2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
44 #include <net/dcbnl.h>
46 #include <net/netprio_cgroup.h>
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
64 /* 802.15.4 specific */
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
72 void netdev_set_default_ethtool_ops(struct net_device *dev,
73 const struct ethtool_ops *ops);
75 /* Backlog congestion levels */
76 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
77 #define NET_RX_DROP 1 /* packet dropped */
80 * Transmit return codes: transmit return codes originate from three different
83 * - qdisc return codes
84 * - driver transmit return codes
87 * Drivers are allowed to return any one of those in their hard_start_xmit()
88 * function. Real network devices commonly used with qdiscs should only return
89 * the driver transmit return codes though - when qdiscs are used, the actual
90 * transmission happens asynchronously, so the value is not propagated to
91 * higher layers. Virtual network devices transmit synchronously; in this case
92 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93 * others are propagated to higher layers.
96 /* qdisc ->enqueue() return codes. */
97 #define NET_XMIT_SUCCESS 0x00
98 #define NET_XMIT_DROP 0x01 /* skb dropped */
99 #define NET_XMIT_CN 0x02 /* congestion notification */
100 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103 * indicates that the device will soon be dropping packets, or already drops
104 * some packets of the same priority; prompting us to send less aggressively. */
105 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
106 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
108 /* Driver transmit return codes */
109 #define NETDEV_TX_MASK 0xf0
112 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
113 NETDEV_TX_OK = 0x00, /* driver took care of packet */
114 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
116 typedef enum netdev_tx netdev_tx_t;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc < NET_XMIT_MASK))
137 * Compute the worst-case header length according to the protocols
141 #if defined(CONFIG_HYPERV_NET)
142 # define LL_MAX_HEADER 128
143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144 # if defined(CONFIG_MAC80211_MESH)
145 # define LL_MAX_HEADER 128
147 # define LL_MAX_HEADER 96
150 # define LL_MAX_HEADER 32
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
161 * Old network device statistics. Fields are native words
162 * (unsigned long) so they can be read and written atomically.
165 struct net_device_stats {
166 unsigned long rx_packets;
167 unsigned long tx_packets;
168 unsigned long rx_bytes;
169 unsigned long tx_bytes;
170 unsigned long rx_errors;
171 unsigned long tx_errors;
172 unsigned long rx_dropped;
173 unsigned long tx_dropped;
174 unsigned long multicast;
175 unsigned long collisions;
176 unsigned long rx_length_errors;
177 unsigned long rx_over_errors;
178 unsigned long rx_crc_errors;
179 unsigned long rx_frame_errors;
180 unsigned long rx_fifo_errors;
181 unsigned long rx_missed_errors;
182 unsigned long tx_aborted_errors;
183 unsigned long tx_carrier_errors;
184 unsigned long tx_fifo_errors;
185 unsigned long tx_heartbeat_errors;
186 unsigned long tx_window_errors;
187 unsigned long rx_compressed;
188 unsigned long tx_compressed;
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
196 #include <linux/static_key.h>
197 extern struct static_key rps_needed;
198 extern struct static_key rfs_needed;
205 struct netdev_hw_addr {
206 struct list_head list;
207 unsigned char addr[MAX_ADDR_LEN];
209 #define NETDEV_HW_ADDR_T_LAN 1
210 #define NETDEV_HW_ADDR_T_SAN 2
211 #define NETDEV_HW_ADDR_T_SLAVE 3
212 #define NETDEV_HW_ADDR_T_UNICAST 4
213 #define NETDEV_HW_ADDR_T_MULTICAST 5
218 struct rcu_head rcu_head;
221 struct netdev_hw_addr_list {
222 struct list_head list;
226 #define netdev_hw_addr_list_count(l) ((l)->count)
227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228 #define netdev_hw_addr_list_for_each(ha, l) \
229 list_for_each_entry(ha, &(l)->list, list)
231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233 #define netdev_for_each_uc_addr(ha, dev) \
234 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238 #define netdev_for_each_mc_addr(ha, dev) \
239 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
245 /* cached hardware header; allow for machine alignment needs. */
246 #define HH_DATA_MOD 16
247 #define HH_DATA_OFF(__len) \
248 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249 #define HH_DATA_ALIGN(__len) \
250 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
256 * dev->hard_header_len ? (dev->hard_header_len +
257 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
259 * We could use other alignment values, but we must maintain the
260 * relationship HH alignment <= LL alignment.
262 #define LL_RESERVED_SPACE(dev) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
268 int (*create) (struct sk_buff *skb, struct net_device *dev,
269 unsigned short type, const void *daddr,
270 const void *saddr, unsigned int len);
271 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 void (*cache_update)(struct hh_cache *hh,
274 const struct net_device *dev,
275 const unsigned char *haddr);
276 bool (*validate)(const char *ll_header, unsigned int len);
277 __be16 (*parse_protocol)(const struct sk_buff *skb);
280 /* These flag bits are private to the generic network queueing
281 * layer; they may not be explicitly referenced by any other
285 enum netdev_state_t {
287 __LINK_STATE_PRESENT,
288 __LINK_STATE_NOCARRIER,
289 __LINK_STATE_LINKWATCH_PENDING,
290 __LINK_STATE_DORMANT,
295 * This structure holds boot-time configured netdevice settings. They
296 * are then used in the device probing.
298 struct netdev_boot_setup {
302 #define NETDEV_BOOT_SETUP_MAX 8
304 int __init netdev_boot_setup(char *str);
307 struct list_head list;
312 * size of gro hash buckets, must less than bit number of
313 * napi_struct::gro_bitmask
315 #define GRO_HASH_BUCKETS 8
318 * Structure for NAPI scheduling similar to tasklet but with weighting
321 /* The poll_list must only be managed by the entity which
322 * changes the state of the NAPI_STATE_SCHED bit. This means
323 * whoever atomically sets that bit can add this napi_struct
324 * to the per-CPU poll_list, and whoever clears that bit
325 * can remove from the list right before clearing the bit.
327 struct list_head poll_list;
331 unsigned long gro_bitmask;
332 int (*poll)(struct napi_struct *, int);
333 #ifdef CONFIG_NETPOLL
336 struct net_device *dev;
337 struct gro_list gro_hash[GRO_HASH_BUCKETS];
339 struct hrtimer timer;
340 struct list_head dev_list;
341 struct hlist_node napi_hash_node;
342 unsigned int napi_id;
346 NAPI_STATE_SCHED, /* Poll is scheduled */
347 NAPI_STATE_MISSED, /* reschedule a napi */
348 NAPI_STATE_DISABLE, /* Disable pending */
349 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
350 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
351 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
352 NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
356 NAPIF_STATE_SCHED = BIT(NAPI_STATE_SCHED),
357 NAPIF_STATE_MISSED = BIT(NAPI_STATE_MISSED),
358 NAPIF_STATE_DISABLE = BIT(NAPI_STATE_DISABLE),
359 NAPIF_STATE_NPSVC = BIT(NAPI_STATE_NPSVC),
360 NAPIF_STATE_HASHED = BIT(NAPI_STATE_HASHED),
361 NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
362 NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
373 typedef enum gro_result gro_result_t;
376 * enum rx_handler_result - Possible return values for rx_handlers.
377 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
379 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
380 * case skb->dev was changed by rx_handler.
381 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
382 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
384 * rx_handlers are functions called from inside __netif_receive_skb(), to do
385 * special processing of the skb, prior to delivery to protocol handlers.
387 * Currently, a net_device can only have a single rx_handler registered. Trying
388 * to register a second rx_handler will return -EBUSY.
390 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
391 * To unregister a rx_handler on a net_device, use
392 * netdev_rx_handler_unregister().
394 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
397 * If the rx_handler consumed the skb in some way, it should return
398 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
399 * the skb to be delivered in some other way.
401 * If the rx_handler changed skb->dev, to divert the skb to another
402 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
403 * new device will be called if it exists.
405 * If the rx_handler decides the skb should be ignored, it should return
406 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
407 * are registered on exact device (ptype->dev == skb->dev).
409 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
410 * delivered, it should return RX_HANDLER_PASS.
412 * A device without a registered rx_handler will behave as if rx_handler
413 * returned RX_HANDLER_PASS.
416 enum rx_handler_result {
422 typedef enum rx_handler_result rx_handler_result_t;
423 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
425 void __napi_schedule(struct napi_struct *n);
426 void __napi_schedule_irqoff(struct napi_struct *n);
428 static inline bool napi_disable_pending(struct napi_struct *n)
430 return test_bit(NAPI_STATE_DISABLE, &n->state);
433 bool napi_schedule_prep(struct napi_struct *n);
436 * napi_schedule - schedule NAPI poll
439 * Schedule NAPI poll routine to be called if it is not already
442 static inline void napi_schedule(struct napi_struct *n)
444 if (napi_schedule_prep(n))
449 * napi_schedule_irqoff - schedule NAPI poll
452 * Variant of napi_schedule(), assuming hard irqs are masked.
454 static inline void napi_schedule_irqoff(struct napi_struct *n)
456 if (napi_schedule_prep(n))
457 __napi_schedule_irqoff(n);
460 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
461 static inline bool napi_reschedule(struct napi_struct *napi)
463 if (napi_schedule_prep(napi)) {
464 __napi_schedule(napi);
470 bool napi_complete_done(struct napi_struct *n, int work_done);
472 * napi_complete - NAPI processing complete
475 * Mark NAPI processing as complete.
476 * Consider using napi_complete_done() instead.
477 * Return false if device should avoid rearming interrupts.
479 static inline bool napi_complete(struct napi_struct *n)
481 return napi_complete_done(n, 0);
485 * napi_hash_del - remove a NAPI from global table
486 * @napi: NAPI context
488 * Warning: caller must observe RCU grace period
489 * before freeing memory containing @napi, if
490 * this function returns true.
491 * Note: core networking stack automatically calls it
492 * from netif_napi_del().
493 * Drivers might want to call this helper to combine all
494 * the needed RCU grace periods into a single one.
496 bool napi_hash_del(struct napi_struct *napi);
499 * napi_disable - prevent NAPI from scheduling
502 * Stop NAPI from being scheduled on this context.
503 * Waits till any outstanding processing completes.
505 void napi_disable(struct napi_struct *n);
508 * napi_enable - enable NAPI scheduling
511 * Resume NAPI from being scheduled on this context.
512 * Must be paired with napi_disable.
514 static inline void napi_enable(struct napi_struct *n)
516 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
517 smp_mb__before_atomic();
518 clear_bit(NAPI_STATE_SCHED, &n->state);
519 clear_bit(NAPI_STATE_NPSVC, &n->state);
523 * napi_synchronize - wait until NAPI is not running
526 * Wait until NAPI is done being scheduled on this context.
527 * Waits till any outstanding processing completes but
528 * does not disable future activations.
530 static inline void napi_synchronize(const struct napi_struct *n)
532 if (IS_ENABLED(CONFIG_SMP))
533 while (test_bit(NAPI_STATE_SCHED, &n->state))
539 enum netdev_queue_state_t {
540 __QUEUE_STATE_DRV_XOFF,
541 __QUEUE_STATE_STACK_XOFF,
542 __QUEUE_STATE_FROZEN,
545 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
546 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
549 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
550 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
552 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
556 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
557 * netif_tx_* functions below are used to manipulate this flag. The
558 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
559 * queue independently. The netif_xmit_*stopped functions below are called
560 * to check if the queue has been stopped by the driver or stack (either
561 * of the XOFF bits are set in the state). Drivers should not need to call
562 * netif_xmit*stopped functions, they should only be using netif_tx_*.
565 struct netdev_queue {
569 struct net_device *dev;
570 struct Qdisc __rcu *qdisc;
571 struct Qdisc *qdisc_sleeping;
575 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
578 unsigned long tx_maxrate;
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
583 unsigned long trans_timeout;
585 /* Subordinate device that the queue has been assigned to */
586 struct net_device *sb_dev;
590 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
593 * Time (in jiffies) of last Tx
595 unsigned long trans_start;
602 } ____cacheline_aligned_in_smp;
604 extern int sysctl_fb_tunnels_only_for_init_net;
606 static inline bool net_has_fallback_tunnels(const struct net *net)
608 return net == &init_net ||
609 !IS_ENABLED(CONFIG_SYSCTL) ||
610 !sysctl_fb_tunnels_only_for_init_net;
613 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
615 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
622 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
624 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
631 * This structure holds an RPS map which can be of variable length. The
632 * map is an array of CPUs.
639 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
642 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
643 * tail pointer for that CPU's input queue at the time of last enqueue, and
644 * a hardware filter index.
646 struct rps_dev_flow {
649 unsigned int last_qtail;
651 #define RPS_NO_FILTER 0xffff
654 * The rps_dev_flow_table structure contains a table of flow mappings.
656 struct rps_dev_flow_table {
659 struct rps_dev_flow flows[0];
661 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
662 ((_num) * sizeof(struct rps_dev_flow)))
665 * The rps_sock_flow_table contains mappings of flows to the last CPU
666 * on which they were processed by the application (set in recvmsg).
667 * Each entry is a 32bit value. Upper part is the high-order bits
668 * of flow hash, lower part is CPU number.
669 * rps_cpu_mask is used to partition the space, depending on number of
670 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
671 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
672 * meaning we use 32-6=26 bits for the hash.
674 struct rps_sock_flow_table {
677 u32 ents[0] ____cacheline_aligned_in_smp;
679 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
681 #define RPS_NO_CPU 0xffff
683 extern u32 rps_cpu_mask;
684 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
686 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
690 unsigned int index = hash & table->mask;
691 u32 val = hash & ~rps_cpu_mask;
693 /* We only give a hint, preemption can change CPU under us */
694 val |= raw_smp_processor_id();
696 if (table->ents[index] != val)
697 table->ents[index] = val;
701 #ifdef CONFIG_RFS_ACCEL
702 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
705 #endif /* CONFIG_RPS */
707 /* This structure contains an instance of an RX queue. */
708 struct netdev_rx_queue {
710 struct rps_map __rcu *rps_map;
711 struct rps_dev_flow_table __rcu *rps_flow_table;
714 struct net_device *dev;
715 struct xdp_rxq_info xdp_rxq;
716 } ____cacheline_aligned_in_smp;
719 * RX queue sysfs structures and functions.
721 struct rx_queue_attribute {
722 struct attribute attr;
723 ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
724 ssize_t (*store)(struct netdev_rx_queue *queue,
725 const char *buf, size_t len);
730 * This structure holds an XPS map which can be of variable length. The
731 * map is an array of queues.
735 unsigned int alloc_len;
739 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
740 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
741 - sizeof(struct xps_map)) / sizeof(u16))
744 * This structure holds all XPS maps for device. Maps are indexed by CPU.
746 struct xps_dev_maps {
748 struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
751 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
752 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
754 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
755 (_rxqs * (_tcs) * sizeof(struct xps_map *)))
757 #endif /* CONFIG_XPS */
759 #define TC_MAX_QUEUE 16
760 #define TC_BITMASK 15
761 /* HW offloaded queuing disciplines txq count and offset maps */
762 struct netdev_tc_txq {
767 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
769 * This structure is to hold information about the device
770 * configured to run FCoE protocol stack.
772 struct netdev_fcoe_hbainfo {
773 char manufacturer[64];
774 char serial_number[64];
775 char hardware_version[64];
776 char driver_version[64];
777 char optionrom_version[64];
778 char firmware_version[64];
780 char model_description[256];
784 #define MAX_PHYS_ITEM_ID_LEN 32
786 /* This structure holds a unique identifier to identify some
787 * physical item (port for example) used by a netdevice.
789 struct netdev_phys_item_id {
790 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
791 unsigned char id_len;
794 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
795 struct netdev_phys_item_id *b)
797 return a->id_len == b->id_len &&
798 memcmp(a->id, b->id, a->id_len) == 0;
801 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
803 struct net_device *sb_dev);
806 TC_SETUP_QDISC_MQPRIO,
809 TC_SETUP_CLSMATCHALL,
819 /* These structures hold the attributes of bpf state that are being passed
820 * to the netdevice through the bpf op.
822 enum bpf_netdev_command {
823 /* Set or clear a bpf program used in the earliest stages of packet
824 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
825 * is responsible for calling bpf_prog_put on any old progs that are
826 * stored. In case of error, the callee need not release the new prog
827 * reference, but on success it takes ownership and must bpf_prog_put
828 * when it is no longer used.
834 /* BPF program for offload callbacks, invoked at program load time. */
835 BPF_OFFLOAD_VERIFIER_PREP,
836 BPF_OFFLOAD_TRANSLATE,
838 BPF_OFFLOAD_MAP_ALLOC,
839 BPF_OFFLOAD_MAP_FREE,
844 struct bpf_prog_offload_ops;
845 struct netlink_ext_ack;
849 enum bpf_netdev_command command;
854 struct bpf_prog *prog;
855 struct netlink_ext_ack *extack;
857 /* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
860 /* flags with which program was installed */
863 /* BPF_OFFLOAD_VERIFIER_PREP */
865 struct bpf_prog *prog;
866 const struct bpf_prog_offload_ops *ops; /* callee set */
868 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
870 struct bpf_prog *prog;
872 /* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
874 struct bpf_offloaded_map *offmap;
876 /* XDP_QUERY_XSK_UMEM, XDP_SETUP_XSK_UMEM */
878 struct xdp_umem *umem; /* out for query*/
879 u16 queue_id; /* in for query */
884 #ifdef CONFIG_XFRM_OFFLOAD
886 int (*xdo_dev_state_add) (struct xfrm_state *x);
887 void (*xdo_dev_state_delete) (struct xfrm_state *x);
888 void (*xdo_dev_state_free) (struct xfrm_state *x);
889 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
890 struct xfrm_state *x);
891 void (*xdo_dev_state_advance_esn) (struct xfrm_state *x);
895 #if IS_ENABLED(CONFIG_TLS_DEVICE)
896 enum tls_offload_ctx_dir {
897 TLS_OFFLOAD_CTX_DIR_RX,
898 TLS_OFFLOAD_CTX_DIR_TX,
901 struct tls_crypto_info;
905 int (*tls_dev_add)(struct net_device *netdev, struct sock *sk,
906 enum tls_offload_ctx_dir direction,
907 struct tls_crypto_info *crypto_info,
908 u32 start_offload_tcp_sn);
909 void (*tls_dev_del)(struct net_device *netdev,
910 struct tls_context *ctx,
911 enum tls_offload_ctx_dir direction);
912 void (*tls_dev_resync_rx)(struct net_device *netdev,
913 struct sock *sk, u32 seq, u64 rcd_sn);
918 struct rcu_head rcuhead;
923 * This structure defines the management hooks for network devices.
924 * The following hooks can be defined; unless noted otherwise, they are
925 * optional and can be filled with a null pointer.
927 * int (*ndo_init)(struct net_device *dev);
928 * This function is called once when a network device is registered.
929 * The network device can use this for any late stage initialization
930 * or semantic validation. It can fail with an error code which will
931 * be propagated back to register_netdev.
933 * void (*ndo_uninit)(struct net_device *dev);
934 * This function is called when device is unregistered or when registration
935 * fails. It is not called if init fails.
937 * int (*ndo_open)(struct net_device *dev);
938 * This function is called when a network device transitions to the up
941 * int (*ndo_stop)(struct net_device *dev);
942 * This function is called when a network device transitions to the down
945 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
946 * struct net_device *dev);
947 * Called when a packet needs to be transmitted.
948 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
949 * the queue before that can happen; it's for obsolete devices and weird
950 * corner cases, but the stack really does a non-trivial amount
951 * of useless work if you return NETDEV_TX_BUSY.
952 * Required; cannot be NULL.
954 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
955 * struct net_device *dev
956 * netdev_features_t features);
957 * Called by core transmit path to determine if device is capable of
958 * performing offload operations on a given packet. This is to give
959 * the device an opportunity to implement any restrictions that cannot
960 * be otherwise expressed by feature flags. The check is called with
961 * the set of features that the stack has calculated and it returns
962 * those the driver believes to be appropriate.
964 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
965 * struct net_device *sb_dev,
966 * select_queue_fallback_t fallback);
967 * Called to decide which queue to use when device supports multiple
970 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
971 * This function is called to allow device receiver to make
972 * changes to configuration when multicast or promiscuous is enabled.
974 * void (*ndo_set_rx_mode)(struct net_device *dev);
975 * This function is called device changes address list filtering.
976 * If driver handles unicast address filtering, it should set
977 * IFF_UNICAST_FLT in its priv_flags.
979 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
980 * This function is called when the Media Access Control address
981 * needs to be changed. If this interface is not defined, the
982 * MAC address can not be changed.
984 * int (*ndo_validate_addr)(struct net_device *dev);
985 * Test if Media Access Control address is valid for the device.
987 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
988 * Called when a user requests an ioctl which can't be handled by
989 * the generic interface code. If not defined ioctls return
990 * not supported error code.
992 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
993 * Used to set network devices bus interface parameters. This interface
994 * is retained for legacy reasons; new devices should use the bus
995 * interface (PCI) for low level management.
997 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
998 * Called when a user wants to change the Maximum Transfer Unit
1001 * void (*ndo_tx_timeout)(struct net_device *dev);
1002 * Callback used when the transmitter has not made any progress
1003 * for dev->watchdog ticks.
1005 * void (*ndo_get_stats64)(struct net_device *dev,
1006 * struct rtnl_link_stats64 *storage);
1007 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1008 * Called when a user wants to get the network device usage
1009 * statistics. Drivers must do one of the following:
1010 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
1011 * rtnl_link_stats64 structure passed by the caller.
1012 * 2. Define @ndo_get_stats to update a net_device_stats structure
1013 * (which should normally be dev->stats) and return a pointer to
1014 * it. The structure may be changed asynchronously only if each
1015 * field is written atomically.
1016 * 3. Update dev->stats asynchronously and atomically, and define
1017 * neither operation.
1019 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1020 * Return true if this device supports offload stats of this attr_id.
1022 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1024 * Get statistics for offload operations by attr_id. Write it into the
1025 * attr_data pointer.
1027 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1028 * If device supports VLAN filtering this function is called when a
1029 * VLAN id is registered.
1031 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1032 * If device supports VLAN filtering this function is called when a
1033 * VLAN id is unregistered.
1035 * void (*ndo_poll_controller)(struct net_device *dev);
1037 * SR-IOV management functions.
1038 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1039 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1040 * u8 qos, __be16 proto);
1041 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1043 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1044 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1045 * int (*ndo_get_vf_config)(struct net_device *dev,
1046 * int vf, struct ifla_vf_info *ivf);
1047 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1048 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1049 * struct nlattr *port[]);
1051 * Enable or disable the VF ability to query its RSS Redirection Table and
1052 * Hash Key. This is needed since on some devices VF share this information
1053 * with PF and querying it may introduce a theoretical security risk.
1054 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1055 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1056 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1058 * Called to setup any 'tc' scheduler, classifier or action on @dev.
1059 * This is always called from the stack with the rtnl lock held and netif
1060 * tx queues stopped. This allows the netdevice to perform queue
1061 * management safely.
1063 * Fiber Channel over Ethernet (FCoE) offload functions.
1064 * int (*ndo_fcoe_enable)(struct net_device *dev);
1065 * Called when the FCoE protocol stack wants to start using LLD for FCoE
1066 * so the underlying device can perform whatever needed configuration or
1067 * initialization to support acceleration of FCoE traffic.
1069 * int (*ndo_fcoe_disable)(struct net_device *dev);
1070 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
1071 * so the underlying device can perform whatever needed clean-ups to
1072 * stop supporting acceleration of FCoE traffic.
1074 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1075 * struct scatterlist *sgl, unsigned int sgc);
1076 * Called when the FCoE Initiator wants to initialize an I/O that
1077 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1078 * perform necessary setup and returns 1 to indicate the device is set up
1079 * successfully to perform DDP on this I/O, otherwise this returns 0.
1081 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1082 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1083 * indicated by the FC exchange id 'xid', so the underlying device can
1084 * clean up and reuse resources for later DDP requests.
1086 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1087 * struct scatterlist *sgl, unsigned int sgc);
1088 * Called when the FCoE Target wants to initialize an I/O that
1089 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1090 * perform necessary setup and returns 1 to indicate the device is set up
1091 * successfully to perform DDP on this I/O, otherwise this returns 0.
1093 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1094 * struct netdev_fcoe_hbainfo *hbainfo);
1095 * Called when the FCoE Protocol stack wants information on the underlying
1096 * device. This information is utilized by the FCoE protocol stack to
1097 * register attributes with Fiber Channel management service as per the
1098 * FC-GS Fabric Device Management Information(FDMI) specification.
1100 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1101 * Called when the underlying device wants to override default World Wide
1102 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1103 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1104 * protocol stack to use.
1107 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1108 * u16 rxq_index, u32 flow_id);
1109 * Set hardware filter for RFS. rxq_index is the target queue index;
1110 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1111 * Return the filter ID on success, or a negative error code.
1113 * Slave management functions (for bridge, bonding, etc).
1114 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1115 * Called to make another netdev an underling.
1117 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1118 * Called to release previously enslaved netdev.
1120 * Feature/offload setting functions.
1121 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1122 * netdev_features_t features);
1123 * Adjusts the requested feature flags according to device-specific
1124 * constraints, and returns the resulting flags. Must not modify
1127 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1128 * Called to update device configuration to new features. Passed
1129 * feature set might be less than what was returned by ndo_fix_features()).
1130 * Must return >0 or -errno if it changed dev->features itself.
1132 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1133 * struct net_device *dev,
1134 * const unsigned char *addr, u16 vid, u16 flags)
1135 * Adds an FDB entry to dev for addr.
1136 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1137 * struct net_device *dev,
1138 * const unsigned char *addr, u16 vid)
1139 * Deletes the FDB entry from dev coresponding to addr.
1140 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1141 * struct net_device *dev, struct net_device *filter_dev,
1143 * Used to add FDB entries to dump requests. Implementers should add
1144 * entries to skb and update idx with the number of entries.
1146 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1148 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1149 * struct net_device *dev, u32 filter_mask,
1151 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1154 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1155 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1156 * which do not represent real hardware may define this to allow their
1157 * userspace components to manage their virtual carrier state. Devices
1158 * that determine carrier state from physical hardware properties (eg
1159 * network cables) or protocol-dependent mechanisms (eg
1160 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1162 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1163 * struct netdev_phys_item_id *ppid);
1164 * Called to get ID of physical port of this device. If driver does
1165 * not implement this, it is assumed that the hw is not able to have
1166 * multiple net devices on single physical port.
1168 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1169 * struct udp_tunnel_info *ti);
1170 * Called by UDP tunnel to notify a driver about the UDP port and socket
1171 * address family that a UDP tunnel is listnening to. It is called only
1172 * when a new port starts listening. The operation is protected by the
1175 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1176 * struct udp_tunnel_info *ti);
1177 * Called by UDP tunnel to notify the driver about a UDP port and socket
1178 * address family that the UDP tunnel is not listening to anymore. The
1179 * operation is protected by the RTNL.
1181 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1182 * struct net_device *dev)
1183 * Called by upper layer devices to accelerate switching or other
1184 * station functionality into hardware. 'pdev is the lowerdev
1185 * to use for the offload and 'dev' is the net device that will
1186 * back the offload. Returns a pointer to the private structure
1187 * the upper layer will maintain.
1188 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1189 * Called by upper layer device to delete the station created
1190 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1191 * the station and priv is the structure returned by the add
1193 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1194 * int queue_index, u32 maxrate);
1195 * Called when a user wants to set a max-rate limitation of specific
1197 * int (*ndo_get_iflink)(const struct net_device *dev);
1198 * Called to get the iflink value of this device.
1199 * void (*ndo_change_proto_down)(struct net_device *dev,
1201 * This function is used to pass protocol port error state information
1202 * to the switch driver. The switch driver can react to the proto_down
1203 * by doing a phys down on the associated switch port.
1204 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1205 * This function is used to get egress tunnel information for given skb.
1206 * This is useful for retrieving outer tunnel header parameters while
1208 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1209 * This function is used to specify the headroom that the skb must
1210 * consider when allocation skb during packet reception. Setting
1211 * appropriate rx headroom value allows avoiding skb head copy on
1212 * forward. Setting a negative value resets the rx headroom to the
1214 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1215 * This function is used to set or query state related to XDP on the
1216 * netdevice and manage BPF offload. See definition of
1217 * enum bpf_netdev_command for details.
1218 * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1220 * This function is used to submit @n XDP packets for transmit on a
1221 * netdevice. Returns number of frames successfully transmitted, frames
1222 * that got dropped are freed/returned via xdp_return_frame().
1223 * Returns negative number, means general error invoking ndo, meaning
1224 * no frames were xmit'ed and core-caller will free all frames.
1226 struct net_device_ops {
1227 int (*ndo_init)(struct net_device *dev);
1228 void (*ndo_uninit)(struct net_device *dev);
1229 int (*ndo_open)(struct net_device *dev);
1230 int (*ndo_stop)(struct net_device *dev);
1231 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1232 struct net_device *dev);
1233 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1234 struct net_device *dev,
1235 netdev_features_t features);
1236 u16 (*ndo_select_queue)(struct net_device *dev,
1237 struct sk_buff *skb,
1238 struct net_device *sb_dev,
1239 select_queue_fallback_t fallback);
1240 void (*ndo_change_rx_flags)(struct net_device *dev,
1242 void (*ndo_set_rx_mode)(struct net_device *dev);
1243 int (*ndo_set_mac_address)(struct net_device *dev,
1245 int (*ndo_validate_addr)(struct net_device *dev);
1246 int (*ndo_do_ioctl)(struct net_device *dev,
1247 struct ifreq *ifr, int cmd);
1248 int (*ndo_set_config)(struct net_device *dev,
1250 int (*ndo_change_mtu)(struct net_device *dev,
1252 int (*ndo_neigh_setup)(struct net_device *dev,
1253 struct neigh_parms *);
1254 void (*ndo_tx_timeout) (struct net_device *dev);
1256 void (*ndo_get_stats64)(struct net_device *dev,
1257 struct rtnl_link_stats64 *storage);
1258 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1259 int (*ndo_get_offload_stats)(int attr_id,
1260 const struct net_device *dev,
1262 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1264 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1265 __be16 proto, u16 vid);
1266 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1267 __be16 proto, u16 vid);
1268 #ifdef CONFIG_NET_POLL_CONTROLLER
1269 void (*ndo_poll_controller)(struct net_device *dev);
1270 int (*ndo_netpoll_setup)(struct net_device *dev,
1271 struct netpoll_info *info);
1272 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1274 int (*ndo_set_vf_mac)(struct net_device *dev,
1275 int queue, u8 *mac);
1276 int (*ndo_set_vf_vlan)(struct net_device *dev,
1277 int queue, u16 vlan,
1278 u8 qos, __be16 proto);
1279 int (*ndo_set_vf_rate)(struct net_device *dev,
1280 int vf, int min_tx_rate,
1282 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1283 int vf, bool setting);
1284 int (*ndo_set_vf_trust)(struct net_device *dev,
1285 int vf, bool setting);
1286 int (*ndo_get_vf_config)(struct net_device *dev,
1288 struct ifla_vf_info *ivf);
1289 int (*ndo_set_vf_link_state)(struct net_device *dev,
1290 int vf, int link_state);
1291 int (*ndo_get_vf_stats)(struct net_device *dev,
1293 struct ifla_vf_stats
1295 int (*ndo_set_vf_port)(struct net_device *dev,
1297 struct nlattr *port[]);
1298 int (*ndo_get_vf_port)(struct net_device *dev,
1299 int vf, struct sk_buff *skb);
1300 int (*ndo_set_vf_guid)(struct net_device *dev,
1303 int (*ndo_set_vf_rss_query_en)(
1304 struct net_device *dev,
1305 int vf, bool setting);
1306 int (*ndo_setup_tc)(struct net_device *dev,
1307 enum tc_setup_type type,
1309 #if IS_ENABLED(CONFIG_FCOE)
1310 int (*ndo_fcoe_enable)(struct net_device *dev);
1311 int (*ndo_fcoe_disable)(struct net_device *dev);
1312 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1314 struct scatterlist *sgl,
1316 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1318 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1320 struct scatterlist *sgl,
1322 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1323 struct netdev_fcoe_hbainfo *hbainfo);
1326 #if IS_ENABLED(CONFIG_LIBFCOE)
1327 #define NETDEV_FCOE_WWNN 0
1328 #define NETDEV_FCOE_WWPN 1
1329 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1330 u64 *wwn, int type);
1333 #ifdef CONFIG_RFS_ACCEL
1334 int (*ndo_rx_flow_steer)(struct net_device *dev,
1335 const struct sk_buff *skb,
1339 int (*ndo_add_slave)(struct net_device *dev,
1340 struct net_device *slave_dev,
1341 struct netlink_ext_ack *extack);
1342 int (*ndo_del_slave)(struct net_device *dev,
1343 struct net_device *slave_dev);
1344 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1345 netdev_features_t features);
1346 int (*ndo_set_features)(struct net_device *dev,
1347 netdev_features_t features);
1348 int (*ndo_neigh_construct)(struct net_device *dev,
1349 struct neighbour *n);
1350 void (*ndo_neigh_destroy)(struct net_device *dev,
1351 struct neighbour *n);
1353 int (*ndo_fdb_add)(struct ndmsg *ndm,
1354 struct nlattr *tb[],
1355 struct net_device *dev,
1356 const unsigned char *addr,
1359 int (*ndo_fdb_del)(struct ndmsg *ndm,
1360 struct nlattr *tb[],
1361 struct net_device *dev,
1362 const unsigned char *addr,
1364 int (*ndo_fdb_dump)(struct sk_buff *skb,
1365 struct netlink_callback *cb,
1366 struct net_device *dev,
1367 struct net_device *filter_dev,
1370 int (*ndo_bridge_setlink)(struct net_device *dev,
1371 struct nlmsghdr *nlh,
1373 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1375 struct net_device *dev,
1378 int (*ndo_bridge_dellink)(struct net_device *dev,
1379 struct nlmsghdr *nlh,
1381 int (*ndo_change_carrier)(struct net_device *dev,
1383 int (*ndo_get_phys_port_id)(struct net_device *dev,
1384 struct netdev_phys_item_id *ppid);
1385 int (*ndo_get_phys_port_name)(struct net_device *dev,
1386 char *name, size_t len);
1387 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1388 struct udp_tunnel_info *ti);
1389 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1390 struct udp_tunnel_info *ti);
1391 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1392 struct net_device *dev);
1393 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1396 int (*ndo_get_lock_subclass)(struct net_device *dev);
1397 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1400 int (*ndo_get_iflink)(const struct net_device *dev);
1401 int (*ndo_change_proto_down)(struct net_device *dev,
1403 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1404 struct sk_buff *skb);
1405 void (*ndo_set_rx_headroom)(struct net_device *dev,
1406 int needed_headroom);
1407 int (*ndo_bpf)(struct net_device *dev,
1408 struct netdev_bpf *bpf);
1409 int (*ndo_xdp_xmit)(struct net_device *dev, int n,
1410 struct xdp_frame **xdp,
1412 int (*ndo_xsk_async_xmit)(struct net_device *dev,
1417 * enum net_device_priv_flags - &struct net_device priv_flags
1419 * These are the &struct net_device, they are only set internally
1420 * by drivers and used in the kernel. These flags are invisible to
1421 * userspace; this means that the order of these flags can change
1422 * during any kernel release.
1424 * You should have a pretty good reason to be extending these flags.
1426 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1427 * @IFF_EBRIDGE: Ethernet bridging device
1428 * @IFF_BONDING: bonding master or slave
1429 * @IFF_ISATAP: ISATAP interface (RFC4214)
1430 * @IFF_WAN_HDLC: WAN HDLC device
1431 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1433 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1434 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1435 * @IFF_MACVLAN_PORT: device used as macvlan port
1436 * @IFF_BRIDGE_PORT: device used as bridge port
1437 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1438 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1439 * @IFF_UNICAST_FLT: Supports unicast filtering
1440 * @IFF_TEAM_PORT: device used as team port
1441 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1442 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1443 * change when it's running
1444 * @IFF_MACVLAN: Macvlan device
1445 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1446 * underlying stacked devices
1447 * @IFF_L3MDEV_MASTER: device is an L3 master device
1448 * @IFF_NO_QUEUE: device can run without qdisc attached
1449 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1450 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1451 * @IFF_TEAM: device is a team device
1452 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1453 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1454 * entity (i.e. the master device for bridged veth)
1455 * @IFF_MACSEC: device is a MACsec device
1456 * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1457 * @IFF_FAILOVER: device is a failover master device
1458 * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1459 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1460 * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1462 enum netdev_priv_flags {
1463 IFF_802_1Q_VLAN = 1<<0,
1467 IFF_WAN_HDLC = 1<<4,
1468 IFF_XMIT_DST_RELEASE = 1<<5,
1469 IFF_DONT_BRIDGE = 1<<6,
1470 IFF_DISABLE_NETPOLL = 1<<7,
1471 IFF_MACVLAN_PORT = 1<<8,
1472 IFF_BRIDGE_PORT = 1<<9,
1473 IFF_OVS_DATAPATH = 1<<10,
1474 IFF_TX_SKB_SHARING = 1<<11,
1475 IFF_UNICAST_FLT = 1<<12,
1476 IFF_TEAM_PORT = 1<<13,
1477 IFF_SUPP_NOFCS = 1<<14,
1478 IFF_LIVE_ADDR_CHANGE = 1<<15,
1479 IFF_MACVLAN = 1<<16,
1480 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1481 IFF_L3MDEV_MASTER = 1<<18,
1482 IFF_NO_QUEUE = 1<<19,
1483 IFF_OPENVSWITCH = 1<<20,
1484 IFF_L3MDEV_SLAVE = 1<<21,
1486 IFF_RXFH_CONFIGURED = 1<<23,
1487 IFF_PHONY_HEADROOM = 1<<24,
1489 IFF_NO_RX_HANDLER = 1<<26,
1490 IFF_FAILOVER = 1<<27,
1491 IFF_FAILOVER_SLAVE = 1<<28,
1492 IFF_L3MDEV_RX_HANDLER = 1<<29,
1493 IFF_LIVE_RENAME_OK = 1<<30,
1496 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1497 #define IFF_EBRIDGE IFF_EBRIDGE
1498 #define IFF_BONDING IFF_BONDING
1499 #define IFF_ISATAP IFF_ISATAP
1500 #define IFF_WAN_HDLC IFF_WAN_HDLC
1501 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1502 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1503 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1504 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1505 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1506 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1507 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1508 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1509 #define IFF_TEAM_PORT IFF_TEAM_PORT
1510 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1511 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1512 #define IFF_MACVLAN IFF_MACVLAN
1513 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1514 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1515 #define IFF_NO_QUEUE IFF_NO_QUEUE
1516 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1517 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1518 #define IFF_TEAM IFF_TEAM
1519 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1520 #define IFF_MACSEC IFF_MACSEC
1521 #define IFF_NO_RX_HANDLER IFF_NO_RX_HANDLER
1522 #define IFF_FAILOVER IFF_FAILOVER
1523 #define IFF_FAILOVER_SLAVE IFF_FAILOVER_SLAVE
1524 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1525 #define IFF_LIVE_RENAME_OK IFF_LIVE_RENAME_OK
1528 * struct net_device - The DEVICE structure.
1530 * Actually, this whole structure is a big mistake. It mixes I/O
1531 * data with strictly "high-level" data, and it has to know about
1532 * almost every data structure used in the INET module.
1534 * @name: This is the first field of the "visible" part of this structure
1535 * (i.e. as seen by users in the "Space.c" file). It is the name
1538 * @name_hlist: Device name hash chain, please keep it close to name[]
1539 * @ifalias: SNMP alias
1540 * @mem_end: Shared memory end
1541 * @mem_start: Shared memory start
1542 * @base_addr: Device I/O address
1543 * @irq: Device IRQ number
1545 * @state: Generic network queuing layer state, see netdev_state_t
1546 * @dev_list: The global list of network devices
1547 * @napi_list: List entry used for polling NAPI devices
1548 * @unreg_list: List entry when we are unregistering the
1549 * device; see the function unregister_netdev
1550 * @close_list: List entry used when we are closing the device
1551 * @ptype_all: Device-specific packet handlers for all protocols
1552 * @ptype_specific: Device-specific, protocol-specific packet handlers
1554 * @adj_list: Directly linked devices, like slaves for bonding
1555 * @features: Currently active device features
1556 * @hw_features: User-changeable features
1558 * @wanted_features: User-requested features
1559 * @vlan_features: Mask of features inheritable by VLAN devices
1561 * @hw_enc_features: Mask of features inherited by encapsulating devices
1562 * This field indicates what encapsulation
1563 * offloads the hardware is capable of doing,
1564 * and drivers will need to set them appropriately.
1566 * @mpls_features: Mask of features inheritable by MPLS
1568 * @ifindex: interface index
1569 * @group: The group the device belongs to
1571 * @stats: Statistics struct, which was left as a legacy, use
1572 * rtnl_link_stats64 instead
1574 * @rx_dropped: Dropped packets by core network,
1575 * do not use this in drivers
1576 * @tx_dropped: Dropped packets by core network,
1577 * do not use this in drivers
1578 * @rx_nohandler: nohandler dropped packets by core network on
1579 * inactive devices, do not use this in drivers
1580 * @carrier_up_count: Number of times the carrier has been up
1581 * @carrier_down_count: Number of times the carrier has been down
1583 * @wireless_handlers: List of functions to handle Wireless Extensions,
1585 * see <net/iw_handler.h> for details.
1586 * @wireless_data: Instance data managed by the core of wireless extensions
1588 * @netdev_ops: Includes several pointers to callbacks,
1589 * if one wants to override the ndo_*() functions
1590 * @ethtool_ops: Management operations
1591 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1592 * discovery handling. Necessary for e.g. 6LoWPAN.
1593 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1594 * of Layer 2 headers.
1596 * @flags: Interface flags (a la BSD)
1597 * @priv_flags: Like 'flags' but invisible to userspace,
1598 * see if.h for the definitions
1599 * @gflags: Global flags ( kept as legacy )
1600 * @padded: How much padding added by alloc_netdev()
1601 * @operstate: RFC2863 operstate
1602 * @link_mode: Mapping policy to operstate
1603 * @if_port: Selectable AUI, TP, ...
1605 * @mtu: Interface MTU value
1606 * @min_mtu: Interface Minimum MTU value
1607 * @max_mtu: Interface Maximum MTU value
1608 * @type: Interface hardware type
1609 * @hard_header_len: Maximum hardware header length.
1610 * @min_header_len: Minimum hardware header length
1612 * @needed_headroom: Extra headroom the hardware may need, but not in all
1613 * cases can this be guaranteed
1614 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1615 * cases can this be guaranteed. Some cases also use
1616 * LL_MAX_HEADER instead to allocate the skb
1618 * interface address info:
1620 * @perm_addr: Permanent hw address
1621 * @addr_assign_type: Hw address assignment type
1622 * @addr_len: Hardware address length
1623 * @upper_level: Maximum depth level of upper devices.
1624 * @lower_level: Maximum depth level of lower devices.
1625 * @neigh_priv_len: Used in neigh_alloc()
1626 * @dev_id: Used to differentiate devices that share
1627 * the same link layer address
1628 * @dev_port: Used to differentiate devices that share
1630 * @addr_list_lock: XXX: need comments on this one
1631 * @uc_promisc: Counter that indicates promiscuous mode
1632 * has been enabled due to the need to listen to
1633 * additional unicast addresses in a device that
1634 * does not implement ndo_set_rx_mode()
1635 * @uc: unicast mac addresses
1636 * @mc: multicast mac addresses
1637 * @dev_addrs: list of device hw addresses
1638 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1639 * @promiscuity: Number of times the NIC is told to work in
1640 * promiscuous mode; if it becomes 0 the NIC will
1641 * exit promiscuous mode
1642 * @allmulti: Counter, enables or disables allmulticast mode
1644 * @vlan_info: VLAN info
1645 * @dsa_ptr: dsa specific data
1646 * @tipc_ptr: TIPC specific data
1647 * @atalk_ptr: AppleTalk link
1648 * @ip_ptr: IPv4 specific data
1649 * @dn_ptr: DECnet specific data
1650 * @ip6_ptr: IPv6 specific data
1651 * @ax25_ptr: AX.25 specific data
1652 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1654 * @dev_addr: Hw address (before bcast,
1655 * because most packets are unicast)
1657 * @_rx: Array of RX queues
1658 * @num_rx_queues: Number of RX queues
1659 * allocated at register_netdev() time
1660 * @real_num_rx_queues: Number of RX queues currently active in device
1662 * @rx_handler: handler for received packets
1663 * @rx_handler_data: XXX: need comments on this one
1664 * @miniq_ingress: ingress/clsact qdisc specific data for
1665 * ingress processing
1666 * @ingress_queue: XXX: need comments on this one
1667 * @broadcast: hw bcast address
1669 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1670 * indexed by RX queue number. Assigned by driver.
1671 * This must only be set if the ndo_rx_flow_steer
1672 * operation is defined
1673 * @index_hlist: Device index hash chain
1675 * @_tx: Array of TX queues
1676 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1677 * @real_num_tx_queues: Number of TX queues currently active in device
1678 * @qdisc: Root qdisc from userspace point of view
1679 * @tx_queue_len: Max frames per queue allowed
1680 * @tx_global_lock: XXX: need comments on this one
1682 * @xps_maps: XXX: need comments on this one
1683 * @miniq_egress: clsact qdisc specific data for
1685 * @watchdog_timeo: Represents the timeout that is used by
1686 * the watchdog (see dev_watchdog())
1687 * @watchdog_timer: List of timers
1689 * @pcpu_refcnt: Number of references to this device
1690 * @todo_list: Delayed register/unregister
1691 * @link_watch_list: XXX: need comments on this one
1693 * @reg_state: Register/unregister state machine
1694 * @dismantle: Device is going to be freed
1695 * @rtnl_link_state: This enum represents the phases of creating
1698 * @needs_free_netdev: Should unregister perform free_netdev?
1699 * @priv_destructor: Called from unregister
1700 * @npinfo: XXX: need comments on this one
1701 * @nd_net: Network namespace this network device is inside
1703 * @ml_priv: Mid-layer private
1704 * @lstats: Loopback statistics
1705 * @tstats: Tunnel statistics
1706 * @dstats: Dummy statistics
1707 * @vstats: Virtual ethernet statistics
1712 * @dev: Class/net/name entry
1713 * @sysfs_groups: Space for optional device, statistics and wireless
1716 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1717 * @rtnl_link_ops: Rtnl_link_ops
1719 * @gso_max_size: Maximum size of generic segmentation offload
1720 * @gso_max_segs: Maximum number of segments that can be passed to the
1723 * @dcbnl_ops: Data Center Bridging netlink ops
1724 * @num_tc: Number of traffic classes in the net device
1725 * @tc_to_txq: XXX: need comments on this one
1726 * @prio_tc_map: XXX: need comments on this one
1728 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1730 * @priomap: XXX: need comments on this one
1731 * @phydev: Physical device may attach itself
1732 * for hardware timestamping
1733 * @sfp_bus: attached &struct sfp_bus structure.
1735 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1736 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1738 * @proto_down: protocol port state information can be sent to the
1739 * switch driver and used to set the phys state of the
1742 * @wol_enabled: Wake-on-LAN is enabled
1744 * FIXME: cleanup struct net_device such that network protocol info
1749 char name[IFNAMSIZ];
1750 struct hlist_node name_hlist;
1751 struct dev_ifalias __rcu *ifalias;
1753 * I/O specific fields
1754 * FIXME: Merge these and struct ifmap into one
1756 unsigned long mem_end;
1757 unsigned long mem_start;
1758 unsigned long base_addr;
1762 * Some hardware also needs these fields (state,dev_list,
1763 * napi_list,unreg_list,close_list) but they are not
1764 * part of the usual set specified in Space.c.
1767 unsigned long state;
1769 struct list_head dev_list;
1770 struct list_head napi_list;
1771 struct list_head unreg_list;
1772 struct list_head close_list;
1773 struct list_head ptype_all;
1774 struct list_head ptype_specific;
1777 struct list_head upper;
1778 struct list_head lower;
1781 netdev_features_t features;
1782 netdev_features_t hw_features;
1783 netdev_features_t wanted_features;
1784 netdev_features_t vlan_features;
1785 netdev_features_t hw_enc_features;
1786 netdev_features_t mpls_features;
1787 netdev_features_t gso_partial_features;
1792 struct net_device_stats stats;
1794 atomic_long_t rx_dropped;
1795 atomic_long_t tx_dropped;
1796 atomic_long_t rx_nohandler;
1798 /* Stats to monitor link on/off, flapping */
1799 atomic_t carrier_up_count;
1800 atomic_t carrier_down_count;
1802 #ifdef CONFIG_WIRELESS_EXT
1803 const struct iw_handler_def *wireless_handlers;
1804 struct iw_public_data *wireless_data;
1806 const struct net_device_ops *netdev_ops;
1807 const struct ethtool_ops *ethtool_ops;
1808 #ifdef CONFIG_NET_SWITCHDEV
1809 const struct switchdev_ops *switchdev_ops;
1811 #ifdef CONFIG_NET_L3_MASTER_DEV
1812 const struct l3mdev_ops *l3mdev_ops;
1814 #if IS_ENABLED(CONFIG_IPV6)
1815 const struct ndisc_ops *ndisc_ops;
1818 #ifdef CONFIG_XFRM_OFFLOAD
1819 const struct xfrmdev_ops *xfrmdev_ops;
1822 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1823 const struct tlsdev_ops *tlsdev_ops;
1826 const struct header_ops *header_ops;
1829 unsigned int priv_flags;
1831 unsigned short gflags;
1832 unsigned short padded;
1834 unsigned char operstate;
1835 unsigned char link_mode;
1837 unsigned char if_port;
1840 /* Note : dev->mtu is often read without holding a lock.
1841 * Writers usually hold RTNL.
1842 * It is recommended to use READ_ONCE() to annotate the reads,
1843 * and to use WRITE_ONCE() to annotate the writes.
1846 unsigned int min_mtu;
1847 unsigned int max_mtu;
1848 unsigned short type;
1849 unsigned short hard_header_len;
1850 unsigned char min_header_len;
1852 unsigned short needed_headroom;
1853 unsigned short needed_tailroom;
1855 /* Interface address info. */
1856 unsigned char perm_addr[MAX_ADDR_LEN];
1857 unsigned char addr_assign_type;
1858 unsigned char addr_len;
1859 unsigned char upper_level;
1860 unsigned char lower_level;
1861 unsigned short neigh_priv_len;
1862 unsigned short dev_id;
1863 unsigned short dev_port;
1864 spinlock_t addr_list_lock;
1865 unsigned char name_assign_type;
1867 struct netdev_hw_addr_list uc;
1868 struct netdev_hw_addr_list mc;
1869 struct netdev_hw_addr_list dev_addrs;
1872 struct kset *queues_kset;
1874 unsigned int promiscuity;
1875 unsigned int allmulti;
1878 /* Protocol-specific pointers */
1880 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1881 struct vlan_info __rcu *vlan_info;
1883 #if IS_ENABLED(CONFIG_NET_DSA)
1884 struct dsa_port *dsa_ptr;
1886 #if IS_ENABLED(CONFIG_TIPC)
1887 struct tipc_bearer __rcu *tipc_ptr;
1889 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1892 struct in_device __rcu *ip_ptr;
1893 #if IS_ENABLED(CONFIG_DECNET)
1894 struct dn_dev __rcu *dn_ptr;
1896 struct inet6_dev __rcu *ip6_ptr;
1897 #if IS_ENABLED(CONFIG_AX25)
1900 struct wireless_dev *ieee80211_ptr;
1901 struct wpan_dev *ieee802154_ptr;
1902 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1903 struct mpls_dev __rcu *mpls_ptr;
1907 * Cache lines mostly used on receive path (including eth_type_trans())
1909 /* Interface address info used in eth_type_trans() */
1910 unsigned char *dev_addr;
1912 struct netdev_rx_queue *_rx;
1913 unsigned int num_rx_queues;
1914 unsigned int real_num_rx_queues;
1916 struct bpf_prog __rcu *xdp_prog;
1917 unsigned long gro_flush_timeout;
1918 rx_handler_func_t __rcu *rx_handler;
1919 void __rcu *rx_handler_data;
1921 #ifdef CONFIG_NET_CLS_ACT
1922 struct mini_Qdisc __rcu *miniq_ingress;
1924 struct netdev_queue __rcu *ingress_queue;
1925 #ifdef CONFIG_NETFILTER_INGRESS
1926 struct nf_hook_entries __rcu *nf_hooks_ingress;
1929 unsigned char broadcast[MAX_ADDR_LEN];
1930 #ifdef CONFIG_RFS_ACCEL
1931 struct cpu_rmap *rx_cpu_rmap;
1933 struct hlist_node index_hlist;
1936 * Cache lines mostly used on transmit path
1938 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1939 unsigned int num_tx_queues;
1940 unsigned int real_num_tx_queues;
1941 struct Qdisc *qdisc;
1942 #ifdef CONFIG_NET_SCHED
1943 DECLARE_HASHTABLE (qdisc_hash, 4);
1945 unsigned int tx_queue_len;
1946 spinlock_t tx_global_lock;
1950 struct xps_dev_maps __rcu *xps_cpus_map;
1951 struct xps_dev_maps __rcu *xps_rxqs_map;
1953 #ifdef CONFIG_NET_CLS_ACT
1954 struct mini_Qdisc __rcu *miniq_egress;
1957 /* These may be needed for future network-power-down code. */
1958 struct timer_list watchdog_timer;
1960 int __percpu *pcpu_refcnt;
1961 struct list_head todo_list;
1963 struct list_head link_watch_list;
1965 enum { NETREG_UNINITIALIZED=0,
1966 NETREG_REGISTERED, /* completed register_netdevice */
1967 NETREG_UNREGISTERING, /* called unregister_netdevice */
1968 NETREG_UNREGISTERED, /* completed unregister todo */
1969 NETREG_RELEASED, /* called free_netdev */
1970 NETREG_DUMMY, /* dummy device for NAPI poll */
1976 RTNL_LINK_INITIALIZED,
1977 RTNL_LINK_INITIALIZING,
1978 } rtnl_link_state:16;
1980 bool needs_free_netdev;
1981 void (*priv_destructor)(struct net_device *dev);
1983 #ifdef CONFIG_NETPOLL
1984 struct netpoll_info __rcu *npinfo;
1987 possible_net_t nd_net;
1989 /* mid-layer private */
1992 struct pcpu_lstats __percpu *lstats;
1993 struct pcpu_sw_netstats __percpu *tstats;
1994 struct pcpu_dstats __percpu *dstats;
1995 struct pcpu_vstats __percpu *vstats;
1998 #if IS_ENABLED(CONFIG_GARP)
1999 struct garp_port __rcu *garp_port;
2001 #if IS_ENABLED(CONFIG_MRP)
2002 struct mrp_port __rcu *mrp_port;
2006 const struct attribute_group *sysfs_groups[4];
2007 const struct attribute_group *sysfs_rx_queue_group;
2009 const struct rtnl_link_ops *rtnl_link_ops;
2011 /* for setting kernel sock attribute on TCP connection setup */
2012 #define GSO_MAX_SIZE 65536
2013 unsigned int gso_max_size;
2014 #define GSO_MAX_SEGS 65535
2018 const struct dcbnl_rtnl_ops *dcbnl_ops;
2021 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
2022 u8 prio_tc_map[TC_BITMASK + 1];
2024 #if IS_ENABLED(CONFIG_FCOE)
2025 unsigned int fcoe_ddp_xid;
2027 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2028 struct netprio_map __rcu *priomap;
2030 struct phy_device *phydev;
2031 struct sfp_bus *sfp_bus;
2032 struct lock_class_key *qdisc_tx_busylock;
2033 struct lock_class_key *qdisc_running_key;
2035 unsigned wol_enabled:1;
2037 #define to_net_dev(d) container_of(d, struct net_device, dev)
2039 static inline bool netif_elide_gro(const struct net_device *dev)
2041 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2046 #define NETDEV_ALIGN 32
2049 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2051 return dev->prio_tc_map[prio & TC_BITMASK];
2055 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2057 if (tc >= dev->num_tc)
2060 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2064 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2065 void netdev_reset_tc(struct net_device *dev);
2066 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2067 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2070 int netdev_get_num_tc(struct net_device *dev)
2075 void netdev_unbind_sb_channel(struct net_device *dev,
2076 struct net_device *sb_dev);
2077 int netdev_bind_sb_channel_queue(struct net_device *dev,
2078 struct net_device *sb_dev,
2079 u8 tc, u16 count, u16 offset);
2080 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2081 static inline int netdev_get_sb_channel(struct net_device *dev)
2083 return max_t(int, -dev->num_tc, 0);
2087 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2090 return &dev->_tx[index];
2093 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2094 const struct sk_buff *skb)
2096 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2099 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2100 void (*f)(struct net_device *,
2101 struct netdev_queue *,
2107 for (i = 0; i < dev->num_tx_queues; i++)
2108 f(dev, &dev->_tx[i], arg);
2111 #define netdev_lockdep_set_classes(dev) \
2113 static struct lock_class_key qdisc_tx_busylock_key; \
2114 static struct lock_class_key qdisc_running_key; \
2115 static struct lock_class_key qdisc_xmit_lock_key; \
2116 static struct lock_class_key dev_addr_list_lock_key; \
2119 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2120 (dev)->qdisc_running_key = &qdisc_running_key; \
2121 lockdep_set_class(&(dev)->addr_list_lock, \
2122 &dev_addr_list_lock_key); \
2123 for (i = 0; i < (dev)->num_tx_queues; i++) \
2124 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2125 &qdisc_xmit_lock_key); \
2128 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2129 struct sk_buff *skb,
2130 struct net_device *sb_dev);
2132 /* returns the headroom that the master device needs to take in account
2133 * when forwarding to this dev
2135 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2137 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2140 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2142 if (dev->netdev_ops->ndo_set_rx_headroom)
2143 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2146 /* set the device rx headroom to the dev's default */
2147 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2149 netdev_set_rx_headroom(dev, -1);
2153 * Net namespace inlines
2156 struct net *dev_net(const struct net_device *dev)
2158 return read_pnet(&dev->nd_net);
2162 void dev_net_set(struct net_device *dev, struct net *net)
2164 write_pnet(&dev->nd_net, net);
2168 * netdev_priv - access network device private data
2169 * @dev: network device
2171 * Get network device private data
2173 static inline void *netdev_priv(const struct net_device *dev)
2175 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2178 /* Set the sysfs physical device reference for the network logical device
2179 * if set prior to registration will cause a symlink during initialization.
2181 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2183 /* Set the sysfs device type for the network logical device to allow
2184 * fine-grained identification of different network device types. For
2185 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2187 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2189 /* Default NAPI poll() weight
2190 * Device drivers are strongly advised to not use bigger value
2192 #define NAPI_POLL_WEIGHT 64
2195 * netif_napi_add - initialize a NAPI context
2196 * @dev: network device
2197 * @napi: NAPI context
2198 * @poll: polling function
2199 * @weight: default weight
2201 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2202 * *any* of the other NAPI-related functions.
2204 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2205 int (*poll)(struct napi_struct *, int), int weight);
2208 * netif_tx_napi_add - initialize a NAPI context
2209 * @dev: network device
2210 * @napi: NAPI context
2211 * @poll: polling function
2212 * @weight: default weight
2214 * This variant of netif_napi_add() should be used from drivers using NAPI
2215 * to exclusively poll a TX queue.
2216 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2218 static inline void netif_tx_napi_add(struct net_device *dev,
2219 struct napi_struct *napi,
2220 int (*poll)(struct napi_struct *, int),
2223 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2224 netif_napi_add(dev, napi, poll, weight);
2228 * netif_napi_del - remove a NAPI context
2229 * @napi: NAPI context
2231 * netif_napi_del() removes a NAPI context from the network device NAPI list
2233 void netif_napi_del(struct napi_struct *napi);
2235 struct napi_gro_cb {
2236 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2239 /* Length of frag0. */
2240 unsigned int frag0_len;
2242 /* This indicates where we are processing relative to skb->data. */
2245 /* This is non-zero if the packet cannot be merged with the new skb. */
2248 /* Save the IP ID here and check when we get to the transport layer */
2251 /* Number of segments aggregated. */
2254 /* Start offset for remote checksum offload */
2255 u16 gro_remcsum_start;
2257 /* jiffies when first packet was created/queued */
2260 /* Used in ipv6_gro_receive() and foo-over-udp */
2263 /* This is non-zero if the packet may be of the same flow. */
2266 /* Used in tunnel GRO receive */
2269 /* GRO checksum is valid */
2272 /* Number of checksums via CHECKSUM_UNNECESSARY */
2277 #define NAPI_GRO_FREE 1
2278 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2280 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2283 /* Used in GRE, set in fou/gue_gro_receive */
2286 /* Used to determine if flush_id can be ignored */
2289 /* Number of gro_receive callbacks this packet already went through */
2290 u8 recursion_counter:4;
2294 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2297 /* used in skb_gro_receive() slow path */
2298 struct sk_buff *last;
2301 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2303 #define GRO_RECURSION_LIMIT 15
2304 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2306 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2309 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2310 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2311 struct list_head *head,
2312 struct sk_buff *skb)
2314 if (unlikely(gro_recursion_inc_test(skb))) {
2315 NAPI_GRO_CB(skb)->flush |= 1;
2319 return cb(head, skb);
2322 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2324 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2326 struct list_head *head,
2327 struct sk_buff *skb)
2329 if (unlikely(gro_recursion_inc_test(skb))) {
2330 NAPI_GRO_CB(skb)->flush |= 1;
2334 return cb(sk, head, skb);
2337 struct packet_type {
2338 __be16 type; /* This is really htons(ether_type). */
2339 struct net_device *dev; /* NULL is wildcarded here */
2340 int (*func) (struct sk_buff *,
2341 struct net_device *,
2342 struct packet_type *,
2343 struct net_device *);
2344 void (*list_func) (struct list_head *,
2345 struct packet_type *,
2346 struct net_device *);
2347 bool (*id_match)(struct packet_type *ptype,
2349 void *af_packet_priv;
2350 struct list_head list;
2353 struct offload_callbacks {
2354 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2355 netdev_features_t features);
2356 struct sk_buff *(*gro_receive)(struct list_head *head,
2357 struct sk_buff *skb);
2358 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2361 struct packet_offload {
2362 __be16 type; /* This is really htons(ether_type). */
2364 struct offload_callbacks callbacks;
2365 struct list_head list;
2368 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2369 struct pcpu_sw_netstats {
2374 struct u64_stats_sync syncp;
2377 #define __netdev_alloc_pcpu_stats(type, gfp) \
2379 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2382 for_each_possible_cpu(__cpu) { \
2383 typeof(type) *stat; \
2384 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2385 u64_stats_init(&stat->syncp); \
2391 #define netdev_alloc_pcpu_stats(type) \
2392 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2394 enum netdev_lag_tx_type {
2395 NETDEV_LAG_TX_TYPE_UNKNOWN,
2396 NETDEV_LAG_TX_TYPE_RANDOM,
2397 NETDEV_LAG_TX_TYPE_BROADCAST,
2398 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2399 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2400 NETDEV_LAG_TX_TYPE_HASH,
2403 enum netdev_lag_hash {
2404 NETDEV_LAG_HASH_NONE,
2406 NETDEV_LAG_HASH_L34,
2407 NETDEV_LAG_HASH_L23,
2408 NETDEV_LAG_HASH_E23,
2409 NETDEV_LAG_HASH_E34,
2410 NETDEV_LAG_HASH_UNKNOWN,
2413 struct netdev_lag_upper_info {
2414 enum netdev_lag_tx_type tx_type;
2415 enum netdev_lag_hash hash_type;
2418 struct netdev_lag_lower_state_info {
2423 #include <linux/notifier.h>
2425 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2426 * and the rtnetlink notification exclusion list in rtnetlink_event() when
2430 NETDEV_UP = 1, /* For now you can't veto a device up/down */
2432 NETDEV_REBOOT, /* Tell a protocol stack a network interface
2433 detected a hardware crash and restarted
2434 - we can use this eg to kick tcp sessions
2436 NETDEV_CHANGE, /* Notify device state change */
2439 NETDEV_CHANGEMTU, /* notify after mtu change happened */
2444 NETDEV_BONDING_FAILOVER,
2446 NETDEV_PRE_TYPE_CHANGE,
2447 NETDEV_POST_TYPE_CHANGE,
2450 NETDEV_NOTIFY_PEERS,
2454 NETDEV_PRECHANGEMTU, /* notify before mtu change happened */
2455 NETDEV_CHANGEINFODATA,
2456 NETDEV_BONDING_INFO,
2457 NETDEV_PRECHANGEUPPER,
2458 NETDEV_CHANGELOWERSTATE,
2459 NETDEV_UDP_TUNNEL_PUSH_INFO,
2460 NETDEV_UDP_TUNNEL_DROP_INFO,
2461 NETDEV_CHANGE_TX_QUEUE_LEN,
2462 NETDEV_CVLAN_FILTER_PUSH_INFO,
2463 NETDEV_CVLAN_FILTER_DROP_INFO,
2464 NETDEV_SVLAN_FILTER_PUSH_INFO,
2465 NETDEV_SVLAN_FILTER_DROP_INFO,
2467 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2469 int register_netdevice_notifier(struct notifier_block *nb);
2470 int unregister_netdevice_notifier(struct notifier_block *nb);
2472 struct netdev_notifier_info {
2473 struct net_device *dev;
2474 struct netlink_ext_ack *extack;
2477 struct netdev_notifier_info_ext {
2478 struct netdev_notifier_info info; /* must be first */
2484 struct netdev_notifier_change_info {
2485 struct netdev_notifier_info info; /* must be first */
2486 unsigned int flags_changed;
2489 struct netdev_notifier_changeupper_info {
2490 struct netdev_notifier_info info; /* must be first */
2491 struct net_device *upper_dev; /* new upper dev */
2492 bool master; /* is upper dev master */
2493 bool linking; /* is the notification for link or unlink */
2494 void *upper_info; /* upper dev info */
2497 struct netdev_notifier_changelowerstate_info {
2498 struct netdev_notifier_info info; /* must be first */
2499 void *lower_state_info; /* is lower dev state */
2502 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2503 struct net_device *dev)
2506 info->extack = NULL;
2509 static inline struct net_device *
2510 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2515 static inline struct netlink_ext_ack *
2516 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2518 return info->extack;
2521 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2524 extern rwlock_t dev_base_lock; /* Device list lock */
2526 #define for_each_netdev(net, d) \
2527 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2528 #define for_each_netdev_reverse(net, d) \
2529 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2530 #define for_each_netdev_rcu(net, d) \
2531 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2532 #define for_each_netdev_safe(net, d, n) \
2533 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2534 #define for_each_netdev_continue(net, d) \
2535 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2536 #define for_each_netdev_continue_rcu(net, d) \
2537 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2538 #define for_each_netdev_in_bond_rcu(bond, slave) \
2539 for_each_netdev_rcu(&init_net, slave) \
2540 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2541 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2543 static inline struct net_device *next_net_device(struct net_device *dev)
2545 struct list_head *lh;
2549 lh = dev->dev_list.next;
2550 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2553 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2555 struct list_head *lh;
2559 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2560 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2563 static inline struct net_device *first_net_device(struct net *net)
2565 return list_empty(&net->dev_base_head) ? NULL :
2566 net_device_entry(net->dev_base_head.next);
2569 static inline struct net_device *first_net_device_rcu(struct net *net)
2571 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2573 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2576 int netdev_boot_setup_check(struct net_device *dev);
2577 unsigned long netdev_boot_base(const char *prefix, int unit);
2578 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2579 const char *hwaddr);
2580 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2581 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2582 void dev_add_pack(struct packet_type *pt);
2583 void dev_remove_pack(struct packet_type *pt);
2584 void __dev_remove_pack(struct packet_type *pt);
2585 void dev_add_offload(struct packet_offload *po);
2586 void dev_remove_offload(struct packet_offload *po);
2588 int dev_get_iflink(const struct net_device *dev);
2589 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2590 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2591 unsigned short mask);
2592 struct net_device *dev_get_by_name(struct net *net, const char *name);
2593 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2594 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2595 int dev_alloc_name(struct net_device *dev, const char *name);
2596 int dev_open(struct net_device *dev);
2597 void dev_close(struct net_device *dev);
2598 void dev_close_many(struct list_head *head, bool unlink);
2599 void dev_disable_lro(struct net_device *dev);
2600 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2601 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2602 struct net_device *sb_dev,
2603 select_queue_fallback_t fallback);
2604 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2605 struct net_device *sb_dev,
2606 select_queue_fallback_t fallback);
2607 int dev_queue_xmit(struct sk_buff *skb);
2608 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2609 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2610 int register_netdevice(struct net_device *dev);
2611 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2612 void unregister_netdevice_many(struct list_head *head);
2613 static inline void unregister_netdevice(struct net_device *dev)
2615 unregister_netdevice_queue(dev, NULL);
2618 int netdev_refcnt_read(const struct net_device *dev);
2619 void free_netdev(struct net_device *dev);
2620 void netdev_freemem(struct net_device *dev);
2621 void synchronize_net(void);
2622 int init_dummy_netdev(struct net_device *dev);
2624 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2625 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2626 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2627 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2628 int netdev_get_name(struct net *net, char *name, int ifindex);
2629 int dev_restart(struct net_device *dev);
2630 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2632 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2634 return NAPI_GRO_CB(skb)->data_offset;
2637 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2639 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2642 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2644 NAPI_GRO_CB(skb)->data_offset += len;
2647 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2648 unsigned int offset)
2650 return NAPI_GRO_CB(skb)->frag0 + offset;
2653 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2655 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2658 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2660 NAPI_GRO_CB(skb)->frag0 = NULL;
2661 NAPI_GRO_CB(skb)->frag0_len = 0;
2664 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2665 unsigned int offset)
2667 if (!pskb_may_pull(skb, hlen))
2670 skb_gro_frag0_invalidate(skb);
2671 return skb->data + offset;
2674 static inline void *skb_gro_network_header(struct sk_buff *skb)
2676 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2677 skb_network_offset(skb);
2680 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2681 const void *start, unsigned int len)
2683 if (NAPI_GRO_CB(skb)->csum_valid)
2684 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2685 csum_partial(start, len, 0));
2688 /* GRO checksum functions. These are logical equivalents of the normal
2689 * checksum functions (in skbuff.h) except that they operate on the GRO
2690 * offsets and fields in sk_buff.
2693 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2695 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2697 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2700 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2704 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2705 skb_checksum_start_offset(skb) <
2706 skb_gro_offset(skb)) &&
2707 !skb_at_gro_remcsum_start(skb) &&
2708 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2709 (!zero_okay || check));
2712 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2715 if (NAPI_GRO_CB(skb)->csum_valid &&
2716 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2719 NAPI_GRO_CB(skb)->csum = psum;
2721 return __skb_gro_checksum_complete(skb);
2724 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2726 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2727 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2728 NAPI_GRO_CB(skb)->csum_cnt--;
2730 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2731 * verified a new top level checksum or an encapsulated one
2732 * during GRO. This saves work if we fallback to normal path.
2734 __skb_incr_checksum_unnecessary(skb);
2738 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2741 __sum16 __ret = 0; \
2742 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2743 __ret = __skb_gro_checksum_validate_complete(skb, \
2744 compute_pseudo(skb, proto)); \
2746 skb_gro_incr_csum_unnecessary(skb); \
2750 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2751 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2753 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2755 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2757 #define skb_gro_checksum_simple_validate(skb) \
2758 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2760 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2762 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2763 !NAPI_GRO_CB(skb)->csum_valid);
2766 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2767 __sum16 check, __wsum pseudo)
2769 NAPI_GRO_CB(skb)->csum = ~pseudo;
2770 NAPI_GRO_CB(skb)->csum_valid = 1;
2773 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2775 if (__skb_gro_checksum_convert_check(skb)) \
2776 __skb_gro_checksum_convert(skb, check, \
2777 compute_pseudo(skb, proto)); \
2780 struct gro_remcsum {
2785 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2791 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2792 unsigned int off, size_t hdrlen,
2793 int start, int offset,
2794 struct gro_remcsum *grc,
2798 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2800 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2803 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2807 ptr = skb_gro_header_fast(skb, off);
2808 if (skb_gro_header_hard(skb, off + plen)) {
2809 ptr = skb_gro_header_slow(skb, off + plen, off);
2814 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2817 /* Adjust skb->csum since we changed the packet */
2818 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2820 grc->offset = off + hdrlen + offset;
2826 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2827 struct gro_remcsum *grc)
2830 size_t plen = grc->offset + sizeof(u16);
2835 ptr = skb_gro_header_fast(skb, grc->offset);
2836 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2837 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2842 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2845 #ifdef CONFIG_XFRM_OFFLOAD
2846 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2848 if (PTR_ERR(pp) != -EINPROGRESS)
2849 NAPI_GRO_CB(skb)->flush |= flush;
2851 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2854 struct gro_remcsum *grc)
2856 if (PTR_ERR(pp) != -EINPROGRESS) {
2857 NAPI_GRO_CB(skb)->flush |= flush;
2858 skb_gro_remcsum_cleanup(skb, grc);
2859 skb->remcsum_offload = 0;
2863 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2865 NAPI_GRO_CB(skb)->flush |= flush;
2867 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2870 struct gro_remcsum *grc)
2872 NAPI_GRO_CB(skb)->flush |= flush;
2873 skb_gro_remcsum_cleanup(skb, grc);
2874 skb->remcsum_offload = 0;
2878 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2879 unsigned short type,
2880 const void *daddr, const void *saddr,
2883 if (!dev->header_ops || !dev->header_ops->create)
2886 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2889 static inline int dev_parse_header(const struct sk_buff *skb,
2890 unsigned char *haddr)
2892 const struct net_device *dev = skb->dev;
2894 if (!dev->header_ops || !dev->header_ops->parse)
2896 return dev->header_ops->parse(skb, haddr);
2899 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2901 const struct net_device *dev = skb->dev;
2903 if (!dev->header_ops || !dev->header_ops->parse_protocol)
2905 return dev->header_ops->parse_protocol(skb);
2908 /* ll_header must have at least hard_header_len allocated */
2909 static inline bool dev_validate_header(const struct net_device *dev,
2910 char *ll_header, int len)
2912 if (likely(len >= dev->hard_header_len))
2914 if (len < dev->min_header_len)
2917 if (capable(CAP_SYS_RAWIO)) {
2918 memset(ll_header + len, 0, dev->hard_header_len - len);
2922 if (dev->header_ops && dev->header_ops->validate)
2923 return dev->header_ops->validate(ll_header, len);
2928 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2930 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2931 static inline int unregister_gifconf(unsigned int family)
2933 return register_gifconf(family, NULL);
2936 #ifdef CONFIG_NET_FLOW_LIMIT
2937 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2938 struct sd_flow_limit {
2940 unsigned int num_buckets;
2941 unsigned int history_head;
2942 u16 history[FLOW_LIMIT_HISTORY];
2946 extern int netdev_flow_limit_table_len;
2947 #endif /* CONFIG_NET_FLOW_LIMIT */
2950 * Incoming packets are placed on per-CPU queues
2952 struct softnet_data {
2953 struct list_head poll_list;
2954 struct sk_buff_head process_queue;
2957 unsigned int processed;
2958 unsigned int time_squeeze;
2959 unsigned int received_rps;
2961 struct softnet_data *rps_ipi_list;
2963 #ifdef CONFIG_NET_FLOW_LIMIT
2964 struct sd_flow_limit __rcu *flow_limit;
2966 struct Qdisc *output_queue;
2967 struct Qdisc **output_queue_tailp;
2968 struct sk_buff *completion_queue;
2969 #ifdef CONFIG_XFRM_OFFLOAD
2970 struct sk_buff_head xfrm_backlog;
2972 /* written and read only by owning cpu: */
2978 /* input_queue_head should be written by cpu owning this struct,
2979 * and only read by other cpus. Worth using a cache line.
2981 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2983 /* Elements below can be accessed between CPUs for RPS/RFS */
2984 call_single_data_t csd ____cacheline_aligned_in_smp;
2985 struct softnet_data *rps_ipi_next;
2987 unsigned int input_queue_tail;
2989 unsigned int dropped;
2990 struct sk_buff_head input_pkt_queue;
2991 struct napi_struct backlog;
2995 static inline void input_queue_head_incr(struct softnet_data *sd)
2998 sd->input_queue_head++;
3002 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3003 unsigned int *qtail)
3006 *qtail = ++sd->input_queue_tail;
3010 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3012 static inline int dev_recursion_level(void)
3014 return this_cpu_read(softnet_data.xmit.recursion);
3017 #define XMIT_RECURSION_LIMIT 8
3018 static inline bool dev_xmit_recursion(void)
3020 return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3021 XMIT_RECURSION_LIMIT);
3024 static inline void dev_xmit_recursion_inc(void)
3026 __this_cpu_inc(softnet_data.xmit.recursion);
3029 static inline void dev_xmit_recursion_dec(void)
3031 __this_cpu_dec(softnet_data.xmit.recursion);
3034 void __netif_schedule(struct Qdisc *q);
3035 void netif_schedule_queue(struct netdev_queue *txq);
3037 static inline void netif_tx_schedule_all(struct net_device *dev)
3041 for (i = 0; i < dev->num_tx_queues; i++)
3042 netif_schedule_queue(netdev_get_tx_queue(dev, i));
3045 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3047 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3051 * netif_start_queue - allow transmit
3052 * @dev: network device
3054 * Allow upper layers to call the device hard_start_xmit routine.
3056 static inline void netif_start_queue(struct net_device *dev)
3058 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3061 static inline void netif_tx_start_all_queues(struct net_device *dev)
3065 for (i = 0; i < dev->num_tx_queues; i++) {
3066 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3067 netif_tx_start_queue(txq);
3071 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3074 * netif_wake_queue - restart transmit
3075 * @dev: network device
3077 * Allow upper layers to call the device hard_start_xmit routine.
3078 * Used for flow control when transmit resources are available.
3080 static inline void netif_wake_queue(struct net_device *dev)
3082 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3085 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3089 for (i = 0; i < dev->num_tx_queues; i++) {
3090 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3091 netif_tx_wake_queue(txq);
3095 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3097 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3101 * netif_stop_queue - stop transmitted packets
3102 * @dev: network device
3104 * Stop upper layers calling the device hard_start_xmit routine.
3105 * Used for flow control when transmit resources are unavailable.
3107 static inline void netif_stop_queue(struct net_device *dev)
3109 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3112 void netif_tx_stop_all_queues(struct net_device *dev);
3114 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3116 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3120 * netif_queue_stopped - test if transmit queue is flowblocked
3121 * @dev: network device
3123 * Test if transmit queue on device is currently unable to send.
3125 static inline bool netif_queue_stopped(const struct net_device *dev)
3127 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3130 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3132 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3136 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3138 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3142 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3144 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3148 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3149 * @dev_queue: pointer to transmit queue
3151 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3152 * to give appropriate hint to the CPU.
3154 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3157 prefetchw(&dev_queue->dql.num_queued);
3162 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3163 * @dev_queue: pointer to transmit queue
3165 * BQL enabled drivers might use this helper in their TX completion path,
3166 * to give appropriate hint to the CPU.
3168 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3171 prefetchw(&dev_queue->dql.limit);
3175 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3179 dql_queued(&dev_queue->dql, bytes);
3181 if (likely(dql_avail(&dev_queue->dql) >= 0))
3184 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3187 * The XOFF flag must be set before checking the dql_avail below,
3188 * because in netdev_tx_completed_queue we update the dql_completed
3189 * before checking the XOFF flag.
3193 /* check again in case another CPU has just made room avail */
3194 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3195 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3200 * netdev_sent_queue - report the number of bytes queued to hardware
3201 * @dev: network device
3202 * @bytes: number of bytes queued to the hardware device queue
3204 * Report the number of bytes queued for sending/completion to the network
3205 * device hardware queue. @bytes should be a good approximation and should
3206 * exactly match netdev_completed_queue() @bytes
3208 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3210 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3213 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3214 unsigned int pkts, unsigned int bytes)
3217 if (unlikely(!bytes))
3220 dql_completed(&dev_queue->dql, bytes);
3223 * Without the memory barrier there is a small possiblity that
3224 * netdev_tx_sent_queue will miss the update and cause the queue to
3225 * be stopped forever
3229 if (dql_avail(&dev_queue->dql) < 0)
3232 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3233 netif_schedule_queue(dev_queue);
3238 * netdev_completed_queue - report bytes and packets completed by device
3239 * @dev: network device
3240 * @pkts: actual number of packets sent over the medium
3241 * @bytes: actual number of bytes sent over the medium
3243 * Report the number of bytes and packets transmitted by the network device
3244 * hardware queue over the physical medium, @bytes must exactly match the
3245 * @bytes amount passed to netdev_sent_queue()
3247 static inline void netdev_completed_queue(struct net_device *dev,
3248 unsigned int pkts, unsigned int bytes)
3250 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3253 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3256 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3262 * netdev_reset_queue - reset the packets and bytes count of a network device
3263 * @dev_queue: network device
3265 * Reset the bytes and packet count of a network device and clear the
3266 * software flow control OFF bit for this network device
3268 static inline void netdev_reset_queue(struct net_device *dev_queue)
3270 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3274 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3275 * @dev: network device
3276 * @queue_index: given tx queue index
3278 * Returns 0 if given tx queue index >= number of device tx queues,
3279 * otherwise returns the originally passed tx queue index.
3281 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3283 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3284 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3285 dev->name, queue_index,
3286 dev->real_num_tx_queues);
3294 * netif_running - test if up
3295 * @dev: network device
3297 * Test if the device has been brought up.
3299 static inline bool netif_running(const struct net_device *dev)
3301 return test_bit(__LINK_STATE_START, &dev->state);
3305 * Routines to manage the subqueues on a device. We only need start,
3306 * stop, and a check if it's stopped. All other device management is
3307 * done at the overall netdevice level.
3308 * Also test the device if we're multiqueue.
3312 * netif_start_subqueue - allow sending packets on subqueue
3313 * @dev: network device
3314 * @queue_index: sub queue index
3316 * Start individual transmit queue of a device with multiple transmit queues.
3318 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3320 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3322 netif_tx_start_queue(txq);
3326 * netif_stop_subqueue - stop sending packets on subqueue
3327 * @dev: network device
3328 * @queue_index: sub queue index
3330 * Stop individual transmit queue of a device with multiple transmit queues.
3332 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3334 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3335 netif_tx_stop_queue(txq);
3339 * netif_subqueue_stopped - test status of subqueue
3340 * @dev: network device
3341 * @queue_index: sub queue index
3343 * Check individual transmit queue of a device with multiple transmit queues.
3345 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3348 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3350 return netif_tx_queue_stopped(txq);
3353 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3354 struct sk_buff *skb)
3356 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3360 * netif_wake_subqueue - allow sending packets on subqueue
3361 * @dev: network device
3362 * @queue_index: sub queue index
3364 * Resume individual transmit queue of a device with multiple transmit queues.
3366 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3368 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3370 netif_tx_wake_queue(txq);
3374 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3376 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3377 u16 index, bool is_rxqs_map);
3380 * netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3381 * @j: CPU/Rx queue index
3382 * @mask: bitmask of all cpus/rx queues
3383 * @nr_bits: number of bits in the bitmask
3385 * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3387 static inline bool netif_attr_test_mask(unsigned long j,
3388 const unsigned long *mask,
3389 unsigned int nr_bits)
3391 cpu_max_bits_warn(j, nr_bits);
3392 return test_bit(j, mask);
3396 * netif_attr_test_online - Test for online CPU/Rx queue
3397 * @j: CPU/Rx queue index
3398 * @online_mask: bitmask for CPUs/Rx queues that are online
3399 * @nr_bits: number of bits in the bitmask
3401 * Returns true if a CPU/Rx queue is online.
3403 static inline bool netif_attr_test_online(unsigned long j,
3404 const unsigned long *online_mask,
3405 unsigned int nr_bits)
3407 cpu_max_bits_warn(j, nr_bits);
3410 return test_bit(j, online_mask);
3412 return (j < nr_bits);
3416 * netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3417 * @n: CPU/Rx queue index
3418 * @srcp: the cpumask/Rx queue mask pointer
3419 * @nr_bits: number of bits in the bitmask
3421 * Returns >= nr_bits if no further CPUs/Rx queues set.
3423 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3424 unsigned int nr_bits)
3426 /* -1 is a legal arg here. */
3428 cpu_max_bits_warn(n, nr_bits);
3431 return find_next_bit(srcp, nr_bits, n + 1);
3437 * netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3438 * @n: CPU/Rx queue index
3439 * @src1p: the first CPUs/Rx queues mask pointer
3440 * @src2p: the second CPUs/Rx queues mask pointer
3441 * @nr_bits: number of bits in the bitmask
3443 * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3445 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3446 const unsigned long *src2p,
3447 unsigned int nr_bits)
3449 /* -1 is a legal arg here. */
3451 cpu_max_bits_warn(n, nr_bits);
3454 return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3456 return find_next_bit(src1p, nr_bits, n + 1);
3458 return find_next_bit(src2p, nr_bits, n + 1);
3463 static inline int netif_set_xps_queue(struct net_device *dev,
3464 const struct cpumask *mask,
3470 static inline int __netif_set_xps_queue(struct net_device *dev,
3471 const unsigned long *mask,
3472 u16 index, bool is_rxqs_map)
3479 * netif_is_multiqueue - test if device has multiple transmit queues
3480 * @dev: network device
3482 * Check if device has multiple transmit queues
3484 static inline bool netif_is_multiqueue(const struct net_device *dev)
3486 return dev->num_tx_queues > 1;
3489 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3492 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3494 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3497 dev->real_num_rx_queues = rxqs;
3502 static inline struct netdev_rx_queue *
3503 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3505 return dev->_rx + rxq;
3509 static inline unsigned int get_netdev_rx_queue_index(
3510 struct netdev_rx_queue *queue)
3512 struct net_device *dev = queue->dev;
3513 int index = queue - dev->_rx;
3515 BUG_ON(index >= dev->num_rx_queues);
3520 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3521 int netif_get_num_default_rss_queues(void);
3523 enum skb_free_reason {
3524 SKB_REASON_CONSUMED,
3528 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3529 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3532 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3533 * interrupt context or with hardware interrupts being disabled.
3534 * (in_irq() || irqs_disabled())
3536 * We provide four helpers that can be used in following contexts :
3538 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3539 * replacing kfree_skb(skb)
3541 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3542 * Typically used in place of consume_skb(skb) in TX completion path
3544 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3545 * replacing kfree_skb(skb)
3547 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3548 * and consumed a packet. Used in place of consume_skb(skb)
3550 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3552 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3555 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3557 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3560 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3562 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3565 static inline void dev_consume_skb_any(struct sk_buff *skb)
3567 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3570 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3571 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3572 int netif_rx(struct sk_buff *skb);
3573 int netif_rx_ni(struct sk_buff *skb);
3574 int netif_receive_skb(struct sk_buff *skb);
3575 int netif_receive_skb_core(struct sk_buff *skb);
3576 void netif_receive_skb_list(struct list_head *head);
3577 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3578 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3579 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3580 gro_result_t napi_gro_frags(struct napi_struct *napi);
3581 struct packet_offload *gro_find_receive_by_type(__be16 type);
3582 struct packet_offload *gro_find_complete_by_type(__be16 type);
3584 static inline void napi_free_frags(struct napi_struct *napi)
3586 kfree_skb(napi->skb);
3590 bool netdev_is_rx_handler_busy(struct net_device *dev);
3591 int netdev_rx_handler_register(struct net_device *dev,
3592 rx_handler_func_t *rx_handler,
3593 void *rx_handler_data);
3594 void netdev_rx_handler_unregister(struct net_device *dev);
3596 bool dev_valid_name(const char *name);
3597 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3599 return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3601 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3602 bool *need_copyout);
3603 int dev_ifconf(struct net *net, struct ifconf *, int);
3604 int dev_ethtool(struct net *net, struct ifreq *);
3605 unsigned int dev_get_flags(const struct net_device *);
3606 int __dev_change_flags(struct net_device *, unsigned int flags);
3607 int dev_change_flags(struct net_device *, unsigned int);
3608 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3609 unsigned int gchanges);
3610 int dev_change_name(struct net_device *, const char *);
3611 int dev_set_alias(struct net_device *, const char *, size_t);
3612 int dev_get_alias(const struct net_device *, char *, size_t);
3613 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3614 int __dev_set_mtu(struct net_device *, int);
3615 int dev_validate_mtu(struct net_device *dev, int mtu,
3616 struct netlink_ext_ack *extack);
3617 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3618 struct netlink_ext_ack *extack);
3619 int dev_set_mtu(struct net_device *, int);
3620 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3621 void dev_set_group(struct net_device *, int);
3622 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3623 int dev_change_carrier(struct net_device *, bool new_carrier);
3624 int dev_get_phys_port_id(struct net_device *dev,
3625 struct netdev_phys_item_id *ppid);
3626 int dev_get_phys_port_name(struct net_device *dev,
3627 char *name, size_t len);
3628 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3629 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3630 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3631 struct netdev_queue *txq, int *ret);
3633 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3634 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3636 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3637 enum bpf_netdev_command cmd);
3638 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3640 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3641 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3642 bool is_skb_forwardable(const struct net_device *dev,
3643 const struct sk_buff *skb);
3645 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3646 struct sk_buff *skb)
3648 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3649 unlikely(!is_skb_forwardable(dev, skb))) {
3650 atomic_long_inc(&dev->rx_dropped);
3655 skb_scrub_packet(skb, true);
3660 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3662 extern int netdev_budget;
3663 extern unsigned int netdev_budget_usecs;
3665 /* Called by rtnetlink.c:rtnl_unlock() */
3666 void netdev_run_todo(void);
3669 * dev_put - release reference to device
3670 * @dev: network device
3672 * Release reference to device to allow it to be freed.
3674 static inline void dev_put(struct net_device *dev)
3676 this_cpu_dec(*dev->pcpu_refcnt);
3680 * dev_hold - get reference to device
3681 * @dev: network device
3683 * Hold reference to device to keep it from being freed.
3685 static inline void dev_hold(struct net_device *dev)
3687 this_cpu_inc(*dev->pcpu_refcnt);
3690 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3691 * and _off may be called from IRQ context, but it is caller
3692 * who is responsible for serialization of these calls.
3694 * The name carrier is inappropriate, these functions should really be
3695 * called netif_lowerlayer_*() because they represent the state of any
3696 * kind of lower layer not just hardware media.
3699 void linkwatch_init_dev(struct net_device *dev);
3700 void linkwatch_fire_event(struct net_device *dev);
3701 void linkwatch_forget_dev(struct net_device *dev);
3704 * netif_carrier_ok - test if carrier present
3705 * @dev: network device
3707 * Check if carrier is present on device
3709 static inline bool netif_carrier_ok(const struct net_device *dev)
3711 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3714 unsigned long dev_trans_start(struct net_device *dev);
3716 void __netdev_watchdog_up(struct net_device *dev);
3718 void netif_carrier_on(struct net_device *dev);
3720 void netif_carrier_off(struct net_device *dev);
3723 * netif_dormant_on - mark device as dormant.
3724 * @dev: network device
3726 * Mark device as dormant (as per RFC2863).
3728 * The dormant state indicates that the relevant interface is not
3729 * actually in a condition to pass packets (i.e., it is not 'up') but is
3730 * in a "pending" state, waiting for some external event. For "on-
3731 * demand" interfaces, this new state identifies the situation where the
3732 * interface is waiting for events to place it in the up state.
3734 static inline void netif_dormant_on(struct net_device *dev)
3736 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3737 linkwatch_fire_event(dev);
3741 * netif_dormant_off - set device as not dormant.
3742 * @dev: network device
3744 * Device is not in dormant state.
3746 static inline void netif_dormant_off(struct net_device *dev)
3748 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3749 linkwatch_fire_event(dev);
3753 * netif_dormant - test if device is dormant
3754 * @dev: network device
3756 * Check if device is dormant.
3758 static inline bool netif_dormant(const struct net_device *dev)
3760 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3765 * netif_oper_up - test if device is operational
3766 * @dev: network device
3768 * Check if carrier is operational
3770 static inline bool netif_oper_up(const struct net_device *dev)
3772 return (dev->operstate == IF_OPER_UP ||
3773 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3777 * netif_device_present - is device available or removed
3778 * @dev: network device
3780 * Check if device has not been removed from system.
3782 static inline bool netif_device_present(struct net_device *dev)
3784 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3787 void netif_device_detach(struct net_device *dev);
3789 void netif_device_attach(struct net_device *dev);
3792 * Network interface message level settings
3796 NETIF_MSG_DRV = 0x0001,
3797 NETIF_MSG_PROBE = 0x0002,
3798 NETIF_MSG_LINK = 0x0004,
3799 NETIF_MSG_TIMER = 0x0008,
3800 NETIF_MSG_IFDOWN = 0x0010,
3801 NETIF_MSG_IFUP = 0x0020,
3802 NETIF_MSG_RX_ERR = 0x0040,
3803 NETIF_MSG_TX_ERR = 0x0080,
3804 NETIF_MSG_TX_QUEUED = 0x0100,
3805 NETIF_MSG_INTR = 0x0200,
3806 NETIF_MSG_TX_DONE = 0x0400,
3807 NETIF_MSG_RX_STATUS = 0x0800,
3808 NETIF_MSG_PKTDATA = 0x1000,
3809 NETIF_MSG_HW = 0x2000,
3810 NETIF_MSG_WOL = 0x4000,
3813 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3814 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3815 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3816 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3817 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3818 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3819 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3820 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3821 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3822 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3823 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3824 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3825 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3826 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3827 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3829 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3832 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3833 return default_msg_enable_bits;
3834 if (debug_value == 0) /* no output */
3836 /* set low N bits */
3837 return (1U << debug_value) - 1;
3840 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3842 spin_lock(&txq->_xmit_lock);
3843 txq->xmit_lock_owner = cpu;
3846 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3848 __acquire(&txq->_xmit_lock);
3852 static inline void __netif_tx_release(struct netdev_queue *txq)
3854 __release(&txq->_xmit_lock);
3857 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3859 spin_lock_bh(&txq->_xmit_lock);
3860 txq->xmit_lock_owner = smp_processor_id();
3863 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3865 bool ok = spin_trylock(&txq->_xmit_lock);
3867 txq->xmit_lock_owner = smp_processor_id();
3871 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3873 txq->xmit_lock_owner = -1;
3874 spin_unlock(&txq->_xmit_lock);
3877 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3879 txq->xmit_lock_owner = -1;
3880 spin_unlock_bh(&txq->_xmit_lock);
3883 static inline void txq_trans_update(struct netdev_queue *txq)
3885 if (txq->xmit_lock_owner != -1)
3886 txq->trans_start = jiffies;
3889 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3890 static inline void netif_trans_update(struct net_device *dev)
3892 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3894 if (txq->trans_start != jiffies)
3895 txq->trans_start = jiffies;
3899 * netif_tx_lock - grab network device transmit lock
3900 * @dev: network device
3902 * Get network device transmit lock
3904 static inline void netif_tx_lock(struct net_device *dev)
3909 spin_lock(&dev->tx_global_lock);
3910 cpu = smp_processor_id();
3911 for (i = 0; i < dev->num_tx_queues; i++) {
3912 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3914 /* We are the only thread of execution doing a
3915 * freeze, but we have to grab the _xmit_lock in
3916 * order to synchronize with threads which are in
3917 * the ->hard_start_xmit() handler and already
3918 * checked the frozen bit.
3920 __netif_tx_lock(txq, cpu);
3921 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3922 __netif_tx_unlock(txq);
3926 static inline void netif_tx_lock_bh(struct net_device *dev)
3932 static inline void netif_tx_unlock(struct net_device *dev)
3936 for (i = 0; i < dev->num_tx_queues; i++) {
3937 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3939 /* No need to grab the _xmit_lock here. If the
3940 * queue is not stopped for another reason, we
3943 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3944 netif_schedule_queue(txq);
3946 spin_unlock(&dev->tx_global_lock);
3949 static inline void netif_tx_unlock_bh(struct net_device *dev)
3951 netif_tx_unlock(dev);
3955 #define HARD_TX_LOCK(dev, txq, cpu) { \
3956 if ((dev->features & NETIF_F_LLTX) == 0) { \
3957 __netif_tx_lock(txq, cpu); \
3959 __netif_tx_acquire(txq); \
3963 #define HARD_TX_TRYLOCK(dev, txq) \
3964 (((dev->features & NETIF_F_LLTX) == 0) ? \
3965 __netif_tx_trylock(txq) : \
3966 __netif_tx_acquire(txq))
3968 #define HARD_TX_UNLOCK(dev, txq) { \
3969 if ((dev->features & NETIF_F_LLTX) == 0) { \
3970 __netif_tx_unlock(txq); \
3972 __netif_tx_release(txq); \
3976 static inline void netif_tx_disable(struct net_device *dev)
3982 cpu = smp_processor_id();
3983 spin_lock(&dev->tx_global_lock);
3984 for (i = 0; i < dev->num_tx_queues; i++) {
3985 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3987 __netif_tx_lock(txq, cpu);
3988 netif_tx_stop_queue(txq);
3989 __netif_tx_unlock(txq);
3991 spin_unlock(&dev->tx_global_lock);
3995 static inline void netif_addr_lock(struct net_device *dev)
3997 spin_lock(&dev->addr_list_lock);
4000 static inline void netif_addr_lock_nested(struct net_device *dev)
4002 int subclass = SINGLE_DEPTH_NESTING;
4004 if (dev->netdev_ops->ndo_get_lock_subclass)
4005 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
4007 spin_lock_nested(&dev->addr_list_lock, subclass);
4010 static inline void netif_addr_lock_bh(struct net_device *dev)
4012 spin_lock_bh(&dev->addr_list_lock);
4015 static inline void netif_addr_unlock(struct net_device *dev)
4017 spin_unlock(&dev->addr_list_lock);
4020 static inline void netif_addr_unlock_bh(struct net_device *dev)
4022 spin_unlock_bh(&dev->addr_list_lock);
4026 * dev_addrs walker. Should be used only for read access. Call with
4027 * rcu_read_lock held.
4029 #define for_each_dev_addr(dev, ha) \
4030 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4032 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4034 void ether_setup(struct net_device *dev);
4036 /* Support for loadable net-drivers */
4037 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4038 unsigned char name_assign_type,
4039 void (*setup)(struct net_device *),
4040 unsigned int txqs, unsigned int rxqs);
4041 int dev_get_valid_name(struct net *net, struct net_device *dev,
4044 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4045 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4047 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4048 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4051 int register_netdev(struct net_device *dev);
4052 void unregister_netdev(struct net_device *dev);
4054 /* General hardware address lists handling functions */
4055 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4056 struct netdev_hw_addr_list *from_list, int addr_len);
4057 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4058 struct netdev_hw_addr_list *from_list, int addr_len);
4059 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4060 struct net_device *dev,
4061 int (*sync)(struct net_device *, const unsigned char *),
4062 int (*unsync)(struct net_device *,
4063 const unsigned char *));
4064 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4065 struct net_device *dev,
4066 int (*unsync)(struct net_device *,
4067 const unsigned char *));
4068 void __hw_addr_init(struct netdev_hw_addr_list *list);
4070 /* Functions used for device addresses handling */
4071 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4072 unsigned char addr_type);
4073 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4074 unsigned char addr_type);
4075 void dev_addr_flush(struct net_device *dev);
4076 int dev_addr_init(struct net_device *dev);
4078 /* Functions used for unicast addresses handling */
4079 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4080 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4081 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4082 int dev_uc_sync(struct net_device *to, struct net_device *from);
4083 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4084 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4085 void dev_uc_flush(struct net_device *dev);
4086 void dev_uc_init(struct net_device *dev);
4089 * __dev_uc_sync - Synchonize device's unicast list
4090 * @dev: device to sync
4091 * @sync: function to call if address should be added
4092 * @unsync: function to call if address should be removed
4094 * Add newly added addresses to the interface, and release
4095 * addresses that have been deleted.
4097 static inline int __dev_uc_sync(struct net_device *dev,
4098 int (*sync)(struct net_device *,
4099 const unsigned char *),
4100 int (*unsync)(struct net_device *,
4101 const unsigned char *))
4103 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4107 * __dev_uc_unsync - Remove synchronized addresses from device
4108 * @dev: device to sync
4109 * @unsync: function to call if address should be removed
4111 * Remove all addresses that were added to the device by dev_uc_sync().
4113 static inline void __dev_uc_unsync(struct net_device *dev,
4114 int (*unsync)(struct net_device *,
4115 const unsigned char *))
4117 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
4120 /* Functions used for multicast addresses handling */
4121 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4122 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4123 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4124 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4125 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4126 int dev_mc_sync(struct net_device *to, struct net_device *from);
4127 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4128 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4129 void dev_mc_flush(struct net_device *dev);
4130 void dev_mc_init(struct net_device *dev);
4133 * __dev_mc_sync - Synchonize device's multicast list
4134 * @dev: device to sync
4135 * @sync: function to call if address should be added
4136 * @unsync: function to call if address should be removed
4138 * Add newly added addresses to the interface, and release
4139 * addresses that have been deleted.
4141 static inline int __dev_mc_sync(struct net_device *dev,
4142 int (*sync)(struct net_device *,
4143 const unsigned char *),
4144 int (*unsync)(struct net_device *,
4145 const unsigned char *))
4147 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4151 * __dev_mc_unsync - Remove synchronized addresses from device
4152 * @dev: device to sync
4153 * @unsync: function to call if address should be removed
4155 * Remove all addresses that were added to the device by dev_mc_sync().
4157 static inline void __dev_mc_unsync(struct net_device *dev,
4158 int (*unsync)(struct net_device *,
4159 const unsigned char *))
4161 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
4164 /* Functions used for secondary unicast and multicast support */
4165 void dev_set_rx_mode(struct net_device *dev);
4166 void __dev_set_rx_mode(struct net_device *dev);
4167 int dev_set_promiscuity(struct net_device *dev, int inc);
4168 int dev_set_allmulti(struct net_device *dev, int inc);
4169 void netdev_state_change(struct net_device *dev);
4170 void netdev_notify_peers(struct net_device *dev);
4171 void netdev_features_change(struct net_device *dev);
4172 /* Load a device via the kmod */
4173 void dev_load(struct net *net, const char *name);
4174 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4175 struct rtnl_link_stats64 *storage);
4176 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4177 const struct net_device_stats *netdev_stats);
4179 extern int netdev_max_backlog;
4180 extern int netdev_tstamp_prequeue;
4181 extern int weight_p;
4182 extern int dev_weight_rx_bias;
4183 extern int dev_weight_tx_bias;
4184 extern int dev_rx_weight;
4185 extern int dev_tx_weight;
4187 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4188 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4189 struct list_head **iter);
4190 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4191 struct list_head **iter);
4193 /* iterate through upper list, must be called under RCU read lock */
4194 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4195 for (iter = &(dev)->adj_list.upper, \
4196 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4198 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4200 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4201 int (*fn)(struct net_device *upper_dev,
4205 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4206 struct net_device *upper_dev);
4208 bool netdev_has_any_upper_dev(struct net_device *dev);
4210 void *netdev_lower_get_next_private(struct net_device *dev,
4211 struct list_head **iter);
4212 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4213 struct list_head **iter);
4215 #define netdev_for_each_lower_private(dev, priv, iter) \
4216 for (iter = (dev)->adj_list.lower.next, \
4217 priv = netdev_lower_get_next_private(dev, &(iter)); \
4219 priv = netdev_lower_get_next_private(dev, &(iter)))
4221 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4222 for (iter = &(dev)->adj_list.lower, \
4223 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4225 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4227 void *netdev_lower_get_next(struct net_device *dev,
4228 struct list_head **iter);
4230 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4231 for (iter = (dev)->adj_list.lower.next, \
4232 ldev = netdev_lower_get_next(dev, &(iter)); \
4234 ldev = netdev_lower_get_next(dev, &(iter)))
4236 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4237 struct list_head **iter);
4238 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4239 struct list_head **iter);
4241 int netdev_walk_all_lower_dev(struct net_device *dev,
4242 int (*fn)(struct net_device *lower_dev,
4245 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4246 int (*fn)(struct net_device *lower_dev,
4250 void *netdev_adjacent_get_private(struct list_head *adj_list);
4251 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4252 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4253 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4254 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4255 struct netlink_ext_ack *extack);
4256 int netdev_master_upper_dev_link(struct net_device *dev,
4257 struct net_device *upper_dev,
4258 void *upper_priv, void *upper_info,
4259 struct netlink_ext_ack *extack);
4260 void netdev_upper_dev_unlink(struct net_device *dev,
4261 struct net_device *upper_dev);
4262 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4263 void *netdev_lower_dev_get_private(struct net_device *dev,
4264 struct net_device *lower_dev);
4265 void netdev_lower_state_changed(struct net_device *lower_dev,
4266 void *lower_state_info);
4268 /* RSS keys are 40 or 52 bytes long */
4269 #define NETDEV_RSS_KEY_LEN 52
4270 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4271 void netdev_rss_key_fill(void *buffer, size_t len);
4273 int dev_get_nest_level(struct net_device *dev);
4274 int skb_checksum_help(struct sk_buff *skb);
4275 int skb_crc32c_csum_help(struct sk_buff *skb);
4276 int skb_csum_hwoffload_help(struct sk_buff *skb,
4277 const netdev_features_t features);
4279 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4280 netdev_features_t features, bool tx_path);
4281 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4282 netdev_features_t features);
4284 struct netdev_bonding_info {
4289 struct netdev_notifier_bonding_info {
4290 struct netdev_notifier_info info; /* must be first */
4291 struct netdev_bonding_info bonding_info;
4294 void netdev_bonding_info_change(struct net_device *dev,
4295 struct netdev_bonding_info *bonding_info);
4298 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4300 return __skb_gso_segment(skb, features, true);
4302 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4304 static inline bool can_checksum_protocol(netdev_features_t features,
4307 if (protocol == htons(ETH_P_FCOE))
4308 return !!(features & NETIF_F_FCOE_CRC);
4310 /* Assume this is an IP checksum (not SCTP CRC) */
4312 if (features & NETIF_F_HW_CSUM) {
4313 /* Can checksum everything */
4318 case htons(ETH_P_IP):
4319 return !!(features & NETIF_F_IP_CSUM);
4320 case htons(ETH_P_IPV6):
4321 return !!(features & NETIF_F_IPV6_CSUM);
4328 void netdev_rx_csum_fault(struct net_device *dev);
4330 static inline void netdev_rx_csum_fault(struct net_device *dev)
4334 /* rx skb timestamps */
4335 void net_enable_timestamp(void);
4336 void net_disable_timestamp(void);
4338 #ifdef CONFIG_PROC_FS
4339 int __init dev_proc_init(void);
4341 #define dev_proc_init() 0
4344 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4345 struct sk_buff *skb, struct net_device *dev,
4348 skb->xmit_more = more ? 1 : 0;
4349 return ops->ndo_start_xmit(skb, dev);
4352 static inline bool netdev_xmit_more(void)
4354 return __this_cpu_read(softnet_data.xmit.more);
4357 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4358 struct netdev_queue *txq, bool more)
4360 const struct net_device_ops *ops = dev->netdev_ops;
4363 rc = __netdev_start_xmit(ops, skb, dev, more);
4364 if (rc == NETDEV_TX_OK)
4365 txq_trans_update(txq);
4370 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4372 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4375 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4377 return netdev_class_create_file_ns(class_attr, NULL);
4380 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4382 netdev_class_remove_file_ns(class_attr, NULL);
4385 extern const struct kobj_ns_type_operations net_ns_type_operations;
4387 const char *netdev_drivername(const struct net_device *dev);
4389 void linkwatch_run_queue(void);
4391 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4392 netdev_features_t f2)
4394 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4395 if (f1 & NETIF_F_HW_CSUM)
4396 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4398 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4404 static inline netdev_features_t netdev_get_wanted_features(
4405 struct net_device *dev)
4407 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4409 netdev_features_t netdev_increment_features(netdev_features_t all,
4410 netdev_features_t one, netdev_features_t mask);
4412 /* Allow TSO being used on stacked device :
4413 * Performing the GSO segmentation before last device
4414 * is a performance improvement.
4416 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4417 netdev_features_t mask)
4419 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4422 int __netdev_update_features(struct net_device *dev);
4423 void netdev_update_features(struct net_device *dev);
4424 void netdev_change_features(struct net_device *dev);
4426 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4427 struct net_device *dev);
4429 netdev_features_t passthru_features_check(struct sk_buff *skb,
4430 struct net_device *dev,
4431 netdev_features_t features);
4432 netdev_features_t netif_skb_features(struct sk_buff *skb);
4434 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4436 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4438 /* check flags correspondence */
4439 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4440 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4441 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4442 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4443 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4444 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4445 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4446 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4447 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4448 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4449 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4450 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4451 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4452 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4453 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4454 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4455 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4456 BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4458 return (features & feature) == feature;
4461 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4463 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4464 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4467 static inline bool netif_needs_gso(struct sk_buff *skb,
4468 netdev_features_t features)
4470 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4471 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4472 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4475 static inline void netif_set_gso_max_size(struct net_device *dev,
4478 dev->gso_max_size = size;
4481 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4482 int pulled_hlen, u16 mac_offset,
4485 skb->protocol = protocol;
4486 skb->encapsulation = 1;
4487 skb_push(skb, pulled_hlen);
4488 skb_reset_transport_header(skb);
4489 skb->mac_header = mac_offset;
4490 skb->network_header = skb->mac_header + mac_len;
4491 skb->mac_len = mac_len;
4494 static inline bool netif_is_macsec(const struct net_device *dev)
4496 return dev->priv_flags & IFF_MACSEC;
4499 static inline bool netif_is_macvlan(const struct net_device *dev)
4501 return dev->priv_flags & IFF_MACVLAN;
4504 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4506 return dev->priv_flags & IFF_MACVLAN_PORT;
4509 static inline bool netif_is_bond_master(const struct net_device *dev)
4511 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4514 static inline bool netif_is_bond_slave(const struct net_device *dev)
4516 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4519 static inline bool netif_supports_nofcs(struct net_device *dev)
4521 return dev->priv_flags & IFF_SUPP_NOFCS;
4524 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4526 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4529 static inline bool netif_is_l3_master(const struct net_device *dev)
4531 return dev->priv_flags & IFF_L3MDEV_MASTER;
4534 static inline bool netif_is_l3_slave(const struct net_device *dev)
4536 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4539 static inline bool netif_is_bridge_master(const struct net_device *dev)
4541 return dev->priv_flags & IFF_EBRIDGE;
4544 static inline bool netif_is_bridge_port(const struct net_device *dev)
4546 return dev->priv_flags & IFF_BRIDGE_PORT;
4549 static inline bool netif_is_ovs_master(const struct net_device *dev)
4551 return dev->priv_flags & IFF_OPENVSWITCH;
4554 static inline bool netif_is_ovs_port(const struct net_device *dev)
4556 return dev->priv_flags & IFF_OVS_DATAPATH;
4559 static inline bool netif_is_team_master(const struct net_device *dev)
4561 return dev->priv_flags & IFF_TEAM;
4564 static inline bool netif_is_team_port(const struct net_device *dev)
4566 return dev->priv_flags & IFF_TEAM_PORT;
4569 static inline bool netif_is_lag_master(const struct net_device *dev)
4571 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4574 static inline bool netif_is_lag_port(const struct net_device *dev)
4576 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4579 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4581 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4584 static inline bool netif_is_failover(const struct net_device *dev)
4586 return dev->priv_flags & IFF_FAILOVER;
4589 static inline bool netif_is_failover_slave(const struct net_device *dev)
4591 return dev->priv_flags & IFF_FAILOVER_SLAVE;
4594 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4595 static inline void netif_keep_dst(struct net_device *dev)
4597 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4600 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4601 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4603 /* TODO: reserve and use an additional IFF bit, if we get more users */
4604 return dev->priv_flags & IFF_MACSEC;
4607 extern struct pernet_operations __net_initdata loopback_net_ops;
4609 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4611 /* netdev_printk helpers, similar to dev_printk */
4613 static inline const char *netdev_name(const struct net_device *dev)
4615 if (!dev->name[0] || strchr(dev->name, '%'))
4616 return "(unnamed net_device)";
4620 static inline bool netdev_unregistering(const struct net_device *dev)
4622 return dev->reg_state == NETREG_UNREGISTERING;
4625 static inline const char *netdev_reg_state(const struct net_device *dev)
4627 switch (dev->reg_state) {
4628 case NETREG_UNINITIALIZED: return " (uninitialized)";
4629 case NETREG_REGISTERED: return "";
4630 case NETREG_UNREGISTERING: return " (unregistering)";
4631 case NETREG_UNREGISTERED: return " (unregistered)";
4632 case NETREG_RELEASED: return " (released)";
4633 case NETREG_DUMMY: return " (dummy)";
4636 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4637 return " (unknown)";
4641 void netdev_printk(const char *level, const struct net_device *dev,
4642 const char *format, ...);
4644 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4646 void netdev_alert(const struct net_device *dev, const char *format, ...);
4648 void netdev_crit(const struct net_device *dev, const char *format, ...);
4650 void netdev_err(const struct net_device *dev, const char *format, ...);
4652 void netdev_warn(const struct net_device *dev, const char *format, ...);
4654 void netdev_notice(const struct net_device *dev, const char *format, ...);
4656 void netdev_info(const struct net_device *dev, const char *format, ...);
4658 #define netdev_level_once(level, dev, fmt, ...) \
4660 static bool __print_once __read_mostly; \
4662 if (!__print_once) { \
4663 __print_once = true; \
4664 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4668 #define netdev_emerg_once(dev, fmt, ...) \
4669 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4670 #define netdev_alert_once(dev, fmt, ...) \
4671 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4672 #define netdev_crit_once(dev, fmt, ...) \
4673 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4674 #define netdev_err_once(dev, fmt, ...) \
4675 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4676 #define netdev_warn_once(dev, fmt, ...) \
4677 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4678 #define netdev_notice_once(dev, fmt, ...) \
4679 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4680 #define netdev_info_once(dev, fmt, ...) \
4681 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4683 #define MODULE_ALIAS_NETDEV(device) \
4684 MODULE_ALIAS("netdev-" device)
4686 #if defined(CONFIG_DYNAMIC_DEBUG)
4687 #define netdev_dbg(__dev, format, args...) \
4689 dynamic_netdev_dbg(__dev, format, ##args); \
4691 #elif defined(DEBUG)
4692 #define netdev_dbg(__dev, format, args...) \
4693 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4695 #define netdev_dbg(__dev, format, args...) \
4698 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4702 #if defined(VERBOSE_DEBUG)
4703 #define netdev_vdbg netdev_dbg
4706 #define netdev_vdbg(dev, format, args...) \
4709 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4715 * netdev_WARN() acts like dev_printk(), but with the key difference
4716 * of using a WARN/WARN_ON to get the message out, including the
4717 * file/line information and a backtrace.
4719 #define netdev_WARN(dev, format, args...) \
4720 WARN(1, "netdevice: %s%s: " format, netdev_name(dev), \
4721 netdev_reg_state(dev), ##args)
4723 #define netdev_WARN_ONCE(dev, format, args...) \
4724 WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev), \
4725 netdev_reg_state(dev), ##args)
4727 /* netif printk helpers, similar to netdev_printk */
4729 #define netif_printk(priv, type, level, dev, fmt, args...) \
4731 if (netif_msg_##type(priv)) \
4732 netdev_printk(level, (dev), fmt, ##args); \
4735 #define netif_level(level, priv, type, dev, fmt, args...) \
4737 if (netif_msg_##type(priv)) \
4738 netdev_##level(dev, fmt, ##args); \
4741 #define netif_emerg(priv, type, dev, fmt, args...) \
4742 netif_level(emerg, priv, type, dev, fmt, ##args)
4743 #define netif_alert(priv, type, dev, fmt, args...) \
4744 netif_level(alert, priv, type, dev, fmt, ##args)
4745 #define netif_crit(priv, type, dev, fmt, args...) \
4746 netif_level(crit, priv, type, dev, fmt, ##args)
4747 #define netif_err(priv, type, dev, fmt, args...) \
4748 netif_level(err, priv, type, dev, fmt, ##args)
4749 #define netif_warn(priv, type, dev, fmt, args...) \
4750 netif_level(warn, priv, type, dev, fmt, ##args)
4751 #define netif_notice(priv, type, dev, fmt, args...) \
4752 netif_level(notice, priv, type, dev, fmt, ##args)
4753 #define netif_info(priv, type, dev, fmt, args...) \
4754 netif_level(info, priv, type, dev, fmt, ##args)
4756 #if defined(CONFIG_DYNAMIC_DEBUG)
4757 #define netif_dbg(priv, type, netdev, format, args...) \
4759 if (netif_msg_##type(priv)) \
4760 dynamic_netdev_dbg(netdev, format, ##args); \
4762 #elif defined(DEBUG)
4763 #define netif_dbg(priv, type, dev, format, args...) \
4764 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4766 #define netif_dbg(priv, type, dev, format, args...) \
4769 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4774 /* if @cond then downgrade to debug, else print at @level */
4775 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4778 netif_dbg(priv, type, netdev, fmt, ##args); \
4780 netif_ ## level(priv, type, netdev, fmt, ##args); \
4783 #if defined(VERBOSE_DEBUG)
4784 #define netif_vdbg netif_dbg
4786 #define netif_vdbg(priv, type, dev, format, args...) \
4789 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4795 * The list of packet types we will receive (as opposed to discard)
4796 * and the routines to invoke.
4798 * Why 16. Because with 16 the only overlap we get on a hash of the
4799 * low nibble of the protocol value is RARP/SNAP/X.25.
4813 #define PTYPE_HASH_SIZE (16)
4814 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4816 #endif /* _LINUX_NETDEVICE_H */