitions to the down
985 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
986 * struct net_device *dev);
987 * Called when a packet needs to be transmitted.
988 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
989 * the queue before that can happen; it's for obsolete devices and weird
990 * corner cases, but the stack really does a non-trivial amount
991 * of useless work if you return NETDEV_TX_BUSY.
992 * Required; cannot be NULL.
994 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
995 * struct net_device *dev
996 * netdev_features_t features);
997 * Called by core transmit path to determine if device is capable of
998 * performing offload operations on a given packet. This is to give
999 * the device an opportunity to implement any restrictions that cannot
1000 * be otherwise expressed by feature flags. The check is called with
1001 * the set of features that the stack has calculated and it returns
1002 * those the driver believes to be appropriate.
1004 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1005 * struct net_device *sb_dev);
1006 * Called to decide which queue to use when device supports multiple
1009 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1010 * This function is called to allow device receiver to make
1011 * changes to configuration when multicast or promiscuous is enabled.
1013 * void (*ndo_set_rx_mode)(struct net_device *dev);
1014 * This function is called device changes address list filtering.
1015 * If driver handles unicast address filtering, it should set
1016 * IFF_UNICAST_FLT in its priv_flags.
1018 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1019 * This function is called when the Media Access Control address
1020 * needs to be changed. If this interface is not defined, the
1021 * MAC address can not be changed.
1023 * int (*ndo_validate_addr)(struct net_device *dev);
1024 * Test if Media Access Control address is valid for the device.
1026 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1027 * Called when a user requests an ioctl which can't be handled by
1028 * the generic interface code. If not defined ioctls return
1029 * not supported error code.
1031 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1032 * Used to set network devices bus interface parameters. This interface
1033 * is retained for legacy reasons; new devices should use the bus
1034 * interface (PCI) for low level management.
1036 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1037 * Called when a user wants to change the Maximum Transfer Unit
1040 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1041 * Callback used when the transmitter has not made any progress
1042 * for dev->watchdog ticks.
1044 * void (*ndo_get_stats64)(struct net_device *dev,
1045 * struct rtnl_link_stats64 *storage);
1046 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1047 * Called when a user wants to get the network device usage
1048 * statistics. Drivers must do one of the following:
1049 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1050 * rtnl_link_stats64 structure passed by the caller.
1051 * 2. Define @ndo_get_stats to update a net_device_stats structure
1052 * (which should normally be dev->stats) and return a pointer to
1053 * it. The structure may be changed asynchronously only if each
1054 * field is written atomically.
1055 * 3. Update dev->stats asynchronously and atomically, and define
1056 * neither operation.
1058 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1059 * Return true if this device supports offload stats of this attr_id.
1061 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1063 * Get statistics for offload operations by attr_id. Write it into the
1064 * attr_data pointer.
1066 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1067 * If device supports VLAN filtering this function is called when a
1068 * VLAN id is registered.
1070 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1071 * If device supports VLAN filtering this function is called when a
1072 * VLAN id is unregistered.
1074 * void (*ndo_poll_controller)(struct net_device *dev);
1076 * SR-IOV management functions.
1077 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1078 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1079 * u8 qos, __be16 proto);
1080 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1082 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1083 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1084 * int (*ndo_get_vf_config)(struct net_device *dev,
1085 * int vf, struct ifla_vf_info *ivf);
1086 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1087 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1088 * struct nlattr *port[]);
1090 * Enable or disable the VF ability to query its RSS Redirection Table and
1091 * Hash Key. This is needed since on some devices VF share this information
1092 * with PF and querying it may introduce a theoretical security risk.
1093 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1094 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1095 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1097 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1098 * This is always called from the stack with the rtnl lock held and netif
1099 * tx queues stopped. This allows the netdevice to perform queue
1100 * management safely.
1102 * Fiber Channel over Ethernet (FCoE) offload functions.
1103 * int (*ndo_fcoe_enable)(struct net_device *dev);
1104 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1105 * so the underlying device can perform whatever needed configuration or
1106 * initialization to support acceleration of FCoE traffic.
1108 * int (*ndo_fcoe_disable)(struct net_device *dev);
1109 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1110 * so the underlying device can perform whatever needed clean-ups to
1111 * stop supporting acceleration of FCoE traffic.
1113 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1114 * struct scatterlist *sgl, unsigned int sgc);
1115 * Called when the FCoE Initiator wants to initialize an I/O that
1116 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1117 * perform necessary setup and returns 1 to indicate the device is set up
1118 * successfully to perform DDP on this I/O, otherwise this returns 0.
1120 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1121 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1122 * indicated by the FC exchange id 'xid', so the underlying device can
1123 * clean up and reuse resources for later DDP requests.
1125 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1126 * struct scatterlist *sgl, unsigned int sgc);
1127 * Called when the FCoE Target wants to initialize an I/O that
1128 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1129 * perform necessary setup and returns 1 to indicate the device is set up
1130 * successfully to perform DDP on this I/O, otherwise this returns 0.
1132 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1133 * struct netdev_fcoe_hbainfo *hbainfo);
1134 * Called when the FCoE Protocol stack wants information on the underlying
1135 * device. This information is utilized by the FCoE protocol stack to
1136 * register attributes with Fiber Channel management service as per the
1137 * FC-GS Fabric Device Management Information(FDMI) specification.
1139 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1140 * Called when the underlying device wants to override default World Wide
1141 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1142 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1143 * protocol stack to use.
1146 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1147 * u16 rxq_index, u32 flow_id);
1148 * Set hardware filter for RFS. rxq_index is the target queue index;
1149 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1150 * Return the filter ID on success, or a negative error code.
1152 * Slave management functions (for bridge, bonding, etc).
1153 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1154 * Called to make another netdev an underling.
1156 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1157 * Called to release previously enslaved netdev.
1159 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1160 * struct sk_buff *skb,
1162 * Get the xmit slave of master device. If all_slaves is true, function
1163 * assume all the slaves can transmit.
1165 * Feature/offload setting functions.
1166 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1167 * netdev_features_t features);
1168 * Adjusts the requested feature flags according to device-specific
1169 * constraints, and returns the resulting flags. Must not modify
1172 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1173 * Called to update device configuration to new features. Passed
1174 * feature set might be less than what was returned by ndo_fix_features()).
1175 * Must return >0 or -errno if it changed dev->features itself.
1177 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1178 * struct net_device *dev,
1179 * const unsigned char *addr, u16 vid, u16 flags,
1180 * struct netlink_ext_ack *extack);
1181 * Adds an FDB entry to dev for addr.
1182 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1183 * struct net_device *dev,
1184 * const unsigned char *addr, u16 vid)
1185 * Deletes the FDB entry from dev coresponding to addr.
1186 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1187 * struct net_device *dev, struct net_device *filter_dev,
1189 * Used to add FDB entries to dump requests. Implementers should add
1190 * entries to skb and update idx with the number of entries.
1192 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1193 * u16 flags, struct netlink_ext_ack *extack)
1194 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1195 * struct net_device *dev, u32 filter_mask,
1197 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1200 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1201 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1202 * which do not represent real hardware may define this to allow their
1203 * userspace components to manage their virtual carrier state. Devices
1204 * that determine carrier state from physical hardware properties (eg
1205 * network cables) or protocol-dependent mechanisms (eg
1206 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1208 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1209 * struct netdev_phys_item_id *ppid);
1210 * Called to get ID of physical port of this device. If driver does
1211 * not implement this, it is assumed that the hw is not able to have
1212 * multiple net devices on single physical port.
1214 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1215 * struct netdev_phys_item_id *ppid)
1216 * Called to get the parent ID of the physical port of this device.
1218 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1219 * struct udp_tunnel_info *ti);
1220 * Called by UDP tunnel to notify a driver about the UDP port and socket
1221 * address family that a UDP tunnel is listnening to. It is called only
1222 * when a new port starts listening. The operation is protected by the
1225 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1226 * struct udp_tunnel_info *ti);
1227 * Called by UDP tunnel to notify the driver about a UDP port and socket
1228 * address family that the UDP tunnel is not listening to anymore. The
1229 * operation is protected by the RTNL.
1231 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1232 * struct net_device *dev)
1233 * Called by upper layer devices to accelerate switching or other
1234 * station functionality into hardware. 'pdev is the lowerdev
1235 * to use for the offload and 'dev' is the net device that will
1236 * back the offload. Returns a pointer to the private structure
1237 * the upper layer will maintain.
1238 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1239 * Called by upper layer device to delete the station created
1240 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1241 * the station and priv is the structure returned by the add
1243 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1244 * int queue_index, u32 maxrate);
1245 * Called when a user wants to set a max-rate limitation of specific
1247 * int (*ndo_get_iflink)(const struct net_device *dev);
1248 * Called to get the iflink value of this device.
1249 * void (*ndo_change_proto_down)(struct net_device *dev,
1251 * This function is used to pass protocol port error state information
1252 * to the switch driver. The switch driver can react to the proto_down
1253 * by doing a phys down on the associated switch port.
1254 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1255 * This function is used to get egress tunnel information for given skb.
1256 * This is useful for retrieving outer tunnel header parameters while
1258 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1259 * This function is used to specify the headroom that the skb must
1260 * consider when allocation skb during packet reception. Setting
1261 * appropriate rx headroom value allows avoiding skb head copy on
1262 * forward. Setting a negative value resets the rx headroom to the
1264 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1265 * This function is used to set or query state related to XDP on the
1266 * netdevice and manage BPF offload. See definition of
1267 * enum bpf_netdev_command for details.
1268 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1270 * This function is used to submit @n XDP packets for transmit on a
1271 * netdevice. Returns number of frames successfully transmitted, frames
1272 * that got dropped are freed/returned via xdp_return_frame().
1273 * Returns negative number, means general error invoking ndo, meaning
1274 * no frames were xmit'ed and core-caller will free all frames.
1275 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1276 * This function is used to wake up the softirq, ksoftirqd or kthread
1277 * responsible for sending and/or receiving packets on a specific
1278 * queue id bound to an AF_XDP socket. The flags field specifies if
1279 * only RX, only Tx, or both should be woken up using the flags
1280 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1281 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1282 * Get devlink port instance associated with a given netdev.
1283 * Called with a reference on the netdevice and devlink locks only,
1284 * rtnl_lock is not held.
1285 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1287 * Add, change, delete or get information on an IPv4 tunnel.
1289 struct net_device_ops {
1290 int (*ndo_init)(struct net_device *dev);
1291 void (*ndo_uninit)(struct net_device *dev);
1292 int (*ndo_open)(struct net_device *dev);
1293 int (*ndo_stop)(struct net_device *dev);
1294 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1295 struct net_device *dev);
1296 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1297 struct net_device *dev,
1298 netdev_features_t features);
1299 u16 (*ndo_select_queue)(struct net_device *dev,
1300 struct sk_buff *skb,
1301 struct net_device *sb_dev);
1302 void (*ndo_change_rx_flags)(struct net_device *dev,
1304 void (*ndo_set_rx_mode)(struct net_device *dev);
1305 int (*ndo_set_mac_address)(struct net_device *dev,
1307 int (*ndo_validate_addr)(struct net_device *dev);
1308 int (*ndo_do_ioctl)(struct net_device *dev,
1309 struct ifreq *ifr, int cmd);
1310 int (*ndo_set_config)(struct net_device *dev,
1312 int (*ndo_change_mtu)(struct net_device *dev,
1314 int (*ndo_neigh_setup)(struct net_device *dev,
1315 struct neigh_parms *);
1316 void (*ndo_tx_timeout) (struct net_device *dev,
1317 unsigned int txqueue);
1319 void (*ndo_get_stats64)(struct net_device *dev,
1320 struct rtnl_link_stats64 *storage);
1321 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1322 int (*ndo_get_offload_stats)(int attr_id,
1323 const struct net_device *dev,
1325 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1327 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1328 __be16 proto, u16 vid);
1329 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1330 __be16 proto, u16 vid);
1331 #ifdef CONFIG_NET_POLL_CONTROLLER
1332 void (*ndo_poll_controller)(struct net_device *dev);
1333 int (*ndo_netpoll_setup)(struct net_device *dev,
1334 struct netpoll_info *info);
1335 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1337 int (*ndo_set_vf_mac)(struct net_device *dev,
1338 int queue, u8 *mac);
1339 int (*ndo_set_vf_vlan)(struct net_device *dev,
1340 int queue, u16 vlan,
1341 u8 qos, __be16 proto);
1342 int (*ndo_set_vf_rate)(struct net_device *dev,
1343 int vf, int min_tx_rate,
1345 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1346 int vf, bool setting);
1347 int (*ndo_set_vf_trust)(struct net_device *dev,
1348 int vf, bool setting);
1349 int (*ndo_get_vf_config)(struct net_device *dev,
1351 struct ifla_vf_info *ivf);
1352 int (*ndo_set_vf_link_state)(struct net_device *dev,
1353 int vf, int link_state);
1354 int (*ndo_get_vf_stats)(struct net_device *dev,
1356 struct ifla_vf_stats
1358 int (*ndo_set_vf_port)(struct net_device *dev,
1360 struct nlattr *port[]);
1361 int (*ndo_get_vf_port)(struct net_device *dev,
1362 int vf, struct sk_buff *skb);
1363 int (*ndo_get_vf_guid)(struct net_device *dev,
1365 struct ifla_vf_guid *node_guid,
1366 struct ifla_vf_guid *port_guid);
1367 int (*ndo_set_vf_guid)(struct net_device *dev,
1370 int (*ndo_set_vf_rss_query_en)(
1371 struct net_device *dev,
1372 int vf, bool setting);
1373 int (*ndo_setup_tc)(struct net_device *dev,
1374 enum tc_setup_type type,
1376 #if IS_ENABLED(CONFIG_FCOE)
1377 int (*ndo_fcoe_enable)(struct net_device *dev);
1378 int (*ndo_fcoe_disable)(struct net_device *dev);
1379 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1381 struct scatterlist *sgl,
1383 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1385 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1387 struct scatterlist *sgl,
1389 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1390 struct netdev_fcoe_hbainfo *hbainfo);
1393 #if IS_ENABLED(CONFIG_LIBFCOE)
1394 #define NETDEV_FCOE_WWNN 0
1395 #define NETDEV_FCOE_WWPN 1
1396 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1397 u64 *wwn, int type);
1400 #ifdef CONFIG_RFS_ACCEL
1401 int (*ndo_rx_flow_steer)(struct net_device *dev,
1402 const struct sk_buff *skb,
1406 int (*ndo_add_slave)(struct net_device *dev,
1407 struct net_device *slave_dev,
1408 struct netlink_ext_ack *extack);
1409 int (*ndo_del_slave)(struct net_device *dev,
1410 struct net_device *slave_dev);
1411 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1412 struct sk_buff *skb,
1414 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1415 netdev_features_t features);
1416 int (*ndo_set_features)(struct net_device *dev,
1417 netdev_features_t features);
1418 int (*ndo_neigh_construct)(struct net_device *dev,
1419 struct neighbour *n);
1420 void (*ndo_neigh_destroy)(struct net_device *dev,
1421 struct neighbour *n);
1423 int (*ndo_fdb_add)(struct ndmsg *ndm,
1424 struct nlattr *tb[],
1425 struct net_device *dev,
1426 const unsigned char *addr,
1429 struct netlink_ext_ack *extack);
1430 int (*ndo_fdb_del)(struct ndmsg *ndm,
1431 struct nlattr *tb[],
1432 struct net_device *dev,
1433 const unsigned char *addr,
1435 int (*ndo_fdb_dump)(struct sk_buff *skb,
1436 struct netlink_callback *cb,
1437 struct net_device *dev,
1438 struct net_device *filter_dev,
1440 int (*ndo_fdb_get)(struct sk_buff *skb,
1441 struct nlattr *tb[],
1442 struct net_device *dev,
1443 const unsigned char *addr,
1444 u16 vid, u32 portid, u32 seq,
1445 struct netlink_ext_ack *extack);
1446 int (*ndo_bridge_setlink)(struct net_device *dev,
1447 struct nlmsghdr *nlh,
1449 struct netlink_ext_ack *extack);
1450 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1452 struct net_device *dev,
1455 int (*ndo_bridge_dellink)(struct net_device *dev,
1456 struct nlmsghdr *nlh,
1458 int (*ndo_change_carrier)(struct net_device *dev,
1460 int (*ndo_get_phys_port_id)(struct net_device *dev,
1461 struct netdev_phys_item_id *ppid);
1462 int (*ndo_get_port_parent_id)(struct net_device *dev,
1463 struct netdev_phys_item_id *ppid);
1464 int (*ndo_get_phys_port_name)(struct net_device *dev,
1465 char *name, size_t len);
1466 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1467 struct udp_tunnel_info *ti);
1468 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1469 struct udp_tunnel_info *ti);
1470 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1471 struct net_device *dev);
1472 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1475 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1478 int (*ndo_get_iflink)(const struct net_device *dev);
1479 int (*ndo_change_proto_down)(struct net_device *dev,
1481 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1482 struct sk_buff *skb);
1483 void (*ndo_set_rx_headroom)(struct net_device *dev,
1484 int needed_headroom);
1485 int (*ndo_bpf)(struct net_device *dev,
1486 struct netdev_bpf *bpf);
1487 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1488 struct xdp_frame **xdp,
1490 int (*ndo_xsk_wakeup)(struct net_device *dev,
1491 u32 queue_id, u32 flags);
1492 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1493 int (*ndo_tunnel_ctl)(struct net_device *dev,
1494 struct ip_tunnel_parm *p, int cmd);
1498 * enum net_device_priv_flags - &struct net_device priv_flags
1500 * These are the &struct net_device, they are only set internally
1501 * by drivers and used in the kernel. These flags are invisible to
1502 * userspace; this means that the order of these flags can change
1503 * during any kernel release.
1505 * You should have a pretty good reason to be extending these flags.
1507 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1508 * @IFF_EBRIDGE: Ethernet bridging device
1509 * @IFF_BONDING: bonding master or slave
1510 * @IFF_ISATAP: ISATAP interface (RFC4214)
1511 * @IFF_WAN_HDLC: WAN HDLC device
1512 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1514 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1515 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1516 * @IFF_MACVLAN_PORT: device used as macvlan port
1517 * @IFF_BRIDGE_PORT: device used as bridge port
1518 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1519 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1520 * @IFF_UNICAST_FLT: Supports unicast filtering
1521 * @IFF_TEAM_PORT: device used as team port
1522 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1523 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1524 * change when it's running
1525 * @IFF_MACVLAN: Macvlan device
1526 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1527 * underlying stacked devices
1528 * @IFF_L3MDEV_MASTER: device is an L3 master device
1529 * @IFF_NO_QUEUE: device can run without qdisc attached
1530 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1531 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1532 * @IFF_TEAM: device is a team device
1533 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1534 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1535 * entity (i.e. the master device for bridged veth)
1536 * @IFF_MACSEC: device is a MACsec device
1537 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1538 * @IFF_FAILOVER: device is a failover master device
1539 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1540 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1541 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1543 enum netdev_priv_flags {
1544 IFF_802_1Q_VLAN = 1<<0,
1548 IFF_WAN_HDLC = 1<<4,
1549 IFF_XMIT_DST_RELEASE = 1<<5,
1550 IFF_DONT_BRIDGE = 1<<6,
1551 IFF_DISABLE_NETPOLL = 1<<7,
1552 IFF_MACVLAN_PORT = 1<<8,
1553 IFF_BRIDGE_PORT = 1<<9,
1554 IFF_OVS_DATAPATH = 1<<10,
1555 IFF_TX_SKB_SHARING = 1<<11,
1556 IFF_UNICAST_FLT = 1<<12,
1557 IFF_TEAM_PORT = 1<<13,
1558 IFF_SUPP_NOFCS = 1<<14,
1559 IFF_LIVE_ADDR_CHANGE = 1<<15,
1560 IFF_MACVLAN = 1<<16,
1561 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1562 IFF_L3MDEV_MASTER = 1<<18,
1563 IFF_NO_QUEUE = 1<<19,
1564 IFF_OPENVSWITCH = 1<<20,
1565 IFF_L3MDEV_SLAVE = 1<<21,
1567 IFF_RXFH_CONFIGURED = 1<<23,
1568 IFF_PHONY_HEADROOM = 1<<24,
1570 IFF_NO_RX_HANDLER = 1<<26,
1571 IFF_FAILOVER = 1<<27,
1572 IFF_FAILOVER_SLAVE = 1<<28,
1573 IFF_L3MDEV_RX_HANDLER = 1<<29,
1574 IFF_LIVE_RENAME_OK = 1<<30,
1577 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1578 #define IFF_EBRIDGE IFF_EBRIDGE
1579 #define IFF_BONDING IFF_BONDING
1580 #define IFF_ISATAP IFF_ISATAP
1581 #define IFF_WAN_HDLC IFF_WAN_HDLC
1582 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1583 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1584 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1585 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1586 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1587 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1588 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1589 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1590 #define IFF_TEAM_PORT IFF_TEAM_PORT
1591 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1592 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1593 #define IFF_MACVLAN IFF_MACVLAN
1594 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1595 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1596 #define IFF_NO_QUEUE IFF_NO_QUEUE
1597 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1598 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1599 #define IFF_TEAM IFF_TEAM
1600 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1601 #define IFF_MACSEC IFF_MACSEC
1602 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1603 #define IFF_FAILOVER IFF_FAILOVER
1604 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1605 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1606 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1609 * struct net_device - The DEVICE structure.
1611 * Actually, this whole structure is a big mistake. It mixes I/O
1612 * data with strictly "high-level" data, and it has to know about
1613 * almost every data structure used in the INET module.
1615 * @name: This is the first field of the "visible" part of this structure
1616 * (i.e. as seen by users in the "Space.c" file). It is the name
1619 * @name_node: Name hashlist node
1620 * @ifalias: SNMP alias
1621 * @mem_end: Shared memory end
1622 * @mem_start: Shared memory start
1623 * @base_addr: Device I/O address
1624 * @irq: Device IRQ number
1626 * @state: Generic network queuing layer state, see netdev_state_t
1627 * @dev_list: The global list of network devices
1628 * @napi_list: List entry used for polling NAPI devices
1629 * @unreg_list: List entry when we are unregistering the
1630 * device; see the function unregister_netdev
1631 * @close_list: List entry used when we are closing the device
1632 * @ptype_all: Device-specific packet handlers for all protocols
1633 * @ptype_specific: Device-specific, protocol-specific packet handlers
1635 * @adj_list: Directly linked devices, like slaves for bonding
1636 * @features: Currently active device features
1637 * @hw_features: User-changeable features
1639 * @wanted_features: User-requested features
1640 * @vlan_features: Mask of features inheritable by VLAN devices
1642 * @hw_enc_features: Mask of features inherited by encapsulating devices
1643 * This field indicates what encapsulation
1644 * offloads the hardware is capable of doing,
1645 * and drivers will need to set them appropriately.
1647 * @mpls_features: Mask of features inheritable by MPLS
1648 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1650 * @ifindex: interface index
1651 * @group: The group the device belongs to
1653 * @stats: Statistics struct, which was left as a legacy, use
1654 * rtnl_link_stats64 instead
1656 * @rx_dropped: Dropped packets by core network,
1657 * do not use this in drivers
1658 * @tx_dropped: Dropped packets by core network,
1659 * do not use this in drivers
1660 * @rx_nohandler: nohandler dropped packets by core network on
1661 * inactive devices, do not use this in drivers
1662 * @carrier_up_count: Number of times the carrier has been up
1663 * @carrier_down_count: Number of times the carrier has been down
1665 * @wireless_handlers: List of functions to handle Wireless Extensions,
1667 * see <net/iw_handler.h> for details.
1668 * @wireless_data: Instance data managed by the core of wireless extensions
1670 * @netdev_ops: Includes several pointers to callbacks,
1671 * if one wants to override the ndo_*() functions
1672 * @ethtool_ops: Management operations
1673 * @l3mdev_ops: Layer 3 master device operations
1674 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1675 * discovery handling. Necessary for e.g. 6LoWPAN.
1676 * @xfrmdev_ops: Transformation offload operations
1677 * @tlsdev_ops: Transport Layer Security offload operations
1678 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1679 * of Layer 2 headers.
1681 * @flags: Interface flags (a la BSD)
1682 * @priv_flags: Like 'flags' but invisible to userspace,
1683 * see if.h for the definitions
1684 * @gflags: Global flags ( kept as legacy )
1685 * @padded: How much padding added by alloc_netdev()
1686 * @operstate: RFC2863 operstate
1687 * @link_mode: Mapping policy to operstate
1688 * @if_port: Selectable AUI, TP, ...
1690 * @mtu: Interface MTU value
1691 * @min_mtu: Interface Minimum MTU value
1692 * @max_mtu: Interface Maximum MTU value
1693 * @type: Interface hardware type
1694 * @hard_header_len: Maximum hardware header length.
1695 * @min_header_len: Minimum hardware header length
1697 * @needed_headroom: Extra headroom the hardware may need, but not in all
1698 * cases can this be guaranteed
1699 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1700 * cases can this be guaranteed. Some cases also use
1701 * LL_MAX_HEADER instead to allocate the skb
1703 * interface address info:
1705 * @perm_addr: Permanent hw address
1706 * @addr_assign_type: Hw address assignment type
1707 * @addr_len: Hardware address length
1708 * @upper_level: Maximum depth level of upper devices.
1709 * @lower_level: Maximum depth level of lower devices.
1710 * @neigh_priv_len: Used in neigh_alloc()
1711 * @dev_id: Used to differentiate devices that share
1712 * the same link layer address
1713 * @dev_port: Used to differentiate devices that share
1715 * @addr_list_lock: XXX: need comments on this one
1716 * @name_assign_type: network interface name assignment type
1717 * @uc_promisc: Counter that indicates promiscuous mode
1718 * has been enabled due to the need to listen to
1719 * additional unicast addresses in a device that
1720 * does not implement ndo_set_rx_mode()
1721 * @uc: unicast mac addresses
1722 * @mc: multicast mac addresses
1723 * @dev_addrs: list of device hw addresses
1724 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1725 * @promiscuity: Number of times the NIC is told to work in
1726 * promiscuous mode; if it becomes 0 the NIC will
1727 * exit promiscuous mode
1728 * @allmulti: Counter, enables or disables allmulticast mode
1730 * @vlan_info: VLAN info
1731 * @dsa_ptr: dsa specific data
1732 * @tipc_ptr: TIPC specific data
1733 * @atalk_ptr: AppleTalk link
1734 * @ip_ptr: IPv4 specific data
1735 * @dn_ptr: DECnet specific data
1736 * @ip6_ptr: IPv6 specific data
1737 * @ax25_ptr: AX.25 specific data
1738 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1739 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1741 * @mpls_ptr: mpls_dev struct pointer
1743 * @dev_addr: Hw address (before bcast,
1744 * because most packets are unicast)
1746 * @_rx: Array of RX queues
1747 * @num_rx_queues: Number of RX queues
1748 * allocated at register_netdev() time
1749 * @real_num_rx_queues: Number of RX queues currently active in device
1750 * @xdp_prog: XDP sockets filter program pointer
1751 * @gro_flush_timeout: timeout for GRO layer in NAPI
1752 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1753 * allow to avoid NIC hard IRQ, on busy queues.
1755 * @rx_handler: handler for received packets
1756 * @rx_handler_data: XXX: need comments on this one
1757 * @miniq_ingress: ingress/clsact qdisc specific data for
1758 * ingress processing
1759 * @ingress_queue: XXX: need comments on this one
1760 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1761 * @broadcast: hw bcast address
1763 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1764 * indexed by RX queue number. Assigned by driver.
1765 * This must only be set if the ndo_rx_flow_steer
1766 * operation is defined
1767 * @index_hlist: Device index hash chain
1769 * @_tx: Array of TX queues
1770 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1771 * @real_num_tx_queues: Number of TX queues currently active in device
1772 * @qdisc: Root qdisc from userspace point of view
1773 * @tx_queue_len: Max frames per queue allowed
1774 * @tx_global_lock: XXX: need comments on this one
1775 * @xdp_bulkq: XDP device bulk queue
1776 * @xps_cpus_map: all CPUs map for XPS device
1777 * @xps_rxqs_map: all RXQs map for XPS device
1779 * @xps_maps: XXX: need comments on this one
1780 * @miniq_egress: clsact qdisc specific data for
1782 * @qdisc_hash: qdisc hash table
1783 * @watchdog_timeo: Represents the timeout that is used by
1784 * the watchdog (see dev_watchdog())
1785 * @watchdog_timer: List of timers
1787 * @proto_down_reason: reason a netdev interface is held down
1788 * @pcpu_refcnt: Number of references to this device
1789 * @todo_list: Delayed register/unregister
1790 * @link_watch_list: XXX: need comments on this one
1792 * @reg_state: Register/unregister state machine
1793 * @dismantle: Device is going to be freed
1794 * @rtnl_link_state: This enum represents the phases of creating
1797 * @needs_free_netdev: Should unregister perform free_netdev?
1798 * @priv_destructor: Called from unregister
1799 * @npinfo: XXX: need comments on this one
1800 * @nd_net: Network namespace this network device is inside
1802 * @ml_priv: Mid-layer private
1803 * @lstats: Loopback statistics
1804 * @tstats: Tunnel statistics
1805 * @dstats: Dummy statistics
1806 * @vstats: Virtual ethernet statistics
1811 * @dev: Class/net/name entry
1812 * @sysfs_groups: Space for optional device, statistics and wireless
1815 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1816 * @rtnl_link_ops: Rtnl_link_ops
1818 * @gso_max_size: Maximum size of generic segmentation offload
1819 * @gso_max_segs: Maximum number of segments that can be passed to the
1822 * @dcbnl_ops: Data Center Bridging netlink ops
1823 * @num_tc: Number of traffic classes in the net device
1824 * @tc_to_txq: XXX: need comments on this one
1825 * @prio_tc_map: XXX: need comments on this one
1827 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1829 * @priomap: XXX: need comments on this one
1830 * @phydev: Physical device may attach itself
1831 * for hardware timestamping
1832 * @sfp_bus: attached &struct sfp_bus structure.
1834 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1835 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1837 * @proto_down: protocol port state information can be sent to the
1838 * switch driver and used to set the phys state of the
1841 * @wol_enabled: Wake-on-LAN is enabled
1843 * @net_notifier_list: List of per-net netdev notifier block
1844 * that follow this device when it is moved
1845 * to another network namespace.
1847 * @macsec_ops: MACsec offloading ops
1849 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1850 * offload capabilities of the device
1851 * @udp_tunnel_nic: UDP tunnel offload state
1852 * @xdp_state: stores info on attached XDP BPF programs
1854 * @nested_level: Used as as a parameter of spin_lock_nested() of
1855 * dev->addr_list_lock.
1856 * @unlink_list: As netif_addr_lock() can be called recursively,
1857 * keep a list of interfaces to be deleted.
1859 * FIXME: cleanup struct net_device such that network protocol info
1864 char name[IFNAMSIZ];
1865 struct netdev_name_node *name_node;
1866 struct dev_ifalias __rcu *ifalias;
1868 * I/O specific fields
1869 * FIXME: Merge these and struct ifmap into one
1871 unsigned long mem_end;
1872 unsigned long mem_start;
1873 unsigned long base_addr;
1877 * Some hardware also needs these fields (state,dev_list,
1878 * napi_list,unreg_list,close_list) but they are not
1879 * part of the usual set specified in Space.c.
1882 unsigned long state;
1884 struct list_head dev_list;
1885 struct list_head napi_list;
1886 struct list_head unreg_list;
1887 struct list_head close_list;
1888 struct list_head ptype_all;
1889 struct list_head ptype_specific;
1892 struct list_head upper;
1893 struct list_head lower;
1896 netdev_features_t features;
1897 netdev_features_t hw_features;
1898 netdev_features_t wanted_features;
1899 netdev_features_t vlan_features;
1900 netdev_features_t hw_enc_features;
1901 netdev_features_t mpls_features;
1902 netdev_features_t gso_partial_features;
1907 struct net_device_stats stats;
1909 atomic_long_t rx_dropped;
1910 atomic_long_t tx_dropped;
1911 atomic_long_t rx_nohandler;
1913 /* Stats to monitor link on/off, flapping */
1914 atomic_t carrier_up_count;
1915 atomic_t carrier_down_count;
1917 #ifdef CONFIG_WIRELESS_EXT
1918 const struct iw_handler_def *wireless_handlers;
1919 struct iw_public_data *wireless_data;
1921 const struct net_device_ops *netdev_ops;
1922 const struct ethtool_ops *ethtool_ops;
1923 #ifdef CONFIG_NET_L3_MASTER_DEV
1924 const struct l3mdev_ops *l3mdev_ops;
1926 #if IS_ENABLED(CONFIG_IPV6)
1927 const struct ndisc_ops *ndisc_ops;
1930 #ifdef CONFIG_XFRM_OFFLOAD
1931 const struct xfrmdev_ops *xfrmdev_ops;
1934 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1935 const struct tlsdev_ops *tlsdev_ops;
1938 const struct header_ops *header_ops;
1941 unsigned int priv_flags;
1943 unsigned short gflags;
1944 unsigned short padded;
1946 unsigned char operstate;
1947 unsigned char link_mode;
1949 unsigned char if_port;
1952 /* Note : dev->mtu is often read without holding a lock.
1953 * Writers usually hold RTNL.
1954 * It is recommended to use READ_ONCE() to annotate the reads,
1955 * and to use WRITE_ONCE() to annotate the writes.
1958 unsigned int min_mtu;
1959 unsigned int max_mtu;
1960 unsigned short type;
1961 unsigned short hard_header_len;
1962 unsigned char min_header_len;
1963 unsigned char name_assign_type;
1965 unsigned short needed_headroom;
1966 unsigned short needed_tailroom;
1968 /* Interface address info. */
1969 unsigned char perm_addr[MAX_ADDR_LEN];
1970 unsigned char addr_assign_type;
1971 unsigned char addr_len;
1972 unsigned char upper_level;
1973 unsigned char lower_level;
1975 unsigned short neigh_priv_len;
1976 unsigned short dev_id;
1977 unsigned short dev_port;
1978 spinlock_t addr_list_lock;
1980 struct netdev_hw_addr_list uc;
1981 struct netdev_hw_addr_list mc;
1982 struct netdev_hw_addr_list dev_addrs;
1985 struct kset *queues_kset;
1987 #ifdef CONFIG_LOCKDEP
1988 struct list_head unlink_list;
1990 unsigned int promiscuity;
1991 unsigned int allmulti;
1993 #ifdef CONFIG_LOCKDEP
1994 unsigned char nested_level;
1998 /* Protocol-specific pointers */
2000 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2001 struct vlan_info __rcu *vlan_info;
2003 #if IS_ENABLED(CONFIG_NET_DSA)
2004 struct dsa_port *dsa_ptr;
2006 #if IS_ENABLED(CONFIG_TIPC)
2007 struct tipc_bearer __rcu *tipc_ptr;
2009 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2012 struct in_device __rcu *ip_ptr;
2013 #if IS_ENABLED(CONFIG_DECNET)
2014 struct dn_dev __rcu *dn_ptr;
2016 struct inet6_dev __rcu *ip6_ptr;
2017 #if IS_ENABLED(CONFIG_AX25)
2020 struct wireless_dev *ieee80211_ptr;
2021 struct wpan_dev *ieee802154_ptr;
2022 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2023 struct mpls_dev __rcu *mpls_ptr;
2027 * Cache lines mostly used on receive path (including eth_type_trans())
2029 /* Interface address info used in eth_type_trans() */
2030 unsigned char *dev_addr;
2032 struct netdev_rx_queue *_rx;
2033 unsigned int num_rx_queues;
2034 unsigned int real_num_rx_queues;
2036 struct bpf_prog __rcu *xdp_prog;
2037 unsigned long gro_flush_timeout;
2038 int napi_defer_hard_irqs;
2039 rx_handler_func_t __rcu *rx_handler;
2040 void __rcu *rx_handler_data;
2042 #ifdef CONFIG_NET_CLS_ACT
2043 struct mini_Qdisc __rcu *miniq_ingress;
2045 struct netdev_queue __rcu *ingress_queue;
2046 #ifdef CONFIG_NETFILTER_INGRESS
2047 struct nf_hook_entries __rcu *nf_hooks_ingress;
2050 unsigned char broadcast[MAX_ADDR_LEN];
2051 #ifdef CONFIG_RFS_ACCEL
2052 struct cpu_rmap *rx_cpu_rmap;
2054 struct hlist_node index_hlist;
2057 * Cache lines mostly used on transmit path
2059 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2060 unsigned int num_tx_queues;
2061 unsigned int real_num_tx_queues;
2062 struct Qdisc *qdisc;
2063 unsigned int tx_queue_len;
2064 spinlock_t tx_global_lock;
2066 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2069 struct xps_dev_maps __rcu *xps_cpus_map;
2070 struct xps_dev_maps __rcu *xps_rxqs_map;
2072 #ifdef CONFIG_NET_CLS_ACT
2073 struct mini_Qdisc __rcu *miniq_egress;
2076 #ifdef CONFIG_NET_SCHED
2077 DECLARE_HASHTABLE (qdisc_hash, 4);
2079 /* These may be needed for future network-power-down code. */
2080 struct timer_list watchdog_timer;
2083 u32 proto_down_reason;
2085 struct list_head todo_list;
2086 int __percpu *pcpu_refcnt;
2088 struct list_head link_watch_list;
2090 enum { NETREG_UNINITIALIZED=0,
2091 NETREG_REGISTERED, /* completed register_netdevice */
2092 NETREG_UNREGISTERING, /* called unregister_netdevice */
2093 NETREG_UNREGISTERED, /* completed unregister todo */
2094 NETREG_RELEASED, /* called free_netdev */
2095 NETREG_DUMMY, /* dummy device for NAPI poll */
2101 RTNL_LINK_INITIALIZED,
2102 RTNL_LINK_INITIALIZING,
2103 } rtnl_link_state:16;
2105 bool needs_free_netdev;
2106 void (*priv_destructor)(struct net_device *dev);
2108 #ifdef CONFIG_NETPOLL
2109 struct netpoll_info __rcu *npinfo;
2112 possible_net_t nd_net;
2114 /* mid-layer private */
2117 struct pcpu_lstats __percpu *lstats;
2118 struct pcpu_sw_netstats __percpu *tstats;
2119 struct pcpu_dstats __percpu *dstats;
2122 #if IS_ENABLED(CONFIG_GARP)
2123 struct garp_port __rcu *garp_port;
2125 #if IS_ENABLED(CONFIG_MRP)
2126 struct mrp_port __rcu *mrp_port;
2130 const struct attribute_group *sysfs_groups[4];
2131 const struct attribute_group *sysfs_rx_queue_group;
2133 const struct rtnl_link_ops *rtnl_link_ops;
2135 /* for setting kernel sock attribute on TCP connection setup */
2136 #define GSO_MAX_SIZE 65536
2137 unsigned int gso_max_size;
2138 #define GSO_MAX_SEGS 65535
2142 const struct dcbnl_rtnl_ops *dcbnl_ops;
2145 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2146 u8 prio_tc_map[TC_BITMASK + 1];
2148 #if IS_ENABLED(CONFIG_FCOE)
2149 unsigned int fcoe_ddp_xid;
2151 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2152 struct netprio_map __rcu *priomap;
2154 struct phy_device *phydev;
2155 struct sfp_bus *sfp_bus;
2156 struct lock_class_key *qdisc_tx_busylock;
2157 struct lock_class_key *qdisc_running_key;
2159 unsigned wol_enabled:1;
2161 struct list_head net_notifier_list;
2163 #if IS_ENABLED(CONFIG_MACSEC)
2164 /* MACsec management functions */
2165 const struct macsec_ops *macsec_ops;
2167 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2168 struct udp_tunnel_nic *udp_tunnel_nic;
2170 /* protected by rtnl_lock */
2171 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2173 #define to_net_dev(d) container_of(d, struct net_device, dev)
2175 static inline bool netif_elide_gro(const struct net_device *dev)
2177 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2182 #define NETDEV_ALIGN 32
2185 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2187 return dev->prio_tc_map[prio & TC_BITMASK];
2191 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2193 if (tc >= dev->num_tc)
2196 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2200 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2201 void netdev_reset_tc(struct net_device *dev);
2202 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2203 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2206 int netdev_get_num_tc(struct net_device *dev)
2211 void netdev_unbind_sb_channel(struct net_device *dev,
2212 struct net_device *sb_dev);
2213 int netdev_bind_sb_channel_queue(struct net_device *dev,
2214 struct net_device *sb_dev,
2215 u8 tc, u16 count, u16 offset);
2216 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2217 static inline int netdev_get_sb_channel(struct net_device *dev)
2219 return max_t(int, -dev->num_tc, 0);
2223 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2226 return &dev->_tx[index];
2229 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2230 const struct sk_buff *skb)
2232 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2235 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2236 void (*f)(struct net_device *,
2237 struct netdev_queue *,
2243 for (i = 0; i < dev->num_tx_queues; i++)
2244 f(dev, &dev->_tx[i], arg);
2247 #define netdev_lockdep_set_classes(dev) \
2249 static struct lock_class_key qdisc_tx_busylock_key; \
2250 static struct lock_class_key qdisc_running_key; \
2251 static struct lock_class_key qdisc_xmit_lock_key; \
2252 static struct lock_class_key dev_addr_list_lock_key; \
2255 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2256 (dev)->qdisc_running_key = &qdisc_running_key; \
2257 lockdep_set_class(&(dev)->addr_list_lock, \
2258 &dev_addr_list_lock_key); \
2259 for (i = 0; i < (dev)->num_tx_queues; i++) \
2260 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2261 &qdisc_xmit_lock_key); \
2264 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2265 struct net_device *sb_dev);
2266 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2267 struct sk_buff *skb,
2268 struct net_device *sb_dev);
2270 /* returns the headroom that the master device needs to take in account
2271 * when forwarding to this dev
2273 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2275 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2278 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2280 if (dev->netdev_ops->ndo_set_rx_headroom)
2281 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2284 /* set the device rx headroom to the dev's default */
2285 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2287 netdev_set_rx_headroom(dev, -1);
2291 * Net namespace inlines
2294 struct net *dev_net(const struct net_device *dev)
2296 return read_pnet(&dev->nd_net);
2300 void dev_net_set(struct net_device *dev, struct net *net)
2302 write_pnet(&dev->nd_net, net);
2306 * netdev_priv - access network device private data
2307 * @dev: network device
2309 * Get network device private data
2311 static inline void *netdev_priv(const struct net_device *dev)
2313 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2316 /* Set the sysfs physical device reference for the network logical device
2317 * if set prior to registration will cause a symlink during initialization.
2319 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2321 /* Set the sysfs device type for the network logical device to allow
2322 * fine-grained identification of different network device types. For
2323 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2325 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2327 /* Default NAPI poll() weight
2328 * Device drivers are strongly advised to not use bigger value
2330 #define NAPI_POLL_WEIGHT 64
2333 * netif_napi_add - initialize a NAPI context
2334 * @dev: network device
2335 * @napi: NAPI context
2336 * @poll: polling function
2337 * @weight: default weight
2339 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2340 * *any* of the other NAPI-related functions.
2342 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2343 int (*poll)(struct napi_struct *, int), int weight);
2346 * netif_tx_napi_add - initialize a NAPI context
2347 * @dev: network device
2348 * @napi: NAPI context
2349 * @poll: polling function
2350 * @weight: default weight
2352 * This variant of netif_napi_add() should be used from drivers using NAPI
2353 * to exclusively poll a TX queue.
2354 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2356 static inline void netif_tx_napi_add(struct net_device *dev,
2357 struct napi_struct *napi,
2358 int (*poll)(struct napi_struct *, int),
2361 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2362 netif_napi_add(dev, napi, poll, weight);
2366 * netif_napi_del - remove a NAPI context
2367 * @napi: NAPI context
2369 * netif_napi_del() removes a NAPI context from the network device NAPI list
2371 void netif_napi_del(struct napi_struct *napi);
2373 struct napi_gro_cb {
2374 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2377 /* Length of frag0. */
2378 unsigned int frag0_len;
2380 /* This indicates where we are processing relative to skb->data. */
2383 /* This is non-zero if the packet cannot be merged with the new skb. */
2386 /* Save the IP ID here and check when we get to the transport layer */
2389 /* Number of segments aggregated. */
2392 /* Start offset for remote checksum offload */
2393 u16 gro_remcsum_start;
2395 /* jiffies when first packet was created/queued */
2398 /* Used in ipv6_gro_receive() and foo-over-udp */
2401 /* This is non-zero if the packet may be of the same flow. */
2404 /* Used in tunnel GRO receive */
2407 /* GRO checksum is valid */
2410 /* Number of checksums via CHECKSUM_UNNECESSARY */
2415 #define NAPI_GRO_FREE 1
2416 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2418 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2421 /* Used in GRE, set in fou/gue_gro_receive */
2424 /* Used to determine if flush_id can be ignored */
2427 /* Number of gro_receive callbacks this packet already went through */
2428 u8 recursion_counter:4;
2430 /* GRO is done by frag_list pointer chaining. */
2433 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2436 /* used in skb_gro_receive() slow path */
2437 struct sk_buff *last;
2440 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2442 #define GRO_RECURSION_LIMIT 15
2443 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2445 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2448 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2449 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2450 struct list_head *head,
2451 struct sk_buff *skb)
2453 if (unlikely(gro_recursion_inc_test(skb))) {
2454 NAPI_GRO_CB(skb)->flush |= 1;
2458 return cb(head, skb);
2461 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2463 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2465 struct list_head *head,
2466 struct sk_buff *skb)
2468 if (unlikely(gro_recursion_inc_test(skb))) {
2469 NAPI_GRO_CB(skb)->flush |= 1;
2473 return cb(sk, head, skb);
2476 struct packet_type {
2477 __be16 type; /* This is really htons(ether_type). */
2478 bool ignore_outgoing;
2479 struct net_device *dev; /* NULL is wildcarded here */
2480 int (*func) (struct sk_buff *,
2481 struct net_device *,
2482 struct packet_type *,
2483 struct net_device *);
2484 void (*list_func) (struct list_head *,
2485 struct packet_type *,
2486 struct net_device *);
2487 bool (*id_match)(struct packet_type *ptype,
2489 void *af_packet_priv;
2490 struct list_head list;
2493 struct offload_callbacks {
2494 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2495 netdev_features_t features);
2496 struct sk_buff *(*gro_receive)(struct list_head *head,
2497 struct sk_buff *skb);
2498 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2501 struct packet_offload {
2502 __be16 type; /* This is really htons(ether_type). */
2504 struct offload_callbacks callbacks;
2505 struct list_head list;
2508 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2509 struct pcpu_sw_netstats {
2514 struct u64_stats_sync syncp;
2515 } __aligned(4 * sizeof(u64));
2517 struct pcpu_lstats {
2518 u64_stats_t packets;
2520 struct u64_stats_sync syncp;
2521 } __aligned(2 * sizeof(u64));
2523 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2525 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2527 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2529 u64_stats_update_begin(&lstats->syncp);
2530 u64_stats_add(&lstats->bytes, len);
2531 u64_stats_inc(&lstats->packets);
2532 u64_stats_update_end(&lstats->syncp);
2535 #define __netdev_alloc_pcpu_stats(type, gfp) \
2537 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2540 for_each_possible_cpu(__cpu) { \
2541 typeof(type) *stat; \
2542 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2543 u64_stats_init(&stat->syncp); \
2549 #define netdev_alloc_pcpu_stats(type) \
2550 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2552 enum netdev_lag_tx_type {
2553 NETDEV_LAG_TX_TYPE_UNKNOWN,
2554 NETDEV_LAG_TX_TYPE_RANDOM,
2555 NETDEV_LAG_TX_TYPE_BROADCAST,
2556 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2557 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2558 NETDEV_LAG_TX_TYPE_HASH,
2561 enum netdev_lag_hash {
2562 NETDEV_LAG_HASH_NONE,
2564 NETDEV_LAG_HASH_L34,
2565 NETDEV_LAG_HASH_L23,
2566 NETDEV_LAG_HASH_E23,
2567 NETDEV_LAG_HASH_E34,
2568 NETDEV_LAG_HASH_UNKNOWN,
2571 struct netdev_lag_upper_info {
2572 enum netdev_lag_tx_type tx_type;
2573 enum netdev_lag_hash hash_type;
2576 struct netdev_lag_lower_state_info {
2581 #include <linux/notifier.h>
2583 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2584 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2588 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2590 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2591 detected a hardware crash and restarted
2592 - we can use this eg to kick tcp sessions
2594 NETDEV_CHANGE, /* Notify device state change */
2597 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2598 NETDEV_CHANGEADDR, /* notify after the address change */
2599 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2603 NETDEV_BONDING_FAILOVER,
2605 NETDEV_PRE_TYPE_CHANGE,
2606 NETDEV_POST_TYPE_CHANGE,
2609 NETDEV_NOTIFY_PEERS,
2613 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2614 NETDEV_CHANGEINFODATA,
2615 NETDEV_BONDING_INFO,
2616 NETDEV_PRECHANGEUPPER,
2617 NETDEV_CHANGELOWERSTATE,
2618 NETDEV_UDP_TUNNEL_PUSH_INFO,
2619 NETDEV_UDP_TUNNEL_DROP_INFO,
2620 NETDEV_CHANGE_TX_QUEUE_LEN,
2621 NETDEV_CVLAN_FILTER_PUSH_INFO,
2622 NETDEV_CVLAN_FILTER_DROP_INFO,
2623 NETDEV_SVLAN_FILTER_PUSH_INFO,
2624 NETDEV_SVLAN_FILTER_DROP_INFO,
2626 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2628 int register_netdevice_notifier(struct notifier_block *nb);
2629 int unregister_netdevice_notifier(struct notifier_block *nb);
2630 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2631 int unregister_netdevice_notifier_net(struct net *net,
2632 struct notifier_block *nb);
2633 int register_netdevice_notifier_dev_net(struct net_device *dev,
2634 struct notifier_block *nb,
2635 struct netdev_net_notifier *nn);
2636 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2637 struct notifier_block *nb,
2638 struct netdev_net_notifier *nn);
2640 struct netdev_notifier_info {
2641 struct net_device *dev;
2642 struct netlink_ext_ack *extack;
2645 struct netdev_notifier_info_ext {
2646 struct netdev_notifier_info info; /* must be first */
2652 struct netdev_notifier_change_info {
2653 struct netdev_notifier_info info; /* must be first */
2654 unsigned int flags_changed;
2657 struct netdev_notifier_changeupper_info {
2658 struct netdev_notifier_info info; /* must be first */
2659 struct net_device *upper_dev; /* new upper dev */
2660 bool master; /* is upper dev master */
2661 bool linking; /* is the notification for link or unlink */
2662 void *upper_info; /* upper dev info */
2665 struct netdev_notifier_changelowerstate_info {
2666 struct netdev_notifier_info info; /* must be first */
2667 void *lower_state_info; /* is lower dev state */
2670 struct netdev_notifier_pre_changeaddr_info {
2671 struct netdev_notifier_info info; /* must be first */
2672 const unsigned char *dev_addr;
2675 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2676 struct net_device *dev)
2679 info->extack = NULL;
2682 static inline struct net_device *
2683 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2688 static inline struct netlink_ext_ack *
2689 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2691 return info->extack;
2694 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2697 extern rwlock_t dev_base_lock; /* Device list lock */
2699 #define for_each_netdev(net, d) \
2700 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2701 #define for_each_netdev_reverse(net, d) \
2702 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2703 #define for_each_netdev_rcu(net, d) \
2704 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2705 #define for_each_netdev_safe(net, d, n) \
2706 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2707 #define for_each_netdev_continue(net, d) \
2708 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2709 #define for_each_netdev_continue_reverse(net, d) \
2710 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2712 #define for_each_netdev_continue_rcu(net, d) \
2713 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2714 #define for_each_netdev_in_bond_rcu(bond, slave) \
2715 for_each_netdev_rcu(&init_net, slave) \
2716 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2717 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2719 static inline struct net_device *next_net_device(struct net_device *dev)
2721 struct list_head *lh;
2725 lh = dev->dev_list.next;
2726 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2729 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2731 struct list_head *lh;
2735 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2736 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2739 static inline struct net_device *first_net_device(struct net *net)
2741 return list_empty(&net->dev_base_head) ? NULL :
2742 net_device_entry(net->dev_base_head.next);
2745 static inline struct net_device *first_net_device_rcu(struct net *net)
2747 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2749 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2752 int netdev_boot_setup_check(struct net_device *dev);
2753 unsigned long netdev_boot_base(const char *prefix, int unit);
2754 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2755 const char *hwaddr);
2756 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2757 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2758 void dev_add_pack(struct packet_type *pt);
2759 void dev_remove_pack(struct packet_type *pt);
2760 void __dev_remove_pack(struct packet_type *pt);
2761 void dev_add_offload(struct packet_offload *po);
2762 void dev_remove_offload(struct packet_offload *po);
2764 int dev_get_iflink(const struct net_device *dev);
2765 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2766 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2767 unsigned short mask);
2768 struct net_device *dev_get_by_name(struct net *net, const char *name);
2769 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2770 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2771 int dev_alloc_name(struct net_device *dev, const char *name);
2772 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2773 void dev_close(struct net_device *dev);
2774 void dev_close_many(struct list_head *head, bool unlink);
2775 void dev_disable_lro(struct net_device *dev);
2776 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2777 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2778 struct net_device *sb_dev);
2779 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2780 struct net_device *sb_dev);
2781 int dev_queue_xmit(struct sk_buff *skb);
2782 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2783 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2784 int register_netdevice(struct net_device *dev);
2785 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2786 void unregister_netdevice_many(struct list_head *head);
2787 static inline void unregister_netdevice(struct net_device *dev)
2789 unregister_netdevice_queue(dev, NULL);
2792 int netdev_refcnt_read(const struct net_device *dev);
2793 void free_netdev(struct net_device *dev);
2794 void netdev_freemem(struct net_device *dev);
2795 void synchronize_net(void);
2796 int init_dummy_netdev(struct net_device *dev);
2798 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2799 struct sk_buff *skb,
2801 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2802 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2803 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2804 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2805 int netdev_get_name(struct net *net, char *name, int ifindex);
2806 int dev_restart(struct net_device *dev);
2807 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2808 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2810 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2812 return NAPI_GRO_CB(skb)->data_offset;
2815 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2817 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2820 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2822 NAPI_GRO_CB(skb)->data_offset += len;
2825 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2826 unsigned int offset)
2828 return NAPI_GRO_CB(skb)->frag0 + offset;
2831 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2833 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2836 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2838 NAPI_GRO_CB(skb)->frag0 = NULL;
2839 NAPI_GRO_CB(skb)->frag0_len = 0;
2842 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2843 unsigned int offset)
2845 if (!pskb_may_pull(skb, hlen))
2848 skb_gro_frag0_invalidate(skb);
2849 return skb->data + offset;
2852 static inline void *skb_gro_network_header(struct sk_buff *skb)
2854 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2855 skb_network_offset(skb);
2858 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2859 const void *start, unsigned int len)
2861 if (NAPI_GRO_CB(skb)->csum_valid)
2862 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2863 csum_partial(start, len, 0));
2866 /* GRO checksum functions. These are logical equivalents of the normal
2867 * checksum functions (in skbuff.h) except that they operate on the GRO
2868 * offsets and fields in sk_buff.
2871 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2873 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2875 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2878 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2882 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2883 skb_checksum_start_offset(skb) <
2884 skb_gro_offset(skb)) &&
2885 !skb_at_gro_remcsum_start(skb) &&
2886 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2887 (!zero_okay || check));
2890 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2893 if (NAPI_GRO_CB(skb)->csum_valid &&
2894 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2897 NAPI_GRO_CB(skb)->csum = psum;
2899 return __skb_gro_checksum_complete(skb);
2902 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2904 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2905 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2906 NAPI_GRO_CB(skb)->csum_cnt--;
2908 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2909 * verified a new top level checksum or an encapsulated one
2910 * during GRO. This saves work if we fallback to normal path.
2912 __skb_incr_checksum_unnecessary(skb);
2916 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2919 __sum16 __ret = 0; \
2920 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2921 __ret = __skb_gro_checksum_validate_complete(skb, \
2922 compute_pseudo(skb, proto)); \
2924 skb_gro_incr_csum_unnecessary(skb); \
2928 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2929 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2931 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2933 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2935 #define skb_gro_checksum_simple_validate(skb) \
2936 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2938 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2940 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2941 !NAPI_GRO_CB(skb)->csum_valid);
2944 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2947 NAPI_GRO_CB(skb)->csum = ~pseudo;
2948 NAPI_GRO_CB(skb)->csum_valid = 1;
2951 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
2953 if (__skb_gro_checksum_convert_check(skb)) \
2954 __skb_gro_checksum_convert(skb, \
2955 compute_pseudo(skb, proto)); \
2958 struct gro_remcsum {
2963 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2969 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2970 unsigned int off, size_t hdrlen,
2971 int start, int offset,
2972 struct gro_remcsum *grc,
2976 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2978 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2981 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2985 ptr = skb_gro_header_fast(skb, off);
2986 if (skb_gro_header_hard(skb, off + plen)) {
2987 ptr = skb_gro_header_slow(skb, off + plen, off);
2992 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2995 /* Adjust skb->csum since we changed the packet */
2996 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2998 grc->offset = off + hdrlen + offset;
3004 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3005 struct gro_remcsum *grc)
3008 size_t plen = grc->offset + sizeof(u16);
3013 ptr = skb_gro_header_fast(skb, grc->offset);
3014 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3015 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3020 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3023 #ifdef CONFIG_XFRM_OFFLOAD
3024 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3026 if (PTR_ERR(pp) != -EINPROGRESS)
3027 NAPI_GRO_CB(skb)->flush |= flush;
3029 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3032 struct gro_remcsum *grc)
3034 if (PTR_ERR(pp) != -EINPROGRESS) {
3035 NAPI_GRO_CB(skb)->flush |= flush;
3036 skb_gro_remcsum_cleanup(skb, grc);
3037 skb->remcsum_offload = 0;
3041 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3043 NAPI_GRO_CB(skb)->flush |= flush;
3045 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3048 struct gro_remcsum *grc)
3050 NAPI_GRO_CB(skb)->flush |= flush;
3051 skb_gro_remcsum_cleanup(skb, grc);
3052 skb->remcsum_offload = 0;
3056 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3057 unsigned short type,
3058 const void *daddr, const void *saddr,
3061 if (!dev->header_ops || !dev->header_ops->create)
3064 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3067 static inline int dev_parse_header(const struct sk_buff *skb,
3068 unsigned char *haddr)
3070 const struct net_device *dev = skb->dev;
3072 if (!dev->header_ops || !dev->header_ops->parse)
3074 return dev->header_ops->parse(skb, haddr);
3077 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3079 const struct net_device *dev = skb->dev;
3081 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3083 return dev->header_ops->parse_protocol(skb);
3086 /* ll_header must have at least hard_header_len allocated */
3087 static inline bool dev_validate_header(const struct net_device *dev,
3088 char *ll_header, int len)
3090 if (likely(len >= dev->hard_header_len))
3092 if (len < dev->min_header_len)
3095 if (capable(CAP_SYS_RAWIO)) {
3096 memset(ll_header + len, 0, dev->hard_header_len - len);
3100 if (dev->header_ops && dev->header_ops->validate)
3101 return dev->header_ops->validate(ll_header, len);
3106 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3108 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3109 static inline int unregister_gifconf(unsigned int family)
3111 return register_gifconf(family, NULL);
3114 #ifdef CONFIG_NET_FLOW_LIMIT
3115 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3116 struct sd_flow_limit {
3118 unsigned int num_buckets;
3119 unsigned int history_head;
3120 u16 history[FLOW_LIMIT_HISTORY];
3124 extern int netdev_flow_limit_table_len;
3125 #endif /* CONFIG_NET_FLOW_LIMIT */
3128 * Incoming packets are placed on per-CPU queues
3130 struct softnet_data {
3131 struct list_head poll_list;
3132 struct sk_buff_head process_queue;
3135 unsigned int processed;
3136 unsigned int time_squeeze;
3137 unsigned int received_rps;
3139 struct softnet_data *rps_ipi_list;
3141 #ifdef CONFIG_NET_FLOW_LIMIT
3142 struct sd_flow_limit __rcu *flow_limit;
3144 struct Qdisc *output_queue;
3145 struct Qdisc **output_queue_tailp;
3146 struct sk_buff *completion_queue;
3147 #ifdef CONFIG_XFRM_OFFLOAD
3148 struct sk_buff_head xfrm_backlog;
3150 /* written and read only by owning cpu: */
3156 /* input_queue_head should be written by cpu owning this struct,
3157 * and only read by other cpus. Worth using a cache line.
3159 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3161 /* Elements below can be accessed between CPUs for RPS/RFS */
3162 call_single_data_t csd ____cacheline_aligned_in_smp;
3163 struct softnet_data *rps_ipi_next;
3165 unsigned int input_queue_tail;
3167 unsigned int dropped;
3168 struct sk_buff_head input_pkt_queue;
3169 struct napi_struct backlog;
3173 static inline void input_queue_head_incr(struct softnet_data *sd)
3176 sd->input_queue_head++;
3180 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3181 unsigned int *qtail)
3184 *qtail = ++sd->input_queue_tail;
3188 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3190 static inline int dev_recursion_level(void)
3192 return this_cpu_read(softnet_data.xmit.recursion);
3195 #define XMIT_RECURSION_LIMIT 8
3196 static inline bool dev_xmit_recursion(void)
3198 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3199 XMIT_RECURSION_LIMIT);
3202 static inline void dev_xmit_recursion_inc(void)
3204 __this_cpu_inc(softnet_data.xmit.recursion);
3207 static inline void dev_xmit_recursion_dec(void)
3209 __this_cpu_dec(softnet_data.xmit.recursion);
3212 void __netif_schedule(struct Qdisc *q);
3213 void netif_schedule_queue(struct netdev_queue *txq);
3215 static inline void netif_tx_schedule_all(struct net_device *dev)
3219 for (i = 0; i < dev->num_tx_queues; i++)
3220 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3223 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3225 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3229 * netif_start_queue - allow transmit
3230 * @dev: network device
3232 * Allow upper layers to call the device hard_start_xmit routine.
3234 static inline void netif_start_queue(struct net_device *dev)
3236 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3239 static inline void netif_tx_start_all_queues(struct net_device *dev)
3243 for (i = 0; i < dev->num_tx_queues; i++) {
3244 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3245 netif_tx_start_queue(txq);
3249 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3252 * netif_wake_queue - restart transmit
3253 * @dev: network device
3255 * Allow upper layers to call the device hard_start_xmit routine.
3256 * Used for flow control when transmit resources are available.
3258 static inline void netif_wake_queue(struct net_device *dev)
3260 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3263 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3267 for (i = 0; i < dev->num_tx_queues; i++) {
3268 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3269 netif_tx_wake_queue(txq);
3273 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3275 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3279 * netif_stop_queue - stop transmitted packets
3280 * @dev: network device
3282 * Stop upper layers calling the device hard_start_xmit routine.
3283 * Used for flow control when transmit resources are unavailable.
3285 static inline void netif_stop_queue(struct net_device *dev)
3287 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3290 void netif_tx_stop_all_queues(struct net_device *dev);
3292 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3294 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3298 * netif_queue_stopped - test if transmit queue is flowblocked
3299 * @dev: network device
3301 * Test if transmit queue on device is currently unable to send.
3303 static inline bool netif_queue_stopped(const struct net_device *dev)
3305 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3308 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3310 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3314 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3316 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3320 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3322 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3326 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3327 * @dev_queue: pointer to transmit queue
3329 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3330 * to give appropriate hint to the CPU.
3332 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3335 prefetchw(&dev_queue->dql.num_queued);
3340 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3341 * @dev_queue: pointer to transmit queue
3343 * BQL enabled drivers might use this helper in their TX completion path,
3344 * to give appropriate hint to the CPU.
3346 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3349 prefetchw(&dev_queue->dql.limit);
3353 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3357 dql_queued(&dev_queue->dql, bytes);
3359 if (likely(dql_avail(&dev_queue->dql) >= 0))
3362 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3365 * The XOFF flag must be set before checking the dql_avail below,
3366 * because in netdev_tx_completed_queue we update the dql_completed
3367 * before checking the XOFF flag.
3371 /* check again in case another CPU has just made room avail */
3372 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3373 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3377 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3378 * that they should not test BQL status themselves.
3379 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3381 * Returns true if the doorbell must be used to kick the NIC.
3383 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3389 dql_queued(&dev_queue->dql, bytes);
3391 return netif_tx_queue_stopped(dev_queue);
3393 netdev_tx_sent_queue(dev_queue, bytes);
3398 * netdev_sent_queue - report the number of bytes queued to hardware
3399 * @dev: network device
3400 * @bytes: number of bytes queued to the hardware device queue
3402 * Report the number of bytes queued for sending/completion to the network
3403 * device hardware queue. @bytes should be a good approximation and should
3404 * exactly match netdev_completed_queue() @bytes
3406 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3408 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3411 static inline bool __netdev_sent_queue(struct net_device *dev,
3415 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3419 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3420 unsigned int pkts, unsigned int bytes)
3423 if (unlikely(!bytes))
3426 dql_completed(&dev_queue->dql, bytes);
3429 * Without the memory barrier there is a small possiblity that
3430 * netdev_tx_sent_queue will miss the update and cause the queue to
3431 * be stopped forever
3435 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3438 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3439 netif_schedule_queue(dev_queue);
3444 * netdev_completed_queue - report bytes and packets completed by device
3445 * @dev: network device
3446 * @pkts: actual number of packets sent over the medium
3447 * @bytes: actual number of bytes sent over the medium
3449 * Report the number of bytes and packets transmitted by the network device
3450 * hardware queue over the physical medium, @bytes must exactly match the
3451 * @bytes amount passed to netdev_sent_queue()
3453 static inline void netdev_completed_queue(struct net_device *dev,
3454 unsigned int pkts, unsigned int bytes)
3456 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3459 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3462 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3468 * netdev_reset_queue - reset the packets and bytes count of a network device
3469 * @dev_queue: network device
3471 * Reset the bytes and packet count of a network device and clear the
3472 * software flow control OFF bit for this network device
3474 static inline void netdev_reset_queue(struct net_device *dev_queue)
3476 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3480 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3481 * @dev: network device
3482 * @queue_index: given tx queue index
3484 * Returns 0 if given tx queue index >= number of device tx queues,
3485 * otherwise returns the originally passed tx queue index.
3487 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3489 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3490 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3491 dev->name, queue_index,
3492 dev->real_num_tx_queues);
3500 * netif_running - test if up
3501 * @dev: network device
3503 * Test if the device has been brought up.
3505 static inline bool netif_running(const struct net_device *dev)
3507 return test_bit(__LINK_STATE_START, &dev->state);
3511 * Routines to manage the subqueues on a device. We only need start,
3512 * stop, and a check if it's stopped. All other device management is
3513 * done at the overall netdevice level.
3514 * Also test the device if we're multiqueue.
3518 * netif_start_subqueue - allow sending packets on subqueue
3519 * @dev: network device
3520 * @queue_index: sub queue index
3522 * Start individual transmit queue of a device with multiple transmit queues.
3524 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3526 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3528 netif_tx_start_queue(txq);
3532 * netif_stop_subqueue - stop sending packets on subqueue
3533 * @dev: network device
3534 * @queue_index: sub queue index
3536 * Stop individual transmit queue of a device with multiple transmit queues.
3538 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3540 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3541 netif_tx_stop_queue(txq);
3545 * netif_subqueue_stopped - test status of subqueue
3546 * @dev: network device
3547 * @queue_index: sub queue index
3549 * Check individual transmit queue of a device with multiple transmit queues.
3551 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3554 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3556 return netif_tx_queue_stopped(txq);
3559 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3560 struct sk_buff *skb)
3562 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3566 * netif_wake_subqueue - allow sending packets on subqueue
3567 * @dev: network device
3568 * @queue_index: sub queue index
3570 * Resume individual transmit queue of a device with multiple transmit queues.
3572 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3574 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3576 netif_tx_wake_queue(txq);
3580 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3582 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3583 u16 index, bool is_rxqs_map);
3586 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3587 * @j: CPU/Rx queue index
3588 * @mask: bitmask of all cpus/rx queues
3589 * @nr_bits: number of bits in the bitmask
3591 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3593 static inline bool netif_attr_test_mask(unsigned long j,
3594 const unsigned long *mask,
3595 unsigned int nr_bits)
3597 cpu_max_bits_warn(j, nr_bits);
3598 return test_bit(j, mask);
3602 * netif_attr_test_online - Test for online CPU/Rx queue
3603 * @j: CPU/Rx queue index
3604 * @online_mask: bitmask for CPUs/Rx queues that are online
3605 * @nr_bits: number of bits in the bitmask
3607 * Returns true if a CPU/Rx queue is online.
3609 static inline bool netif_attr_test_online(unsigned long j,
3610 const unsigned long *online_mask,
3611 unsigned int nr_bits)
3613 cpu_max_bits_warn(j, nr_bits);
3616 return test_bit(j, online_mask);
3618 return (j < nr_bits);
3622 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3623 * @n: CPU/Rx queue index
3624 * @srcp: the cpumask/Rx queue mask pointer
3625 * @nr_bits: number of bits in the bitmask
3627 * Returns >= nr_bits if no further CPUs/Rx queues set.
3629 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3630 unsigned int nr_bits)
3632 /* -1 is a legal arg here. */
3634 cpu_max_bits_warn(n, nr_bits);
3637 return find_next_bit(srcp, nr_bits, n + 1);
3643 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3644 * @n: CPU/Rx queue index
3645 * @src1p: the first CPUs/Rx queues mask pointer
3646 * @src2p: the second CPUs/Rx queues mask pointer
3647 * @nr_bits: number of bits in the bitmask
3649 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3651 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3652 const unsigned long *src2p,
3653 unsigned int nr_bits)
3655 /* -1 is a legal arg here. */
3657 cpu_max_bits_warn(n, nr_bits);
3660 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3662 return find_next_bit(src1p, nr_bits, n + 1);
3664 return find_next_bit(src2p, nr_bits, n + 1);
3669 static inline int netif_set_xps_queue(struct net_device *dev,
3670 const struct cpumask *mask,
3676 static inline int __netif_set_xps_queue(struct net_device *dev,
3677 const unsigned long *mask,
3678 u16 index, bool is_rxqs_map)
3685 * netif_is_multiqueue - test if device has multiple transmit queues
3686 * @dev: network device
3688 * Check if device has multiple transmit queues
3690 static inline bool netif_is_multiqueue(const struct net_device *dev)
3692 return dev->num_tx_queues > 1;
3695 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3698 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3700 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3703 dev->real_num_rx_queues = rxqs;
3708 static inline struct netdev_rx_queue *
3709 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3711 return dev->_rx + rxq;
3715 static inline unsigned int get_netdev_rx_queue_index(
3716 struct netdev_rx_queue *queue)
3718 struct net_device *dev = queue->dev;
3719 int index = queue - dev->_rx;
3721 BUG_ON(index >= dev->num_rx_queues);
3726 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3727 int netif_get_num_default_rss_queues(void);
3729 enum skb_free_reason {
3730 SKB_REASON_CONSUMED,
3734 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3735 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3738 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3739 * interrupt context or with hardware interrupts being disabled.
3740 * (in_irq() || irqs_disabled())
3742 * We provide four helpers that can be used in following contexts :
3744 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3745 * replacing kfree_skb(skb)
3747 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3748 * Typically used in place of consume_skb(skb) in TX completion path
3750 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3751 * replacing kfree_skb(skb)
3753 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3754 * and consumed a packet. Used in place of consume_skb(skb)
3756 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3758 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3761 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3763 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3766 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3768 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3771 static inline void dev_consume_skb_any(struct sk_buff *skb)
3773 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3776 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3777 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3778 int netif_rx(struct sk_buff *skb);
3779 int netif_rx_ni(struct sk_buff *skb);
3780 int netif_receive_skb(struct sk_buff *skb);
3781 int netif_receive_skb_core(struct sk_buff *skb);
3782 void netif_receive_skb_list(struct list_head *head);
3783 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3784 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3785 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3786 gro_result_t napi_gro_frags(struct napi_struct *napi);
3787 struct packet_offload *gro_find_receive_by_type(__be16 type);
3788 struct packet_offload *gro_find_complete_by_type(__be16 type);
3790 static inline void napi_free_frags(struct napi_struct *napi)
3792 kfree_skb(napi->skb);
3796 bool netdev_is_rx_handler_busy(struct net_device *dev);
3797 int netdev_rx_handler_register(struct net_device *dev,
3798 rx_handler_func_t *rx_handler,
3799 void *rx_handler_data);
3800 void netdev_rx_handler_unregister(struct net_device *dev);
3802 bool dev_valid_name(const char *name);
3803 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3804 bool *need_copyout);
3805 int dev_ifconf(struct net *net, struct ifconf *, int);
3806 int dev_ethtool(struct net *net, struct ifreq *);
3807 unsigned int dev_get_flags(const struct net_device *);
3808 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3809 struct netlink_ext_ack *extack);
3810 int dev_change_flags(struct net_device *dev, unsigned int flags,
3811 struct netlink_ext_ack *extack);
3812 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3813 unsigned int gchanges);
3814 int dev_change_name(struct net_device *, const char *);
3815 int dev_set_alias(struct net_device *, const char *, size_t);
3816 int dev_get_alias(const struct net_device *, char *, size_t);
3817 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3818 int __dev_set_mtu(struct net_device *, int);
3819 int dev_validate_mtu(struct net_device *dev, int mtu,
3820 struct netlink_ext_ack *extack);
3821 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3822 struct netlink_ext_ack *extack);
3823 int dev_set_mtu(struct net_device *, int);
3824 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3825 void dev_set_group(struct net_device *, int);
3826 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3827 struct netlink_ext_ack *extack);
3828 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3829 struct netlink_ext_ack *extack);
3830 int dev_change_carrier(struct net_device *, bool new_carrier);
3831 int dev_get_phys_port_id(struct net_device *dev,
3832 struct netdev_phys_item_id *ppid);
3833 int dev_get_phys_port_name(struct net_device *dev,
3834 char *name, size_t len);
3835 int dev_get_port_parent_id(struct net_device *dev,
3836 struct netdev_phys_item_id *ppid, bool recurse);
3837 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3838 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3839 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3840 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3842 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3843 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3844 struct netdev_queue *txq, int *ret);
3846 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3847 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3848 int fd, int expected_fd, u32 flags);
3849 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3850 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3852 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3854 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3855 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3856 bool is_skb_forwardable(const struct net_device *dev,
3857 const struct sk_buff *skb);
3859 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3860 struct sk_buff *skb)
3862 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3863 unlikely(!is_skb_forwardable(dev, skb))) {
3864 atomic_long_inc(&dev->rx_dropped);
3869 skb_scrub_packet(skb, true);
3874 bool dev_nit_active(struct net_device *dev);
3875 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3877 extern int netdev_budget;
3878 extern unsigned int netdev_budget_usecs;
3880 /* Called by rtnetlink.c:rtnl_unlock() */
3881 void netdev_run_todo(void);
3884 * dev_put - release reference to device
3885 * @dev: network device
3887 * Release reference to device to allow it to be freed.
3889 static inline void dev_put(struct net_device *dev)
3891 this_cpu_dec(*dev->pcpu_refcnt);
3895 * dev_hold - get reference to device
3896 * @dev: network device
3898 * Hold reference to device to keep it from being freed.
3900 static inline void dev_hold(struct net_device *dev)
3902 this_cpu_inc(*dev->pcpu_refcnt);
3905 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3906 * and _off may be called from IRQ context, but it is caller
3907 * who is responsible for serialization of these calls.
3909 * The name carrier is inappropriate, these functions should really be
3910 * called netif_lowerlayer_*() because they represent the state of any
3911 * kind of lower layer not just hardware media.
3914 void linkwatch_init_dev(struct net_device *dev);
3915 void linkwatch_fire_event(struct net_device *dev);
3916 void linkwatch_forget_dev(struct net_device *dev);
3919 * netif_carrier_ok - test if carrier present
3920 * @dev: network device
3922 * Check if carrier is present on device
3924 static inline bool netif_carrier_ok(const struct net_device *dev)
3926 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3929 unsigned long dev_trans_start(struct net_device *dev);
3931 void __netdev_watchdog_up(struct net_device *dev);
3933 void netif_carrier_on(struct net_device *dev);
3935 void netif_carrier_off(struct net_device *dev);
3938 * netif_dormant_on - mark device as dormant.
3939 * @dev: network device
3941 * Mark device as dormant (as per RFC2863).
3943 * The dormant state indicates that the relevant interface is not
3944 * actually in a condition to pass packets (i.e., it is not 'up') but is
3945 * in a "pending" state, waiting for some external event. For "on-
3946 * demand" interfaces, this new state identifies the situation where the
3947 * interface is waiting for events to place it in the up state.
3949 static inline void netif_dormant_on(struct net_device *dev)
3951 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3952 linkwatch_fire_event(dev);
3956 * netif_dormant_off - set device as not dormant.
3957 * @dev: network device
3959 * Device is not in dormant state.
3961 static inline void netif_dormant_off(struct net_device *dev)
3963 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3964 linkwatch_fire_event(dev);
3968 * netif_dormant - test if device is dormant
3969 * @dev: network device
3971 * Check if device is dormant.
3973 static inline bool netif_dormant(const struct net_device *dev)
3975 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3980 * netif_testing_on - mark device as under test.
3981 * @dev: network device
3983 * Mark device as under test (as per RFC2863).
3985 * The testing state indicates that some test(s) must be performed on
3986 * the interface. After completion, of the test, the interface state
3987 * will change to up, dormant, or down, as appropriate.
3989 static inline void netif_testing_on(struct net_device *dev)
3991 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
3992 linkwatch_fire_event(dev);
3996 * netif_testing_off - set device as not under test.
3997 * @dev: network device
3999 * Device is not in testing state.
4001 static inline void netif_testing_off(struct net_device *dev)
4003 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4004 linkwatch_fire_event(dev);
4008 * netif_testing - test if device is under test
4009 * @dev: network device
4011 * Check if device is under test
4013 static inline bool netif_testing(const struct net_device *dev)
4015 return test_bit(__LINK_STATE_TESTING, &dev->state);
4020 * netif_oper_up - test if device is operational
4021 * @dev: network device
4023 * Check if carrier is operational
4025 static inline bool netif_oper_up(const struct net_device *dev)
4027 return (dev->operstate == IF_OPER_UP ||
4028 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4032 * netif_device_present - is device available or removed
4033 * @dev: network device
4035 * Check if device has not been removed from system.
4037 static inline bool netif_device_present(struct net_device *dev)
4039 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4042 void netif_device_detach(struct net_device *dev);
4044 void netif_device_attach(struct net_device *dev);
4047 * Network interface message level settings
4052 NETIF_MSG_PROBE_BIT,
4054 NETIF_MSG_TIMER_BIT,
4055 NETIF_MSG_IFDOWN_BIT,
4057 NETIF_MSG_RX_ERR_BIT,
4058 NETIF_MSG_TX_ERR_BIT,
4059 NETIF_MSG_TX_QUEUED_BIT,
4061 NETIF_MSG_TX_DONE_BIT,
4062 NETIF_MSG_RX_STATUS_BIT,
4063 NETIF_MSG_PKTDATA_BIT,
4067 /* When you add a new bit above, update netif_msg_class_names array
4068 * in net/ethtool/common.c
4070 NETIF_MSG_CLASS_COUNT,
4072 /* Both ethtool_ops interface and internal driver implementation use u32 */
4073 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4075 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4076 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4078 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4079 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4080 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4081 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4082 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4083 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4084 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4085 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4086 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4087 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4088 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4089 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4090 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4091 #define NETIF_MSG_HW __NETIF_MSG(HW)
4092 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4094 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4095 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4096 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4097 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4098 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4099 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4100 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4101 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4102 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4103 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4104 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4105 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4106 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4107 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4108 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4110 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4113 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4114 return default_msg_enable_bits;
4115 if (debug_value == 0) /* no output */
4117 /* set low N bits */
4118 return (1U << debug_value) - 1;
4121 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4123 spin_lock(&txq->_xmit_lock);
4124 txq->xmit_lock_owner = cpu;
4127 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4129 __acquire(&txq->_xmit_lock);
4133 static inline void __netif_tx_release(struct netdev_queue *txq)
4135 __release(&txq->_xmit_lock);
4138 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4140 spin_lock_bh(&txq->_xmit_lock);
4141 txq->xmit_lock_owner = smp_processor_id();
4144 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4146 bool ok = spin_trylock(&txq->_xmit_lock);
4148 txq->xmit_lock_owner = smp_processor_id();
4152 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4154 txq->xmit_lock_owner = -1;
4155 spin_unlock(&txq->_xmit_lock);
4158 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4160 txq->xmit_lock_owner = -1;
4161 spin_unlock_bh(&txq->_xmit_lock);
4164 static inline void txq_trans_update(struct netdev_queue *txq)
4166 if (txq->xmit_lock_owner != -1)
4167 txq->trans_start = jiffies;
4170 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4171 static inline void netif_trans_update(struct net_device *dev)
4173 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4175 if (txq->trans_start != jiffies)
4176 txq->trans_start = jiffies;
4180 * netif_tx_lock - grab network device transmit lock
4181 * @dev: network device
4183 * Get network device transmit lock
4185 static inline void netif_tx_lock(struct net_device *dev)
4190 spin_lock(&dev->tx_global_lock);
4191 cpu = smp_processor_id();
4192 for (i = 0; i < dev->num_tx_queues; i++) {
4193 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4195 /* We are the only thread of execution doing a
4196 * freeze, but we have to grab the _xmit_lock in
4197 * order to synchronize with threads which are in
4198 * the ->hard_start_xmit() handler and already
4199 * checked the frozen bit.
4201 __netif_tx_lock(txq, cpu);
4202 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4203 __netif_tx_unlock(txq);
4207 static inline void netif_tx_lock_bh(struct net_device *dev)
4213 static inline void netif_tx_unlock(struct net_device *dev)
4217 for (i = 0; i < dev->num_tx_queues; i++) {
4218 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4220 /* No need to grab the _xmit_lock here. If the
4221 * queue is not stopped for another reason, we
4224 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4225 netif_schedule_queue(txq);
4227 spin_unlock(&dev->tx_global_lock);
4230 static inline void netif_tx_unlock_bh(struct net_device *dev)
4232 netif_tx_unlock(dev);
4236 #define HARD_TX_LOCK(dev, txq, cpu) { \
4237 if ((dev->features & NETIF_F_LLTX) == 0) { \
4238 __netif_tx_lock(txq, cpu); \
4240 __netif_tx_acquire(txq); \
4244 #define HARD_TX_TRYLOCK(dev, txq) \
4245 (((dev->features & NETIF_F_LLTX) == 0) ? \
4246 __netif_tx_trylock(txq) : \
4247 __netif_tx_acquire(txq))
4249 #define HARD_TX_UNLOCK(dev, txq) { \
4250 if ((dev->features & NETIF_F_LLTX) == 0) { \
4251 __netif_tx_unlock(txq); \
4253 __netif_tx_release(txq); \
4257 static inline void netif_tx_disable(struct net_device *dev)
4263 cpu = smp_processor_id();
4264 for (i = 0; i < dev->num_tx_queues; i++) {
4265 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4267 __netif_tx_lock(txq, cpu);
4268 netif_tx_stop_queue(txq);
4269 __netif_tx_unlock(txq);
4274 static inline void netif_addr_lock(struct net_device *dev)
4276 unsigned char nest_level = 0;
4278 #ifdef CONFIG_LOCKDEP
4279 nest_level = dev->nested_level;
4281 spin_lock_nested(&dev->addr_list_lock, nest_level);
4284 static inline void netif_addr_lock_bh(struct net_device *dev)
4286 unsigned char nest_level = 0;
4288 #ifdef CONFIG_LOCKDEP
4289 nest_level = dev->nested_level;
4292 spin_lock_nested(&dev->addr_list_lock, nest_level);
4295 static inline void netif_addr_unlock(struct net_device *dev)
4297 spin_unlock(&dev->addr_list_lock);
4300 static inline void netif_addr_unlock_bh(struct net_device *dev)
4302 spin_unlock_bh(&dev->addr_list_lock);
4306 * dev_addrs walker. Should be used only for read access. Call with
4307 * rcu_read_lock held.
4309 #define for_each_dev_addr(dev, ha) \
4310 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4312 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4314 void ether_setup(struct net_device *dev);
4316 /* Support for loadable net-drivers */
4317 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4318 unsigned char name_assign_type,
4319 void (*setup)(struct net_device *),
4320 unsigned int txqs, unsigned int rxqs);
4321 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4322 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4324 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4325 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4328 int register_netdev(struct net_device *dev);
4329 void unregister_netdev(struct net_device *dev);
4331 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4333 /* General hardware address lists handling functions */
4334 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4335 struct netdev_hw_addr_list *from_list, int addr_len);
4336 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4337 struct netdev_hw_addr_list *from_list, int addr_len);
4338 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4339 struct net_device *dev,
4340 int (*sync)(struct net_device *, const unsigned char *),
4341 int (*unsync)(struct net_device *,
4342 const unsigned char *));
4343 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4344 struct net_device *dev,
4345 int (*sync)(struct net_device *,
4346 const unsigned char *, int),
4347 int (*unsync)(struct net_device *,
4348 const unsigned char *, int));
4349 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4350 struct net_device *dev,
4351 int (*unsync)(struct net_device *,
4352 const unsigned char *, int));
4353 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4354 struct net_device *dev,
4355 int (*unsync)(struct net_device *,
4356 const unsigned char *));
4357 void __hw_addr_init(struct netdev_hw_addr_list *list);
4359 /* Functions used for device addresses handling */
4360 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4361 unsigned char addr_type);
4362 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4363 unsigned char addr_type);
4364 void dev_addr_flush(struct net_device *dev);
4365 int dev_addr_init(struct net_device *dev);
4367 /* Functions used for unicast addresses handling */
4368 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4369 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4370 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4371 int dev_uc_sync(struct net_device *to, struct net_device *from);
4372 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4373 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4374 void dev_uc_flush(struct net_device *dev);
4375 void dev_uc_init(struct net_device *dev);
4378 * __dev_uc_sync - Synchonize device's unicast list
4379 * @dev: device to sync
4380 * @sync: function to call if address should be added
4381 * @unsync: function to call if address should be removed
4383 * Add newly added addresses to the interface, and release
4384 * addresses that have been deleted.
4386 static inline int __dev_uc_sync(struct net_device *dev,
4387 int (*sync)(struct net_device *,
4388 const unsigned char *),
4389 int (*unsync)(struct net_device *,
4390 const unsigned char *))
4392 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4396 * __dev_uc_unsync - Remove synchronized addresses from device
4397 * @dev: device to sync
4398 * @unsync: function to call if address should be removed
4400 * Remove all addresses that were added to the device by dev_uc_sync().
4402 static inline void __dev_uc_unsync(struct net_device *dev,
4403 int (*unsync)(struct net_device *,
4404 const unsigned char *))
4406 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4409 /* Functions used for multicast addresses handling */
4410 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4411 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4412 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4413 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4414 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4415 int dev_mc_sync(struct net_device *to, struct net_device *from);
4416 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4417 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4418 void dev_mc_flush(struct net_device *dev);
4419 void dev_mc_init(struct net_device *dev);
4422 * __dev_mc_sync - Synchonize device's multicast list
4423 * @dev: device to sync
4424 * @sync: function to call if address should be added
4425 * @unsync: function to call if address should be removed
4427 * Add newly added addresses to the interface, and release
4428 * addresses that have been deleted.
4430 static inline int __dev_mc_sync(struct net_device *dev,
4431 int (*sync)(struct net_device *,
4432 const unsigned char *),
4433 int (*unsync)(struct net_device *,
4434 const unsigned char *))
4436 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4440 * __dev_mc_unsync - Remove synchronized addresses from device
4441 * @dev: device to sync
4442 * @unsync: function to call if address should be removed
4444 * Remove all addresses that were added to the device by dev_mc_sync().
4446 static inline void __dev_mc_unsync(struct net_device *dev,
4447 int (*unsync)(struct net_device *,
4448 const unsigned char *))
4450 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4453 /* Functions used for secondary unicast and multicast support */
4454 void dev_set_rx_mode(struct net_device *dev);
4455 void __dev_set_rx_mode(struct net_device *dev);
4456 int dev_set_promiscuity(struct net_device *dev, int inc);
4457 int dev_set_allmulti(struct net_device *dev, int inc);
4458 void netdev_state_change(struct net_device *dev);
4459 void netdev_notify_peers(struct net_device *dev);
4460 void netdev_features_change(struct net_device *dev);
4461 /* Load a device via the kmod */
4462 void dev_load(struct net *net, const char *name);
4463 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4464 struct rtnl_link_stats64 *storage);
4465 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4466 const struct net_device_stats *netdev_stats);
4468 extern int netdev_max_backlog;
4469 extern int netdev_tstamp_prequeue;
4470 extern int weight_p;
4471 extern int dev_weight_rx_bias;
4472 extern int dev_weight_tx_bias;
4473 extern int dev_rx_weight;
4474 extern int dev_tx_weight;
4475 extern int gro_normal_batch;
4478 NESTED_SYNC_IMM_BIT,
4479 NESTED_SYNC_TODO_BIT,
4482 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4483 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4485 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4486 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4488 struct netdev_nested_priv {
4489 unsigned char flags;
4493 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4494 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4495 struct list_head **iter);
4496 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4497 struct list_head **iter);
4499 #ifdef CONFIG_LOCKDEP
4500 static LIST_HEAD(net_unlink_list);
4502 static inline void net_unlink_todo(struct net_device *dev)
4504 if (list_empty(&dev->unlink_list))
4505 list_add_tail(&dev->unlink_list, &net_unlink_list);
4509 /* iterate through upper list, must be called under RCU read lock */
4510 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4511 for (iter = &(dev)->adj_list.upper, \
4512 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4514 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4516 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4517 int (*fn)(struct net_device *upper_dev,
4518 struct netdev_nested_priv *priv),
4519 struct netdev_nested_priv *priv);
4521 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4522 struct net_device *upper_dev);
4524 bool netdev_has_any_upper_dev(struct net_device *dev);
4526 void *netdev_lower_get_next_private(struct net_device *dev,
4527 struct list_head **iter);
4528 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4529 struct list_head **iter);
4531 #define netdev_for_each_lower_private(dev, priv, iter) \
4532 for (iter = (dev)->adj_list.lower.next, \
4533 priv = netdev_lower_get_next_private(dev, &(iter)); \
4535 priv = netdev_lower_get_next_private(dev, &(iter)))
4537 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4538 for (iter = &(dev)->adj_list.lower, \
4539 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4541 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4543 void *netdev_lower_get_next(struct net_device *dev,
4544 struct list_head **iter);
4546 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4547 for (iter = (dev)->adj_list.lower.next, \
4548 ldev = netdev_lower_get_next(dev, &(iter)); \
4550 ldev = netdev_lower_get_next(dev, &(iter)))
4552 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4553 struct list_head **iter);
4554 int netdev_walk_all_lower_dev(struct net_device *dev,
4555 int (*fn)(struct net_device *lower_dev,
4556 struct netdev_nested_priv *priv),
4557 struct netdev_nested_priv *priv);
4558 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4559 int (*fn)(struct net_device *lower_dev,
4560 struct netdev_nested_priv *priv),
4561 struct netdev_nested_priv *priv);
4563 void *netdev_adjacent_get_private(struct list_head *adj_list);
4564 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4565 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4566 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4567 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4568 struct netlink_ext_ack *extack);
4569 int netdev_master_upper_dev_link(struct net_device *dev,
4570 struct net_device *upper_dev,
4571 void *upper_priv, void *upper_info,
4572 struct netlink_ext_ack *extack);
4573 void netdev_upper_dev_unlink(struct net_device *dev,
4574 struct net_device *upper_dev);
4575 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4576 struct net_device *new_dev,
4577 struct net_device *dev,
4578 struct netlink_ext_ack *extack);
4579 void netdev_adjacent_change_commit(struct net_device *old_dev,
4580 struct net_device *new_dev,
4581 struct net_device *dev);
4582 void netdev_adjacent_change_abort(struct net_device *old_dev,
4583 struct net_device *new_dev,
4584 struct net_device *dev);
4585 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4586 void *netdev_lower_dev_get_private(struct net_device *dev,
4587 struct net_device *lower_dev);
4588 void netdev_lower_state_changed(struct net_device *lower_dev,
4589 void *lower_state_info);
4591 /* RSS keys are 40 or 52 bytes long */
4592 #define NETDEV_RSS_KEY_LEN 52
4593 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4594 void netdev_rss_key_fill(void *buffer, size_t len);
4596 int skb_checksum_help(struct sk_buff *skb);
4597 int skb_crc32c_csum_help(struct sk_buff *skb);
4598 int skb_csum_hwoffload_help(struct sk_buff *skb,
4599 const netdev_features_t features);
4601 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4602 netdev_features_t features, bool tx_path);
4603 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4604 netdev_features_t features);
4606 struct netdev_bonding_info {
4611 struct netdev_notifier_bonding_info {
4612 struct netdev_notifier_info info; /* must be first */
4613 struct netdev_bonding_info bonding_info;
4616 void netdev_bonding_info_change(struct net_device *dev,
4617 struct netdev_bonding_info *bonding_info);
4619 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4620 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4622 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4629 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4631 return __skb_gso_segment(skb, features, true);
4633 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4635 static inline bool can_checksum_protocol(netdev_features_t features,
4638 if (protocol == htons(ETH_P_FCOE))
4639 return !!(features & NETIF_F_FCOE_CRC);
4641 /* Assume this is an IP checksum (not SCTP CRC) */
4643 if (features & NETIF_F_HW_CSUM) {
4644 /* Can checksum everything */
4649 case htons(ETH_P_IP):
4650 return !!(features & NETIF_F_IP_CSUM);
4651 case htons(ETH_P_IPV6):
4652 return !!(features & NETIF_F_IPV6_CSUM);
4659 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4661 static inline void netdev_rx_csum_fault(struct net_device *dev,
4662 struct sk_buff *skb)
4666 /* rx skb timestamps */
4667 void net_enable_timestamp(void);
4668 void net_disable_timestamp(void);
4670 #ifdef CONFIG_PROC_FS
4671 int __init dev_proc_init(void);
4673 #define dev_proc_init() 0
4676 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4677 struct sk_buff *skb, struct net_device *dev,
4680 __this_cpu_write(softnet_data.xmit.more, more);
4681 return ops->ndo_start_xmit(skb, dev);
4684 static inline bool netdev_xmit_more(void)
4686 return __this_cpu_read(softnet_data.xmit.more);
4689 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4690 struct netdev_queue *txq, bool more)
4692 const struct net_device_ops *ops = dev->netdev_ops;
4695 rc = __netdev_start_xmit(ops, skb, dev, more);
4696 if (rc == NETDEV_TX_OK)
4697 txq_trans_update(txq);
4702 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4704 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4707 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4709 return netdev_class_create_file_ns(class_attr, NULL);
4712 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4714 netdev_class_remove_file_ns(class_attr, NULL);
4717 extern const struct kobj_ns_type_operations net_ns_type_operations;
4719 const char *netdev_drivername(const struct net_device *dev);
4721 void linkwatch_run_queue(void);
4723 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4724 netdev_features_t f2)
4726 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4727 if (f1 & NETIF_F_HW_CSUM)
4728 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4730 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4736 static inline netdev_features_t netdev_get_wanted_features(
4737 struct net_device *dev)
4739 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4741 netdev_features_t netdev_increment_features(netdev_features_t all,
4742 netdev_features_t one, netdev_features_t mask);
4744 /* Allow TSO being used on stacked device :
4745 * Performing the GSO segmentation before last device
4746 * is a performance improvement.
4748 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4749 netdev_features_t mask)
4751 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4754 int __netdev_update_features(struct net_device *dev);
4755 void netdev_update_features(struct net_device *dev);
4756 void netdev_change_features(struct net_device *dev);
4758 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4759 struct net_device *dev);
4761 netdev_features_t passthru_features_check(struct sk_buff *skb,
4762 struct net_device *dev,
4763 netdev_features_t features);
4764 netdev_features_t netif_skb_features(struct sk_buff *skb);
4766 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4768 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4770 /* check flags correspondence */
4771 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4772 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4773 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4774 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4775 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4776 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4777 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4778 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4779 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4780 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4781 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4782 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4783 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4784 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4785 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4786 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4787 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4788 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4789 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4791 return (features & feature) == feature;
4794 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4796 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4797 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4800 static inline bool netif_needs_gso(struct sk_buff *skb,
4801 netdev_features_t features)
4803 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4804 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4805 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4808 static inline void netif_set_gso_max_size(struct net_device *dev,
4811 dev->gso_max_size = size;
4814 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4815 int pulled_hlen, u16 mac_offset,
4818 skb->protocol = protocol;
4819 skb->encapsulation = 1;
4820 skb_push(skb, pulled_hlen);
4821 skb_reset_transport_header(skb);
4822 skb->mac_header = mac_offset;
4823 skb->network_header = skb->mac_header + mac_len;
4824 skb->mac_len = mac_len;
4827 static inline bool netif_is_macsec(const struct net_device *dev)
4829 return dev->priv_flags & IFF_MACSEC;
4832 static inline bool netif_is_macvlan(const struct net_device *dev)
4834 return dev->priv_flags & IFF_MACVLAN;
4837 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4839 return dev->priv_flags & IFF_MACVLAN_PORT;
4842 static inline bool netif_is_bond_master(const struct net_device *dev)
4844 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4847 static inline bool netif_is_bond_slave(const struct net_device *dev)
4849 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4852 static inline bool netif_supports_nofcs(struct net_device *dev)
4854 return dev->priv_flags & IFF_SUPP_NOFCS;
4857 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4859 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4862 static inline bool netif_is_l3_master(const struct net_device *dev)
4864 return dev->priv_flags & IFF_L3MDEV_MASTER;
4867 static inline bool netif_is_l3_slave(const struct net_device *dev)
4869 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4872 static inline bool netif_is_bridge_master(const struct net_device *dev)
4874 return dev->priv_flags & IFF_EBRIDGE;
4877 static inline bool netif_is_bridge_port(const struct net_device *dev)
4879 return dev->priv_flags & IFF_BRIDGE_PORT;
4882 static inline bool netif_is_ovs_master(const struct net_device *dev)
4884 return dev->priv_flags & IFF_OPENVSWITCH;
4887 static inline bool netif_is_ovs_port(const struct net_device *dev)
4889 return dev->priv_flags & IFF_OVS_DATAPATH;
4892 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4894 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4897 static inline bool netif_is_team_master(const struct net_device *dev)
4899 return dev->priv_flags & IFF_TEAM;
4902 static inline bool netif_is_team_port(const struct net_device *dev)
4904 return dev->priv_flags & IFF_TEAM_PORT;
4907 static inline bool netif_is_lag_master(const struct net_device *dev)
4909 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4912 static inline bool netif_is_lag_port(const struct net_device *dev)
4914 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4917 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4919 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4922 static inline bool netif_is_failover(const struct net_device *dev)
4924 return dev->priv_flags & IFF_FAILOVER;
4927 static inline bool netif_is_failover_slave(const struct net_device *dev)
4929 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4932 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4933 static inline void netif_keep_dst(struct net_device *dev)
4935 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4938 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4939 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4941 /* TODO: reserve and use an additional IFF bit, if we get more users */
4942 return dev->priv_flags & IFF_MACSEC;
4945 extern struct pernet_operations __net_initdata loopback_net_ops;
4947 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4949 /* netdev_printk helpers, similar to dev_printk */
4951 static inline const char *netdev_name(const struct net_device *dev)
4953 if (!dev->name[0] || strchr(dev->name, '%'))
4954 return "(unnamed net_device)";
4958 static inline bool netdev_unregistering(const struct net_device *dev)
4960 return dev->reg_state == NETREG_UNREGISTERING;
4963 static inline const char *netdev_reg_state(const struct net_device *dev)
4965 switch (dev->reg_state) {
4966 case NETREG_UNINITIALIZED: return " (uninitialized)";
4967 case NETREG_REGISTERED: return "";
4968 case NETREG_UNREGISTERING: return " (unregistering)";
4969 case NETREG_UNREGISTERED: return " (unregistered)";
4970 case NETREG_RELEASED: return " (released)";
4971 case NETREG_DUMMY: return " (dummy)";
4974 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4975 return " (unknown)";
4978 __printf(3, 4) __cold
4979 void netdev_printk(const char *level, const struct net_device *dev,
4980 const char *format, ...);
4981 __printf(2, 3) __cold
4982 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4983 __printf(2, 3) __cold
4984 void netdev_alert(const struct net_device *dev, const char *format, ...);
4985 __printf(2, 3) __cold
4986 void netdev_crit(const struct net_device *dev, const char *format, ...);
4987 __printf(2, 3) __cold
4988 void netdev_err(const struct net_device *dev, const char *format, ...);
4989 __printf(2, 3) __cold
4990 void netdev_warn(const struct net_device *dev, const char *format, ...);
4991 __printf(2, 3) __cold
4992 void netdev_notice(const struct net_device *dev, const char *format, ...);
4993 __printf(2, 3) __cold
4994 void netdev_info(const struct net_device *dev, const char *format, ...);
4996 #define netdev_level_once(level, dev, fmt, ...) \
4998 static bool __print_once __read_mostly; \
5000 if (!__print_once) { \
5001 __print_once = true; \
5002 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5006 #define netdev_emerg_once(dev, fmt, ...) \
5007 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5008 #define netdev_alert_once(dev, fmt, ...) \
5009 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5010 #define netdev_crit_once(dev, fmt, ...) \
5011 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5012 #define netdev_err_once(dev, fmt, ...) \
5013 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5014 #define netdev_warn_once(dev, fmt, ...) \
5015 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5016 #define netdev_notice_once(dev, fmt, ...) \
5017 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5018 #define netdev_info_once(dev, fmt, ...) \
5019 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5021 #define MODULE_ALIAS_NETDEV(device) \
5022 MODULE_ALIAS("netdev-" device)
5024 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5025 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5026 #define netdev_dbg(__dev, format, args...) \
5028 dynamic_netdev_dbg(__dev, format, ##args); \
5030 #elif defined(DEBUG)
5031 #define netdev_dbg(__dev, format, args...) \
5032 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5034 #define netdev_dbg(__dev, format, args...) \
5037 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5041 #if defined(VERBOSE_DEBUG)
5042 #define netdev_vdbg netdev_dbg
5045 #define netdev_vdbg(dev, format, args...) \
5048 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5054 * netdev_WARN() acts like dev_printk(), but with the key difference
5055 * of using a WARN/WARN_ON to get the message out, including the
5056 * file/line information and a backtrace.
5058 #define netdev_WARN(dev, format, args...) \
5059 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5060 netdev_reg_state(dev), ##args)
5062 #define netdev_WARN_ONCE(dev, format, args...) \
5063 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5064 netdev_reg_state(dev), ##args)
5066 /* netif printk helpers, similar to netdev_printk */
5068 #define netif_printk(priv, type, level, dev, fmt, args...) \
5070 if (netif_msg_##type(priv)) \
5071 netdev_printk(level, (dev), fmt, ##args); \
5074 #define netif_level(level, priv, type, dev, fmt, args...) \
5076 if (netif_msg_##type(priv)) \
5077 netdev_##level(dev, fmt, ##args); \
5080 #define netif_emerg(priv, type, dev, fmt, args...) \
5081 netif_level(emerg, priv, type, dev, fmt, ##args)
5082 #define netif_alert(priv, type, dev, fmt, args...) \
5083 netif_level(alert, priv, type, dev, fmt, ##args)
5084 #define netif_crit(priv, type, dev, fmt, args...) \
5085 netif_level(crit, priv, type, dev, fmt, ##args)
5086 #define netif_err(priv, type, dev, fmt, args...) \
5087 netif_level(err, priv, type, dev, fmt, ##args)
5088 #define netif_warn(priv, type, dev, fmt, args...) \
5089 netif_level(warn, priv, type, dev, fmt, ##args)
5090 #define netif_notice(priv, type, dev, fmt, args...) \
5091 netif_level(notice, priv, type, dev, fmt, ##args)
5092 #define netif_info(priv, type, dev, fmt, args...) \
5093 netif_level(info, priv, type, dev, fmt, ##args)
5095 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5096 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5097 #define netif_dbg(priv, type, netdev, format, args...) \
5099 if (netif_msg_##type(priv)) \
5100 dynamic_netdev_dbg(netdev, format, ##args); \
5102 #elif defined(DEBUG)
5103 #define netif_dbg(priv, type, dev, format, args...) \
5104 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5106 #define netif_dbg(priv, type, dev, format, args...) \
5109 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5114 /* if @cond then downgrade to debug, else print at @level */
5115 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5118 netif_dbg(priv, type, netdev, fmt, ##args); \
5120 netif_ ## level(priv, type, netdev, fmt, ##args); \
5123 #if defined(VERBOSE_DEBUG)
5124 #define netif_vdbg netif_dbg
5126 #define netif_vdbg(priv, type, dev, format, args...) \
5129 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5135 * The list of packet types we will receive (as opposed to discard)
5136 * and the routines to invoke.
5138 * Why 16. Because with 16 the only overlap we get on a hash of the
5139 * low nibble of the protocol value is RARP/SNAP/X.25.
5153 #define PTYPE_HASH_SIZE (16)
5154 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5156 extern struct net_device *blackhole_netdev;
5158 #endif /* _LINUX_NETDEVICE_H */