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 synchronize_net(void);
74 void netdev_set_default_ethtool_ops(struct net_device *dev,
75 const struct ethtool_ops *ops);
77 /* Backlog congestion levels */
78 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
79 #define NET_RX_DROP 1 /* packet dropped */
81 #define MAX_NEST_DEV 8
84 * Transmit return codes: transmit return codes originate from three different
87 * - qdisc return codes
88 * - driver transmit return codes
91 * Drivers are allowed to return any one of those in their hard_start_xmit()
92 * function. Real network devices commonly used with qdiscs should only return
93 * the driver transmit return codes though - when qdiscs are used, the actual
94 * transmission happens asynchronously, so the value is not propagated to
95 * higher layers. Virtual network devices transmit synchronously; in this case
96 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
97 * others are propagated to higher layers.
100 /* qdisc ->enqueue() return codes. */
101 #define NET_XMIT_SUCCESS 0x00
102 #define NET_XMIT_DROP 0x01 /* skb dropped */
103 #define NET_XMIT_CN 0x02 /* congestion notification */
104 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
106 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
107 * indicates that the device will soon be dropping packets, or already drops
108 * some packets of the same priority; prompting us to send less aggressively. */
109 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
110 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
112 /* Driver transmit return codes */
113 #define NETDEV_TX_MASK 0xf0
116 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
117 NETDEV_TX_OK = 0x00, /* driver took care of packet */
118 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
120 typedef enum netdev_tx netdev_tx_t;
123 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
124 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
126 static inline bool dev_xmit_complete(int rc)
129 * Positive cases with an skb consumed by a driver:
130 * - successful transmission (rc == NETDEV_TX_OK)
131 * - error while transmitting (rc < 0)
132 * - error while queueing to a different device (rc & NET_XMIT_MASK)
134 if (likely(rc < NET_XMIT_MASK))
141 * Compute the worst-case header length according to the protocols
145 #if defined(CONFIG_HYPERV_NET)
146 # define LL_MAX_HEADER 128
147 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
148 # if defined(CONFIG_MAC80211_MESH)
149 # define LL_MAX_HEADER 128
151 # define LL_MAX_HEADER 96
154 # define LL_MAX_HEADER 32
157 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
158 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
159 #define MAX_HEADER LL_MAX_HEADER
161 #define MAX_HEADER (LL_MAX_HEADER + 48)
165 * Old network device statistics. Fields are native words
166 * (unsigned long) so they can be read and written atomically.
169 struct net_device_stats {
170 unsigned long rx_packets;
171 unsigned long tx_packets;
172 unsigned long rx_bytes;
173 unsigned long tx_bytes;
174 unsigned long rx_errors;
175 unsigned long tx_errors;
176 unsigned long rx_dropped;
177 unsigned long tx_dropped;
178 unsigned long multicast;
179 unsigned long collisions;
180 unsigned long rx_length_errors;
181 unsigned long rx_over_errors;
182 unsigned long rx_crc_errors;
183 unsigned long rx_frame_errors;
184 unsigned long rx_fifo_errors;
185 unsigned long rx_missed_errors;
186 unsigned long tx_aborted_errors;
187 unsigned long tx_carrier_errors;
188 unsigned long tx_fifo_errors;
189 unsigned long tx_heartbeat_errors;
190 unsigned long tx_window_errors;
191 unsigned long rx_compressed;
192 unsigned long tx_compressed;
196 #include <linux/cache.h>
197 #include <linux/skbuff.h>
200 #include <linux/static_key.h>
201 extern struct static_key_false rps_needed;
202 extern struct static_key_false rfs_needed;
209 struct netdev_hw_addr {
210 struct list_head list;
211 unsigned char addr[MAX_ADDR_LEN];
213 #define NETDEV_HW_ADDR_T_LAN 1
214 #define NETDEV_HW_ADDR_T_SAN 2
215 #define NETDEV_HW_ADDR_T_UNICAST 3
216 #define NETDEV_HW_ADDR_T_MULTICAST 4
221 struct rcu_head rcu_head;
224 struct netdev_hw_addr_list {
225 struct list_head list;
229 #define netdev_hw_addr_list_count(l) ((l)->count)
230 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
231 #define netdev_hw_addr_list_for_each(ha, l) \
232 list_for_each_entry(ha, &(l)->list, list)
234 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
235 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
236 #define netdev_for_each_uc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
239 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
240 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
241 #define netdev_for_each_mc_addr(ha, dev) \
242 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
248 /* cached hardware header; allow for machine alignment needs. */
249 #define HH_DATA_MOD 16
250 #define HH_DATA_OFF(__len) \
251 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
252 #define HH_DATA_ALIGN(__len) \
253 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
254 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
257 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
259 * dev->hard_header_len ? (dev->hard_header_len +
260 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
262 * We could use other alignment values, but we must maintain the
263 * relationship HH alignment <= LL alignment.
265 #define LL_RESERVED_SPACE(dev) \
266 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
267 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
268 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
271 int (*create) (struct sk_buff *skb, struct net_device *dev,
272 unsigned short type, const void *daddr,
273 const void *saddr, unsigned int len);
274 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
275 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
276 void (*cache_update)(struct hh_cache *hh,
277 const struct net_device *dev,
278 const unsigned char *haddr);
279 bool (*validate)(const char *ll_header, unsigned int len);
280 __be16 (*parse_protocol)(const struct sk_buff *skb);
283 /* These flag bits are private to the generic network queueing
284 * layer; they may not be explicitly referenced by any other
288 enum netdev_state_t {
290 __LINK_STATE_PRESENT,
291 __LINK_STATE_NOCARRIER,
292 __LINK_STATE_LINKWATCH_PENDING,
293 __LINK_STATE_DORMANT,
294 __LINK_STATE_TESTING,
299 * This structure holds boot-time configured netdevice settings. They
300 * are then used in the device probing.
302 struct netdev_boot_setup {
306 #define NETDEV_BOOT_SETUP_MAX 8
308 int __init netdev_boot_setup(char *str);
311 struct list_head list;
316 * size of gro hash buckets, must less than bit number of
317 * napi_struct::gro_bitmask
319 #define GRO_HASH_BUCKETS 8
322 * Structure for NAPI scheduling similar to tasklet but with weighting
325 /* The poll_list must only be managed by the entity which
326 * changes the state of the NAPI_STATE_SCHED bit. This means
327 * whoever atomically sets that bit can add this napi_struct
328 * to the per-CPU poll_list, and whoever clears that bit
329 * can remove from the list right before clearing the bit.
331 struct list_head poll_list;
335 int defer_hard_irqs_count;
336 unsigned long gro_bitmask;
337 int (*poll)(struct napi_struct *, int);
338 #ifdef CONFIG_NETPOLL
341 struct net_device *dev;
342 struct gro_list gro_hash[GRO_HASH_BUCKETS];
344 struct list_head rx_list; /* Pending GRO_NORMAL skbs */
345 int rx_count; /* length of rx_list */
346 struct hrtimer timer;
347 struct list_head dev_list;
348 struct hlist_node napi_hash_node;
349 unsigned int napi_id;
353 NAPI_STATE_SCHED, /* Poll is scheduled */
354 NAPI_STATE_MISSED, /* reschedule a napi */
355 NAPI_STATE_DISABLE, /* Disable pending */
356 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
357 NAPI_STATE_LISTED, /* NAPI added to system lists */
358 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
359 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
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_LISTED = BIT(NAPI_STATE_LISTED),
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_disable - prevent NAPI from scheduling
495 * Stop NAPI from being scheduled on this context.
496 * Waits till any outstanding processing completes.
498 void napi_disable(struct napi_struct *n);
501 * napi_enable - enable NAPI scheduling
504 * Resume NAPI from being scheduled on this context.
505 * Must be paired with napi_disable.
507 static inline void napi_enable(struct napi_struct *n)
509 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
510 smp_mb__before_atomic();
511 clear_bit(NAPI_STATE_SCHED, &n->state);
512 clear_bit(NAPI_STATE_NPSVC, &n->state);
516 * napi_synchronize - wait until NAPI is not running
519 * Wait until NAPI is done being scheduled on this context.
520 * Waits till any outstanding processing completes but
521 * does not disable future activations.
523 static inline void napi_synchronize(const struct napi_struct *n)
525 if (IS_ENABLED(CONFIG_SMP))
526 while (test_bit(NAPI_STATE_SCHED, &n->state))
533 * napi_if_scheduled_mark_missed - if napi is running, set the
537 * If napi is running, set the NAPIF_STATE_MISSED, and return true if
540 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
542 unsigned long val, new;
545 val = READ_ONCE(n->state);
546 if (val & NAPIF_STATE_DISABLE)
549 if (!(val & NAPIF_STATE_SCHED))
552 new = val | NAPIF_STATE_MISSED;
553 } while (cmpxchg(&n->state, val, new) != val);
558 enum netdev_queue_state_t {
559 __QUEUE_STATE_DRV_XOFF,
560 __QUEUE_STATE_STACK_XOFF,
561 __QUEUE_STATE_FROZEN,
564 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
565 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
566 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
568 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
569 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
571 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
575 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
576 * netif_tx_* functions below are used to manipulate this flag. The
577 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
578 * queue independently. The netif_xmit_*stopped functions below are called
579 * to check if the queue has been stopped by the driver or stack (either
580 * of the XOFF bits are set in the state). Drivers should not need to call
581 * netif_xmit*stopped functions, they should only be using netif_tx_*.
584 struct netdev_queue {
588 struct net_device *dev;
589 struct Qdisc __rcu *qdisc;
590 struct Qdisc *qdisc_sleeping;
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
597 unsigned long tx_maxrate;
599 * Number of TX timeouts for this queue
600 * (/sys/class/net/DEV/Q/trans_timeout)
602 unsigned long trans_timeout;
604 /* Subordinate device that the queue has been assigned to */
605 struct net_device *sb_dev;
606 #ifdef CONFIG_XDP_SOCKETS
607 struct xsk_buff_pool *pool;
612 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
615 * Time (in jiffies) of last Tx
617 unsigned long trans_start;
624 } ____cacheline_aligned_in_smp;
626 extern int sysctl_fb_tunnels_only_for_init_net;
627 extern int sysctl_devconf_inherit_init_net;
630 * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
631 * == 1 : For initns only
634 static inline bool net_has_fallback_tunnels(const struct net *net)
636 return !IS_ENABLED(CONFIG_SYSCTL) ||
637 !sysctl_fb_tunnels_only_for_init_net ||
638 (net == &init_net && sysctl_fb_tunnels_only_for_init_net == 1);
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
659 * This structure holds an RPS map which can be of variable length. The
660 * map is an array of CPUs.
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
670 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671 * tail pointer for that CPU's input queue at the time of last enqueue, and
672 * a hardware filter index.
674 struct rps_dev_flow {
677 unsigned int last_qtail;
679 #define RPS_NO_FILTER 0xffff
682 * The rps_dev_flow_table structure contains a table of flow mappings.
684 struct rps_dev_flow_table {
687 struct rps_dev_flow flows[];
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690 ((_num) * sizeof(struct rps_dev_flow)))
693 * The rps_sock_flow_table contains mappings of flows to the last CPU
694 * on which they were processed by the application (set in recvmsg).
695 * Each entry is a 32bit value. Upper part is the high-order bits
696 * of flow hash, lower part is CPU number.
697 * rps_cpu_mask is used to partition the space, depending on number of
698 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700 * meaning we use 32-6=26 bits for the hash.
702 struct rps_sock_flow_table {
705 u32 ents[] ____cacheline_aligned_in_smp;
707 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
709 #define RPS_NO_CPU 0xffff
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
718 unsigned int index = hash & table->mask;
719 u32 val = hash & ~rps_cpu_mask;
721 /* We only give a hint, preemption can change CPU under us */
722 val |= raw_smp_processor_id();
724 if (table->ents[index] != val)
725 table->ents[index] = val;
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
733 #endif /* CONFIG_RPS */
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
738 struct rps_map __rcu *rps_map;
739 struct rps_dev_flow_table __rcu *rps_flow_table;
742 struct net_device *dev;
743 struct xdp_rxq_info xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 struct xsk_buff_pool *pool;
747 } ____cacheline_aligned_in_smp;
750 * RX queue sysfs structures and functions.
752 struct rx_queue_attribute {
753 struct attribute attr;
754 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 ssize_t (*store)(struct netdev_rx_queue *queue,
756 const char *buf, size_t len);
761 * This structure holds an XPS map which can be of variable length. The
762 * map is an array of queues.
766 unsigned int alloc_len;
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772 - sizeof(struct xps_map)) / sizeof(u16))
775 * This structure holds all XPS maps for device. Maps are indexed by CPU.
777 struct xps_dev_maps {
779 struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
783 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
788 #endif /* CONFIG_XPS */
790 #define TC_MAX_QUEUE 16
791 #define TC_BITMASK 15
792 /* HW offloaded queuing disciplines txq count and offset maps */
793 struct netdev_tc_txq {
798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
800 * This structure is to hold information about the device
801 * configured to run FCoE protocol stack.
803 struct netdev_fcoe_hbainfo {
804 char manufacturer[64];
805 char serial_number[64];
806 char hardware_version[64];
807 char driver_version[64];
808 char optionrom_version[64];
809 char firmware_version[64];
811 char model_description[256];
815 #define MAX_PHYS_ITEM_ID_LEN 32
817 /* This structure holds a unique identifier to identify some
818 * physical item (port for example) used by a netdevice.
820 struct netdev_phys_item_id {
821 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 unsigned char id_len;
825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 struct netdev_phys_item_id *b)
828 return a->id_len == b->id_len &&
829 memcmp(a->id, b->id, a->id_len) == 0;
832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
834 struct net_device *sb_dev);
837 TC_SETUP_QDISC_MQPRIO,
840 TC_SETUP_CLSMATCHALL,
850 TC_SETUP_QDISC_TAPRIO,
857 /* These structures hold the attributes of bpf state that are being passed
858 * to the netdevice through the bpf op.
860 enum bpf_netdev_command {
861 /* Set or clear a bpf program used in the earliest stages of packet
862 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
863 * is responsible for calling bpf_prog_put on any old progs that are
864 * stored. In case of error, the callee need not release the new prog
865 * reference, but on success it takes ownership and must bpf_prog_put
866 * when it is no longer used.
870 /* BPF program for offload callbacks, invoked at program load time. */
871 BPF_OFFLOAD_MAP_ALLOC,
872 BPF_OFFLOAD_MAP_FREE,
876 struct bpf_prog_offload_ops;
877 struct netlink_ext_ack;
879 struct xdp_dev_bulk_queue;
889 struct bpf_xdp_entity {
890 struct bpf_prog *prog;
891 struct bpf_xdp_link *link;
895 enum bpf_netdev_command command;
900 struct bpf_prog *prog;
901 struct netlink_ext_ack *extack;
903 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
905 struct bpf_offloaded_map *offmap;
907 /* XDP_SETUP_XSK_POOL */
909 struct xsk_buff_pool *pool;
915 /* Flags for ndo_xsk_wakeup. */
916 #define XDP_WAKEUP_RX (1 << 0)
917 #define XDP_WAKEUP_TX (1 << 1)
919 #ifdef CONFIG_XFRM_OFFLOAD
921 int (*xdo_dev_state_add) (struct xfrm_state *x);
922 void (*xdo_dev_state_delete) (struct xfrm_state *x);
923 void (*xdo_dev_state_free) (struct xfrm_state *x);
924 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
925 struct xfrm_state *x);
926 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
931 struct rcu_head rcuhead;
938 struct netdev_name_node {
939 struct hlist_node hlist;
940 struct list_head list;
941 struct net_device *dev;
945 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
946 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
948 struct netdev_net_notifier {
949 struct list_head list;
950 struct notifier_block *nb;
954 * This structure defines the management hooks for network devices.
955 * The following hooks can be defined; unless noted otherwise, they are
956 * optional and can be filled with a null pointer.
958 * int (*ndo_init)(struct net_device *dev);
959 * This function is called once when a network device is registered.
960 * The network device can use this for any late stage initialization
961 * or semantic validation. It can fail with an error code which will
962 * be propagated back to register_netdev.
964 * void (*ndo_uninit)(struct net_device *dev);
965 * This function is called when device is unregistered or when registration
966 * fails. It is not called if init fails.
968 * int (*ndo_open)(struct net_device *dev);
969 * This function is called when a network device transitions to the up
972 * int (*ndo_stop)(struct net_device *dev);
973 * This function is called when a network device transitions to the down
976 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
977 * struct net_device *dev);
978 * Called when a packet needs to be transmitted.
979 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
980 * the queue before that can happen; it's for obsolete devices and weird
981 * corner cases, but the stack really does a non-trivial amount
982 * of useless work if you return NETDEV_TX_BUSY.
983 * Required; cannot be NULL.
985 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
986 * struct net_device *dev
987 * netdev_features_t features);
988 * Called by core transmit path to determine if device is capable of
989 * performing offload operations on a given packet. This is to give
990 * the device an opportunity to implement any restrictions that cannot
991 * be otherwise expressed by feature flags. The check is called with
992 * the set of features that the stack has calculated and it returns
993 * those the driver believes to be appropriate.
995 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
996 * struct net_device *sb_dev);
997 * Called to decide which queue to use when device supports multiple
1000 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1001 * This function is called to allow device receiver to make
1002 * changes to configuration when multicast or promiscuous is enabled.
1004 * void (*ndo_set_rx_mode)(struct net_device *dev);
1005 * This function is called device changes address list filtering.
1006 * If driver handles unicast address filtering, it should set
1007 * IFF_UNICAST_FLT in its priv_flags.
1009 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1010 * This function is called when the Media Access Control address
1011 * needs to be changed. If this interface is not defined, the
1012 * MAC address can not be changed.
1014 * int (*ndo_validate_addr)(struct net_device *dev);
1015 * Test if Media Access Control address is valid for the device.
1017 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1018 * Called when a user requests an ioctl which can't be handled by
1019 * the generic interface code. If not defined ioctls return
1020 * not supported error code.
1022 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1023 * Used to set network devices bus interface parameters. This interface
1024 * is retained for legacy reasons; new devices should use the bus
1025 * interface (PCI) for low level management.
1027 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1028 * Called when a user wants to change the Maximum Transfer Unit
1031 * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1032 * Callback used when the transmitter has not made any progress
1033 * for dev->watchdog ticks.
1035 * void (*ndo_get_stats64)(struct net_device *dev,
1036 * struct rtnl_link_stats64 *storage);
1037 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1038 * Called when a user wants to get the network device usage
1039 * statistics. Drivers must do one of the following:
1040 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1041 * rtnl_link_stats64 structure passed by the caller.
1042 * 2. Define @ndo_get_stats to update a net_device_stats structure
1043 * (which should normally be dev->stats) and return a pointer to
1044 * it. The structure may be changed asynchronously only if each
1045 * field is written atomically.
1046 * 3. Update dev->stats asynchronously and atomically, and define
1047 * neither operation.
1049 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1050 * Return true if this device supports offload stats of this attr_id.
1052 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1054 * Get statistics for offload operations by attr_id. Write it into the
1055 * attr_data pointer.
1057 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1058 * If device supports VLAN filtering this function is called when a
1059 * VLAN id is registered.
1061 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1062 * If device supports VLAN filtering this function is called when a
1063 * VLAN id is unregistered.
1065 * void (*ndo_poll_controller)(struct net_device *dev);
1067 * SR-IOV management functions.
1068 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1069 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1070 * u8 qos, __be16 proto);
1071 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1073 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1074 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1075 * int (*ndo_get_vf_config)(struct net_device *dev,
1076 * int vf, struct ifla_vf_info *ivf);
1077 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1078 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1079 * struct nlattr *port[]);
1081 * Enable or disable the VF ability to query its RSS Redirection Table and
1082 * Hash Key. This is needed since on some devices VF share this information
1083 * with PF and querying it may introduce a theoretical security risk.
1084 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1085 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1086 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1088 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1089 * This is always called from the stack with the rtnl lock held and netif
1090 * tx queues stopped. This allows the netdevice to perform queue
1091 * management safely.
1093 * Fiber Channel over Ethernet (FCoE) offload functions.
1094 * int (*ndo_fcoe_enable)(struct net_device *dev);
1095 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1096 * so the underlying device can perform whatever needed configuration or
1097 * initialization to support acceleration of FCoE traffic.
1099 * int (*ndo_fcoe_disable)(struct net_device *dev);
1100 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1101 * so the underlying device can perform whatever needed clean-ups to
1102 * stop supporting acceleration of FCoE traffic.
1104 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1105 * struct scatterlist *sgl, unsigned int sgc);
1106 * Called when the FCoE Initiator wants to initialize an I/O that
1107 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1108 * perform necessary setup and returns 1 to indicate the device is set up
1109 * successfully to perform DDP on this I/O, otherwise this returns 0.
1111 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1112 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1113 * indicated by the FC exchange id 'xid', so the underlying device can
1114 * clean up and reuse resources for later DDP requests.
1116 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1117 * struct scatterlist *sgl, unsigned int sgc);
1118 * Called when the FCoE Target wants to initialize an I/O that
1119 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1120 * perform necessary setup and returns 1 to indicate the device is set up
1121 * successfully to perform DDP on this I/O, otherwise this returns 0.
1123 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1124 * struct netdev_fcoe_hbainfo *hbainfo);
1125 * Called when the FCoE Protocol stack wants information on the underlying
1126 * device. This information is utilized by the FCoE protocol stack to
1127 * register attributes with Fiber Channel management service as per the
1128 * FC-GS Fabric Device Management Information(FDMI) specification.
1130 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1131 * Called when the underlying device wants to override default World Wide
1132 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1133 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1134 * protocol stack to use.
1137 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1138 * u16 rxq_index, u32 flow_id);
1139 * Set hardware filter for RFS. rxq_index is the target queue index;
1140 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1141 * Return the filter ID on success, or a negative error code.
1143 * Slave management functions (for bridge, bonding, etc).
1144 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1145 * Called to make another netdev an underling.
1147 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1148 * Called to release previously enslaved netdev.
1150 * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1151 * struct sk_buff *skb,
1153 * Get the xmit slave of master device. If all_slaves is true, function
1154 * assume all the slaves can transmit.
1156 * Feature/offload setting functions.
1157 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1158 * netdev_features_t features);
1159 * Adjusts the requested feature flags according to device-specific
1160 * constraints, and returns the resulting flags. Must not modify
1163 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1164 * Called to update device configuration to new features. Passed
1165 * feature set might be less than what was returned by ndo_fix_features()).
1166 * Must return >0 or -errno if it changed dev->features itself.
1168 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1169 * struct net_device *dev,
1170 * const unsigned char *addr, u16 vid, u16 flags,
1171 * struct netlink_ext_ack *extack);
1172 * Adds an FDB entry to dev for addr.
1173 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1174 * struct net_device *dev,
1175 * const unsigned char *addr, u16 vid)
1176 * Deletes the FDB entry from dev coresponding to addr.
1177 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1178 * struct net_device *dev, struct net_device *filter_dev,
1180 * Used to add FDB entries to dump requests. Implementers should add
1181 * entries to skb and update idx with the number of entries.
1183 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1184 * u16 flags, struct netlink_ext_ack *extack)
1185 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1186 * struct net_device *dev, u32 filter_mask,
1188 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1191 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1192 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1193 * which do not represent real hardware may define this to allow their
1194 * userspace components to manage their virtual carrier state. Devices
1195 * that determine carrier state from physical hardware properties (eg
1196 * network cables) or protocol-dependent mechanisms (eg
1197 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1199 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1200 * struct netdev_phys_item_id *ppid);
1201 * Called to get ID of physical port of this device. If driver does
1202 * not implement this, it is assumed that the hw is not able to have
1203 * multiple net devices on single physical port.
1205 * int (*ndo_get_port_parent_id)(struct net_device *dev,
1206 * struct netdev_phys_item_id *ppid)
1207 * Called to get the parent ID of the physical port of this device.
1209 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1210 * struct udp_tunnel_info *ti);
1211 * Called by UDP tunnel to notify a driver about the UDP port and socket
1212 * address family that a UDP tunnel is listnening to. It is called only
1213 * when a new port starts listening. The operation is protected by the
1216 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1217 * struct udp_tunnel_info *ti);
1218 * Called by UDP tunnel to notify the driver about a UDP port and socket
1219 * address family that the UDP tunnel is not listening to anymore. The
1220 * operation is protected by the RTNL.
1222 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1223 * struct net_device *dev)
1224 * Called by upper layer devices to accelerate switching or other
1225 * station functionality into hardware. 'pdev is the lowerdev
1226 * to use for the offload and 'dev' is the net device that will
1227 * back the offload. Returns a pointer to the private structure
1228 * the upper layer will maintain.
1229 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1230 * Called by upper layer device to delete the station created
1231 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1232 * the station and priv is the structure returned by the add
1234 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1235 * int queue_index, u32 maxrate);
1236 * Called when a user wants to set a max-rate limitation of specific
1238 * int (*ndo_get_iflink)(const struct net_device *dev);
1239 * Called to get the iflink value of this device.
1240 * void (*ndo_change_proto_down)(struct net_device *dev,
1242 * This function is used to pass protocol port error state information
1243 * to the switch driver. The switch driver can react to the proto_down
1244 * by doing a phys down on the associated switch port.
1245 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1246 * This function is used to get egress tunnel information for given skb.
1247 * This is useful for retrieving outer tunnel header parameters while
1249 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1250 * This function is used to specify the headroom that the skb must
1251 * consider when allocation skb during packet reception. Setting
1252 * appropriate rx headroom value allows avoiding skb head copy on
1253 * forward. Setting a negative value resets the rx headroom to the
1255 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1256 * This function is used to set or query state related to XDP on the
1257 * netdevice and manage BPF offload. See definition of
1258 * enum bpf_netdev_command for details.
1259 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1261 * This function is used to submit @n XDP packets for transmit on a
1262 * netdevice. Returns number of frames successfully transmitted, frames
1263 * that got dropped are freed/returned via xdp_return_frame().
1264 * Returns negative number, means general error invoking ndo, meaning
1265 * no frames were xmit'ed and core-caller will free all frames.
1266 * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1267 * This function is used to wake up the softirq, ksoftirqd or kthread
1268 * responsible for sending and/or receiving packets on a specific
1269 * queue id bound to an AF_XDP socket. The flags field specifies if
1270 * only RX, only Tx, or both should be woken up using the flags
1271 * XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1272 * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1273 * Get devlink port instance associated with a given netdev.
1274 * Called with a reference on the netdevice and devlink locks only,
1275 * rtnl_lock is not held.
1276 * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1278 * Add, change, delete or get information on an IPv4 tunnel.
1279 * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1280 * If a device is paired with a peer device, return the peer instance.
1281 * The caller must be under RCU read context.
1283 struct net_device_ops {
1284 int (*ndo_init)(struct net_device *dev);
1285 void (*ndo_uninit)(struct net_device *dev);
1286 int (*ndo_open)(struct net_device *dev);
1287 int (*ndo_stop)(struct net_device *dev);
1288 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1289 struct net_device *dev);
1290 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1291 struct net_device *dev,
1292 netdev_features_t features);
1293 u16 (*ndo_select_queue)(struct net_device *dev,
1294 struct sk_buff *skb,
1295 struct net_device *sb_dev);
1296 void (*ndo_change_rx_flags)(struct net_device *dev,
1298 void (*ndo_set_rx_mode)(struct net_device *dev);
1299 int (*ndo_set_mac_address)(struct net_device *dev,
1301 int (*ndo_validate_addr)(struct net_device *dev);
1302 int (*ndo_do_ioctl)(struct net_device *dev,
1303 struct ifreq *ifr, int cmd);
1304 int (*ndo_set_config)(struct net_device *dev,
1306 int (*ndo_change_mtu)(struct net_device *dev,
1308 int (*ndo_neigh_setup)(struct net_device *dev,
1309 struct neigh_parms *);
1310 void (*ndo_tx_timeout) (struct net_device *dev,
1311 unsigned int txqueue);
1313 void (*ndo_get_stats64)(struct net_device *dev,
1314 struct rtnl_link_stats64 *storage);
1315 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1316 int (*ndo_get_offload_stats)(int attr_id,
1317 const struct net_device *dev,
1319 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1321 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1322 __be16 proto, u16 vid);
1323 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1324 __be16 proto, u16 vid);
1325 #ifdef CONFIG_NET_POLL_CONTROLLER
1326 void (*ndo_poll_controller)(struct net_device *dev);
1327 int (*ndo_netpoll_setup)(struct net_device *dev,
1328 struct netpoll_info *info);
1329 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1331 int (*ndo_set_vf_mac)(struct net_device *dev,
1332 int queue, u8 *mac);
1333 int (*ndo_set_vf_vlan)(struct net_device *dev,
1334 int queue, u16 vlan,
1335 u8 qos, __be16 proto);
1336 int (*ndo_set_vf_rate)(struct net_device *dev,
1337 int vf, int min_tx_rate,
1339 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1340 int vf, bool setting);
1341 int (*ndo_set_vf_trust)(struct net_device *dev,
1342 int vf, bool setting);
1343 int (*ndo_get_vf_config)(struct net_device *dev,
1345 struct ifla_vf_info *ivf);
1346 int (*ndo_set_vf_link_state)(struct net_device *dev,
1347 int vf, int link_state);
1348 int (*ndo_get_vf_stats)(struct net_device *dev,
1350 struct ifla_vf_stats
1352 int (*ndo_set_vf_port)(struct net_device *dev,
1354 struct nlattr *port[]);
1355 int (*ndo_get_vf_port)(struct net_device *dev,
1356 int vf, struct sk_buff *skb);
1357 int (*ndo_get_vf_guid)(struct net_device *dev,
1359 struct ifla_vf_guid *node_guid,
1360 struct ifla_vf_guid *port_guid);
1361 int (*ndo_set_vf_guid)(struct net_device *dev,
1364 int (*ndo_set_vf_rss_query_en)(
1365 struct net_device *dev,
1366 int vf, bool setting);
1367 int (*ndo_setup_tc)(struct net_device *dev,
1368 enum tc_setup_type type,
1370 #if IS_ENABLED(CONFIG_FCOE)
1371 int (*ndo_fcoe_enable)(struct net_device *dev);
1372 int (*ndo_fcoe_disable)(struct net_device *dev);
1373 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1375 struct scatterlist *sgl,
1377 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1379 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1381 struct scatterlist *sgl,
1383 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1384 struct netdev_fcoe_hbainfo *hbainfo);
1387 #if IS_ENABLED(CONFIG_LIBFCOE)
1388 #define NETDEV_FCOE_WWNN 0
1389 #define NETDEV_FCOE_WWPN 1
1390 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1391 u64 *wwn, int type);
1394 #ifdef CONFIG_RFS_ACCEL
1395 int (*ndo_rx_flow_steer)(struct net_device *dev,
1396 const struct sk_buff *skb,
1400 int (*ndo_add_slave)(struct net_device *dev,
1401 struct net_device *slave_dev,
1402 struct netlink_ext_ack *extack);
1403 int (*ndo_del_slave)(struct net_device *dev,
1404 struct net_device *slave_dev);
1405 struct net_device* (*ndo_get_xmit_slave)(struct net_device *dev,
1406 struct sk_buff *skb,
1408 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1409 netdev_features_t features);
1410 int (*ndo_set_features)(struct net_device *dev,
1411 netdev_features_t features);
1412 int (*ndo_neigh_construct)(struct net_device *dev,
1413 struct neighbour *n);
1414 void (*ndo_neigh_destroy)(struct net_device *dev,
1415 struct neighbour *n);
1417 int (*ndo_fdb_add)(struct ndmsg *ndm,
1418 struct nlattr *tb[],
1419 struct net_device *dev,
1420 const unsigned char *addr,
1423 struct netlink_ext_ack *extack);
1424 int (*ndo_fdb_del)(struct ndmsg *ndm,
1425 struct nlattr *tb[],
1426 struct net_device *dev,
1427 const unsigned char *addr,
1429 int (*ndo_fdb_dump)(struct sk_buff *skb,
1430 struct netlink_callback *cb,
1431 struct net_device *dev,
1432 struct net_device *filter_dev,
1434 int (*ndo_fdb_get)(struct sk_buff *skb,
1435 struct nlattr *tb[],
1436 struct net_device *dev,
1437 const unsigned char *addr,
1438 u16 vid, u32 portid, u32 seq,
1439 struct netlink_ext_ack *extack);
1440 int (*ndo_bridge_setlink)(struct net_device *dev,
1441 struct nlmsghdr *nlh,
1443 struct netlink_ext_ack *extack);
1444 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1446 struct net_device *dev,
1449 int (*ndo_bridge_dellink)(struct net_device *dev,
1450 struct nlmsghdr *nlh,
1452 int (*ndo_change_carrier)(struct net_device *dev,
1454 int (*ndo_get_phys_port_id)(struct net_device *dev,
1455 struct netdev_phys_item_id *ppid);
1456 int (*ndo_get_port_parent_id)(struct net_device *dev,
1457 struct netdev_phys_item_id *ppid);
1458 int (*ndo_get_phys_port_name)(struct net_device *dev,
1459 char *name, size_t len);
1460 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1461 struct udp_tunnel_info *ti);
1462 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1463 struct udp_tunnel_info *ti);
1464 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1465 struct net_device *dev);
1466 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1469 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1472 int (*ndo_get_iflink)(const struct net_device *dev);
1473 int (*ndo_change_proto_down)(struct net_device *dev,
1475 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1476 struct sk_buff *skb);
1477 void (*ndo_set_rx_headroom)(struct net_device *dev,
1478 int needed_headroom);
1479 int (*ndo_bpf)(struct net_device *dev,
1480 struct netdev_bpf *bpf);
1481 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1482 struct xdp_frame **xdp,
1484 int (*ndo_xsk_wakeup)(struct net_device *dev,
1485 u32 queue_id, u32 flags);
1486 struct devlink_port * (*ndo_get_devlink_port)(struct net_device *dev);
1487 int (*ndo_tunnel_ctl)(struct net_device *dev,
1488 struct ip_tunnel_parm *p, int cmd);
1489 struct net_device * (*ndo_get_peer_dev)(struct net_device *dev);
1493 * enum net_device_priv_flags - &struct net_device priv_flags
1495 * These are the &struct net_device, they are only set internally
1496 * by drivers and used in the kernel. These flags are invisible to
1497 * userspace; this means that the order of these flags can change
1498 * during any kernel release.
1500 * You should have a pretty good reason to be extending these flags.
1502 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1503 * @IFF_EBRIDGE: Ethernet bridging device
1504 * @IFF_BONDING: bonding master or slave
1505 * @IFF_ISATAP: ISATAP interface (RFC4214)
1506 * @IFF_WAN_HDLC: WAN HDLC device
1507 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1509 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1510 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1511 * @IFF_MACVLAN_PORT: device used as macvlan port
1512 * @IFF_BRIDGE_PORT: device used as bridge port
1513 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1514 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1515 * @IFF_UNICAST_FLT: Supports unicast filtering
1516 * @IFF_TEAM_PORT: device used as team port
1517 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1518 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1519 * change when it's running
1520 * @IFF_MACVLAN: Macvlan device
1521 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1522 * underlying stacked devices
1523 * @IFF_L3MDEV_MASTER: device is an L3 master device
1524 * @IFF_NO_QUEUE: device can run without qdisc attached
1525 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1526 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1527 * @IFF_TEAM: device is a team device
1528 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1529 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1530 * entity (i.e. the master device for bridged veth)
1531 * @IFF_MACSEC: device is a MACsec device
1532 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1533 * @IFF_FAILOVER: device is a failover master device
1534 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1535 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1536 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1538 enum netdev_priv_flags {
1539 IFF_802_1Q_VLAN = 1<<0,
1543 IFF_WAN_HDLC = 1<<4,
1544 IFF_XMIT_DST_RELEASE = 1<<5,
1545 IFF_DONT_BRIDGE = 1<<6,
1546 IFF_DISABLE_NETPOLL = 1<<7,
1547 IFF_MACVLAN_PORT = 1<<8,
1548 IFF_BRIDGE_PORT = 1<<9,
1549 IFF_OVS_DATAPATH = 1<<10,
1550 IFF_TX_SKB_SHARING = 1<<11,
1551 IFF_UNICAST_FLT = 1<<12,
1552 IFF_TEAM_PORT = 1<<13,
1553 IFF_SUPP_NOFCS = 1<<14,
1554 IFF_LIVE_ADDR_CHANGE = 1<<15,
1555 IFF_MACVLAN = 1<<16,
1556 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1557 IFF_L3MDEV_MASTER = 1<<18,
1558 IFF_NO_QUEUE = 1<<19,
1559 IFF_OPENVSWITCH = 1<<20,
1560 IFF_L3MDEV_SLAVE = 1<<21,
1562 IFF_RXFH_CONFIGURED = 1<<23,
1563 IFF_PHONY_HEADROOM = 1<<24,
1565 IFF_NO_RX_HANDLER = 1<<26,
1566 IFF_FAILOVER = 1<<27,
1567 IFF_FAILOVER_SLAVE = 1<<28,
1568 IFF_L3MDEV_RX_HANDLER = 1<<29,
1569 IFF_LIVE_RENAME_OK = 1<<30,
1572 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1573 #define IFF_EBRIDGE IFF_EBRIDGE
1574 #define IFF_BONDING IFF_BONDING
1575 #define IFF_ISATAP IFF_ISATAP
1576 #define IFF_WAN_HDLC IFF_WAN_HDLC
1577 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1578 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1579 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1580 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1581 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1582 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1583 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1584 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1585 #define IFF_TEAM_PORT IFF_TEAM_PORT
1586 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1587 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1588 #define IFF_MACVLAN IFF_MACVLAN
1589 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1590 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1591 #define IFF_NO_QUEUE IFF_NO_QUEUE
1592 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1593 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1594 #define IFF_TEAM IFF_TEAM
1595 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1596 #define IFF_MACSEC IFF_MACSEC
1597 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1598 #define IFF_FAILOVER IFF_FAILOVER
1599 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1600 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1601 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1604 * struct net_device - The DEVICE structure.
1606 * Actually, this whole structure is a big mistake. It mixes I/O
1607 * data with strictly "high-level" data, and it has to know about
1608 * almost every data structure used in the INET module.
1610 * @name: This is the first field of the "visible" part of this structure
1611 * (i.e. as seen by users in the "Space.c" file). It is the name
1614 * @name_node: Name hashlist node
1615 * @ifalias: SNMP alias
1616 * @mem_end: Shared memory end
1617 * @mem_start: Shared memory start
1618 * @base_addr: Device I/O address
1619 * @irq: Device IRQ number
1621 * @state: Generic network queuing layer state, see netdev_state_t
1622 * @dev_list: The global list of network devices
1623 * @napi_list: List entry used for polling NAPI devices
1624 * @unreg_list: List entry when we are unregistering the
1625 * device; see the function unregister_netdev
1626 * @close_list: List entry used when we are closing the device
1627 * @ptype_all: Device-specific packet handlers for all protocols
1628 * @ptype_specific: Device-specific, protocol-specific packet handlers
1630 * @adj_list: Directly linked devices, like slaves for bonding
1631 * @features: Currently active device features
1632 * @hw_features: User-changeable features
1634 * @wanted_features: User-requested features
1635 * @vlan_features: Mask of features inheritable by VLAN devices
1637 * @hw_enc_features: Mask of features inherited by encapsulating devices
1638 * This field indicates what encapsulation
1639 * offloads the hardware is capable of doing,
1640 * and drivers will need to set them appropriately.
1642 * @mpls_features: Mask of features inheritable by MPLS
1643 * @gso_partial_features: value(s) from NETIF_F_GSO\*
1645 * @ifindex: interface index
1646 * @group: The group the device belongs to
1648 * @stats: Statistics struct, which was left as a legacy, use
1649 * rtnl_link_stats64 instead
1651 * @rx_dropped: Dropped packets by core network,
1652 * do not use this in drivers
1653 * @tx_dropped: Dropped packets by core network,
1654 * do not use this in drivers
1655 * @rx_nohandler: nohandler dropped packets by core network on
1656 * inactive devices, do not use this in drivers
1657 * @carrier_up_count: Number of times the carrier has been up
1658 * @carrier_down_count: Number of times the carrier has been down
1660 * @wireless_handlers: List of functions to handle Wireless Extensions,
1662 * see <net/iw_handler.h> for details.
1663 * @wireless_data: Instance data managed by the core of wireless extensions
1665 * @netdev_ops: Includes several pointers to callbacks,
1666 * if one wants to override the ndo_*() functions
1667 * @ethtool_ops: Management operations
1668 * @l3mdev_ops: Layer 3 master device operations
1669 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1670 * discovery handling. Necessary for e.g. 6LoWPAN.
1671 * @xfrmdev_ops: Transformation offload operations
1672 * @tlsdev_ops: Transport Layer Security offload operations
1673 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1674 * of Layer 2 headers.
1676 * @flags: Interface flags (a la BSD)
1677 * @priv_flags: Like 'flags' but invisible to userspace,
1678 * see if.h for the definitions
1679 * @gflags: Global flags ( kept as legacy )
1680 * @padded: How much padding added by alloc_netdev()
1681 * @operstate: RFC2863 operstate
1682 * @link_mode: Mapping policy to operstate
1683 * @if_port: Selectable AUI, TP, ...
1685 * @mtu: Interface MTU value
1686 * @min_mtu: Interface Minimum MTU value
1687 * @max_mtu: Interface Maximum MTU value
1688 * @type: Interface hardware type
1689 * @hard_header_len: Maximum hardware header length.
1690 * @min_header_len: Minimum hardware header length
1692 * @needed_headroom: Extra headroom the hardware may need, but not in all
1693 * cases can this be guaranteed
1694 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1695 * cases can this be guaranteed. Some cases also use
1696 * LL_MAX_HEADER instead to allocate the skb
1698 * interface address info:
1700 * @perm_addr: Permanent hw address
1701 * @addr_assign_type: Hw address assignment type
1702 * @addr_len: Hardware address length
1703 * @upper_level: Maximum depth level of upper devices.
1704 * @lower_level: Maximum depth level of lower devices.
1705 * @neigh_priv_len: Used in neigh_alloc()
1706 * @dev_id: Used to differentiate devices that share
1707 * the same link layer address
1708 * @dev_port: Used to differentiate devices that share
1710 * @addr_list_lock: XXX: need comments on this one
1711 * @name_assign_type: network interface name assignment type
1712 * @uc_promisc: Counter that indicates promiscuous mode
1713 * has been enabled due to the need to listen to
1714 * additional unicast addresses in a device that
1715 * does not implement ndo_set_rx_mode()
1716 * @uc: unicast mac addresses
1717 * @mc: multicast mac addresses
1718 * @dev_addrs: list of device hw addresses
1719 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1720 * @promiscuity: Number of times the NIC is told to work in
1721 * promiscuous mode; if it becomes 0 the NIC will
1722 * exit promiscuous mode
1723 * @allmulti: Counter, enables or disables allmulticast mode
1725 * @vlan_info: VLAN info
1726 * @dsa_ptr: dsa specific data
1727 * @tipc_ptr: TIPC specific data
1728 * @atalk_ptr: AppleTalk link
1729 * @ip_ptr: IPv4 specific data
1730 * @dn_ptr: DECnet specific data
1731 * @ip6_ptr: IPv6 specific data
1732 * @ax25_ptr: AX.25 specific data
1733 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1734 * @ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1736 * @mpls_ptr: mpls_dev struct pointer
1738 * @dev_addr: Hw address (before bcast,
1739 * because most packets are unicast)
1741 * @_rx: Array of RX queues
1742 * @num_rx_queues: Number of RX queues
1743 * allocated at register_netdev() time
1744 * @real_num_rx_queues: Number of RX queues currently active in device
1745 * @xdp_prog: XDP sockets filter program pointer
1746 * @gro_flush_timeout: timeout for GRO layer in NAPI
1747 * @napi_defer_hard_irqs: If not zero, provides a counter that would
1748 * allow to avoid NIC hard IRQ, on busy queues.
1750 * @rx_handler: handler for received packets
1751 * @rx_handler_data: XXX: need comments on this one
1752 * @miniq_ingress: ingress/clsact qdisc specific data for
1753 * ingress processing
1754 * @ingress_queue: XXX: need comments on this one
1755 * @nf_hooks_ingress: netfilter hooks executed for ingress packets
1756 * @broadcast: hw bcast address
1758 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1759 * indexed by RX queue number. Assigned by driver.
1760 * This must only be set if the ndo_rx_flow_steer
1761 * operation is defined
1762 * @index_hlist: Device index hash chain
1764 * @_tx: Array of TX queues
1765 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1766 * @real_num_tx_queues: Number of TX queues currently active in device
1767 * @qdisc: Root qdisc from userspace point of view
1768 * @tx_queue_len: Max frames per queue allowed
1769 * @tx_global_lock: XXX: need comments on this one
1770 * @xdp_bulkq: XDP device bulk queue
1771 * @xps_cpus_map: all CPUs map for XPS device
1772 * @xps_rxqs_map: all RXQs map for XPS device
1774 * @xps_maps: XXX: need comments on this one
1775 * @miniq_egress: clsact qdisc specific data for
1777 * @qdisc_hash: qdisc hash table
1778 * @watchdog_timeo: Represents the timeout that is used by
1779 * the watchdog (see dev_watchdog())
1780 * @watchdog_timer: List of timers
1782 * @proto_down_reason: reason a netdev interface is held down
1783 * @pcpu_refcnt: Number of references to this device
1784 * @todo_list: Delayed register/unregister
1785 * @link_watch_list: XXX: need comments on this one
1787 * @reg_state: Register/unregister state machine
1788 * @dismantle: Device is going to be freed
1789 * @rtnl_link_state: This enum represents the phases of creating
1792 * @needs_free_netdev: Should unregister perform free_netdev?
1793 * @priv_destructor: Called from unregister
1794 * @npinfo: XXX: need comments on this one
1795 * @nd_net: Network namespace this network device is inside
1797 * @ml_priv: Mid-layer private
1798 * @lstats: Loopback statistics
1799 * @tstats: Tunnel statistics
1800 * @dstats: Dummy statistics
1801 * @vstats: Virtual ethernet statistics
1806 * @dev: Class/net/name entry
1807 * @sysfs_groups: Space for optional device, statistics and wireless
1810 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1811 * @rtnl_link_ops: Rtnl_link_ops
1813 * @gso_max_size: Maximum size of generic segmentation offload
1814 * @gso_max_segs: Maximum number of segments that can be passed to the
1817 * @dcbnl_ops: Data Center Bridging netlink ops
1818 * @num_tc: Number of traffic classes in the net device
1819 * @tc_to_txq: XXX: need comments on this one
1820 * @prio_tc_map: XXX: need comments on this one
1822 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1824 * @priomap: XXX: need comments on this one
1825 * @phydev: Physical device may attach itself
1826 * for hardware timestamping
1827 * @sfp_bus: attached &struct sfp_bus structure.
1829 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1830 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1832 * @proto_down: protocol port state information can be sent to the
1833 * switch driver and used to set the phys state of the
1836 * @wol_enabled: Wake-on-LAN is enabled
1838 * @net_notifier_list: List of per-net netdev notifier block
1839 * that follow this device when it is moved
1840 * to another network namespace.
1842 * @macsec_ops: MACsec offloading ops
1844 * @udp_tunnel_nic_info: static structure describing the UDP tunnel
1845 * offload capabilities of the device
1846 * @udp_tunnel_nic: UDP tunnel offload state
1847 * @xdp_state: stores info on attached XDP BPF programs
1849 * @nested_level: Used as as a parameter of spin_lock_nested() of
1850 * dev->addr_list_lock.
1851 * @unlink_list: As netif_addr_lock() can be called recursively,
1852 * keep a list of interfaces to be deleted.
1854 * FIXME: cleanup struct net_device such that network protocol info
1859 char name[IFNAMSIZ];
1860 struct netdev_name_node *name_node;
1861 struct dev_ifalias __rcu *ifalias;
1863 * I/O specific fields
1864 * FIXME: Merge these and struct ifmap into one
1866 unsigned long mem_end;
1867 unsigned long mem_start;
1868 unsigned long base_addr;
1872 * Some hardware also needs these fields (state,dev_list,
1873 * napi_list,unreg_list,close_list) but they are not
1874 * part of the usual set specified in Space.c.
1877 unsigned long state;
1879 struct list_head dev_list;
1880 struct list_head napi_list;
1881 struct list_head unreg_list;
1882 struct list_head close_list;
1883 struct list_head ptype_all;
1884 struct list_head ptype_specific;
1887 struct list_head upper;
1888 struct list_head lower;
1891 netdev_features_t features;
1892 netdev_features_t hw_features;
1893 netdev_features_t wanted_features;
1894 netdev_features_t vlan_features;
1895 netdev_features_t hw_enc_features;
1896 netdev_features_t mpls_features;
1897 netdev_features_t gso_partial_features;
1902 struct net_device_stats stats;
1904 atomic_long_t rx_dropped;
1905 atomic_long_t tx_dropped;
1906 atomic_long_t rx_nohandler;
1908 /* Stats to monitor link on/off, flapping */
1909 atomic_t carrier_up_count;
1910 atomic_t carrier_down_count;
1912 #ifdef CONFIG_WIRELESS_EXT
1913 const struct iw_handler_def *wireless_handlers;
1914 struct iw_public_data *wireless_data;
1916 const struct net_device_ops *netdev_ops;
1917 const struct ethtool_ops *ethtool_ops;
1918 #ifdef CONFIG_NET_L3_MASTER_DEV
1919 const struct l3mdev_ops *l3mdev_ops;
1921 #if IS_ENABLED(CONFIG_IPV6)
1922 const struct ndisc_ops *ndisc_ops;
1925 #ifdef CONFIG_XFRM_OFFLOAD
1926 const struct xfrmdev_ops *xfrmdev_ops;
1929 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1930 const struct tlsdev_ops *tlsdev_ops;
1933 const struct header_ops *header_ops;
1936 unsigned int priv_flags;
1938 unsigned short gflags;
1939 unsigned short padded;
1941 unsigned char operstate;
1942 unsigned char link_mode;
1944 unsigned char if_port;
1947 /* Note : dev->mtu is often read without holding a lock.
1948 * Writers usually hold RTNL.
1949 * It is recommended to use READ_ONCE() to annotate the reads,
1950 * and to use WRITE_ONCE() to annotate the writes.
1953 unsigned int min_mtu;
1954 unsigned int max_mtu;
1955 unsigned short type;
1956 unsigned short hard_header_len;
1957 unsigned char min_header_len;
1958 unsigned char name_assign_type;
1960 unsigned short needed_headroom;
1961 unsigned short needed_tailroom;
1963 /* Interface address info. */
1964 unsigned char perm_addr[MAX_ADDR_LEN];
1965 unsigned char addr_assign_type;
1966 unsigned char addr_len;
1967 unsigned char upper_level;
1968 unsigned char lower_level;
1970 unsigned short neigh_priv_len;
1971 unsigned short dev_id;
1972 unsigned short dev_port;
1973 spinlock_t addr_list_lock;
1975 struct netdev_hw_addr_list uc;
1976 struct netdev_hw_addr_list mc;
1977 struct netdev_hw_addr_list dev_addrs;
1980 struct kset *queues_kset;
1982 #ifdef CONFIG_LOCKDEP
1983 struct list_head unlink_list;
1985 unsigned int promiscuity;
1986 unsigned int allmulti;
1988 #ifdef CONFIG_LOCKDEP
1989 unsigned char nested_level;
1993 /* Protocol-specific pointers */
1995 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1996 struct vlan_info __rcu *vlan_info;
1998 #if IS_ENABLED(CONFIG_NET_DSA)
1999 struct dsa_port *dsa_ptr;
2001 #if IS_ENABLED(CONFIG_TIPC)
2002 struct tipc_bearer __rcu *tipc_ptr;
2004 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
2007 struct in_device __rcu *ip_ptr;
2008 #if IS_ENABLED(CONFIG_DECNET)
2009 struct dn_dev __rcu *dn_ptr;
2011 struct inet6_dev __rcu *ip6_ptr;
2012 #if IS_ENABLED(CONFIG_AX25)
2015 struct wireless_dev *ieee80211_ptr;
2016 struct wpan_dev *ieee802154_ptr;
2017 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2018 struct mpls_dev __rcu *mpls_ptr;
2022 * Cache lines mostly used on receive path (including eth_type_trans())
2024 /* Interface address info used in eth_type_trans() */
2025 unsigned char *dev_addr;
2027 struct netdev_rx_queue *_rx;
2028 unsigned int num_rx_queues;
2029 unsigned int real_num_rx_queues;
2031 struct bpf_prog __rcu *xdp_prog;
2032 unsigned long gro_flush_timeout;
2033 int napi_defer_hard_irqs;
2034 rx_handler_func_t __rcu *rx_handler;
2035 void __rcu *rx_handler_data;
2037 #ifdef CONFIG_NET_CLS_ACT
2038 struct mini_Qdisc __rcu *miniq_ingress;
2040 struct netdev_queue __rcu *ingress_queue;
2041 #ifdef CONFIG_NETFILTER_INGRESS
2042 struct nf_hook_entries __rcu *nf_hooks_ingress;
2045 unsigned char broadcast[MAX_ADDR_LEN];
2046 #ifdef CONFIG_RFS_ACCEL
2047 struct cpu_rmap *rx_cpu_rmap;
2049 struct hlist_node index_hlist;
2052 * Cache lines mostly used on transmit path
2054 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
2055 unsigned int num_tx_queues;
2056 unsigned int real_num_tx_queues;
2057 struct Qdisc *qdisc;
2058 unsigned int tx_queue_len;
2059 spinlock_t tx_global_lock;
2061 struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2064 struct xps_dev_maps __rcu *xps_cpus_map;
2065 struct xps_dev_maps __rcu *xps_rxqs_map;
2067 #ifdef CONFIG_NET_CLS_ACT
2068 struct mini_Qdisc __rcu *miniq_egress;
2071 #ifdef CONFIG_NET_SCHED
2072 DECLARE_HASHTABLE (qdisc_hash, 4);
2074 /* These may be needed for future network-power-down code. */
2075 struct timer_list watchdog_timer;
2078 u32 proto_down_reason;
2080 struct list_head todo_list;
2081 int __percpu *pcpu_refcnt;
2083 struct list_head link_watch_list;
2085 enum { NETREG_UNINITIALIZED=0,
2086 NETREG_REGISTERED, /* completed register_netdevice */
2087 NETREG_UNREGISTERING, /* called unregister_netdevice */
2088 NETREG_UNREGISTERED, /* completed unregister todo */
2089 NETREG_RELEASED, /* called free_netdev */
2090 NETREG_DUMMY, /* dummy device for NAPI poll */
2096 RTNL_LINK_INITIALIZED,
2097 RTNL_LINK_INITIALIZING,
2098 } rtnl_link_state:16;
2100 bool needs_free_netdev;
2101 void (*priv_destructor)(struct net_device *dev);
2103 #ifdef CONFIG_NETPOLL
2104 struct netpoll_info __rcu *npinfo;
2107 possible_net_t nd_net;
2109 /* mid-layer private */
2112 struct pcpu_lstats __percpu *lstats;
2113 struct pcpu_sw_netstats __percpu *tstats;
2114 struct pcpu_dstats __percpu *dstats;
2117 #if IS_ENABLED(CONFIG_GARP)
2118 struct garp_port __rcu *garp_port;
2120 #if IS_ENABLED(CONFIG_MRP)
2121 struct mrp_port __rcu *mrp_port;
2125 const struct attribute_group *sysfs_groups[4];
2126 const struct attribute_group *sysfs_rx_queue_group;
2128 const struct rtnl_link_ops *rtnl_link_ops;
2130 /* for setting kernel sock attribute on TCP connection setup */
2131 #define GSO_MAX_SIZE 65536
2132 unsigned int gso_max_size;
2133 #define GSO_MAX_SEGS 65535
2137 const struct dcbnl_rtnl_ops *dcbnl_ops;
2140 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2141 u8 prio_tc_map[TC_BITMASK + 1];
2143 #if IS_ENABLED(CONFIG_FCOE)
2144 unsigned int fcoe_ddp_xid;
2146 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2147 struct netprio_map __rcu *priomap;
2149 struct phy_device *phydev;
2150 struct sfp_bus *sfp_bus;
2151 struct lock_class_key *qdisc_tx_busylock;
2152 struct lock_class_key *qdisc_running_key;
2154 unsigned wol_enabled:1;
2156 struct list_head net_notifier_list;
2158 #if IS_ENABLED(CONFIG_MACSEC)
2159 /* MACsec management functions */
2160 const struct macsec_ops *macsec_ops;
2162 const struct udp_tunnel_nic_info *udp_tunnel_nic_info;
2163 struct udp_tunnel_nic *udp_tunnel_nic;
2165 /* protected by rtnl_lock */
2166 struct bpf_xdp_entity xdp_state[__MAX_XDP_MODE];
2168 #define to_net_dev(d) container_of(d, struct net_device, dev)
2170 static inline bool netif_elide_gro(const struct net_device *dev)
2172 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2177 #define NETDEV_ALIGN 32
2180 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2182 return dev->prio_tc_map[prio & TC_BITMASK];
2186 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2188 if (tc >= dev->num_tc)
2191 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2195 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2196 void netdev_reset_tc(struct net_device *dev);
2197 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2198 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2201 int netdev_get_num_tc(struct net_device *dev)
2206 static inline void net_prefetch(void *p)
2209 #if L1_CACHE_BYTES < 128
2210 prefetch((u8 *)p + L1_CACHE_BYTES);
2214 static inline void net_prefetchw(void *p)
2217 #if L1_CACHE_BYTES < 128
2218 prefetchw((u8 *)p + L1_CACHE_BYTES);
2222 void netdev_unbind_sb_channel(struct net_device *dev,
2223 struct net_device *sb_dev);
2224 int netdev_bind_sb_channel_queue(struct net_device *dev,
2225 struct net_device *sb_dev,
2226 u8 tc, u16 count, u16 offset);
2227 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2228 static inline int netdev_get_sb_channel(struct net_device *dev)
2230 return max_t(int, -dev->num_tc, 0);
2234 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2237 return &dev->_tx[index];
2240 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2241 const struct sk_buff *skb)
2243 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2246 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2247 void (*f)(struct net_device *,
2248 struct netdev_queue *,
2254 for (i = 0; i < dev->num_tx_queues; i++)
2255 f(dev, &dev->_tx[i], arg);
2258 #define netdev_lockdep_set_classes(dev) \
2260 static struct lock_class_key qdisc_tx_busylock_key; \
2261 static struct lock_class_key qdisc_running_key; \
2262 static struct lock_class_key qdisc_xmit_lock_key; \
2263 static struct lock_class_key dev_addr_list_lock_key; \
2266 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2267 (dev)->qdisc_running_key = &qdisc_running_key; \
2268 lockdep_set_class(&(dev)->addr_list_lock, \
2269 &dev_addr_list_lock_key); \
2270 for (i = 0; i < (dev)->num_tx_queues; i++) \
2271 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2272 &qdisc_xmit_lock_key); \
2275 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2276 struct net_device *sb_dev);
2277 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2278 struct sk_buff *skb,
2279 struct net_device *sb_dev);
2281 /* returns the headroom that the master device needs to take in account
2282 * when forwarding to this dev
2284 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2286 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2289 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2291 if (dev->netdev_ops->ndo_set_rx_headroom)
2292 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2295 /* set the device rx headroom to the dev's default */
2296 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2298 netdev_set_rx_headroom(dev, -1);
2302 * Net namespace inlines
2305 struct net *dev_net(const struct net_device *dev)
2307 return read_pnet(&dev->nd_net);
2311 void dev_net_set(struct net_device *dev, struct net *net)
2313 write_pnet(&dev->nd_net, net);
2317 * netdev_priv - access network device private data
2318 * @dev: network device
2320 * Get network device private data
2322 static inline void *netdev_priv(const struct net_device *dev)
2324 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2327 /* Set the sysfs physical device reference for the network logical device
2328 * if set prior to registration will cause a symlink during initialization.
2330 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2332 /* Set the sysfs device type for the network logical device to allow
2333 * fine-grained identification of different network device types. For
2334 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2336 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2338 /* Default NAPI poll() weight
2339 * Device drivers are strongly advised to not use bigger value
2341 #define NAPI_POLL_WEIGHT 64
2344 * netif_napi_add - initialize a NAPI context
2345 * @dev: network device
2346 * @napi: NAPI context
2347 * @poll: polling function
2348 * @weight: default weight
2350 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2351 * *any* of the other NAPI-related functions.
2353 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2354 int (*poll)(struct napi_struct *, int), int weight);
2357 * netif_tx_napi_add - initialize a NAPI context
2358 * @dev: network device
2359 * @napi: NAPI context
2360 * @poll: polling function
2361 * @weight: default weight
2363 * This variant of netif_napi_add() should be used from drivers using NAPI
2364 * to exclusively poll a TX queue.
2365 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2367 static inline void netif_tx_napi_add(struct net_device *dev,
2368 struct napi_struct *napi,
2369 int (*poll)(struct napi_struct *, int),
2372 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2373 netif_napi_add(dev, napi, poll, weight);
2377 * __netif_napi_del - remove a NAPI context
2378 * @napi: NAPI context
2380 * Warning: caller must observe RCU grace period before freeing memory
2381 * containing @napi. Drivers might want to call this helper to combine
2382 * all the needed RCU grace periods into a single one.
2384 void __netif_napi_del(struct napi_struct *napi);
2387 * netif_napi_del - remove a NAPI context
2388 * @napi: NAPI context
2390 * netif_napi_del() removes a NAPI context from the network device NAPI list
2392 static inline void netif_napi_del(struct napi_struct *napi)
2394 __netif_napi_del(napi);
2398 struct napi_gro_cb {
2399 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2402 /* Length of frag0. */
2403 unsigned int frag0_len;
2405 /* This indicates where we are processing relative to skb->data. */
2408 /* This is non-zero if the packet cannot be merged with the new skb. */
2411 /* Save the IP ID here and check when we get to the transport layer */
2414 /* Number of segments aggregated. */
2417 /* Start offset for remote checksum offload */
2418 u16 gro_remcsum_start;
2420 /* jiffies when first packet was created/queued */
2423 /* Used in ipv6_gro_receive() and foo-over-udp */
2426 /* This is non-zero if the packet may be of the same flow. */
2429 /* Used in tunnel GRO receive */
2432 /* GRO checksum is valid */
2435 /* Number of checksums via CHECKSUM_UNNECESSARY */
2440 #define NAPI_GRO_FREE 1
2441 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2443 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2446 /* Used in GRE, set in fou/gue_gro_receive */
2449 /* Used to determine if flush_id can be ignored */
2452 /* Number of gro_receive callbacks this packet already went through */
2453 u8 recursion_counter:4;
2455 /* GRO is done by frag_list pointer chaining. */
2458 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2461 /* used in skb_gro_receive() slow path */
2462 struct sk_buff *last;
2465 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2467 #define GRO_RECURSION_LIMIT 15
2468 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2470 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2473 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2474 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2475 struct list_head *head,
2476 struct sk_buff *skb)
2478 if (unlikely(gro_recursion_inc_test(skb))) {
2479 NAPI_GRO_CB(skb)->flush |= 1;
2483 return cb(head, skb);
2486 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2488 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2490 struct list_head *head,
2491 struct sk_buff *skb)
2493 if (unlikely(gro_recursion_inc_test(skb))) {
2494 NAPI_GRO_CB(skb)->flush |= 1;
2498 return cb(sk, head, skb);
2501 struct packet_type {
2502 __be16 type; /* This is really htons(ether_type). */
2503 bool ignore_outgoing;
2504 struct net_device *dev; /* NULL is wildcarded here */
2505 int (*func) (struct sk_buff *,
2506 struct net_device *,
2507 struct packet_type *,
2508 struct net_device *);
2509 void (*list_func) (struct list_head *,
2510 struct packet_type *,
2511 struct net_device *);
2512 bool (*id_match)(struct packet_type *ptype,
2514 void *af_packet_priv;
2515 struct list_head list;
2518 struct offload_callbacks {
2519 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2520 netdev_features_t features);
2521 struct sk_buff *(*gro_receive)(struct list_head *head,
2522 struct sk_buff *skb);
2523 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2526 struct packet_offload {
2527 __be16 type; /* This is really htons(ether_type). */
2529 struct offload_callbacks callbacks;
2530 struct list_head list;
2533 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2534 struct pcpu_sw_netstats {
2539 struct u64_stats_sync syncp;
2540 } __aligned(4 * sizeof(u64));
2542 struct pcpu_lstats {
2543 u64_stats_t packets;
2545 struct u64_stats_sync syncp;
2546 } __aligned(2 * sizeof(u64));
2548 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2550 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2552 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2554 u64_stats_update_begin(&tstats->syncp);
2555 tstats->rx_bytes += len;
2556 tstats->rx_packets++;
2557 u64_stats_update_end(&tstats->syncp);
2560 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2562 struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2564 u64_stats_update_begin(&lstats->syncp);
2565 u64_stats_add(&lstats->bytes, len);
2566 u64_stats_inc(&lstats->packets);
2567 u64_stats_update_end(&lstats->syncp);
2570 #define __netdev_alloc_pcpu_stats(type, gfp) \
2572 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2575 for_each_possible_cpu(__cpu) { \
2576 typeof(type) *stat; \
2577 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2578 u64_stats_init(&stat->syncp); \
2584 #define netdev_alloc_pcpu_stats(type) \
2585 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2587 enum netdev_lag_tx_type {
2588 NETDEV_LAG_TX_TYPE_UNKNOWN,
2589 NETDEV_LAG_TX_TYPE_RANDOM,
2590 NETDEV_LAG_TX_TYPE_BROADCAST,
2591 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2592 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2593 NETDEV_LAG_TX_TYPE_HASH,
2596 enum netdev_lag_hash {
2597 NETDEV_LAG_HASH_NONE,
2599 NETDEV_LAG_HASH_L34,
2600 NETDEV_LAG_HASH_L23,
2601 NETDEV_LAG_HASH_E23,
2602 NETDEV_LAG_HASH_E34,
2603 NETDEV_LAG_HASH_UNKNOWN,
2606 struct netdev_lag_upper_info {
2607 enum netdev_lag_tx_type tx_type;
2608 enum netdev_lag_hash hash_type;
2611 struct netdev_lag_lower_state_info {
2616 #include <linux/notifier.h>
2618 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2619 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2623 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2625 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2626 detected a hardware crash and restarted
2627 - we can use this eg to kick tcp sessions
2629 NETDEV_CHANGE, /* Notify device state change */
2632 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2633 NETDEV_CHANGEADDR, /* notify after the address change */
2634 NETDEV_PRE_CHANGEADDR, /* notify before the address change */
2638 NETDEV_BONDING_FAILOVER,
2640 NETDEV_PRE_TYPE_CHANGE,
2641 NETDEV_POST_TYPE_CHANGE,
2644 NETDEV_NOTIFY_PEERS,
2648 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2649 NETDEV_CHANGEINFODATA,
2650 NETDEV_BONDING_INFO,
2651 NETDEV_PRECHANGEUPPER,
2652 NETDEV_CHANGELOWERSTATE,
2653 NETDEV_UDP_TUNNEL_PUSH_INFO,
2654 NETDEV_UDP_TUNNEL_DROP_INFO,
2655 NETDEV_CHANGE_TX_QUEUE_LEN,
2656 NETDEV_CVLAN_FILTER_PUSH_INFO,
2657 NETDEV_CVLAN_FILTER_DROP_INFO,
2658 NETDEV_SVLAN_FILTER_PUSH_INFO,
2659 NETDEV_SVLAN_FILTER_DROP_INFO,
2661 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2663 int register_netdevice_notifier(struct notifier_block *nb);
2664 int unregister_netdevice_notifier(struct notifier_block *nb);
2665 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2666 int unregister_netdevice_notifier_net(struct net *net,
2667 struct notifier_block *nb);
2668 int register_netdevice_notifier_dev_net(struct net_device *dev,
2669 struct notifier_block *nb,
2670 struct netdev_net_notifier *nn);
2671 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2672 struct notifier_block *nb,
2673 struct netdev_net_notifier *nn);
2675 struct netdev_notifier_info {
2676 struct net_device *dev;
2677 struct netlink_ext_ack *extack;
2680 struct netdev_notifier_info_ext {
2681 struct netdev_notifier_info info; /* must be first */
2687 struct netdev_notifier_change_info {
2688 struct netdev_notifier_info info; /* must be first */
2689 unsigned int flags_changed;
2692 struct netdev_notifier_changeupper_info {
2693 struct netdev_notifier_info info; /* must be first */
2694 struct net_device *upper_dev; /* new upper dev */
2695 bool master; /* is upper dev master */
2696 bool linking; /* is the notification for link or unlink */
2697 void *upper_info; /* upper dev info */
2700 struct netdev_notifier_changelowerstate_info {
2701 struct netdev_notifier_info info; /* must be first */
2702 void *lower_state_info; /* is lower dev state */
2705 struct netdev_notifier_pre_changeaddr_info {
2706 struct netdev_notifier_info info; /* must be first */
2707 const unsigned char *dev_addr;
2710 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2711 struct net_device *dev)
2714 info->extack = NULL;
2717 static inline struct net_device *
2718 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2723 static inline struct netlink_ext_ack *
2724 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2726 return info->extack;
2729 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2732 extern rwlock_t dev_base_lock; /* Device list lock */
2734 #define for_each_netdev(net, d) \
2735 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2736 #define for_each_netdev_reverse(net, d) \
2737 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2738 #define for_each_netdev_rcu(net, d) \
2739 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2740 #define for_each_netdev_safe(net, d, n) \
2741 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2742 #define for_each_netdev_continue(net, d) \
2743 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2744 #define for_each_netdev_continue_reverse(net, d) \
2745 list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2747 #define for_each_netdev_continue_rcu(net, d) \
2748 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2749 #define for_each_netdev_in_bond_rcu(bond, slave) \
2750 for_each_netdev_rcu(&init_net, slave) \
2751 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2752 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2754 static inline struct net_device *next_net_device(struct net_device *dev)
2756 struct list_head *lh;
2760 lh = dev->dev_list.next;
2761 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2764 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2766 struct list_head *lh;
2770 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2771 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2774 static inline struct net_device *first_net_device(struct net *net)
2776 return list_empty(&net->dev_base_head) ? NULL :
2777 net_device_entry(net->dev_base_head.next);
2780 static inline struct net_device *first_net_device_rcu(struct net *net)
2782 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2784 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2787 int netdev_boot_setup_check(struct net_device *dev);
2788 unsigned long netdev_boot_base(const char *prefix, int unit);
2789 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2790 const char *hwaddr);
2791 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2792 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2793 void dev_add_pack(struct packet_type *pt);
2794 void dev_remove_pack(struct packet_type *pt);
2795 void __dev_remove_pack(struct packet_type *pt);
2796 void dev_add_offload(struct packet_offload *po);
2797 void dev_remove_offload(struct packet_offload *po);
2799 int dev_get_iflink(const struct net_device *dev);
2800 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2801 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2802 unsigned short mask);
2803 struct net_device *dev_get_by_name(struct net *net, const char *name);
2804 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2805 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2806 int dev_alloc_name(struct net_device *dev, const char *name);
2807 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2808 void dev_close(struct net_device *dev);
2809 void dev_close_many(struct list_head *head, bool unlink);
2810 void dev_disable_lro(struct net_device *dev);
2811 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2812 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2813 struct net_device *sb_dev);
2814 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2815 struct net_device *sb_dev);
2816 int dev_queue_xmit(struct sk_buff *skb);
2817 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2818 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2819 int register_netdevice(struct net_device *dev);
2820 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2821 void unregister_netdevice_many(struct list_head *head);
2822 static inline void unregister_netdevice(struct net_device *dev)
2824 unregister_netdevice_queue(dev, NULL);
2827 int netdev_refcnt_read(const struct net_device *dev);
2828 void free_netdev(struct net_device *dev);
2829 void netdev_freemem(struct net_device *dev);
2830 int init_dummy_netdev(struct net_device *dev);
2832 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
2833 struct sk_buff *skb,
2835 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2836 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2837 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2838 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2839 int netdev_get_name(struct net *net, char *name, int ifindex);
2840 int dev_restart(struct net_device *dev);
2841 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2842 int skb_gro_receive_list(struct sk_buff *p, struct sk_buff *skb);
2844 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2846 return NAPI_GRO_CB(skb)->data_offset;
2849 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2851 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2854 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2856 NAPI_GRO_CB(skb)->data_offset += len;
2859 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2860 unsigned int offset)
2862 return NAPI_GRO_CB(skb)->frag0 + offset;
2865 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2867 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2870 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2872 NAPI_GRO_CB(skb)->frag0 = NULL;
2873 NAPI_GRO_CB(skb)->frag0_len = 0;
2876 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2877 unsigned int offset)
2879 if (!pskb_may_pull(skb, hlen))
2882 skb_gro_frag0_invalidate(skb);
2883 return skb->data + offset;
2886 static inline void *skb_gro_network_header(struct sk_buff *skb)
2888 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2889 skb_network_offset(skb);
2892 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2893 const void *start, unsigned int len)
2895 if (NAPI_GRO_CB(skb)->csum_valid)
2896 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2897 csum_partial(start, len, 0));
2900 /* GRO checksum functions. These are logical equivalents of the normal
2901 * checksum functions (in skbuff.h) except that they operate on the GRO
2902 * offsets and fields in sk_buff.
2905 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2907 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2909 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2912 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2916 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2917 skb_checksum_start_offset(skb) <
2918 skb_gro_offset(skb)) &&
2919 !skb_at_gro_remcsum_start(skb) &&
2920 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2921 (!zero_okay || check));
2924 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2927 if (NAPI_GRO_CB(skb)->csum_valid &&
2928 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2931 NAPI_GRO_CB(skb)->csum = psum;
2933 return __skb_gro_checksum_complete(skb);
2936 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2938 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2939 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2940 NAPI_GRO_CB(skb)->csum_cnt--;
2942 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2943 * verified a new top level checksum or an encapsulated one
2944 * during GRO. This saves work if we fallback to normal path.
2946 __skb_incr_checksum_unnecessary(skb);
2950 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2953 __sum16 __ret = 0; \
2954 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2955 __ret = __skb_gro_checksum_validate_complete(skb, \
2956 compute_pseudo(skb, proto)); \
2958 skb_gro_incr_csum_unnecessary(skb); \
2962 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2963 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2965 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2967 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2969 #define skb_gro_checksum_simple_validate(skb) \
2970 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2972 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2974 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2975 !NAPI_GRO_CB(skb)->csum_valid);
2978 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2981 NAPI_GRO_CB(skb)->csum = ~pseudo;
2982 NAPI_GRO_CB(skb)->csum_valid = 1;
2985 #define skb_gro_checksum_try_convert(skb, proto, compute_pseudo) \
2987 if (__skb_gro_checksum_convert_check(skb)) \
2988 __skb_gro_checksum_convert(skb, \
2989 compute_pseudo(skb, proto)); \
2992 struct gro_remcsum {
2997 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
3003 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
3004 unsigned int off, size_t hdrlen,
3005 int start, int offset,
3006 struct gro_remcsum *grc,
3010 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
3012 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
3015 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
3019 ptr = skb_gro_header_fast(skb, off);
3020 if (skb_gro_header_hard(skb, off + plen)) {
3021 ptr = skb_gro_header_slow(skb, off + plen, off);
3026 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
3029 /* Adjust skb->csum since we changed the packet */
3030 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
3032 grc->offset = off + hdrlen + offset;
3038 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
3039 struct gro_remcsum *grc)
3042 size_t plen = grc->offset + sizeof(u16);
3047 ptr = skb_gro_header_fast(skb, grc->offset);
3048 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
3049 ptr = skb_gro_header_slow(skb, plen, grc->offset);
3054 remcsum_unadjust((__sum16 *)ptr, grc->delta);
3057 #ifdef CONFIG_XFRM_OFFLOAD
3058 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3060 if (PTR_ERR(pp) != -EINPROGRESS)
3061 NAPI_GRO_CB(skb)->flush |= flush;
3063 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3066 struct gro_remcsum *grc)
3068 if (PTR_ERR(pp) != -EINPROGRESS) {
3069 NAPI_GRO_CB(skb)->flush |= flush;
3070 skb_gro_remcsum_cleanup(skb, grc);
3071 skb->remcsum_offload = 0;
3075 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
3077 NAPI_GRO_CB(skb)->flush |= flush;
3079 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
3082 struct gro_remcsum *grc)
3084 NAPI_GRO_CB(skb)->flush |= flush;
3085 skb_gro_remcsum_cleanup(skb, grc);
3086 skb->remcsum_offload = 0;
3090 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3091 unsigned short type,
3092 const void *daddr, const void *saddr,
3095 if (!dev->header_ops || !dev->header_ops->create)
3098 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3101 static inline int dev_parse_header(const struct sk_buff *skb,
3102 unsigned char *haddr)
3104 const struct net_device *dev = skb->dev;
3106 if (!dev->header_ops || !dev->header_ops->parse)
3108 return dev->header_ops->parse(skb, haddr);
3111 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3113 const struct net_device *dev = skb->dev;
3115 if (!dev->header_ops || !dev->header_ops->parse_protocol)
3117 return dev->header_ops->parse_protocol(skb);
3120 /* ll_header must have at least hard_header_len allocated */
3121 static inline bool dev_validate_header(const struct net_device *dev,
3122 char *ll_header, int len)
3124 if (likely(len >= dev->hard_header_len))
3126 if (len < dev->min_header_len)
3129 if (capable(CAP_SYS_RAWIO)) {
3130 memset(ll_header + len, 0, dev->hard_header_len - len);
3134 if (dev->header_ops && dev->header_ops->validate)
3135 return dev->header_ops->validate(ll_header, len);
3140 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
3142 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
3143 static inline int unregister_gifconf(unsigned int family)
3145 return register_gifconf(family, NULL);
3148 #ifdef CONFIG_NET_FLOW_LIMIT
3149 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
3150 struct sd_flow_limit {
3152 unsigned int num_buckets;
3153 unsigned int history_head;
3154 u16 history[FLOW_LIMIT_HISTORY];
3158 extern int netdev_flow_limit_table_len;
3159 #endif /* CONFIG_NET_FLOW_LIMIT */
3162 * Incoming packets are placed on per-CPU queues
3164 struct softnet_data {
3165 struct list_head poll_list;
3166 struct sk_buff_head process_queue;
3169 unsigned int processed;
3170 unsigned int time_squeeze;
3171 unsigned int received_rps;
3173 struct softnet_data *rps_ipi_list;
3175 #ifdef CONFIG_NET_FLOW_LIMIT
3176 struct sd_flow_limit __rcu *flow_limit;
3178 struct Qdisc *output_queue;
3179 struct Qdisc **output_queue_tailp;
3180 struct sk_buff *completion_queue;
3181 #ifdef CONFIG_XFRM_OFFLOAD
3182 struct sk_buff_head xfrm_backlog;
3184 /* written and read only by owning cpu: */
3190 /* input_queue_head should be written by cpu owning this struct,
3191 * and only read by other cpus. Worth using a cache line.
3193 unsigned int input_queue_head ____cacheline_aligned_in_smp;
3195 /* Elements below can be accessed between CPUs for RPS/RFS */
3196 call_single_data_t csd ____cacheline_aligned_in_smp;
3197 struct softnet_data *rps_ipi_next;
3199 unsigned int input_queue_tail;
3201 unsigned int dropped;
3202 struct sk_buff_head input_pkt_queue;
3203 struct napi_struct backlog;
3207 static inline void input_queue_head_incr(struct softnet_data *sd)
3210 sd->input_queue_head++;
3214 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3215 unsigned int *qtail)
3218 *qtail = ++sd->input_queue_tail;
3222 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3224 static inline int dev_recursion_level(void)
3226 return this_cpu_read(softnet_data.xmit.recursion);
3229 #define XMIT_RECURSION_LIMIT 8
3230 static inline bool dev_xmit_recursion(void)
3232 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3233 XMIT_RECURSION_LIMIT);
3236 static inline void dev_xmit_recursion_inc(void)
3238 __this_cpu_inc(softnet_data.xmit.recursion);
3241 static inline void dev_xmit_recursion_dec(void)
3243 __this_cpu_dec(softnet_data.xmit.recursion);
3246 void __netif_schedule(struct Qdisc *q);
3247 void netif_schedule_queue(struct netdev_queue *txq);
3249 static inline void netif_tx_schedule_all(struct net_device *dev)
3253 for (i = 0; i < dev->num_tx_queues; i++)
3254 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3257 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3259 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3263 * netif_start_queue - allow transmit
3264 * @dev: network device
3266 * Allow upper layers to call the device hard_start_xmit routine.
3268 static inline void netif_start_queue(struct net_device *dev)
3270 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3273 static inline void netif_tx_start_all_queues(struct net_device *dev)
3277 for (i = 0; i < dev->num_tx_queues; i++) {
3278 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3279 netif_tx_start_queue(txq);
3283 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3286 * netif_wake_queue - restart transmit
3287 * @dev: network device
3289 * Allow upper layers to call the device hard_start_xmit routine.
3290 * Used for flow control when transmit resources are available.
3292 static inline void netif_wake_queue(struct net_device *dev)
3294 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3297 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3301 for (i = 0; i < dev->num_tx_queues; i++) {
3302 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3303 netif_tx_wake_queue(txq);
3307 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3309 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3313 * netif_stop_queue - stop transmitted packets
3314 * @dev: network device
3316 * Stop upper layers calling the device hard_start_xmit routine.
3317 * Used for flow control when transmit resources are unavailable.
3319 static inline void netif_stop_queue(struct net_device *dev)
3321 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3324 void netif_tx_stop_all_queues(struct net_device *dev);
3326 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3328 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3332 * netif_queue_stopped - test if transmit queue is flowblocked
3333 * @dev: network device
3335 * Test if transmit queue on device is currently unable to send.
3337 static inline bool netif_queue_stopped(const struct net_device *dev)
3339 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3342 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3344 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3348 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3350 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3354 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3356 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3360 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3361 * @dev_queue: pointer to transmit queue
3363 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3364 * to give appropriate hint to the CPU.
3366 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3369 prefetchw(&dev_queue->dql.num_queued);
3374 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3375 * @dev_queue: pointer to transmit queue
3377 * BQL enabled drivers might use this helper in their TX completion path,
3378 * to give appropriate hint to the CPU.
3380 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3383 prefetchw(&dev_queue->dql.limit);
3387 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3391 dql_queued(&dev_queue->dql, bytes);
3393 if (likely(dql_avail(&dev_queue->dql) >= 0))
3396 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3399 * The XOFF flag must be set before checking the dql_avail below,
3400 * because in netdev_tx_completed_queue we update the dql_completed
3401 * before checking the XOFF flag.
3405 /* check again in case another CPU has just made room avail */
3406 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3407 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3411 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3412 * that they should not test BQL status themselves.
3413 * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3415 * Returns true if the doorbell must be used to kick the NIC.
3417 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3423 dql_queued(&dev_queue->dql, bytes);
3425 return netif_tx_queue_stopped(dev_queue);
3427 netdev_tx_sent_queue(dev_queue, bytes);
3432 * netdev_sent_queue - report the number of bytes queued to hardware
3433 * @dev: network device
3434 * @bytes: number of bytes queued to the hardware device queue
3436 * Report the number of bytes queued for sending/completion to the network
3437 * device hardware queue. @bytes should be a good approximation and should
3438 * exactly match netdev_completed_queue() @bytes
3440 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3442 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3445 static inline bool __netdev_sent_queue(struct net_device *dev,
3449 return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3453 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3454 unsigned int pkts, unsigned int bytes)
3457 if (unlikely(!bytes))
3460 dql_completed(&dev_queue->dql, bytes);
3463 * Without the memory barrier there is a small possiblity that
3464 * netdev_tx_sent_queue will miss the update and cause the queue to
3465 * be stopped forever
3469 if (unlikely(dql_avail(&dev_queue->dql) < 0))
3472 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3473 netif_schedule_queue(dev_queue);
3478 * netdev_completed_queue - report bytes and packets completed by device
3479 * @dev: network device
3480 * @pkts: actual number of packets sent over the medium
3481 * @bytes: actual number of bytes sent over the medium
3483 * Report the number of bytes and packets transmitted by the network device
3484 * hardware queue over the physical medium, @bytes must exactly match the
3485 * @bytes amount passed to netdev_sent_queue()
3487 static inline void netdev_completed_queue(struct net_device *dev,
3488 unsigned int pkts, unsigned int bytes)
3490 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3493 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3496 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3502 * netdev_reset_queue - reset the packets and bytes count of a network device
3503 * @dev_queue: network device
3505 * Reset the bytes and packet count of a network device and clear the
3506 * software flow control OFF bit for this network device
3508 static inline void netdev_reset_queue(struct net_device *dev_queue)
3510 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3514 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3515 * @dev: network device
3516 * @queue_index: given tx queue index
3518 * Returns 0 if given tx queue index >= number of device tx queues,
3519 * otherwise returns the originally passed tx queue index.
3521 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3523 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3524 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3525 dev->name, queue_index,
3526 dev->real_num_tx_queues);
3534 * netif_running - test if up
3535 * @dev: network device
3537 * Test if the device has been brought up.
3539 static inline bool netif_running(const struct net_device *dev)
3541 return test_bit(__LINK_STATE_START, &dev->state);
3545 * Routines to manage the subqueues on a device. We only need start,
3546 * stop, and a check if it's stopped. All other device management is
3547 * done at the overall netdevice level.
3548 * Also test the device if we're multiqueue.
3552 * netif_start_subqueue - allow sending packets on subqueue
3553 * @dev: network device
3554 * @queue_index: sub queue index
3556 * Start individual transmit queue of a device with multiple transmit queues.
3558 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3560 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3562 netif_tx_start_queue(txq);
3566 * netif_stop_subqueue - stop sending packets on subqueue
3567 * @dev: network device
3568 * @queue_index: sub queue index
3570 * Stop individual transmit queue of a device with multiple transmit queues.
3572 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3574 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3575 netif_tx_stop_queue(txq);
3579 * netif_subqueue_stopped - test status of subqueue
3580 * @dev: network device
3581 * @queue_index: sub queue index
3583 * Check individual transmit queue of a device with multiple transmit queues.
3585 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3588 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3590 return netif_tx_queue_stopped(txq);
3593 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3594 struct sk_buff *skb)
3596 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3600 * netif_wake_subqueue - allow sending packets on subqueue
3601 * @dev: network device
3602 * @queue_index: sub queue index
3604 * Resume individual transmit queue of a device with multiple transmit queues.
3606 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3608 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3610 netif_tx_wake_queue(txq);
3614 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3616 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3617 u16 index, bool is_rxqs_map);
3620 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3621 * @j: CPU/Rx queue index
3622 * @mask: bitmask of all cpus/rx queues
3623 * @nr_bits: number of bits in the bitmask
3625 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3627 static inline bool netif_attr_test_mask(unsigned long j,
3628 const unsigned long *mask,
3629 unsigned int nr_bits)
3631 cpu_max_bits_warn(j, nr_bits);
3632 return test_bit(j, mask);
3636 * netif_attr_test_online - Test for online CPU/Rx queue
3637 * @j: CPU/Rx queue index
3638 * @online_mask: bitmask for CPUs/Rx queues that are online
3639 * @nr_bits: number of bits in the bitmask
3641 * Returns true if a CPU/Rx queue is online.
3643 static inline bool netif_attr_test_online(unsigned long j,
3644 const unsigned long *online_mask,
3645 unsigned int nr_bits)
3647 cpu_max_bits_warn(j, nr_bits);
3650 return test_bit(j, online_mask);
3652 return (j < nr_bits);
3656 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3657 * @n: CPU/Rx queue index
3658 * @srcp: the cpumask/Rx queue mask pointer
3659 * @nr_bits: number of bits in the bitmask
3661 * Returns >= nr_bits if no further CPUs/Rx queues set.
3663 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3664 unsigned int nr_bits)
3666 /* -1 is a legal arg here. */
3668 cpu_max_bits_warn(n, nr_bits);
3671 return find_next_bit(srcp, nr_bits, n + 1);
3677 * netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3678 * @n: CPU/Rx queue index
3679 * @src1p: the first CPUs/Rx queues mask pointer
3680 * @src2p: the second CPUs/Rx queues mask pointer
3681 * @nr_bits: number of bits in the bitmask
3683 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3685 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3686 const unsigned long *src2p,
3687 unsigned int nr_bits)
3689 /* -1 is a legal arg here. */
3691 cpu_max_bits_warn(n, nr_bits);
3694 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3696 return find_next_bit(src1p, nr_bits, n + 1);
3698 return find_next_bit(src2p, nr_bits, n + 1);
3703 static inline int netif_set_xps_queue(struct net_device *dev,
3704 const struct cpumask *mask,
3710 static inline int __netif_set_xps_queue(struct net_device *dev,
3711 const unsigned long *mask,
3712 u16 index, bool is_rxqs_map)
3719 * netif_is_multiqueue - test if device has multiple transmit queues
3720 * @dev: network device
3722 * Check if device has multiple transmit queues
3724 static inline bool netif_is_multiqueue(const struct net_device *dev)
3726 return dev->num_tx_queues > 1;
3729 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3732 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3734 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3737 dev->real_num_rx_queues = rxqs;
3742 static inline struct netdev_rx_queue *
3743 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3745 return dev->_rx + rxq;
3749 static inline unsigned int get_netdev_rx_queue_index(
3750 struct netdev_rx_queue *queue)
3752 struct net_device *dev = queue->dev;
3753 int index = queue - dev->_rx;
3755 BUG_ON(index >= dev->num_rx_queues);
3760 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3761 int netif_get_num_default_rss_queues(void);
3763 enum skb_free_reason {
3764 SKB_REASON_CONSUMED,
3768 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3769 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3772 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3773 * interrupt context or with hardware interrupts being disabled.
3774 * (in_irq() || irqs_disabled())
3776 * We provide four helpers that can be used in following contexts :
3778 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3779 * replacing kfree_skb(skb)
3781 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3782 * Typically used in place of consume_skb(skb) in TX completion path
3784 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3785 * replacing kfree_skb(skb)
3787 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3788 * and consumed a packet. Used in place of consume_skb(skb)
3790 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3792 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3795 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3797 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3800 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3802 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3805 static inline void dev_consume_skb_any(struct sk_buff *skb)
3807 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3810 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3811 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3812 int netif_rx(struct sk_buff *skb);
3813 int netif_rx_ni(struct sk_buff *skb);
3814 int netif_rx_any_context(struct sk_buff *skb);
3815 int netif_receive_skb(struct sk_buff *skb);
3816 int netif_receive_skb_core(struct sk_buff *skb);
3817 void netif_receive_skb_list(struct list_head *head);
3818 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3819 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3820 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3821 gro_result_t napi_gro_frags(struct napi_struct *napi);
3822 struct packet_offload *gro_find_receive_by_type(__be16 type);
3823 struct packet_offload *gro_find_complete_by_type(__be16 type);
3825 static inline void napi_free_frags(struct napi_struct *napi)
3827 kfree_skb(napi->skb);
3831 bool netdev_is_rx_handler_busy(struct net_device *dev);
3832 int netdev_rx_handler_register(struct net_device *dev,
3833 rx_handler_func_t *rx_handler,
3834 void *rx_handler_data);
3835 void netdev_rx_handler_unregister(struct net_device *dev);
3837 bool dev_valid_name(const char *name);
3838 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3839 bool *need_copyout);
3840 int dev_ifconf(struct net *net, struct ifconf *, int);
3841 int dev_ethtool(struct net *net, struct ifreq *);
3842 unsigned int dev_get_flags(const struct net_device *);
3843 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3844 struct netlink_ext_ack *extack);
3845 int dev_change_flags(struct net_device *dev, unsigned int flags,
3846 struct netlink_ext_ack *extack);
3847 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3848 unsigned int gchanges);
3849 int dev_change_name(struct net_device *, const char *);
3850 int dev_set_alias(struct net_device *, const char *, size_t);
3851 int dev_get_alias(const struct net_device *, char *, size_t);
3852 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3853 int __dev_set_mtu(struct net_device *, int);
3854 int dev_validate_mtu(struct net_device *dev, int mtu,
3855 struct netlink_ext_ack *extack);
3856 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3857 struct netlink_ext_ack *extack);
3858 int dev_set_mtu(struct net_device *, int);
3859 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3860 void dev_set_group(struct net_device *, int);
3861 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3862 struct netlink_ext_ack *extack);
3863 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3864 struct netlink_ext_ack *extack);
3865 int dev_change_carrier(struct net_device *, bool new_carrier);
3866 int dev_get_phys_port_id(struct net_device *dev,
3867 struct netdev_phys_item_id *ppid);
3868 int dev_get_phys_port_name(struct net_device *dev,
3869 char *name, size_t len);
3870 int dev_get_port_parent_id(struct net_device *dev,
3871 struct netdev_phys_item_id *ppid, bool recurse);
3872 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3873 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3874 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3875 void dev_change_proto_down_reason(struct net_device *dev, unsigned long mask,
3877 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3878 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3879 struct netdev_queue *txq, int *ret);
3881 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3882 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3883 int fd, int expected_fd, u32 flags);
3884 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3885 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3887 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3889 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3890 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3891 bool is_skb_forwardable(const struct net_device *dev,
3892 const struct sk_buff *skb);
3894 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3895 struct sk_buff *skb)
3897 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3898 unlikely(!is_skb_forwardable(dev, skb))) {
3899 atomic_long_inc(&dev->rx_dropped);
3904 skb_scrub_packet(skb, true);
3909 bool dev_nit_active(struct net_device *dev);
3910 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3912 extern int netdev_budget;
3913 extern unsigned int netdev_budget_usecs;
3915 /* Called by rtnetlink.c:rtnl_unlock() */
3916 void netdev_run_todo(void);
3919 * dev_put - release reference to device
3920 * @dev: network device
3922 * Release reference to device to allow it to be freed.
3924 static inline void dev_put(struct net_device *dev)
3926 this_cpu_dec(*dev->pcpu_refcnt);
3930 * dev_hold - get reference to device
3931 * @dev: network device
3933 * Hold reference to device to keep it from being freed.
3935 static inline void dev_hold(struct net_device *dev)
3937 this_cpu_inc(*dev->pcpu_refcnt);
3940 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3941 * and _off may be called from IRQ context, but it is caller
3942 * who is responsible for serialization of these calls.
3944 * The name carrier is inappropriate, these functions should really be
3945 * called netif_lowerlayer_*() because they represent the state of any
3946 * kind of lower layer not just hardware media.
3949 void linkwatch_init_dev(struct net_device *dev);
3950 void linkwatch_fire_event(struct net_device *dev);
3951 void linkwatch_forget_dev(struct net_device *dev);
3954 * netif_carrier_ok - test if carrier present
3955 * @dev: network device
3957 * Check if carrier is present on device
3959 static inline bool netif_carrier_ok(const struct net_device *dev)
3961 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3964 unsigned long dev_trans_start(struct net_device *dev);
3966 void __netdev_watchdog_up(struct net_device *dev);
3968 void netif_carrier_on(struct net_device *dev);
3970 void netif_carrier_off(struct net_device *dev);
3973 * netif_dormant_on - mark device as dormant.
3974 * @dev: network device
3976 * Mark device as dormant (as per RFC2863).
3978 * The dormant state indicates that the relevant interface is not
3979 * actually in a condition to pass packets (i.e., it is not 'up') but is
3980 * in a "pending" state, waiting for some external event. For "on-
3981 * demand" interfaces, this new state identifies the situation where the
3982 * interface is waiting for events to place it in the up state.
3984 static inline void netif_dormant_on(struct net_device *dev)
3986 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3987 linkwatch_fire_event(dev);
3991 * netif_dormant_off - set device as not dormant.
3992 * @dev: network device
3994 * Device is not in dormant state.
3996 static inline void netif_dormant_off(struct net_device *dev)
3998 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3999 linkwatch_fire_event(dev);
4003 * netif_dormant - test if device is dormant
4004 * @dev: network device
4006 * Check if device is dormant.
4008 static inline bool netif_dormant(const struct net_device *dev)
4010 return test_bit(__LINK_STATE_DORMANT, &dev->state);
4015 * netif_testing_on - mark device as under test.
4016 * @dev: network device
4018 * Mark device as under test (as per RFC2863).
4020 * The testing state indicates that some test(s) must be performed on
4021 * the interface. After completion, of the test, the interface state
4022 * will change to up, dormant, or down, as appropriate.
4024 static inline void netif_testing_on(struct net_device *dev)
4026 if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4027 linkwatch_fire_event(dev);
4031 * netif_testing_off - set device as not under test.
4032 * @dev: network device
4034 * Device is not in testing state.
4036 static inline void netif_testing_off(struct net_device *dev)
4038 if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4039 linkwatch_fire_event(dev);
4043 * netif_testing - test if device is under test
4044 * @dev: network device
4046 * Check if device is under test
4048 static inline bool netif_testing(const struct net_device *dev)
4050 return test_bit(__LINK_STATE_TESTING, &dev->state);
4055 * netif_oper_up - test if device is operational
4056 * @dev: network device
4058 * Check if carrier is operational
4060 static inline bool netif_oper_up(const struct net_device *dev)
4062 return (dev->operstate == IF_OPER_UP ||
4063 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4067 * netif_device_present - is device available or removed
4068 * @dev: network device
4070 * Check if device has not been removed from system.
4072 static inline bool netif_device_present(struct net_device *dev)
4074 return test_bit(__LINK_STATE_PRESENT, &dev->state);
4077 void netif_device_detach(struct net_device *dev);
4079 void netif_device_attach(struct net_device *dev);
4082 * Network interface message level settings
4087 NETIF_MSG_PROBE_BIT,
4089 NETIF_MSG_TIMER_BIT,
4090 NETIF_MSG_IFDOWN_BIT,
4092 NETIF_MSG_RX_ERR_BIT,
4093 NETIF_MSG_TX_ERR_BIT,
4094 NETIF_MSG_TX_QUEUED_BIT,
4096 NETIF_MSG_TX_DONE_BIT,
4097 NETIF_MSG_RX_STATUS_BIT,
4098 NETIF_MSG_PKTDATA_BIT,
4102 /* When you add a new bit above, update netif_msg_class_names array
4103 * in net/ethtool/common.c
4105 NETIF_MSG_CLASS_COUNT,
4107 /* Both ethtool_ops interface and internal driver implementation use u32 */
4108 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4110 #define __NETIF_MSG_BIT(bit) ((u32)1 << (bit))
4111 #define __NETIF_MSG(name) __NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4113 #define NETIF_MSG_DRV __NETIF_MSG(DRV)
4114 #define NETIF_MSG_PROBE __NETIF_MSG(PROBE)
4115 #define NETIF_MSG_LINK __NETIF_MSG(LINK)
4116 #define NETIF_MSG_TIMER __NETIF_MSG(TIMER)
4117 #define NETIF_MSG_IFDOWN __NETIF_MSG(IFDOWN)
4118 #define NETIF_MSG_IFUP __NETIF_MSG(IFUP)
4119 #define NETIF_MSG_RX_ERR __NETIF_MSG(RX_ERR)
4120 #define NETIF_MSG_TX_ERR __NETIF_MSG(TX_ERR)
4121 #define NETIF_MSG_TX_QUEUED __NETIF_MSG(TX_QUEUED)
4122 #define NETIF_MSG_INTR __NETIF_MSG(INTR)
4123 #define NETIF_MSG_TX_DONE __NETIF_MSG(TX_DONE)
4124 #define NETIF_MSG_RX_STATUS __NETIF_MSG(RX_STATUS)
4125 #define NETIF_MSG_PKTDATA __NETIF_MSG(PKTDATA)
4126 #define NETIF_MSG_HW __NETIF_MSG(HW)
4127 #define NETIF_MSG_WOL __NETIF_MSG(WOL)
4129 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
4130 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
4131 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
4132 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
4133 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
4134 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
4135 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
4136 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
4137 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4138 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
4139 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
4140 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
4141 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
4142 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
4143 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
4145 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4148 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4149 return default_msg_enable_bits;
4150 if (debug_value == 0) /* no output */
4152 /* set low N bits */
4153 return (1U << debug_value) - 1;
4156 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4158 spin_lock(&txq->_xmit_lock);
4159 txq->xmit_lock_owner = cpu;
4162 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4164 __acquire(&txq->_xmit_lock);
4168 static inline void __netif_tx_release(struct netdev_queue *txq)
4170 __release(&txq->_xmit_lock);
4173 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4175 spin_lock_bh(&txq->_xmit_lock);
4176 txq->xmit_lock_owner = smp_processor_id();
4179 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4181 bool ok = spin_trylock(&txq->_xmit_lock);
4183 txq->xmit_lock_owner = smp_processor_id();
4187 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4189 txq->xmit_lock_owner = -1;
4190 spin_unlock(&txq->_xmit_lock);
4193 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4195 txq->xmit_lock_owner = -1;
4196 spin_unlock_bh(&txq->_xmit_lock);
4199 static inline void txq_trans_update(struct netdev_queue *txq)
4201 if (txq->xmit_lock_owner != -1)
4202 txq->trans_start = jiffies;
4205 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4206 static inline void netif_trans_update(struct net_device *dev)
4208 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4210 if (txq->trans_start != jiffies)
4211 txq->trans_start = jiffies;
4215 * netif_tx_lock - grab network device transmit lock
4216 * @dev: network device
4218 * Get network device transmit lock
4220 static inline void netif_tx_lock(struct net_device *dev)
4225 spin_lock(&dev->tx_global_lock);
4226 cpu = smp_processor_id();
4227 for (i = 0; i < dev->num_tx_queues; i++) {
4228 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4230 /* We are the only thread of execution doing a
4231 * freeze, but we have to grab the _xmit_lock in
4232 * order to synchronize with threads which are in
4233 * the ->hard_start_xmit() handler and already
4234 * checked the frozen bit.
4236 __netif_tx_lock(txq, cpu);
4237 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
4238 __netif_tx_unlock(txq);
4242 static inline void netif_tx_lock_bh(struct net_device *dev)
4248 static inline void netif_tx_unlock(struct net_device *dev)
4252 for (i = 0; i < dev->num_tx_queues; i++) {
4253 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4255 /* No need to grab the _xmit_lock here. If the
4256 * queue is not stopped for another reason, we
4259 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4260 netif_schedule_queue(txq);
4262 spin_unlock(&dev->tx_global_lock);
4265 static inline void netif_tx_unlock_bh(struct net_device *dev)
4267 netif_tx_unlock(dev);
4271 #define HARD_TX_LOCK(dev, txq, cpu) { \
4272 if ((dev->features & NETIF_F_LLTX) == 0) { \
4273 __netif_tx_lock(txq, cpu); \
4275 __netif_tx_acquire(txq); \
4279 #define HARD_TX_TRYLOCK(dev, txq) \
4280 (((dev->features & NETIF_F_LLTX) == 0) ? \
4281 __netif_tx_trylock(txq) : \
4282 __netif_tx_acquire(txq))
4284 #define HARD_TX_UNLOCK(dev, txq) { \
4285 if ((dev->features & NETIF_F_LLTX) == 0) { \
4286 __netif_tx_unlock(txq); \
4288 __netif_tx_release(txq); \
4292 static inline void netif_tx_disable(struct net_device *dev)
4298 cpu = smp_processor_id();
4299 for (i = 0; i < dev->num_tx_queues; i++) {
4300 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4302 __netif_tx_lock(txq, cpu);
4303 netif_tx_stop_queue(txq);
4304 __netif_tx_unlock(txq);
4309 static inline void netif_addr_lock(struct net_device *dev)
4311 unsigned char nest_level = 0;
4313 #ifdef CONFIG_LOCKDEP
4314 nest_level = dev->nested_level;
4316 spin_lock_nested(&dev->addr_list_lock, nest_level);
4319 static inline void netif_addr_lock_bh(struct net_device *dev)
4321 unsigned char nest_level = 0;
4323 #ifdef CONFIG_LOCKDEP
4324 nest_level = dev->nested_level;
4327 spin_lock_nested(&dev->addr_list_lock, nest_level);
4330 static inline void netif_addr_unlock(struct net_device *dev)
4332 spin_unlock(&dev->addr_list_lock);
4335 static inline void netif_addr_unlock_bh(struct net_device *dev)
4337 spin_unlock_bh(&dev->addr_list_lock);
4341 * dev_addrs walker. Should be used only for read access. Call with
4342 * rcu_read_lock held.
4344 #define for_each_dev_addr(dev, ha) \
4345 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4347 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4349 void ether_setup(struct net_device *dev);
4351 /* Support for loadable net-drivers */
4352 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4353 unsigned char name_assign_type,
4354 void (*setup)(struct net_device *),
4355 unsigned int txqs, unsigned int rxqs);
4356 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4357 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4359 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4360 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4363 int register_netdev(struct net_device *dev);
4364 void unregister_netdev(struct net_device *dev);
4366 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4368 /* General hardware address lists handling functions */
4369 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4370 struct netdev_hw_addr_list *from_list, int addr_len);
4371 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4372 struct netdev_hw_addr_list *from_list, int addr_len);
4373 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4374 struct net_device *dev,
4375 int (*sync)(struct net_device *, const unsigned char *),
4376 int (*unsync)(struct net_device *,
4377 const unsigned char *));
4378 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4379 struct net_device *dev,
4380 int (*sync)(struct net_device *,
4381 const unsigned char *, int),
4382 int (*unsync)(struct net_device *,
4383 const unsigned char *, int));
4384 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4385 struct net_device *dev,
4386 int (*unsync)(struct net_device *,
4387 const unsigned char *, int));
4388 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4389 struct net_device *dev,
4390 int (*unsync)(struct net_device *,
4391 const unsigned char *));
4392 void __hw_addr_init(struct netdev_hw_addr_list *list);
4394 /* Functions used for device addresses handling */
4395 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4396 unsigned char addr_type);
4397 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4398 unsigned char addr_type);
4399 void dev_addr_flush(struct net_device *dev);
4400 int dev_addr_init(struct net_device *dev);
4402 /* Functions used for unicast addresses handling */
4403 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4404 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4405 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4406 int dev_uc_sync(struct net_device *to, struct net_device *from);
4407 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4408 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4409 void dev_uc_flush(struct net_device *dev);
4410 void dev_uc_init(struct net_device *dev);
4413 * __dev_uc_sync - Synchonize device's unicast list
4414 * @dev: device to sync
4415 * @sync: function to call if address should be added
4416 * @unsync: function to call if address should be removed
4418 * Add newly added addresses to the interface, and release
4419 * addresses that have been deleted.
4421 static inline int __dev_uc_sync(struct net_device *dev,
4422 int (*sync)(struct net_device *,
4423 const unsigned char *),
4424 int (*unsync)(struct net_device *,
4425 const unsigned char *))
4427 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4431 * __dev_uc_unsync - Remove synchronized addresses from device
4432 * @dev: device to sync
4433 * @unsync: function to call if address should be removed
4435 * Remove all addresses that were added to the device by dev_uc_sync().
4437 static inline void __dev_uc_unsync(struct net_device *dev,
4438 int (*unsync)(struct net_device *,
4439 const unsigned char *))
4441 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4444 /* Functions used for multicast addresses handling */
4445 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4446 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4447 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4448 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4449 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4450 int dev_mc_sync(struct net_device *to, struct net_device *from);
4451 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4452 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4453 void dev_mc_flush(struct net_device *dev);
4454 void dev_mc_init(struct net_device *dev);
4457 * __dev_mc_sync - Synchonize device's multicast list
4458 * @dev: device to sync
4459 * @sync: function to call if address should be added
4460 * @unsync: function to call if address should be removed
4462 * Add newly added addresses to the interface, and release
4463 * addresses that have been deleted.
4465 static inline int __dev_mc_sync(struct net_device *dev,
4466 int (*sync)(struct net_device *,
4467 const unsigned char *),
4468 int (*unsync)(struct net_device *,
4469 const unsigned char *))
4471 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4475 * __dev_mc_unsync - Remove synchronized addresses from device
4476 * @dev: device to sync
4477 * @unsync: function to call if address should be removed
4479 * Remove all addresses that were added to the device by dev_mc_sync().
4481 static inline void __dev_mc_unsync(struct net_device *dev,
4482 int (*unsync)(struct net_device *,
4483 const unsigned char *))
4485 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4488 /* Functions used for secondary unicast and multicast support */
4489 void dev_set_rx_mode(struct net_device *dev);
4490 void __dev_set_rx_mode(struct net_device *dev);
4491 int dev_set_promiscuity(struct net_device *dev, int inc);
4492 int dev_set_allmulti(struct net_device *dev, int inc);
4493 void netdev_state_change(struct net_device *dev);
4494 void netdev_notify_peers(struct net_device *dev);
4495 void netdev_features_change(struct net_device *dev);
4496 /* Load a device via the kmod */
4497 void dev_load(struct net *net, const char *name);
4498 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4499 struct rtnl_link_stats64 *storage);
4500 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4501 const struct net_device_stats *netdev_stats);
4502 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4503 const struct pcpu_sw_netstats __percpu *netstats);
4505 extern int netdev_max_backlog;
4506 extern int netdev_tstamp_prequeue;
4507 extern int weight_p;
4508 extern int dev_weight_rx_bias;
4509 extern int dev_weight_tx_bias;
4510 extern int dev_rx_weight;
4511 extern int dev_tx_weight;
4512 extern int gro_normal_batch;
4515 NESTED_SYNC_IMM_BIT,
4516 NESTED_SYNC_TODO_BIT,
4519 #define __NESTED_SYNC_BIT(bit) ((u32)1 << (bit))
4520 #define __NESTED_SYNC(name) __NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4522 #define NESTED_SYNC_IMM __NESTED_SYNC(IMM)
4523 #define NESTED_SYNC_TODO __NESTED_SYNC(TODO)
4525 struct netdev_nested_priv {
4526 unsigned char flags;
4530 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4531 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4532 struct list_head **iter);
4533 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4534 struct list_head **iter);
4536 #ifdef CONFIG_LOCKDEP
4537 static LIST_HEAD(net_unlink_list);
4539 static inline void net_unlink_todo(struct net_device *dev)
4541 if (list_empty(&dev->unlink_list))
4542 list_add_tail(&dev->unlink_list, &net_unlink_list);
4546 /* iterate through upper list, must be called under RCU read lock */
4547 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4548 for (iter = &(dev)->adj_list.upper, \
4549 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4551 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4553 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4554 int (*fn)(struct net_device *upper_dev,
4555 struct netdev_nested_priv *priv),
4556 struct netdev_nested_priv *priv);
4558 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4559 struct net_device *upper_dev);
4561 bool netdev_has_any_upper_dev(struct net_device *dev);
4563 void *netdev_lower_get_next_private(struct net_device *dev,
4564 struct list_head **iter);
4565 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4566 struct list_head **iter);
4568 #define netdev_for_each_lower_private(dev, priv, iter) \
4569 for (iter = (dev)->adj_list.lower.next, \
4570 priv = netdev_lower_get_next_private(dev, &(iter)); \
4572 priv = netdev_lower_get_next_private(dev, &(iter)))
4574 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4575 for (iter = &(dev)->adj_list.lower, \
4576 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4578 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4580 void *netdev_lower_get_next(struct net_device *dev,
4581 struct list_head **iter);
4583 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4584 for (iter = (dev)->adj_list.lower.next, \
4585 ldev = netdev_lower_get_next(dev, &(iter)); \
4587 ldev = netdev_lower_get_next(dev, &(iter)))
4589 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4590 struct list_head **iter);
4591 int netdev_walk_all_lower_dev(struct net_device *dev,
4592 int (*fn)(struct net_device *lower_dev,
4593 struct netdev_nested_priv *priv),
4594 struct netdev_nested_priv *priv);
4595 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4596 int (*fn)(struct net_device *lower_dev,
4597 struct netdev_nested_priv *priv),
4598 struct netdev_nested_priv *priv);
4600 void *netdev_adjacent_get_private(struct list_head *adj_list);
4601 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4602 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4603 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4604 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4605 struct netlink_ext_ack *extack);
4606 int netdev_master_upper_dev_link(struct net_device *dev,
4607 struct net_device *upper_dev,
4608 void *upper_priv, void *upper_info,
4609 struct netlink_ext_ack *extack);
4610 void netdev_upper_dev_unlink(struct net_device *dev,
4611 struct net_device *upper_dev);
4612 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4613 struct net_device *new_dev,
4614 struct net_device *dev,
4615 struct netlink_ext_ack *extack);
4616 void netdev_adjacent_change_commit(struct net_device *old_dev,
4617 struct net_device *new_dev,
4618 struct net_device *dev);
4619 void netdev_adjacent_change_abort(struct net_device *old_dev,
4620 struct net_device *new_dev,
4621 struct net_device *dev);
4622 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4623 void *netdev_lower_dev_get_private(struct net_device *dev,
4624 struct net_device *lower_dev);
4625 void netdev_lower_state_changed(struct net_device *lower_dev,
4626 void *lower_state_info);
4628 /* RSS keys are 40 or 52 bytes long */
4629 #define NETDEV_RSS_KEY_LEN 52
4630 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4631 void netdev_rss_key_fill(void *buffer, size_t len);
4633 int skb_checksum_help(struct sk_buff *skb);
4634 int skb_crc32c_csum_help(struct sk_buff *skb);
4635 int skb_csum_hwoffload_help(struct sk_buff *skb,
4636 const netdev_features_t features);
4638 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4639 netdev_features_t features, bool tx_path);
4640 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4641 netdev_features_t features);
4643 struct netdev_bonding_info {
4648 struct netdev_notifier_bonding_info {
4649 struct netdev_notifier_info info; /* must be first */
4650 struct netdev_bonding_info bonding_info;
4653 void netdev_bonding_info_change(struct net_device *dev,
4654 struct netdev_bonding_info *bonding_info);
4656 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4657 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4659 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4666 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4668 return __skb_gso_segment(skb, features, true);
4670 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4672 static inline bool can_checksum_protocol(netdev_features_t features,
4675 if (protocol == htons(ETH_P_FCOE))
4676 return !!(features & NETIF_F_FCOE_CRC);
4678 /* Assume this is an IP checksum (not SCTP CRC) */
4680 if (features & NETIF_F_HW_CSUM) {
4681 /* Can checksum everything */
4686 case htons(ETH_P_IP):
4687 return !!(features & NETIF_F_IP_CSUM);
4688 case htons(ETH_P_IPV6):
4689 return !!(features & NETIF_F_IPV6_CSUM);
4696 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4698 static inline void netdev_rx_csum_fault(struct net_device *dev,
4699 struct sk_buff *skb)
4703 /* rx skb timestamps */
4704 void net_enable_timestamp(void);
4705 void net_disable_timestamp(void);
4707 #ifdef CONFIG_PROC_FS
4708 int __init dev_proc_init(void);
4710 #define dev_proc_init() 0
4713 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4714 struct sk_buff *skb, struct net_device *dev,
4717 __this_cpu_write(softnet_data.xmit.more, more);
4718 return ops->ndo_start_xmit(skb, dev);
4721 static inline bool netdev_xmit_more(void)
4723 return __this_cpu_read(softnet_data.xmit.more);
4726 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4727 struct netdev_queue *txq, bool more)
4729 const struct net_device_ops *ops = dev->netdev_ops;
4732 rc = __netdev_start_xmit(ops, skb, dev, more);
4733 if (rc == NETDEV_TX_OK)
4734 txq_trans_update(txq);
4739 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4741 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4744 extern const struct kobj_ns_type_operations net_ns_type_operations;
4746 const char *netdev_drivername(const struct net_device *dev);
4748 void linkwatch_run_queue(void);
4750 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4751 netdev_features_t f2)
4753 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4754 if (f1 & NETIF_F_HW_CSUM)
4755 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4757 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4763 static inline netdev_features_t netdev_get_wanted_features(
4764 struct net_device *dev)
4766 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4768 netdev_features_t netdev_increment_features(netdev_features_t all,
4769 netdev_features_t one, netdev_features_t mask);
4771 /* Allow TSO being used on stacked device :
4772 * Performing the GSO segmentation before last device
4773 * is a performance improvement.
4775 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4776 netdev_features_t mask)
4778 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4781 int __netdev_update_features(struct net_device *dev);
4782 void netdev_update_features(struct net_device *dev);
4783 void netdev_change_features(struct net_device *dev);
4785 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4786 struct net_device *dev);
4788 netdev_features_t passthru_features_check(struct sk_buff *skb,
4789 struct net_device *dev,
4790 netdev_features_t features);
4791 netdev_features_t netif_skb_features(struct sk_buff *skb);
4793 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4795 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4797 /* check flags correspondence */
4798 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4799 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4800 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4801 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4802 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4803 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4804 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4805 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4806 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4807 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4808 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4809 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4810 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4811 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4812 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4813 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4814 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4815 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4816 BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
4818 return (features & feature) == feature;
4821 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4823 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4824 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4827 static inline bool netif_needs_gso(struct sk_buff *skb,
4828 netdev_features_t features)
4830 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4831 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4832 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4835 static inline void netif_set_gso_max_size(struct net_device *dev,
4838 dev->gso_max_size = size;
4841 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4842 int pulled_hlen, u16 mac_offset,
4845 skb->protocol = protocol;
4846 skb->encapsulation = 1;
4847 skb_push(skb, pulled_hlen);
4848 skb_reset_transport_header(skb);
4849 skb->mac_header = mac_offset;
4850 skb->network_header = skb->mac_header + mac_len;
4851 skb->mac_len = mac_len;
4854 static inline bool netif_is_macsec(const struct net_device *dev)
4856 return dev->priv_flags & IFF_MACSEC;
4859 static inline bool netif_is_macvlan(const struct net_device *dev)
4861 return dev->priv_flags & IFF_MACVLAN;
4864 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4866 return dev->priv_flags & IFF_MACVLAN_PORT;
4869 static inline bool netif_is_bond_master(const struct net_device *dev)
4871 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4874 static inline bool netif_is_bond_slave(const struct net_device *dev)
4876 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4879 static inline bool netif_supports_nofcs(struct net_device *dev)
4881 return dev->priv_flags & IFF_SUPP_NOFCS;
4884 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4886 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4889 static inline bool netif_is_l3_master(const struct net_device *dev)
4891 return dev->priv_flags & IFF_L3MDEV_MASTER;
4894 static inline bool netif_is_l3_slave(const struct net_device *dev)
4896 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4899 static inline bool netif_is_bridge_master(const struct net_device *dev)
4901 return dev->priv_flags & IFF_EBRIDGE;
4904 static inline bool netif_is_bridge_port(const struct net_device *dev)
4906 return dev->priv_flags & IFF_BRIDGE_PORT;
4909 static inline bool netif_is_ovs_master(const struct net_device *dev)
4911 return dev->priv_flags & IFF_OPENVSWITCH;
4914 static inline bool netif_is_ovs_port(const struct net_device *dev)
4916 return dev->priv_flags & IFF_OVS_DATAPATH;
4919 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
4921 return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
4924 static inline bool netif_is_team_master(const struct net_device *dev)
4926 return dev->priv_flags & IFF_TEAM;
4929 static inline bool netif_is_team_port(const struct net_device *dev)
4931 return dev->priv_flags & IFF_TEAM_PORT;
4934 static inline bool netif_is_lag_master(const struct net_device *dev)
4936 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4939 static inline bool netif_is_lag_port(const struct net_device *dev)
4941 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4944 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4946 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4949 static inline bool netif_is_failover(const struct net_device *dev)
4951 return dev->priv_flags & IFF_FAILOVER;
4954 static inline bool netif_is_failover_slave(const struct net_device *dev)
4956 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4959 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4960 static inline void netif_keep_dst(struct net_device *dev)
4962 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4965 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4966 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4968 /* TODO: reserve and use an additional IFF bit, if we get more users */
4969 return dev->priv_flags & IFF_MACSEC;
4972 extern struct pernet_operations __net_initdata loopback_net_ops;
4974 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4976 /* netdev_printk helpers, similar to dev_printk */
4978 static inline const char *netdev_name(const struct net_device *dev)
4980 if (!dev->name[0] || strchr(dev->name, '%'))
4981 return "(unnamed net_device)";
4985 static inline bool netdev_unregistering(const struct net_device *dev)
4987 return dev->reg_state == NETREG_UNREGISTERING;
4990 static inline const char *netdev_reg_state(const struct net_device *dev)
4992 switch (dev->reg_state) {
4993 case NETREG_UNINITIALIZED: return " (uninitialized)";
4994 case NETREG_REGISTERED: return "";
4995 case NETREG_UNREGISTERING: return " (unregistering)";
4996 case NETREG_UNREGISTERED: return " (unregistered)";
4997 case NETREG_RELEASED: return " (released)";
4998 case NETREG_DUMMY: return " (dummy)";
5001 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5002 return " (unknown)";
5005 __printf(3, 4) __cold
5006 void netdev_printk(const char *level, const struct net_device *dev,
5007 const char *format, ...);
5008 __printf(2, 3) __cold
5009 void netdev_emerg(const struct net_device *dev, const char *format, ...);
5010 __printf(2, 3) __cold
5011 void netdev_alert(const struct net_device *dev, const char *format, ...);
5012 __printf(2, 3) __cold
5013 void netdev_crit(const struct net_device *dev, const char *format, ...);
5014 __printf(2, 3) __cold
5015 void netdev_err(const struct net_device *dev, const char *format, ...);
5016 __printf(2, 3) __cold
5017 void netdev_warn(const struct net_device *dev, const char *format, ...);
5018 __printf(2, 3) __cold
5019 void netdev_notice(const struct net_device *dev, const char *format, ...);
5020 __printf(2, 3) __cold
5021 void netdev_info(const struct net_device *dev, const char *format, ...);
5023 #define netdev_level_once(level, dev, fmt, ...) \
5025 static bool __print_once __read_mostly; \
5027 if (!__print_once) { \
5028 __print_once = true; \
5029 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
5033 #define netdev_emerg_once(dev, fmt, ...) \
5034 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
5035 #define netdev_alert_once(dev, fmt, ...) \
5036 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
5037 #define netdev_crit_once(dev, fmt, ...) \
5038 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
5039 #define netdev_err_once(dev, fmt, ...) \
5040 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
5041 #define netdev_warn_once(dev, fmt, ...) \
5042 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
5043 #define netdev_notice_once(dev, fmt, ...) \
5044 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
5045 #define netdev_info_once(dev, fmt, ...) \
5046 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
5048 #define MODULE_ALIAS_NETDEV(device) \
5049 MODULE_ALIAS("netdev-" device)
5051 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5052 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5053 #define netdev_dbg(__dev, format, args...) \
5055 dynamic_netdev_dbg(__dev, format, ##args); \
5057 #elif defined(DEBUG)
5058 #define netdev_dbg(__dev, format, args...) \
5059 netdev_printk(KERN_DEBUG, __dev, format, ##args)
5061 #define netdev_dbg(__dev, format, args...) \
5064 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
5068 #if defined(VERBOSE_DEBUG)
5069 #define netdev_vdbg netdev_dbg
5072 #define netdev_vdbg(dev, format, args...) \
5075 netdev_printk(KERN_DEBUG, dev, format, ##args); \
5081 * netdev_WARN() acts like dev_printk(), but with the key difference
5082 * of using a WARN/WARN_ON to get the message out, including the
5083 * file/line information and a backtrace.
5085 #define netdev_WARN(dev, format, args...) \
5086 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
5087 netdev_reg_state(dev), ##args)
5089 #define netdev_WARN_ONCE(dev, format, args...) \
5090 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
5091 netdev_reg_state(dev), ##args)
5093 /* netif printk helpers, similar to netdev_printk */
5095 #define netif_printk(priv, type, level, dev, fmt, args...) \
5097 if (netif_msg_##type(priv)) \
5098 netdev_printk(level, (dev), fmt, ##args); \
5101 #define netif_level(level, priv, type, dev, fmt, args...) \
5103 if (netif_msg_##type(priv)) \
5104 netdev_##level(dev, fmt, ##args); \
5107 #define netif_emerg(priv, type, dev, fmt, args...) \
5108 netif_level(emerg, priv, type, dev, fmt, ##args)
5109 #define netif_alert(priv, type, dev, fmt, args...) \
5110 netif_level(alert, priv, type, dev, fmt, ##args)
5111 #define netif_crit(priv, type, dev, fmt, args...) \
5112 netif_level(crit, priv, type, dev, fmt, ##args)
5113 #define netif_err(priv, type, dev, fmt, args...) \
5114 netif_level(err, priv, type, dev, fmt, ##args)
5115 #define netif_warn(priv, type, dev, fmt, args...) \
5116 netif_level(warn, priv, type, dev, fmt, ##args)
5117 #define netif_notice(priv, type, dev, fmt, args...) \
5118 netif_level(notice, priv, type, dev, fmt, ##args)
5119 #define netif_info(priv, type, dev, fmt, args...) \
5120 netif_level(info, priv, type, dev, fmt, ##args)
5122 #if defined(CONFIG_DYNAMIC_DEBUG) || \
5123 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
5124 #define netif_dbg(priv, type, netdev, format, args...) \
5126 if (netif_msg_##type(priv)) \
5127 dynamic_netdev_dbg(netdev, format, ##args); \
5129 #elif defined(DEBUG)
5130 #define netif_dbg(priv, type, dev, format, args...) \
5131 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
5133 #define netif_dbg(priv, type, dev, format, args...) \
5136 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5141 /* if @cond then downgrade to debug, else print at @level */
5142 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
5145 netif_dbg(priv, type, netdev, fmt, ##args); \
5147 netif_ ## level(priv, type, netdev, fmt, ##args); \
5150 #if defined(VERBOSE_DEBUG)
5151 #define netif_vdbg netif_dbg
5153 #define netif_vdbg(priv, type, dev, format, args...) \
5156 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
5162 * The list of packet types we will receive (as opposed to discard)
5163 * and the routines to invoke.
5165 * Why 16. Because with 16 the only overlap we get on a hash of the
5166 * low nibble of the protocol value is RARP/SNAP/X.25.
5180 #define PTYPE_HASH_SIZE (16)
5181 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
5183 extern struct net_device *blackhole_netdev;
5185 #endif /* _LINUX_NETDEVICE_H */