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>
48 #include <linux/netdev_features.h>
49 #include <linux/neighbour.h>
50 #include <uapi/linux/netdevice.h>
51 #include <uapi/linux/if_bonding.h>
52 #include <uapi/linux/pkt_cls.h>
53 #include <linux/hashtable.h>
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info
;
70 void netdev_set_default_ethtool_ops(struct net_device
*dev
,
71 const struct ethtool_ops
*ops
);
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
110 __NETDEV_TX_MIN
= INT_MIN
, /* make sure enum is signed */
111 NETDEV_TX_OK
= 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY
= 0x10, /* driver tx path was busy*/
114 typedef enum netdev_tx netdev_tx_t
;
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 static inline bool dev_xmit_complete(int rc
)
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 if (likely(rc
< NET_XMIT_MASK
))
135 * Compute the worst-case header length according to the protocols
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats
{
164 unsigned long rx_packets
;
165 unsigned long tx_packets
;
166 unsigned long rx_bytes
;
167 unsigned long tx_bytes
;
168 unsigned long rx_errors
;
169 unsigned long tx_errors
;
170 unsigned long rx_dropped
;
171 unsigned long tx_dropped
;
172 unsigned long multicast
;
173 unsigned long collisions
;
174 unsigned long rx_length_errors
;
175 unsigned long rx_over_errors
;
176 unsigned long rx_crc_errors
;
177 unsigned long rx_frame_errors
;
178 unsigned long rx_fifo_errors
;
179 unsigned long rx_missed_errors
;
180 unsigned long tx_aborted_errors
;
181 unsigned long tx_carrier_errors
;
182 unsigned long tx_fifo_errors
;
183 unsigned long tx_heartbeat_errors
;
184 unsigned long tx_window_errors
;
185 unsigned long rx_compressed
;
186 unsigned long tx_compressed
;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed
;
196 extern struct static_key rfs_needed
;
203 struct netdev_hw_addr
{
204 struct list_head list
;
205 unsigned char addr
[MAX_ADDR_LEN
];
207 #define NETDEV_HW_ADDR_T_LAN 1
208 #define NETDEV_HW_ADDR_T_SAN 2
209 #define NETDEV_HW_ADDR_T_SLAVE 3
210 #define NETDEV_HW_ADDR_T_UNICAST 4
211 #define NETDEV_HW_ADDR_T_MULTICAST 5
216 struct rcu_head rcu_head
;
219 struct netdev_hw_addr_list
{
220 struct list_head list
;
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data
[HH_DATA_ALIGN(LL_MAX_HEADER
) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create
) (struct sk_buff
*skb
, struct net_device
*dev
,
267 unsigned short type
, const void *daddr
,
268 const void *saddr
, unsigned int len
);
269 int (*parse
)(const struct sk_buff
*skb
, unsigned char *haddr
);
270 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
271 void (*cache_update
)(struct hh_cache
*hh
,
272 const struct net_device
*dev
,
273 const unsigned char *haddr
);
274 bool (*validate
)(const char *ll_header
, unsigned int len
);
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
282 enum netdev_state_t
{
284 __LINK_STATE_PRESENT
,
285 __LINK_STATE_NOCARRIER
,
286 __LINK_STATE_LINKWATCH_PENDING
,
287 __LINK_STATE_DORMANT
,
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup
{
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init
netdev_boot_setup(char *str
);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list
;
317 unsigned int gro_count
;
318 int (*poll
)(struct napi_struct
*, int);
319 #ifdef CONFIG_NETPOLL
322 struct net_device
*dev
;
323 struct sk_buff
*gro_list
;
325 struct hrtimer timer
;
326 struct list_head dev_list
;
327 struct hlist_node napi_hash_node
;
328 unsigned int napi_id
;
332 NAPI_STATE_SCHED
, /* Poll is scheduled */
333 NAPI_STATE_MISSED
, /* reschedule a napi */
334 NAPI_STATE_DISABLE
, /* Disable pending */
335 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
336 NAPI_STATE_HASHED
, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL
,/* Do not add in napi_hash, no busy polling */
338 NAPI_STATE_IN_BUSY_POLL
,/* sk_busy_loop() owns this NAPI */
342 NAPIF_STATE_SCHED
= BIT(NAPI_STATE_SCHED
),
343 NAPIF_STATE_MISSED
= BIT(NAPI_STATE_MISSED
),
344 NAPIF_STATE_DISABLE
= BIT(NAPI_STATE_DISABLE
),
345 NAPIF_STATE_NPSVC
= BIT(NAPI_STATE_NPSVC
),
346 NAPIF_STATE_HASHED
= BIT(NAPI_STATE_HASHED
),
347 NAPIF_STATE_NO_BUSY_POLL
= BIT(NAPI_STATE_NO_BUSY_POLL
),
348 NAPIF_STATE_IN_BUSY_POLL
= BIT(NAPI_STATE_IN_BUSY_POLL
),
359 typedef enum gro_result gro_result_t
;
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
383 * If the rx_handler consumed the skb in some way, it should return
384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385 * the skb to be delivered in some other way.
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
391 * If the rx_handler decides the skb should be ignored, it should return
392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393 * are registered on exact device (ptype->dev == skb->dev).
395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396 * delivered, it should return RX_HANDLER_PASS.
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
402 enum rx_handler_result
{
408 typedef enum rx_handler_result rx_handler_result_t
;
409 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
411 void __napi_schedule(struct napi_struct
*n
);
412 void __napi_schedule_irqoff(struct napi_struct
*n
);
414 static inline bool napi_disable_pending(struct napi_struct
*n
)
416 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
419 bool napi_schedule_prep(struct napi_struct
*n
);
422 * napi_schedule - schedule NAPI poll
425 * Schedule NAPI poll routine to be called if it is not already
428 static inline void napi_schedule(struct napi_struct
*n
)
430 if (napi_schedule_prep(n
))
435 * napi_schedule_irqoff - schedule NAPI poll
438 * Variant of napi_schedule(), assuming hard irqs are masked.
440 static inline void napi_schedule_irqoff(struct napi_struct
*n
)
442 if (napi_schedule_prep(n
))
443 __napi_schedule_irqoff(n
);
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct
*napi
)
449 if (napi_schedule_prep(napi
)) {
450 __napi_schedule(napi
);
456 bool napi_complete_done(struct napi_struct
*n
, int work_done
);
458 * napi_complete - NAPI processing complete
461 * Mark NAPI processing as complete.
462 * Consider using napi_complete_done() instead.
463 * Return false if device should avoid rearming interrupts.
465 static inline bool napi_complete(struct napi_struct
*n
)
467 return napi_complete_done(n
, 0);
471 * napi_hash_del - remove a NAPI from global table
472 * @napi: NAPI context
474 * Warning: caller must observe RCU grace period
475 * before freeing memory containing @napi, if
476 * this function returns true.
477 * Note: core networking stack automatically calls it
478 * from netif_napi_del().
479 * Drivers might want to call this helper to combine all
480 * the needed RCU grace periods into a single one.
482 bool napi_hash_del(struct napi_struct
*napi
);
485 * napi_disable - prevent NAPI from scheduling
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
491 void napi_disable(struct napi_struct
*n
);
494 * napi_enable - enable NAPI scheduling
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct
*n
)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
503 smp_mb__before_atomic();
504 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
505 clear_bit(NAPI_STATE_NPSVC
, &n
->state
);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct
*n
)
518 if (IS_ENABLED(CONFIG_SMP
))
519 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
525 enum netdev_queue_state_t
{
526 __QUEUE_STATE_DRV_XOFF
,
527 __QUEUE_STATE_STACK_XOFF
,
528 __QUEUE_STATE_FROZEN
,
531 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
551 struct netdev_queue
{
555 struct net_device
*dev
;
556 struct Qdisc __rcu
*qdisc
;
557 struct Qdisc
*qdisc_sleeping
;
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
564 unsigned long tx_maxrate
;
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
569 unsigned long trans_timeout
;
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
576 * Time (in jiffies) of last Tx
578 unsigned long trans_start
;
585 } ____cacheline_aligned_in_smp
;
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue
*q
)
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 static inline void netdev_queue_numa_node_write(struct netdev_queue
*q
, int node
)
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
620 struct rps_dev_flow
{
623 unsigned int last_qtail
;
625 #define RPS_NO_FILTER 0xffff
628 * The rps_dev_flow_table structure contains a table of flow mappings.
630 struct rps_dev_flow_table
{
633 struct rps_dev_flow flows
[0];
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636 ((_num) * sizeof(struct rps_dev_flow)))
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
643 * rps_cpu_mask is used to partition the space, depending on number of
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646 * meaning we use 32-6=26 bits for the hash.
648 struct rps_sock_flow_table
{
651 u32 ents
[0] ____cacheline_aligned_in_smp
;
653 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655 #define RPS_NO_CPU 0xffff
657 extern u32 rps_cpu_mask
;
658 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
664 unsigned int index
= hash
& table
->mask
;
665 u32 val
= hash
& ~rps_cpu_mask
;
667 /* We only give a hint, preemption can change CPU under us */
668 val
|= raw_smp_processor_id();
670 if (table
->ents
[index
] != val
)
671 table
->ents
[index
] = val
;
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
679 #endif /* CONFIG_RPS */
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue
{
684 struct rps_map __rcu
*rps_map
;
685 struct rps_dev_flow_table __rcu
*rps_flow_table
;
688 struct net_device
*dev
;
689 } ____cacheline_aligned_in_smp
;
692 * RX queue sysfs structures and functions.
694 struct rx_queue_attribute
{
695 struct attribute attr
;
696 ssize_t (*show
)(struct netdev_rx_queue
*queue
, char *buf
);
697 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
698 const char *buf
, size_t len
);
703 * This structure holds an XPS map which can be of variable length. The
704 * map is an array of queues.
708 unsigned int alloc_len
;
712 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
713 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
714 - sizeof(struct xps_map)) / sizeof(u16))
717 * This structure holds all XPS maps for device. Maps are indexed by CPU.
719 struct xps_dev_maps
{
721 struct xps_map __rcu
*cpu_map
[0];
723 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
724 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
725 #endif /* CONFIG_XPS */
727 #define TC_MAX_QUEUE 16
728 #define TC_BITMASK 15
729 /* HW offloaded queuing disciplines txq count and offset maps */
730 struct netdev_tc_txq
{
735 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
737 * This structure is to hold information about the device
738 * configured to run FCoE protocol stack.
740 struct netdev_fcoe_hbainfo
{
741 char manufacturer
[64];
742 char serial_number
[64];
743 char hardware_version
[64];
744 char driver_version
[64];
745 char optionrom_version
[64];
746 char firmware_version
[64];
748 char model_description
[256];
752 #define MAX_PHYS_ITEM_ID_LEN 32
754 /* This structure holds a unique identifier to identify some
755 * physical item (port for example) used by a netdevice.
757 struct netdev_phys_item_id
{
758 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
759 unsigned char id_len
;
762 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id
*a
,
763 struct netdev_phys_item_id
*b
)
765 return a
->id_len
== b
->id_len
&&
766 memcmp(a
->id
, b
->id
, a
->id_len
) == 0;
769 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
770 struct sk_buff
*skb
);
773 TC_SETUP_QDISC_MQPRIO
,
776 TC_SETUP_CLSMATCHALL
,
783 /* These structures hold the attributes of bpf state that are being passed
784 * to the netdevice through the bpf op.
786 enum bpf_netdev_command
{
787 /* Set or clear a bpf program used in the earliest stages of packet
788 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
789 * is responsible for calling bpf_prog_put on any old progs that are
790 * stored. In case of error, the callee need not release the new prog
791 * reference, but on success it takes ownership and must bpf_prog_put
792 * when it is no longer used.
796 /* Check if a bpf program is set on the device. The callee should
797 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
798 * is equivalent to XDP_ATTACHED_DRV.
801 /* BPF program for offload callbacks, invoked at program load time. */
802 BPF_OFFLOAD_VERIFIER_PREP
,
803 BPF_OFFLOAD_TRANSLATE
,
807 struct bpf_ext_analyzer_ops
;
808 struct netlink_ext_ack
;
811 enum bpf_netdev_command command
;
816 struct bpf_prog
*prog
;
817 struct netlink_ext_ack
*extack
;
824 /* BPF_OFFLOAD_VERIFIER_PREP */
826 struct bpf_prog
*prog
;
827 const struct bpf_ext_analyzer_ops
*ops
; /* callee set */
829 /* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
831 struct bpf_prog
*prog
;
836 #ifdef CONFIG_XFRM_OFFLOAD
838 int (*xdo_dev_state_add
) (struct xfrm_state
*x
);
839 void (*xdo_dev_state_delete
) (struct xfrm_state
*x
);
840 void (*xdo_dev_state_free
) (struct xfrm_state
*x
);
841 bool (*xdo_dev_offload_ok
) (struct sk_buff
*skb
,
842 struct xfrm_state
*x
);
847 struct rcu_head rcuhead
;
852 * This structure defines the management hooks for network devices.
853 * The following hooks can be defined; unless noted otherwise, they are
854 * optional and can be filled with a null pointer.
856 * int (*ndo_init)(struct net_device *dev);
857 * This function is called once when a network device is registered.
858 * The network device can use this for any late stage initialization
859 * or semantic validation. It can fail with an error code which will
860 * be propagated back to register_netdev.
862 * void (*ndo_uninit)(struct net_device *dev);
863 * This function is called when device is unregistered or when registration
864 * fails. It is not called if init fails.
866 * int (*ndo_open)(struct net_device *dev);
867 * This function is called when a network device transitions to the up
870 * int (*ndo_stop)(struct net_device *dev);
871 * This function is called when a network device transitions to the down
874 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
875 * struct net_device *dev);
876 * Called when a packet needs to be transmitted.
877 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
878 * the queue before that can happen; it's for obsolete devices and weird
879 * corner cases, but the stack really does a non-trivial amount
880 * of useless work if you return NETDEV_TX_BUSY.
881 * Required; cannot be NULL.
883 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
884 * struct net_device *dev
885 * netdev_features_t features);
886 * Called by core transmit path to determine if device is capable of
887 * performing offload operations on a given packet. This is to give
888 * the device an opportunity to implement any restrictions that cannot
889 * be otherwise expressed by feature flags. The check is called with
890 * the set of features that the stack has calculated and it returns
891 * those the driver believes to be appropriate.
893 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
894 * void *accel_priv, select_queue_fallback_t fallback);
895 * Called to decide which queue to use when device supports multiple
898 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
899 * This function is called to allow device receiver to make
900 * changes to configuration when multicast or promiscuous is enabled.
902 * void (*ndo_set_rx_mode)(struct net_device *dev);
903 * This function is called device changes address list filtering.
904 * If driver handles unicast address filtering, it should set
905 * IFF_UNICAST_FLT in its priv_flags.
907 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
908 * This function is called when the Media Access Control address
909 * needs to be changed. If this interface is not defined, the
910 * MAC address can not be changed.
912 * int (*ndo_validate_addr)(struct net_device *dev);
913 * Test if Media Access Control address is valid for the device.
915 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
916 * Called when a user requests an ioctl which can't be handled by
917 * the generic interface code. If not defined ioctls return
918 * not supported error code.
920 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
921 * Used to set network devices bus interface parameters. This interface
922 * is retained for legacy reasons; new devices should use the bus
923 * interface (PCI) for low level management.
925 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
926 * Called when a user wants to change the Maximum Transfer Unit
929 * void (*ndo_tx_timeout)(struct net_device *dev);
930 * Callback used when the transmitter has not made any progress
931 * for dev->watchdog ticks.
933 * void (*ndo_get_stats64)(struct net_device *dev,
934 * struct rtnl_link_stats64 *storage);
935 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
936 * Called when a user wants to get the network device usage
937 * statistics. Drivers must do one of the following:
938 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
939 * rtnl_link_stats64 structure passed by the caller.
940 * 2. Define @ndo_get_stats to update a net_device_stats structure
941 * (which should normally be dev->stats) and return a pointer to
942 * it. The structure may be changed asynchronously only if each
943 * field is written atomically.
944 * 3. Update dev->stats asynchronously and atomically, and define
947 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
948 * Return true if this device supports offload stats of this attr_id.
950 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
952 * Get statistics for offload operations by attr_id. Write it into the
955 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
956 * If device supports VLAN filtering this function is called when a
957 * VLAN id is registered.
959 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
960 * If device supports VLAN filtering this function is called when a
961 * VLAN id is unregistered.
963 * void (*ndo_poll_controller)(struct net_device *dev);
965 * SR-IOV management functions.
966 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
967 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
968 * u8 qos, __be16 proto);
969 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
971 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
972 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
973 * int (*ndo_get_vf_config)(struct net_device *dev,
974 * int vf, struct ifla_vf_info *ivf);
975 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
976 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
977 * struct nlattr *port[]);
979 * Enable or disable the VF ability to query its RSS Redirection Table and
980 * Hash Key. This is needed since on some devices VF share this information
981 * with PF and querying it may introduce a theoretical security risk.
982 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
983 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
984 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
986 * Called to setup any 'tc' scheduler, classifier or action on @dev.
987 * This is always called from the stack with the rtnl lock held and netif
988 * tx queues stopped. This allows the netdevice to perform queue
991 * Fiber Channel over Ethernet (FCoE) offload functions.
992 * int (*ndo_fcoe_enable)(struct net_device *dev);
993 * Called when the FCoE protocol stack wants to start using LLD for FCoE
994 * so the underlying device can perform whatever needed configuration or
995 * initialization to support acceleration of FCoE traffic.
997 * int (*ndo_fcoe_disable)(struct net_device *dev);
998 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
999 * so the underlying device can perform whatever needed clean-ups to
1000 * stop supporting acceleration of FCoE traffic.
1002 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1003 * struct scatterlist *sgl, unsigned int sgc);
1004 * Called when the FCoE Initiator wants to initialize an I/O that
1005 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1006 * perform necessary setup and returns 1 to indicate the device is set up
1007 * successfully to perform DDP on this I/O, otherwise this returns 0.
1009 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
1010 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
1011 * indicated by the FC exchange id 'xid', so the underlying device can
1012 * clean up and reuse resources for later DDP requests.
1014 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1015 * struct scatterlist *sgl, unsigned int sgc);
1016 * Called when the FCoE Target wants to initialize an I/O that
1017 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1018 * perform necessary setup and returns 1 to indicate the device is set up
1019 * successfully to perform DDP on this I/O, otherwise this returns 0.
1021 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1022 * struct netdev_fcoe_hbainfo *hbainfo);
1023 * Called when the FCoE Protocol stack wants information on the underlying
1024 * device. This information is utilized by the FCoE protocol stack to
1025 * register attributes with Fiber Channel management service as per the
1026 * FC-GS Fabric Device Management Information(FDMI) specification.
1028 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1029 * Called when the underlying device wants to override default World Wide
1030 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1031 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1032 * protocol stack to use.
1035 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1036 * u16 rxq_index, u32 flow_id);
1037 * Set hardware filter for RFS. rxq_index is the target queue index;
1038 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1039 * Return the filter ID on success, or a negative error code.
1041 * Slave management functions (for bridge, bonding, etc).
1042 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1043 * Called to make another netdev an underling.
1045 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1046 * Called to release previously enslaved netdev.
1048 * Feature/offload setting functions.
1049 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1050 * netdev_features_t features);
1051 * Adjusts the requested feature flags according to device-specific
1052 * constraints, and returns the resulting flags. Must not modify
1055 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1056 * Called to update device configuration to new features. Passed
1057 * feature set might be less than what was returned by ndo_fix_features()).
1058 * Must return >0 or -errno if it changed dev->features itself.
1060 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1061 * struct net_device *dev,
1062 * const unsigned char *addr, u16 vid, u16 flags)
1063 * Adds an FDB entry to dev for addr.
1064 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1065 * struct net_device *dev,
1066 * const unsigned char *addr, u16 vid)
1067 * Deletes the FDB entry from dev coresponding to addr.
1068 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1069 * struct net_device *dev, struct net_device *filter_dev,
1071 * Used to add FDB entries to dump requests. Implementers should add
1072 * entries to skb and update idx with the number of entries.
1074 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1076 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1077 * struct net_device *dev, u32 filter_mask,
1079 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1082 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1083 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1084 * which do not represent real hardware may define this to allow their
1085 * userspace components to manage their virtual carrier state. Devices
1086 * that determine carrier state from physical hardware properties (eg
1087 * network cables) or protocol-dependent mechanisms (eg
1088 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1090 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1091 * struct netdev_phys_item_id *ppid);
1092 * Called to get ID of physical port of this device. If driver does
1093 * not implement this, it is assumed that the hw is not able to have
1094 * multiple net devices on single physical port.
1096 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1097 * struct udp_tunnel_info *ti);
1098 * Called by UDP tunnel to notify a driver about the UDP port and socket
1099 * address family that a UDP tunnel is listnening to. It is called only
1100 * when a new port starts listening. The operation is protected by the
1103 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1104 * struct udp_tunnel_info *ti);
1105 * Called by UDP tunnel to notify the driver about a UDP port and socket
1106 * address family that the UDP tunnel is not listening to anymore. The
1107 * operation is protected by the RTNL.
1109 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1110 * struct net_device *dev)
1111 * Called by upper layer devices to accelerate switching or other
1112 * station functionality into hardware. 'pdev is the lowerdev
1113 * to use for the offload and 'dev' is the net device that will
1114 * back the offload. Returns a pointer to the private structure
1115 * the upper layer will maintain.
1116 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1117 * Called by upper layer device to delete the station created
1118 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1119 * the station and priv is the structure returned by the add
1121 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1122 * int queue_index, u32 maxrate);
1123 * Called when a user wants to set a max-rate limitation of specific
1125 * int (*ndo_get_iflink)(const struct net_device *dev);
1126 * Called to get the iflink value of this device.
1127 * void (*ndo_change_proto_down)(struct net_device *dev,
1129 * This function is used to pass protocol port error state information
1130 * to the switch driver. The switch driver can react to the proto_down
1131 * by doing a phys down on the associated switch port.
1132 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1133 * This function is used to get egress tunnel information for given skb.
1134 * This is useful for retrieving outer tunnel header parameters while
1136 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1137 * This function is used to specify the headroom that the skb must
1138 * consider when allocation skb during packet reception. Setting
1139 * appropriate rx headroom value allows avoiding skb head copy on
1140 * forward. Setting a negative value resets the rx headroom to the
1142 * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1143 * This function is used to set or query state related to XDP on the
1144 * netdevice and manage BPF offload. See definition of
1145 * enum bpf_netdev_command for details.
1146 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1147 * This function is used to submit a XDP packet for transmit on a
1149 * void (*ndo_xdp_flush)(struct net_device *dev);
1150 * This function is used to inform the driver to flush a particular
1151 * xdp tx queue. Must be called on same CPU as xdp_xmit.
1153 struct net_device_ops
{
1154 int (*ndo_init
)(struct net_device
*dev
);
1155 void (*ndo_uninit
)(struct net_device
*dev
);
1156 int (*ndo_open
)(struct net_device
*dev
);
1157 int (*ndo_stop
)(struct net_device
*dev
);
1158 netdev_tx_t (*ndo_start_xmit
)(struct sk_buff
*skb
,
1159 struct net_device
*dev
);
1160 netdev_features_t (*ndo_features_check
)(struct sk_buff
*skb
,
1161 struct net_device
*dev
,
1162 netdev_features_t features
);
1163 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1164 struct sk_buff
*skb
,
1166 select_queue_fallback_t fallback
);
1167 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1169 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1170 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1172 int (*ndo_validate_addr
)(struct net_device
*dev
);
1173 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1174 struct ifreq
*ifr
, int cmd
);
1175 int (*ndo_set_config
)(struct net_device
*dev
,
1177 int (*ndo_change_mtu
)(struct net_device
*dev
,
1179 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1180 struct neigh_parms
*);
1181 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1183 void (*ndo_get_stats64
)(struct net_device
*dev
,
1184 struct rtnl_link_stats64
*storage
);
1185 bool (*ndo_has_offload_stats
)(const struct net_device
*dev
, int attr_id
);
1186 int (*ndo_get_offload_stats
)(int attr_id
,
1187 const struct net_device
*dev
,
1189 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1191 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1192 __be16 proto
, u16 vid
);
1193 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1194 __be16 proto
, u16 vid
);
1195 #ifdef CONFIG_NET_POLL_CONTROLLER
1196 void (*ndo_poll_controller
)(struct net_device
*dev
);
1197 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1198 struct netpoll_info
*info
);
1199 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1201 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1202 int queue
, u8
*mac
);
1203 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1204 int queue
, u16 vlan
,
1205 u8 qos
, __be16 proto
);
1206 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1207 int vf
, int min_tx_rate
,
1209 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1210 int vf
, bool setting
);
1211 int (*ndo_set_vf_trust
)(struct net_device
*dev
,
1212 int vf
, bool setting
);
1213 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1215 struct ifla_vf_info
*ivf
);
1216 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1217 int vf
, int link_state
);
1218 int (*ndo_get_vf_stats
)(struct net_device
*dev
,
1220 struct ifla_vf_stats
1222 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1224 struct nlattr
*port
[]);
1225 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1226 int vf
, struct sk_buff
*skb
);
1227 int (*ndo_set_vf_guid
)(struct net_device
*dev
,
1230 int (*ndo_set_vf_rss_query_en
)(
1231 struct net_device
*dev
,
1232 int vf
, bool setting
);
1233 int (*ndo_setup_tc
)(struct net_device
*dev
,
1234 enum tc_setup_type type
,
1236 #if IS_ENABLED(CONFIG_FCOE)
1237 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1238 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1239 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1241 struct scatterlist
*sgl
,
1243 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1245 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1247 struct scatterlist
*sgl
,
1249 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1250 struct netdev_fcoe_hbainfo
*hbainfo
);
1253 #if IS_ENABLED(CONFIG_LIBFCOE)
1254 #define NETDEV_FCOE_WWNN 0
1255 #define NETDEV_FCOE_WWPN 1
1256 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1257 u64
*wwn
, int type
);
1260 #ifdef CONFIG_RFS_ACCEL
1261 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1262 const struct sk_buff
*skb
,
1266 int (*ndo_add_slave
)(struct net_device
*dev
,
1267 struct net_device
*slave_dev
,
1268 struct netlink_ext_ack
*extack
);
1269 int (*ndo_del_slave
)(struct net_device
*dev
,
1270 struct net_device
*slave_dev
);
1271 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1272 netdev_features_t features
);
1273 int (*ndo_set_features
)(struct net_device
*dev
,
1274 netdev_features_t features
);
1275 int (*ndo_neigh_construct
)(struct net_device
*dev
,
1276 struct neighbour
*n
);
1277 void (*ndo_neigh_destroy
)(struct net_device
*dev
,
1278 struct neighbour
*n
);
1280 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1281 struct nlattr
*tb
[],
1282 struct net_device
*dev
,
1283 const unsigned char *addr
,
1286 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1287 struct nlattr
*tb
[],
1288 struct net_device
*dev
,
1289 const unsigned char *addr
,
1291 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1292 struct netlink_callback
*cb
,
1293 struct net_device
*dev
,
1294 struct net_device
*filter_dev
,
1297 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1298 struct nlmsghdr
*nlh
,
1300 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1302 struct net_device
*dev
,
1305 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1306 struct nlmsghdr
*nlh
,
1308 int (*ndo_change_carrier
)(struct net_device
*dev
,
1310 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1311 struct netdev_phys_item_id
*ppid
);
1312 int (*ndo_get_phys_port_name
)(struct net_device
*dev
,
1313 char *name
, size_t len
);
1314 void (*ndo_udp_tunnel_add
)(struct net_device
*dev
,
1315 struct udp_tunnel_info
*ti
);
1316 void (*ndo_udp_tunnel_del
)(struct net_device
*dev
,
1317 struct udp_tunnel_info
*ti
);
1318 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1319 struct net_device
*dev
);
1320 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1323 int (*ndo_get_lock_subclass
)(struct net_device
*dev
);
1324 int (*ndo_set_tx_maxrate
)(struct net_device
*dev
,
1327 int (*ndo_get_iflink
)(const struct net_device
*dev
);
1328 int (*ndo_change_proto_down
)(struct net_device
*dev
,
1330 int (*ndo_fill_metadata_dst
)(struct net_device
*dev
,
1331 struct sk_buff
*skb
);
1332 void (*ndo_set_rx_headroom
)(struct net_device
*dev
,
1333 int needed_headroom
);
1334 int (*ndo_bpf
)(struct net_device
*dev
,
1335 struct netdev_bpf
*bpf
);
1336 int (*ndo_xdp_xmit
)(struct net_device
*dev
,
1337 struct xdp_buff
*xdp
);
1338 void (*ndo_xdp_flush
)(struct net_device
*dev
);
1342 * enum net_device_priv_flags - &struct net_device priv_flags
1344 * These are the &struct net_device, they are only set internally
1345 * by drivers and used in the kernel. These flags are invisible to
1346 * userspace; this means that the order of these flags can change
1347 * during any kernel release.
1349 * You should have a pretty good reason to be extending these flags.
1351 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1352 * @IFF_EBRIDGE: Ethernet bridging device
1353 * @IFF_BONDING: bonding master or slave
1354 * @IFF_ISATAP: ISATAP interface (RFC4214)
1355 * @IFF_WAN_HDLC: WAN HDLC device
1356 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1358 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1359 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1360 * @IFF_MACVLAN_PORT: device used as macvlan port
1361 * @IFF_BRIDGE_PORT: device used as bridge port
1362 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1363 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1364 * @IFF_UNICAST_FLT: Supports unicast filtering
1365 * @IFF_TEAM_PORT: device used as team port
1366 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1367 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1368 * change when it's running
1369 * @IFF_MACVLAN: Macvlan device
1370 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1371 * underlying stacked devices
1372 * @IFF_IPVLAN_MASTER: IPvlan master device
1373 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1374 * @IFF_L3MDEV_MASTER: device is an L3 master device
1375 * @IFF_NO_QUEUE: device can run without qdisc attached
1376 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1377 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1378 * @IFF_TEAM: device is a team device
1379 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1380 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1381 * entity (i.e. the master device for bridged veth)
1382 * @IFF_MACSEC: device is a MACsec device
1383 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1385 enum netdev_priv_flags
{
1386 IFF_802_1Q_VLAN
= 1<<0,
1390 IFF_WAN_HDLC
= 1<<4,
1391 IFF_XMIT_DST_RELEASE
= 1<<5,
1392 IFF_DONT_BRIDGE
= 1<<6,
1393 IFF_DISABLE_NETPOLL
= 1<<7,
1394 IFF_MACVLAN_PORT
= 1<<8,
1395 IFF_BRIDGE_PORT
= 1<<9,
1396 IFF_OVS_DATAPATH
= 1<<10,
1397 IFF_TX_SKB_SHARING
= 1<<11,
1398 IFF_UNICAST_FLT
= 1<<12,
1399 IFF_TEAM_PORT
= 1<<13,
1400 IFF_SUPP_NOFCS
= 1<<14,
1401 IFF_LIVE_ADDR_CHANGE
= 1<<15,
1402 IFF_MACVLAN
= 1<<16,
1403 IFF_XMIT_DST_RELEASE_PERM
= 1<<17,
1404 IFF_IPVLAN_MASTER
= 1<<18,
1405 IFF_IPVLAN_SLAVE
= 1<<19,
1406 IFF_L3MDEV_MASTER
= 1<<20,
1407 IFF_NO_QUEUE
= 1<<21,
1408 IFF_OPENVSWITCH
= 1<<22,
1409 IFF_L3MDEV_SLAVE
= 1<<23,
1411 IFF_RXFH_CONFIGURED
= 1<<25,
1412 IFF_PHONY_HEADROOM
= 1<<26,
1414 IFF_L3MDEV_RX_HANDLER
= 1<<28,
1417 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1418 #define IFF_EBRIDGE IFF_EBRIDGE
1419 #define IFF_BONDING IFF_BONDING
1420 #define IFF_ISATAP IFF_ISATAP
1421 #define IFF_WAN_HDLC IFF_WAN_HDLC
1422 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1423 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1424 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1425 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1426 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1427 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1428 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1429 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1430 #define IFF_TEAM_PORT IFF_TEAM_PORT
1431 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1432 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1433 #define IFF_MACVLAN IFF_MACVLAN
1434 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1435 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1436 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1437 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1438 #define IFF_NO_QUEUE IFF_NO_QUEUE
1439 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1440 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1441 #define IFF_TEAM IFF_TEAM
1442 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1443 #define IFF_MACSEC IFF_MACSEC
1444 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1447 * struct net_device - The DEVICE structure.
1449 * Actually, this whole structure is a big mistake. It mixes I/O
1450 * data with strictly "high-level" data, and it has to know about
1451 * almost every data structure used in the INET module.
1453 * @name: This is the first field of the "visible" part of this structure
1454 * (i.e. as seen by users in the "Space.c" file). It is the name
1457 * @name_hlist: Device name hash chain, please keep it close to name[]
1458 * @ifalias: SNMP alias
1459 * @mem_end: Shared memory end
1460 * @mem_start: Shared memory start
1461 * @base_addr: Device I/O address
1462 * @irq: Device IRQ number
1464 * @carrier_changes: Stats to monitor carrier on<->off transitions
1466 * @state: Generic network queuing layer state, see netdev_state_t
1467 * @dev_list: The global list of network devices
1468 * @napi_list: List entry used for polling NAPI devices
1469 * @unreg_list: List entry when we are unregistering the
1470 * device; see the function unregister_netdev
1471 * @close_list: List entry used when we are closing the device
1472 * @ptype_all: Device-specific packet handlers for all protocols
1473 * @ptype_specific: Device-specific, protocol-specific packet handlers
1475 * @adj_list: Directly linked devices, like slaves for bonding
1476 * @features: Currently active device features
1477 * @hw_features: User-changeable features
1479 * @wanted_features: User-requested features
1480 * @vlan_features: Mask of features inheritable by VLAN devices
1482 * @hw_enc_features: Mask of features inherited by encapsulating devices
1483 * This field indicates what encapsulation
1484 * offloads the hardware is capable of doing,
1485 * and drivers will need to set them appropriately.
1487 * @mpls_features: Mask of features inheritable by MPLS
1489 * @ifindex: interface index
1490 * @group: The group the device belongs to
1492 * @stats: Statistics struct, which was left as a legacy, use
1493 * rtnl_link_stats64 instead
1495 * @rx_dropped: Dropped packets by core network,
1496 * do not use this in drivers
1497 * @tx_dropped: Dropped packets by core network,
1498 * do not use this in drivers
1499 * @rx_nohandler: nohandler dropped packets by core network on
1500 * inactive devices, do not use this in drivers
1502 * @wireless_handlers: List of functions to handle Wireless Extensions,
1504 * see <net/iw_handler.h> for details.
1505 * @wireless_data: Instance data managed by the core of wireless extensions
1507 * @netdev_ops: Includes several pointers to callbacks,
1508 * if one wants to override the ndo_*() functions
1509 * @ethtool_ops: Management operations
1510 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1511 * discovery handling. Necessary for e.g. 6LoWPAN.
1512 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1513 * of Layer 2 headers.
1515 * @flags: Interface flags (a la BSD)
1516 * @priv_flags: Like 'flags' but invisible to userspace,
1517 * see if.h for the definitions
1518 * @gflags: Global flags ( kept as legacy )
1519 * @padded: How much padding added by alloc_netdev()
1520 * @operstate: RFC2863 operstate
1521 * @link_mode: Mapping policy to operstate
1522 * @if_port: Selectable AUI, TP, ...
1524 * @mtu: Interface MTU value
1525 * @min_mtu: Interface Minimum MTU value
1526 * @max_mtu: Interface Maximum MTU value
1527 * @type: Interface hardware type
1528 * @hard_header_len: Maximum hardware header length.
1529 * @min_header_len: Minimum hardware header length
1531 * @needed_headroom: Extra headroom the hardware may need, but not in all
1532 * cases can this be guaranteed
1533 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1534 * cases can this be guaranteed. Some cases also use
1535 * LL_MAX_HEADER instead to allocate the skb
1537 * interface address info:
1539 * @perm_addr: Permanent hw address
1540 * @addr_assign_type: Hw address assignment type
1541 * @addr_len: Hardware address length
1542 * @neigh_priv_len: Used in neigh_alloc()
1543 * @dev_id: Used to differentiate devices that share
1544 * the same link layer address
1545 * @dev_port: Used to differentiate devices that share
1547 * @addr_list_lock: XXX: need comments on this one
1548 * @uc_promisc: Counter that indicates promiscuous mode
1549 * has been enabled due to the need to listen to
1550 * additional unicast addresses in a device that
1551 * does not implement ndo_set_rx_mode()
1552 * @uc: unicast mac addresses
1553 * @mc: multicast mac addresses
1554 * @dev_addrs: list of device hw addresses
1555 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1556 * @promiscuity: Number of times the NIC is told to work in
1557 * promiscuous mode; if it becomes 0 the NIC will
1558 * exit promiscuous mode
1559 * @allmulti: Counter, enables or disables allmulticast mode
1561 * @vlan_info: VLAN info
1562 * @dsa_ptr: dsa specific data
1563 * @tipc_ptr: TIPC specific data
1564 * @atalk_ptr: AppleTalk link
1565 * @ip_ptr: IPv4 specific data
1566 * @dn_ptr: DECnet specific data
1567 * @ip6_ptr: IPv6 specific data
1568 * @ax25_ptr: AX.25 specific data
1569 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1571 * @dev_addr: Hw address (before bcast,
1572 * because most packets are unicast)
1574 * @_rx: Array of RX queues
1575 * @num_rx_queues: Number of RX queues
1576 * allocated at register_netdev() time
1577 * @real_num_rx_queues: Number of RX queues currently active in device
1579 * @rx_handler: handler for received packets
1580 * @rx_handler_data: XXX: need comments on this one
1581 * @miniq_ingress: ingress/clsact qdisc specific data for
1582 * ingress processing
1583 * @ingress_queue: XXX: need comments on this one
1584 * @broadcast: hw bcast address
1586 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1587 * indexed by RX queue number. Assigned by driver.
1588 * This must only be set if the ndo_rx_flow_steer
1589 * operation is defined
1590 * @index_hlist: Device index hash chain
1592 * @_tx: Array of TX queues
1593 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1594 * @real_num_tx_queues: Number of TX queues currently active in device
1595 * @qdisc: Root qdisc from userspace point of view
1596 * @tx_queue_len: Max frames per queue allowed
1597 * @tx_global_lock: XXX: need comments on this one
1599 * @xps_maps: XXX: need comments on this one
1600 * @miniq_egress: clsact qdisc specific data for
1602 * @watchdog_timeo: Represents the timeout that is used by
1603 * the watchdog (see dev_watchdog())
1604 * @watchdog_timer: List of timers
1606 * @pcpu_refcnt: Number of references to this device
1607 * @todo_list: Delayed register/unregister
1608 * @link_watch_list: XXX: need comments on this one
1610 * @reg_state: Register/unregister state machine
1611 * @dismantle: Device is going to be freed
1612 * @rtnl_link_state: This enum represents the phases of creating
1615 * @needs_free_netdev: Should unregister perform free_netdev?
1616 * @priv_destructor: Called from unregister
1617 * @npinfo: XXX: need comments on this one
1618 * @nd_net: Network namespace this network device is inside
1620 * @ml_priv: Mid-layer private
1621 * @lstats: Loopback statistics
1622 * @tstats: Tunnel statistics
1623 * @dstats: Dummy statistics
1624 * @vstats: Virtual ethernet statistics
1629 * @dev: Class/net/name entry
1630 * @sysfs_groups: Space for optional device, statistics and wireless
1633 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1634 * @rtnl_link_ops: Rtnl_link_ops
1636 * @gso_max_size: Maximum size of generic segmentation offload
1637 * @gso_max_segs: Maximum number of segments that can be passed to the
1640 * @dcbnl_ops: Data Center Bridging netlink ops
1641 * @num_tc: Number of traffic classes in the net device
1642 * @tc_to_txq: XXX: need comments on this one
1643 * @prio_tc_map: XXX: need comments on this one
1645 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1647 * @priomap: XXX: need comments on this one
1648 * @phydev: Physical device may attach itself
1649 * for hardware timestamping
1651 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1652 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1654 * @proto_down: protocol port state information can be sent to the
1655 * switch driver and used to set the phys state of the
1658 * FIXME: cleanup struct net_device such that network protocol info
1663 char name
[IFNAMSIZ
];
1664 struct hlist_node name_hlist
;
1665 struct dev_ifalias __rcu
*ifalias
;
1667 * I/O specific fields
1668 * FIXME: Merge these and struct ifmap into one
1670 unsigned long mem_end
;
1671 unsigned long mem_start
;
1672 unsigned long base_addr
;
1676 * Some hardware also needs these fields (state,dev_list,
1677 * napi_list,unreg_list,close_list) but they are not
1678 * part of the usual set specified in Space.c.
1681 unsigned long state
;
1683 struct list_head dev_list
;
1684 struct list_head napi_list
;
1685 struct list_head unreg_list
;
1686 struct list_head close_list
;
1687 struct list_head ptype_all
;
1688 struct list_head ptype_specific
;
1691 struct list_head upper
;
1692 struct list_head lower
;
1695 netdev_features_t features
;
1696 netdev_features_t hw_features
;
1697 netdev_features_t wanted_features
;
1698 netdev_features_t vlan_features
;
1699 netdev_features_t hw_enc_features
;
1700 netdev_features_t mpls_features
;
1701 netdev_features_t gso_partial_features
;
1706 struct net_device_stats stats
;
1708 atomic_long_t rx_dropped
;
1709 atomic_long_t tx_dropped
;
1710 atomic_long_t rx_nohandler
;
1712 /* Stats to monitor link on/off, flapping */
1713 atomic_t carrier_up_count
;
1714 atomic_t carrier_down_count
;
1716 #ifdef CONFIG_WIRELESS_EXT
1717 const struct iw_handler_def
*wireless_handlers
;
1718 struct iw_public_data
*wireless_data
;
1720 const struct net_device_ops
*netdev_ops
;
1721 const struct ethtool_ops
*ethtool_ops
;
1722 #ifdef CONFIG_NET_SWITCHDEV
1723 const struct switchdev_ops
*switchdev_ops
;
1725 #ifdef CONFIG_NET_L3_MASTER_DEV
1726 const struct l3mdev_ops
*l3mdev_ops
;
1728 #if IS_ENABLED(CONFIG_IPV6)
1729 const struct ndisc_ops
*ndisc_ops
;
1733 const struct xfrmdev_ops
*xfrmdev_ops
;
1736 const struct header_ops
*header_ops
;
1739 unsigned int priv_flags
;
1741 unsigned short gflags
;
1742 unsigned short padded
;
1744 unsigned char operstate
;
1745 unsigned char link_mode
;
1747 unsigned char if_port
;
1751 unsigned int min_mtu
;
1752 unsigned int max_mtu
;
1753 unsigned short type
;
1754 unsigned short hard_header_len
;
1755 unsigned char min_header_len
;
1757 unsigned short needed_headroom
;
1758 unsigned short needed_tailroom
;
1760 /* Interface address info. */
1761 unsigned char perm_addr
[MAX_ADDR_LEN
];
1762 unsigned char addr_assign_type
;
1763 unsigned char addr_len
;
1764 unsigned short neigh_priv_len
;
1765 unsigned short dev_id
;
1766 unsigned short dev_port
;
1767 spinlock_t addr_list_lock
;
1768 unsigned char name_assign_type
;
1770 struct netdev_hw_addr_list uc
;
1771 struct netdev_hw_addr_list mc
;
1772 struct netdev_hw_addr_list dev_addrs
;
1775 struct kset
*queues_kset
;
1777 unsigned int promiscuity
;
1778 unsigned int allmulti
;
1781 /* Protocol-specific pointers */
1783 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1784 struct vlan_info __rcu
*vlan_info
;
1786 #if IS_ENABLED(CONFIG_NET_DSA)
1787 struct dsa_port
*dsa_ptr
;
1789 #if IS_ENABLED(CONFIG_TIPC)
1790 struct tipc_bearer __rcu
*tipc_ptr
;
1793 struct in_device __rcu
*ip_ptr
;
1794 struct dn_dev __rcu
*dn_ptr
;
1795 struct inet6_dev __rcu
*ip6_ptr
;
1797 struct wireless_dev
*ieee80211_ptr
;
1798 struct wpan_dev
*ieee802154_ptr
;
1799 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1800 struct mpls_dev __rcu
*mpls_ptr
;
1804 * Cache lines mostly used on receive path (including eth_type_trans())
1806 /* Interface address info used in eth_type_trans() */
1807 unsigned char *dev_addr
;
1810 struct netdev_rx_queue
*_rx
;
1812 unsigned int num_rx_queues
;
1813 unsigned int real_num_rx_queues
;
1816 struct bpf_prog __rcu
*xdp_prog
;
1817 unsigned long gro_flush_timeout
;
1818 rx_handler_func_t __rcu
*rx_handler
;
1819 void __rcu
*rx_handler_data
;
1821 #ifdef CONFIG_NET_CLS_ACT
1822 struct mini_Qdisc __rcu
*miniq_ingress
;
1824 struct netdev_queue __rcu
*ingress_queue
;
1825 #ifdef CONFIG_NETFILTER_INGRESS
1826 struct nf_hook_entries __rcu
*nf_hooks_ingress
;
1829 unsigned char broadcast
[MAX_ADDR_LEN
];
1830 #ifdef CONFIG_RFS_ACCEL
1831 struct cpu_rmap
*rx_cpu_rmap
;
1833 struct hlist_node index_hlist
;
1836 * Cache lines mostly used on transmit path
1838 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1839 unsigned int num_tx_queues
;
1840 unsigned int real_num_tx_queues
;
1841 struct Qdisc
*qdisc
;
1842 #ifdef CONFIG_NET_SCHED
1843 DECLARE_HASHTABLE (qdisc_hash
, 4);
1845 unsigned int tx_queue_len
;
1846 spinlock_t tx_global_lock
;
1850 struct xps_dev_maps __rcu
*xps_maps
;
1852 #ifdef CONFIG_NET_CLS_ACT
1853 struct mini_Qdisc __rcu
*miniq_egress
;
1856 /* These may be needed for future network-power-down code. */
1857 struct timer_list watchdog_timer
;
1859 int __percpu
*pcpu_refcnt
;
1860 struct list_head todo_list
;
1862 struct list_head link_watch_list
;
1864 enum { NETREG_UNINITIALIZED
=0,
1865 NETREG_REGISTERED
, /* completed register_netdevice */
1866 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1867 NETREG_UNREGISTERED
, /* completed unregister todo */
1868 NETREG_RELEASED
, /* called free_netdev */
1869 NETREG_DUMMY
, /* dummy device for NAPI poll */
1875 RTNL_LINK_INITIALIZED
,
1876 RTNL_LINK_INITIALIZING
,
1877 } rtnl_link_state
:16;
1879 bool needs_free_netdev
;
1880 void (*priv_destructor
)(struct net_device
*dev
);
1882 #ifdef CONFIG_NETPOLL
1883 struct netpoll_info __rcu
*npinfo
;
1886 possible_net_t nd_net
;
1888 /* mid-layer private */
1891 struct pcpu_lstats __percpu
*lstats
;
1892 struct pcpu_sw_netstats __percpu
*tstats
;
1893 struct pcpu_dstats __percpu
*dstats
;
1894 struct pcpu_vstats __percpu
*vstats
;
1897 #if IS_ENABLED(CONFIG_GARP)
1898 struct garp_port __rcu
*garp_port
;
1900 #if IS_ENABLED(CONFIG_MRP)
1901 struct mrp_port __rcu
*mrp_port
;
1905 const struct attribute_group
*sysfs_groups
[4];
1906 const struct attribute_group
*sysfs_rx_queue_group
;
1908 const struct rtnl_link_ops
*rtnl_link_ops
;
1910 /* for setting kernel sock attribute on TCP connection setup */
1911 #define GSO_MAX_SIZE 65536
1912 unsigned int gso_max_size
;
1913 #define GSO_MAX_SEGS 65535
1917 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1920 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1921 u8 prio_tc_map
[TC_BITMASK
+ 1];
1923 #if IS_ENABLED(CONFIG_FCOE)
1924 unsigned int fcoe_ddp_xid
;
1926 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1927 struct netprio_map __rcu
*priomap
;
1929 struct phy_device
*phydev
;
1930 struct lock_class_key
*qdisc_tx_busylock
;
1931 struct lock_class_key
*qdisc_running_key
;
1934 #define to_net_dev(d) container_of(d, struct net_device, dev)
1936 static inline bool netif_elide_gro(const struct net_device
*dev
)
1938 if (!(dev
->features
& NETIF_F_GRO
) || dev
->xdp_prog
)
1943 #define NETDEV_ALIGN 32
1946 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1948 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1952 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1954 if (tc
>= dev
->num_tc
)
1957 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1961 int netdev_txq_to_tc(struct net_device
*dev
, unsigned int txq
);
1962 void netdev_reset_tc(struct net_device
*dev
);
1963 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
);
1964 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
);
1967 int netdev_get_num_tc(struct net_device
*dev
)
1973 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1976 return &dev
->_tx
[index
];
1979 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1980 const struct sk_buff
*skb
)
1982 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1985 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1986 void (*f
)(struct net_device
*,
1987 struct netdev_queue
*,
1993 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1994 f(dev
, &dev
->_tx
[i
], arg
);
1997 #define netdev_lockdep_set_classes(dev) \
1999 static struct lock_class_key qdisc_tx_busylock_key; \
2000 static struct lock_class_key qdisc_running_key; \
2001 static struct lock_class_key qdisc_xmit_lock_key; \
2002 static struct lock_class_key dev_addr_list_lock_key; \
2005 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2006 (dev)->qdisc_running_key = &qdisc_running_key; \
2007 lockdep_set_class(&(dev)->addr_list_lock, \
2008 &dev_addr_list_lock_key); \
2009 for (i = 0; i < (dev)->num_tx_queues; i++) \
2010 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2011 &qdisc_xmit_lock_key); \
2014 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
2015 struct sk_buff
*skb
,
2018 /* returns the headroom that the master device needs to take in account
2019 * when forwarding to this dev
2021 static inline unsigned netdev_get_fwd_headroom(struct net_device
*dev
)
2023 return dev
->priv_flags
& IFF_PHONY_HEADROOM
? 0 : dev
->needed_headroom
;
2026 static inline void netdev_set_rx_headroom(struct net_device
*dev
, int new_hr
)
2028 if (dev
->netdev_ops
->ndo_set_rx_headroom
)
2029 dev
->netdev_ops
->ndo_set_rx_headroom(dev
, new_hr
);
2032 /* set the device rx headroom to the dev's default */
2033 static inline void netdev_reset_rx_headroom(struct net_device
*dev
)
2035 netdev_set_rx_headroom(dev
, -1);
2039 * Net namespace inlines
2042 struct net
*dev_net(const struct net_device
*dev
)
2044 return read_pnet(&dev
->nd_net
);
2048 void dev_net_set(struct net_device
*dev
, struct net
*net
)
2050 write_pnet(&dev
->nd_net
, net
);
2054 * netdev_priv - access network device private data
2055 * @dev: network device
2057 * Get network device private data
2059 static inline void *netdev_priv(const struct net_device
*dev
)
2061 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
2064 /* Set the sysfs physical device reference for the network logical device
2065 * if set prior to registration will cause a symlink during initialization.
2067 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2069 /* Set the sysfs device type for the network logical device to allow
2070 * fine-grained identification of different network device types. For
2071 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2073 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2075 /* Default NAPI poll() weight
2076 * Device drivers are strongly advised to not use bigger value
2078 #define NAPI_POLL_WEIGHT 64
2081 * netif_napi_add - initialize a NAPI context
2082 * @dev: network device
2083 * @napi: NAPI context
2084 * @poll: polling function
2085 * @weight: default weight
2087 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2088 * *any* of the other NAPI-related functions.
2090 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
2091 int (*poll
)(struct napi_struct
*, int), int weight
);
2094 * netif_tx_napi_add - initialize a NAPI context
2095 * @dev: network device
2096 * @napi: NAPI context
2097 * @poll: polling function
2098 * @weight: default weight
2100 * This variant of netif_napi_add() should be used from drivers using NAPI
2101 * to exclusively poll a TX queue.
2102 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2104 static inline void netif_tx_napi_add(struct net_device
*dev
,
2105 struct napi_struct
*napi
,
2106 int (*poll
)(struct napi_struct
*, int),
2109 set_bit(NAPI_STATE_NO_BUSY_POLL
, &napi
->state
);
2110 netif_napi_add(dev
, napi
, poll
, weight
);
2114 * netif_napi_del - remove a NAPI context
2115 * @napi: NAPI context
2117 * netif_napi_del() removes a NAPI context from the network device NAPI list
2119 void netif_napi_del(struct napi_struct
*napi
);
2121 struct napi_gro_cb
{
2122 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2125 /* Length of frag0. */
2126 unsigned int frag0_len
;
2128 /* This indicates where we are processing relative to skb->data. */
2131 /* This is non-zero if the packet cannot be merged with the new skb. */
2134 /* Save the IP ID here and check when we get to the transport layer */
2137 /* Number of segments aggregated. */
2140 /* Start offset for remote checksum offload */
2141 u16 gro_remcsum_start
;
2143 /* jiffies when first packet was created/queued */
2146 /* Used in ipv6_gro_receive() and foo-over-udp */
2149 /* This is non-zero if the packet may be of the same flow. */
2152 /* Used in tunnel GRO receive */
2155 /* GRO checksum is valid */
2158 /* Number of checksums via CHECKSUM_UNNECESSARY */
2163 #define NAPI_GRO_FREE 1
2164 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2166 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2169 /* Used in GRE, set in fou/gue_gro_receive */
2172 /* Used to determine if flush_id can be ignored */
2175 /* Number of gro_receive callbacks this packet already went through */
2176 u8 recursion_counter
:4;
2180 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2183 /* used in skb_gro_receive() slow path */
2184 struct sk_buff
*last
;
2187 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2189 #define GRO_RECURSION_LIMIT 15
2190 static inline int gro_recursion_inc_test(struct sk_buff
*skb
)
2192 return ++NAPI_GRO_CB(skb
)->recursion_counter
== GRO_RECURSION_LIMIT
;
2195 typedef struct sk_buff
**(*gro_receive_t
)(struct sk_buff
**, struct sk_buff
*);
2196 static inline struct sk_buff
**call_gro_receive(gro_receive_t cb
,
2197 struct sk_buff
**head
,
2198 struct sk_buff
*skb
)
2200 if (unlikely(gro_recursion_inc_test(skb
))) {
2201 NAPI_GRO_CB(skb
)->flush
|= 1;
2205 return cb(head
, skb
);
2208 typedef struct sk_buff
**(*gro_receive_sk_t
)(struct sock
*, struct sk_buff
**,
2210 static inline struct sk_buff
**call_gro_receive_sk(gro_receive_sk_t cb
,
2212 struct sk_buff
**head
,
2213 struct sk_buff
*skb
)
2215 if (unlikely(gro_recursion_inc_test(skb
))) {
2216 NAPI_GRO_CB(skb
)->flush
|= 1;
2220 return cb(sk
, head
, skb
);
2223 struct packet_type
{
2224 __be16 type
; /* This is really htons(ether_type). */
2225 struct net_device
*dev
; /* NULL is wildcarded here */
2226 int (*func
) (struct sk_buff
*,
2227 struct net_device
*,
2228 struct packet_type
*,
2229 struct net_device
*);
2230 bool (*id_match
)(struct packet_type
*ptype
,
2232 void *af_packet_priv
;
2233 struct list_head list
;
2236 struct offload_callbacks
{
2237 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
2238 netdev_features_t features
);
2239 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
2240 struct sk_buff
*skb
);
2241 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
2244 struct packet_offload
{
2245 __be16 type
; /* This is really htons(ether_type). */
2247 struct offload_callbacks callbacks
;
2248 struct list_head list
;
2251 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2252 struct pcpu_sw_netstats
{
2257 struct u64_stats_sync syncp
;
2260 #define __netdev_alloc_pcpu_stats(type, gfp) \
2262 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2265 for_each_possible_cpu(__cpu) { \
2266 typeof(type) *stat; \
2267 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2268 u64_stats_init(&stat->syncp); \
2274 #define netdev_alloc_pcpu_stats(type) \
2275 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2277 enum netdev_lag_tx_type
{
2278 NETDEV_LAG_TX_TYPE_UNKNOWN
,
2279 NETDEV_LAG_TX_TYPE_RANDOM
,
2280 NETDEV_LAG_TX_TYPE_BROADCAST
,
2281 NETDEV_LAG_TX_TYPE_ROUNDROBIN
,
2282 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP
,
2283 NETDEV_LAG_TX_TYPE_HASH
,
2286 struct netdev_lag_upper_info
{
2287 enum netdev_lag_tx_type tx_type
;
2290 struct netdev_lag_lower_state_info
{
2295 #include <linux/notifier.h>
2297 /* netdevice notifier chain. Please remember to update the rtnetlink
2298 * notification exclusion list in rtnetlink_event() when adding new
2301 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2302 #define NETDEV_DOWN 0x0002
2303 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2304 detected a hardware crash and restarted
2305 - we can use this eg to kick tcp sessions
2307 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2308 #define NETDEV_REGISTER 0x0005
2309 #define NETDEV_UNREGISTER 0x0006
2310 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2311 #define NETDEV_CHANGEADDR 0x0008
2312 #define NETDEV_GOING_DOWN 0x0009
2313 #define NETDEV_CHANGENAME 0x000A
2314 #define NETDEV_FEAT_CHANGE 0x000B
2315 #define NETDEV_BONDING_FAILOVER 0x000C
2316 #define NETDEV_PRE_UP 0x000D
2317 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2318 #define NETDEV_POST_TYPE_CHANGE 0x000F
2319 #define NETDEV_POST_INIT 0x0010
2320 #define NETDEV_UNREGISTER_FINAL 0x0011
2321 #define NETDEV_RELEASE 0x0012
2322 #define NETDEV_NOTIFY_PEERS 0x0013
2323 #define NETDEV_JOIN 0x0014
2324 #define NETDEV_CHANGEUPPER 0x0015
2325 #define NETDEV_RESEND_IGMP 0x0016
2326 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2327 #define NETDEV_CHANGEINFODATA 0x0018
2328 #define NETDEV_BONDING_INFO 0x0019
2329 #define NETDEV_PRECHANGEUPPER 0x001A
2330 #define NETDEV_CHANGELOWERSTATE 0x001B
2331 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2332 #define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
2333 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2335 int register_netdevice_notifier(struct notifier_block
*nb
);
2336 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2338 struct netdev_notifier_info
{
2339 struct net_device
*dev
;
2340 struct netlink_ext_ack
*extack
;
2343 struct netdev_notifier_info_ext
{
2344 struct netdev_notifier_info info
; /* must be first */
2350 struct netdev_notifier_change_info
{
2351 struct netdev_notifier_info info
; /* must be first */
2352 unsigned int flags_changed
;
2355 struct netdev_notifier_changeupper_info
{
2356 struct netdev_notifier_info info
; /* must be first */
2357 struct net_device
*upper_dev
; /* new upper dev */
2358 bool master
; /* is upper dev master */
2359 bool linking
; /* is the notification for link or unlink */
2360 void *upper_info
; /* upper dev info */
2363 struct netdev_notifier_changelowerstate_info
{
2364 struct netdev_notifier_info info
; /* must be first */
2365 void *lower_state_info
; /* is lower dev state */
2368 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2369 struct net_device
*dev
)
2372 info
->extack
= NULL
;
2375 static inline struct net_device
*
2376 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2381 static inline struct netlink_ext_ack
*
2382 netdev_notifier_info_to_extack(const struct netdev_notifier_info
*info
)
2384 return info
->extack
;
2387 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2390 extern rwlock_t dev_base_lock
; /* Device list lock */
2392 #define for_each_netdev(net, d) \
2393 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2394 #define for_each_netdev_reverse(net, d) \
2395 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2396 #define for_each_netdev_rcu(net, d) \
2397 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2398 #define for_each_netdev_safe(net, d, n) \
2399 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2400 #define for_each_netdev_continue(net, d) \
2401 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2402 #define for_each_netdev_continue_rcu(net, d) \
2403 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2404 #define for_each_netdev_in_bond_rcu(bond, slave) \
2405 for_each_netdev_rcu(&init_net, slave) \
2406 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2407 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2409 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2411 struct list_head
*lh
;
2415 lh
= dev
->dev_list
.next
;
2416 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2419 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2421 struct list_head
*lh
;
2425 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2426 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2429 static inline struct net_device
*first_net_device(struct net
*net
)
2431 return list_empty(&net
->dev_base_head
) ? NULL
:
2432 net_device_entry(net
->dev_base_head
.next
);
2435 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2437 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2439 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2442 int netdev_boot_setup_check(struct net_device
*dev
);
2443 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2444 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2445 const char *hwaddr
);
2446 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2447 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2448 void dev_add_pack(struct packet_type
*pt
);
2449 void dev_remove_pack(struct packet_type
*pt
);
2450 void __dev_remove_pack(struct packet_type
*pt
);
2451 void dev_add_offload(struct packet_offload
*po
);
2452 void dev_remove_offload(struct packet_offload
*po
);
2454 int dev_get_iflink(const struct net_device
*dev
);
2455 int dev_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
);
2456 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2457 unsigned short mask
);
2458 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2459 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2460 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2461 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2462 int dev_open(struct net_device
*dev
);
2463 void dev_close(struct net_device
*dev
);
2464 void dev_close_many(struct list_head
*head
, bool unlink
);
2465 void dev_disable_lro(struct net_device
*dev
);
2466 int dev_loopback_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*newskb
);
2467 int dev_queue_xmit(struct sk_buff
*skb
);
2468 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2469 int register_netdevice(struct net_device
*dev
);
2470 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2471 void unregister_netdevice_many(struct list_head
*head
);
2472 static inline void unregister_netdevice(struct net_device
*dev
)
2474 unregister_netdevice_queue(dev
, NULL
);
2477 int netdev_refcnt_read(const struct net_device
*dev
);
2478 void free_netdev(struct net_device
*dev
);
2479 void netdev_freemem(struct net_device
*dev
);
2480 void synchronize_net(void);
2481 int init_dummy_netdev(struct net_device
*dev
);
2483 DECLARE_PER_CPU(int, xmit_recursion
);
2484 #define XMIT_RECURSION_LIMIT 10
2486 static inline int dev_recursion_level(void)
2488 return this_cpu_read(xmit_recursion
);
2491 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2492 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2493 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2494 struct net_device
*dev_get_by_napi_id(unsigned int napi_id
);
2495 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2496 int dev_restart(struct net_device
*dev
);
2497 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2499 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2501 return NAPI_GRO_CB(skb
)->data_offset
;
2504 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2506 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2509 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2511 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2514 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2515 unsigned int offset
)
2517 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2520 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2522 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2525 static inline void skb_gro_frag0_invalidate(struct sk_buff
*skb
)
2527 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2528 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2531 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2532 unsigned int offset
)
2534 if (!pskb_may_pull(skb
, hlen
))
2537 skb_gro_frag0_invalidate(skb
);
2538 return skb
->data
+ offset
;
2541 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2543 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2544 skb_network_offset(skb
);
2547 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2548 const void *start
, unsigned int len
)
2550 if (NAPI_GRO_CB(skb
)->csum_valid
)
2551 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2552 csum_partial(start
, len
, 0));
2555 /* GRO checksum functions. These are logical equivalents of the normal
2556 * checksum functions (in skbuff.h) except that they operate on the GRO
2557 * offsets and fields in sk_buff.
2560 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2562 static inline bool skb_at_gro_remcsum_start(struct sk_buff
*skb
)
2564 return (NAPI_GRO_CB(skb
)->gro_remcsum_start
== skb_gro_offset(skb
));
2567 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2571 return ((skb
->ip_summed
!= CHECKSUM_PARTIAL
||
2572 skb_checksum_start_offset(skb
) <
2573 skb_gro_offset(skb
)) &&
2574 !skb_at_gro_remcsum_start(skb
) &&
2575 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2576 (!zero_okay
|| check
));
2579 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2582 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2583 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2586 NAPI_GRO_CB(skb
)->csum
= psum
;
2588 return __skb_gro_checksum_complete(skb
);
2591 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2593 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2594 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2595 NAPI_GRO_CB(skb
)->csum_cnt
--;
2597 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2598 * verified a new top level checksum or an encapsulated one
2599 * during GRO. This saves work if we fallback to normal path.
2601 __skb_incr_checksum_unnecessary(skb
);
2605 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2608 __sum16 __ret = 0; \
2609 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2610 __ret = __skb_gro_checksum_validate_complete(skb, \
2611 compute_pseudo(skb, proto)); \
2613 skb_gro_incr_csum_unnecessary(skb); \
2617 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2618 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2620 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2622 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2624 #define skb_gro_checksum_simple_validate(skb) \
2625 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2627 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2629 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2630 !NAPI_GRO_CB(skb
)->csum_valid
);
2633 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2634 __sum16 check
, __wsum pseudo
)
2636 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2637 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2640 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2642 if (__skb_gro_checksum_convert_check(skb)) \
2643 __skb_gro_checksum_convert(skb, check, \
2644 compute_pseudo(skb, proto)); \
2647 struct gro_remcsum
{
2652 static inline void skb_gro_remcsum_init(struct gro_remcsum
*grc
)
2658 static inline void *skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2659 unsigned int off
, size_t hdrlen
,
2660 int start
, int offset
,
2661 struct gro_remcsum
*grc
,
2665 size_t plen
= hdrlen
+ max_t(size_t, offset
+ sizeof(u16
), start
);
2667 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2670 NAPI_GRO_CB(skb
)->gro_remcsum_start
= off
+ hdrlen
+ start
;
2674 ptr
= skb_gro_header_fast(skb
, off
);
2675 if (skb_gro_header_hard(skb
, off
+ plen
)) {
2676 ptr
= skb_gro_header_slow(skb
, off
+ plen
, off
);
2681 delta
= remcsum_adjust(ptr
+ hdrlen
, NAPI_GRO_CB(skb
)->csum
,
2684 /* Adjust skb->csum since we changed the packet */
2685 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2687 grc
->offset
= off
+ hdrlen
+ offset
;
2693 static inline void skb_gro_remcsum_cleanup(struct sk_buff
*skb
,
2694 struct gro_remcsum
*grc
)
2697 size_t plen
= grc
->offset
+ sizeof(u16
);
2702 ptr
= skb_gro_header_fast(skb
, grc
->offset
);
2703 if (skb_gro_header_hard(skb
, grc
->offset
+ sizeof(u16
))) {
2704 ptr
= skb_gro_header_slow(skb
, plen
, grc
->offset
);
2709 remcsum_unadjust((__sum16
*)ptr
, grc
->delta
);
2712 #ifdef CONFIG_XFRM_OFFLOAD
2713 static inline void skb_gro_flush_final(struct sk_buff
*skb
, struct sk_buff
**pp
, int flush
)
2715 if (PTR_ERR(pp
) != -EINPROGRESS
)
2716 NAPI_GRO_CB(skb
)->flush
|= flush
;
2718 static inline void skb_gro_flush_final_remcsum(struct sk_buff
*skb
,
2719 struct sk_buff
**pp
,
2721 struct gro_remcsum
*grc
)
2723 if (PTR_ERR(pp
) != -EINPROGRESS
) {
2724 NAPI_GRO_CB(skb
)->flush
|= flush
;
2725 skb_gro_remcsum_cleanup(skb
, grc
);
2726 skb
->remcsum_offload
= 0;
2730 static inline void skb_gro_flush_final(struct sk_buff
*skb
, struct sk_buff
**pp
, int flush
)
2732 NAPI_GRO_CB(skb
)->flush
|= flush
;
2734 static inline void skb_gro_flush_final_remcsum(struct sk_buff
*skb
,
2735 struct sk_buff
**pp
,
2737 struct gro_remcsum
*grc
)
2739 NAPI_GRO_CB(skb
)->flush
|= flush
;
2740 skb_gro_remcsum_cleanup(skb
, grc
);
2741 skb
->remcsum_offload
= 0;
2745 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2746 unsigned short type
,
2747 const void *daddr
, const void *saddr
,
2750 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2753 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2756 static inline int dev_parse_header(const struct sk_buff
*skb
,
2757 unsigned char *haddr
)
2759 const struct net_device
*dev
= skb
->dev
;
2761 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2763 return dev
->header_ops
->parse(skb
, haddr
);
2766 /* ll_header must have at least hard_header_len allocated */
2767 static inline bool dev_validate_header(const struct net_device
*dev
,
2768 char *ll_header
, int len
)
2770 if (likely(len
>= dev
->hard_header_len
))
2772 if (len
< dev
->min_header_len
)
2775 if (capable(CAP_SYS_RAWIO
)) {
2776 memset(ll_header
+ len
, 0, dev
->hard_header_len
- len
);
2780 if (dev
->header_ops
&& dev
->header_ops
->validate
)
2781 return dev
->header_ops
->validate(ll_header
, len
);
2786 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2787 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2788 static inline int unregister_gifconf(unsigned int family
)
2790 return register_gifconf(family
, NULL
);
2793 #ifdef CONFIG_NET_FLOW_LIMIT
2794 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2795 struct sd_flow_limit
{
2797 unsigned int num_buckets
;
2798 unsigned int history_head
;
2799 u16 history
[FLOW_LIMIT_HISTORY
];
2803 extern int netdev_flow_limit_table_len
;
2804 #endif /* CONFIG_NET_FLOW_LIMIT */
2807 * Incoming packets are placed on per-CPU queues
2809 struct softnet_data
{
2810 struct list_head poll_list
;
2811 struct sk_buff_head process_queue
;
2814 unsigned int processed
;
2815 unsigned int time_squeeze
;
2816 unsigned int received_rps
;
2818 struct softnet_data
*rps_ipi_list
;
2820 #ifdef CONFIG_NET_FLOW_LIMIT
2821 struct sd_flow_limit __rcu
*flow_limit
;
2823 struct Qdisc
*output_queue
;
2824 struct Qdisc
**output_queue_tailp
;
2825 struct sk_buff
*completion_queue
;
2828 /* input_queue_head should be written by cpu owning this struct,
2829 * and only read by other cpus. Worth using a cache line.
2831 unsigned int input_queue_head ____cacheline_aligned_in_smp
;
2833 /* Elements below can be accessed between CPUs for RPS/RFS */
2834 call_single_data_t csd ____cacheline_aligned_in_smp
;
2835 struct softnet_data
*rps_ipi_next
;
2837 unsigned int input_queue_tail
;
2839 unsigned int dropped
;
2840 struct sk_buff_head input_pkt_queue
;
2841 struct napi_struct backlog
;
2845 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2848 sd
->input_queue_head
++;
2852 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2853 unsigned int *qtail
)
2856 *qtail
= ++sd
->input_queue_tail
;
2860 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2862 void __netif_schedule(struct Qdisc
*q
);
2863 void netif_schedule_queue(struct netdev_queue
*txq
);
2865 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2869 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2870 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2873 static __always_inline
void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2875 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2879 * netif_start_queue - allow transmit
2880 * @dev: network device
2882 * Allow upper layers to call the device hard_start_xmit routine.
2884 static inline void netif_start_queue(struct net_device
*dev
)
2886 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2889 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2893 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2894 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2895 netif_tx_start_queue(txq
);
2899 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2902 * netif_wake_queue - restart transmit
2903 * @dev: network device
2905 * Allow upper layers to call the device hard_start_xmit routine.
2906 * Used for flow control when transmit resources are available.
2908 static inline void netif_wake_queue(struct net_device
*dev
)
2910 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2913 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2917 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2918 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2919 netif_tx_wake_queue(txq
);
2923 static __always_inline
void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2925 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2929 * netif_stop_queue - stop transmitted packets
2930 * @dev: network device
2932 * Stop upper layers calling the device hard_start_xmit routine.
2933 * Used for flow control when transmit resources are unavailable.
2935 static inline void netif_stop_queue(struct net_device
*dev
)
2937 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2940 void netif_tx_stop_all_queues(struct net_device
*dev
);
2942 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2944 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2948 * netif_queue_stopped - test if transmit queue is flowblocked
2949 * @dev: network device
2951 * Test if transmit queue on device is currently unable to send.
2953 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2955 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2958 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2960 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2964 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2966 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2970 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2972 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2976 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2977 * @dev_queue: pointer to transmit queue
2979 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2980 * to give appropriate hint to the CPU.
2982 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2985 prefetchw(&dev_queue
->dql
.num_queued
);
2990 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2991 * @dev_queue: pointer to transmit queue
2993 * BQL enabled drivers might use this helper in their TX completion path,
2994 * to give appropriate hint to the CPU.
2996 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2999 prefetchw(&dev_queue
->dql
.limit
);
3003 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
3007 dql_queued(&dev_queue
->dql
, bytes
);
3009 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
3012 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
3015 * The XOFF flag must be set before checking the dql_avail below,
3016 * because in netdev_tx_completed_queue we update the dql_completed
3017 * before checking the XOFF flag.
3021 /* check again in case another CPU has just made room avail */
3022 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
3023 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
3028 * netdev_sent_queue - report the number of bytes queued to hardware
3029 * @dev: network device
3030 * @bytes: number of bytes queued to the hardware device queue
3032 * Report the number of bytes queued for sending/completion to the network
3033 * device hardware queue. @bytes should be a good approximation and should
3034 * exactly match netdev_completed_queue() @bytes
3036 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
3038 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
3041 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
3042 unsigned int pkts
, unsigned int bytes
)
3045 if (unlikely(!bytes
))
3048 dql_completed(&dev_queue
->dql
, bytes
);
3051 * Without the memory barrier there is a small possiblity that
3052 * netdev_tx_sent_queue will miss the update and cause the queue to
3053 * be stopped forever
3057 if (dql_avail(&dev_queue
->dql
) < 0)
3060 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
3061 netif_schedule_queue(dev_queue
);
3066 * netdev_completed_queue - report bytes and packets completed by device
3067 * @dev: network device
3068 * @pkts: actual number of packets sent over the medium
3069 * @bytes: actual number of bytes sent over the medium
3071 * Report the number of bytes and packets transmitted by the network device
3072 * hardware queue over the physical medium, @bytes must exactly match the
3073 * @bytes amount passed to netdev_sent_queue()
3075 static inline void netdev_completed_queue(struct net_device
*dev
,
3076 unsigned int pkts
, unsigned int bytes
)
3078 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
3081 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
3084 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
3090 * netdev_reset_queue - reset the packets and bytes count of a network device
3091 * @dev_queue: network device
3093 * Reset the bytes and packet count of a network device and clear the
3094 * software flow control OFF bit for this network device
3096 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
3098 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
3102 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3103 * @dev: network device
3104 * @queue_index: given tx queue index
3106 * Returns 0 if given tx queue index >= number of device tx queues,
3107 * otherwise returns the originally passed tx queue index.
3109 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
3111 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
3112 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3113 dev
->name
, queue_index
,
3114 dev
->real_num_tx_queues
);
3122 * netif_running - test if up
3123 * @dev: network device
3125 * Test if the device has been brought up.
3127 static inline bool netif_running(const struct net_device
*dev
)
3129 return test_bit(__LINK_STATE_START
, &dev
->state
);
3133 * Routines to manage the subqueues on a device. We only need start,
3134 * stop, and a check if it's stopped. All other device management is
3135 * done at the overall netdevice level.
3136 * Also test the device if we're multiqueue.
3140 * netif_start_subqueue - allow sending packets on subqueue
3141 * @dev: network device
3142 * @queue_index: sub queue index
3144 * Start individual transmit queue of a device with multiple transmit queues.
3146 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
3148 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3150 netif_tx_start_queue(txq
);
3154 * netif_stop_subqueue - stop sending packets on subqueue
3155 * @dev: network device
3156 * @queue_index: sub queue index
3158 * Stop individual transmit queue of a device with multiple transmit queues.
3160 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
3162 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3163 netif_tx_stop_queue(txq
);
3167 * netif_subqueue_stopped - test status of subqueue
3168 * @dev: network device
3169 * @queue_index: sub queue index
3171 * Check individual transmit queue of a device with multiple transmit queues.
3173 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
3176 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3178 return netif_tx_queue_stopped(txq
);
3181 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
3182 struct sk_buff
*skb
)
3184 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
3188 * netif_wake_subqueue - allow sending packets on subqueue
3189 * @dev: network device
3190 * @queue_index: sub queue index
3192 * Resume individual transmit queue of a device with multiple transmit queues.
3194 static inline void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
)
3196 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3198 netif_tx_wake_queue(txq
);
3202 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
3205 static inline int netif_set_xps_queue(struct net_device
*dev
,
3206 const struct cpumask
*mask
,
3213 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3214 unsigned int num_tx_queues
);
3217 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3218 * as a distribution range limit for the returned value.
3220 static inline u16
skb_tx_hash(const struct net_device
*dev
,
3221 struct sk_buff
*skb
)
3223 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
3227 * netif_is_multiqueue - test if device has multiple transmit queues
3228 * @dev: network device
3230 * Check if device has multiple transmit queues
3232 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
3234 return dev
->num_tx_queues
> 1;
3237 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
3240 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
3242 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
3250 static inline unsigned int get_netdev_rx_queue_index(
3251 struct netdev_rx_queue
*queue
)
3253 struct net_device
*dev
= queue
->dev
;
3254 int index
= queue
- dev
->_rx
;
3256 BUG_ON(index
>= dev
->num_rx_queues
);
3261 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3262 int netif_get_num_default_rss_queues(void);
3264 enum skb_free_reason
{
3265 SKB_REASON_CONSUMED
,
3269 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
3270 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
3273 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3274 * interrupt context or with hardware interrupts being disabled.
3275 * (in_irq() || irqs_disabled())
3277 * We provide four helpers that can be used in following contexts :
3279 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3280 * replacing kfree_skb(skb)
3282 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3283 * Typically used in place of consume_skb(skb) in TX completion path
3285 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3286 * replacing kfree_skb(skb)
3288 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3289 * and consumed a packet. Used in place of consume_skb(skb)
3291 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
3293 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
3296 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
3298 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
3301 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
3303 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
3306 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
3308 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
3311 void generic_xdp_tx(struct sk_buff
*skb
, struct bpf_prog
*xdp_prog
);
3312 int do_xdp_generic(struct bpf_prog
*xdp_prog
, struct sk_buff
*skb
);
3313 int netif_rx(struct sk_buff
*skb
);
3314 int netif_rx_ni(struct sk_buff
*skb
);
3315 int netif_receive_skb(struct sk_buff
*skb
);
3316 int netif_receive_skb_core(struct sk_buff
*skb
);
3317 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
3318 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
3319 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
3320 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
3321 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
3322 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
3324 static inline void napi_free_frags(struct napi_struct
*napi
)
3326 kfree_skb(napi
->skb
);
3330 bool netdev_is_rx_handler_busy(struct net_device
*dev
);
3331 int netdev_rx_handler_register(struct net_device
*dev
,
3332 rx_handler_func_t
*rx_handler
,
3333 void *rx_handler_data
);
3334 void netdev_rx_handler_unregister(struct net_device
*dev
);
3336 bool dev_valid_name(const char *name
);
3337 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
3338 int dev_ethtool(struct net
*net
, struct ifreq
*);
3339 unsigned int dev_get_flags(const struct net_device
*);
3340 int __dev_change_flags(struct net_device
*, unsigned int flags
);
3341 int dev_change_flags(struct net_device
*, unsigned int);
3342 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
3343 unsigned int gchanges
);
3344 int dev_change_name(struct net_device
*, const char *);
3345 int dev_set_alias(struct net_device
*, const char *, size_t);
3346 int dev_get_alias(const struct net_device
*, char *, size_t);
3347 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
3348 int __dev_set_mtu(struct net_device
*, int);
3349 int dev_set_mtu(struct net_device
*, int);
3350 void dev_set_group(struct net_device
*, int);
3351 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
3352 int dev_change_carrier(struct net_device
*, bool new_carrier
);
3353 int dev_get_phys_port_id(struct net_device
*dev
,
3354 struct netdev_phys_item_id
*ppid
);
3355 int dev_get_phys_port_name(struct net_device
*dev
,
3356 char *name
, size_t len
);
3357 int dev_change_proto_down(struct net_device
*dev
, bool proto_down
);
3358 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
3359 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3360 struct netdev_queue
*txq
, int *ret
);
3362 typedef int (*bpf_op_t
)(struct net_device
*dev
, struct netdev_bpf
*bpf
);
3363 int dev_change_xdp_fd(struct net_device
*dev
, struct netlink_ext_ack
*extack
,
3365 u8
__dev_xdp_attached(struct net_device
*dev
, bpf_op_t xdp_op
, u32
*prog_id
);
3367 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3368 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3369 bool is_skb_forwardable(const struct net_device
*dev
,
3370 const struct sk_buff
*skb
);
3372 static __always_inline
int ____dev_forward_skb(struct net_device
*dev
,
3373 struct sk_buff
*skb
)
3375 if (skb_orphan_frags(skb
, GFP_ATOMIC
) ||
3376 unlikely(!is_skb_forwardable(dev
, skb
))) {
3377 atomic_long_inc(&dev
->rx_dropped
);
3382 skb_scrub_packet(skb
, true);
3387 void dev_queue_xmit_nit(struct sk_buff
*skb
, struct net_device
*dev
);
3389 extern int netdev_budget
;
3390 extern unsigned int netdev_budget_usecs
;
3392 /* Called by rtnetlink.c:rtnl_unlock() */
3393 void netdev_run_todo(void);
3396 * dev_put - release reference to device
3397 * @dev: network device
3399 * Release reference to device to allow it to be freed.
3401 static inline void dev_put(struct net_device
*dev
)
3403 this_cpu_dec(*dev
->pcpu_refcnt
);
3407 * dev_hold - get reference to device
3408 * @dev: network device
3410 * Hold reference to device to keep it from being freed.
3412 static inline void dev_hold(struct net_device
*dev
)
3414 this_cpu_inc(*dev
->pcpu_refcnt
);
3417 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3418 * and _off may be called from IRQ context, but it is caller
3419 * who is responsible for serialization of these calls.
3421 * The name carrier is inappropriate, these functions should really be
3422 * called netif_lowerlayer_*() because they represent the state of any
3423 * kind of lower layer not just hardware media.
3426 void linkwatch_init_dev(struct net_device
*dev
);
3427 void linkwatch_fire_event(struct net_device
*dev
);
3428 void linkwatch_forget_dev(struct net_device
*dev
);
3431 * netif_carrier_ok - test if carrier present
3432 * @dev: network device
3434 * Check if carrier is present on device
3436 static inline bool netif_carrier_ok(const struct net_device
*dev
)
3438 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
3441 unsigned long dev_trans_start(struct net_device
*dev
);
3443 void __netdev_watchdog_up(struct net_device
*dev
);
3445 void netif_carrier_on(struct net_device
*dev
);
3447 void netif_carrier_off(struct net_device
*dev
);
3450 * netif_dormant_on - mark device as dormant.
3451 * @dev: network device
3453 * Mark device as dormant (as per RFC2863).
3455 * The dormant state indicates that the relevant interface is not
3456 * actually in a condition to pass packets (i.e., it is not 'up') but is
3457 * in a "pending" state, waiting for some external event. For "on-
3458 * demand" interfaces, this new state identifies the situation where the
3459 * interface is waiting for events to place it in the up state.
3461 static inline void netif_dormant_on(struct net_device
*dev
)
3463 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3464 linkwatch_fire_event(dev
);
3468 * netif_dormant_off - set device as not dormant.
3469 * @dev: network device
3471 * Device is not in dormant state.
3473 static inline void netif_dormant_off(struct net_device
*dev
)
3475 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3476 linkwatch_fire_event(dev
);
3480 * netif_dormant - test if device is dormant
3481 * @dev: network device
3483 * Check if device is dormant.
3485 static inline bool netif_dormant(const struct net_device
*dev
)
3487 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3492 * netif_oper_up - test if device is operational
3493 * @dev: network device
3495 * Check if carrier is operational
3497 static inline bool netif_oper_up(const struct net_device
*dev
)
3499 return (dev
->operstate
== IF_OPER_UP
||
3500 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3504 * netif_device_present - is device available or removed
3505 * @dev: network device
3507 * Check if device has not been removed from system.
3509 static inline bool netif_device_present(struct net_device
*dev
)
3511 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3514 void netif_device_detach(struct net_device
*dev
);
3516 void netif_device_attach(struct net_device
*dev
);
3519 * Network interface message level settings
3523 NETIF_MSG_DRV
= 0x0001,
3524 NETIF_MSG_PROBE
= 0x0002,
3525 NETIF_MSG_LINK
= 0x0004,
3526 NETIF_MSG_TIMER
= 0x0008,
3527 NETIF_MSG_IFDOWN
= 0x0010,
3528 NETIF_MSG_IFUP
= 0x0020,
3529 NETIF_MSG_RX_ERR
= 0x0040,
3530 NETIF_MSG_TX_ERR
= 0x0080,
3531 NETIF_MSG_TX_QUEUED
= 0x0100,
3532 NETIF_MSG_INTR
= 0x0200,
3533 NETIF_MSG_TX_DONE
= 0x0400,
3534 NETIF_MSG_RX_STATUS
= 0x0800,
3535 NETIF_MSG_PKTDATA
= 0x1000,
3536 NETIF_MSG_HW
= 0x2000,
3537 NETIF_MSG_WOL
= 0x4000,
3540 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3541 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3542 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3543 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3544 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3545 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3546 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3547 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3548 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3549 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3550 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3551 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3552 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3553 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3554 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3556 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3559 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3560 return default_msg_enable_bits
;
3561 if (debug_value
== 0) /* no output */
3563 /* set low N bits */
3564 return (1 << debug_value
) - 1;
3567 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3569 spin_lock(&txq
->_xmit_lock
);
3570 txq
->xmit_lock_owner
= cpu
;
3573 static inline bool __netif_tx_acquire(struct netdev_queue
*txq
)
3575 __acquire(&txq
->_xmit_lock
);
3579 static inline void __netif_tx_release(struct netdev_queue
*txq
)
3581 __release(&txq
->_xmit_lock
);
3584 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3586 spin_lock_bh(&txq
->_xmit_lock
);
3587 txq
->xmit_lock_owner
= smp_processor_id();
3590 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3592 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3594 txq
->xmit_lock_owner
= smp_processor_id();
3598 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3600 txq
->xmit_lock_owner
= -1;
3601 spin_unlock(&txq
->_xmit_lock
);
3604 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3606 txq
->xmit_lock_owner
= -1;
3607 spin_unlock_bh(&txq
->_xmit_lock
);
3610 static inline void txq_trans_update(struct netdev_queue
*txq
)
3612 if (txq
->xmit_lock_owner
!= -1)
3613 txq
->trans_start
= jiffies
;
3616 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3617 static inline void netif_trans_update(struct net_device
*dev
)
3619 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, 0);
3621 if (txq
->trans_start
!= jiffies
)
3622 txq
->trans_start
= jiffies
;
3626 * netif_tx_lock - grab network device transmit lock
3627 * @dev: network device
3629 * Get network device transmit lock
3631 static inline void netif_tx_lock(struct net_device
*dev
)
3636 spin_lock(&dev
->tx_global_lock
);
3637 cpu
= smp_processor_id();
3638 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3639 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3641 /* We are the only thread of execution doing a
3642 * freeze, but we have to grab the _xmit_lock in
3643 * order to synchronize with threads which are in
3644 * the ->hard_start_xmit() handler and already
3645 * checked the frozen bit.
3647 __netif_tx_lock(txq
, cpu
);
3648 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3649 __netif_tx_unlock(txq
);
3653 static inline void netif_tx_lock_bh(struct net_device
*dev
)
3659 static inline void netif_tx_unlock(struct net_device
*dev
)
3663 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3664 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3666 /* No need to grab the _xmit_lock here. If the
3667 * queue is not stopped for another reason, we
3670 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3671 netif_schedule_queue(txq
);
3673 spin_unlock(&dev
->tx_global_lock
);
3676 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
3678 netif_tx_unlock(dev
);
3682 #define HARD_TX_LOCK(dev, txq, cpu) { \
3683 if ((dev->features & NETIF_F_LLTX) == 0) { \
3684 __netif_tx_lock(txq, cpu); \
3686 __netif_tx_acquire(txq); \
3690 #define HARD_TX_TRYLOCK(dev, txq) \
3691 (((dev->features & NETIF_F_LLTX) == 0) ? \
3692 __netif_tx_trylock(txq) : \
3693 __netif_tx_acquire(txq))
3695 #define HARD_TX_UNLOCK(dev, txq) { \
3696 if ((dev->features & NETIF_F_LLTX) == 0) { \
3697 __netif_tx_unlock(txq); \
3699 __netif_tx_release(txq); \
3703 static inline void netif_tx_disable(struct net_device
*dev
)
3709 cpu
= smp_processor_id();
3710 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3711 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3713 __netif_tx_lock(txq
, cpu
);
3714 netif_tx_stop_queue(txq
);
3715 __netif_tx_unlock(txq
);
3720 static inline void netif_addr_lock(struct net_device
*dev
)
3722 spin_lock(&dev
->addr_list_lock
);
3725 static inline void netif_addr_lock_nested(struct net_device
*dev
)
3727 int subclass
= SINGLE_DEPTH_NESTING
;
3729 if (dev
->netdev_ops
->ndo_get_lock_subclass
)
3730 subclass
= dev
->netdev_ops
->ndo_get_lock_subclass(dev
);
3732 spin_lock_nested(&dev
->addr_list_lock
, subclass
);
3735 static inline void netif_addr_lock_bh(struct net_device
*dev
)
3737 spin_lock_bh(&dev
->addr_list_lock
);
3740 static inline void netif_addr_unlock(struct net_device
*dev
)
3742 spin_unlock(&dev
->addr_list_lock
);
3745 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
3747 spin_unlock_bh(&dev
->addr_list_lock
);
3751 * dev_addrs walker. Should be used only for read access. Call with
3752 * rcu_read_lock held.
3754 #define for_each_dev_addr(dev, ha) \
3755 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3757 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3759 void ether_setup(struct net_device
*dev
);
3761 /* Support for loadable net-drivers */
3762 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
3763 unsigned char name_assign_type
,
3764 void (*setup
)(struct net_device
*),
3765 unsigned int txqs
, unsigned int rxqs
);
3766 int dev_get_valid_name(struct net
*net
, struct net_device
*dev
,
3769 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3770 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3772 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3773 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3776 int register_netdev(struct net_device
*dev
);
3777 void unregister_netdev(struct net_device
*dev
);
3779 /* General hardware address lists handling functions */
3780 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
3781 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3782 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
3783 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3784 int __hw_addr_sync_dev(struct netdev_hw_addr_list
*list
,
3785 struct net_device
*dev
,
3786 int (*sync
)(struct net_device
*, const unsigned char *),
3787 int (*unsync
)(struct net_device
*,
3788 const unsigned char *));
3789 void __hw_addr_unsync_dev(struct netdev_hw_addr_list
*list
,
3790 struct net_device
*dev
,
3791 int (*unsync
)(struct net_device
*,
3792 const unsigned char *));
3793 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
3795 /* Functions used for device addresses handling */
3796 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
3797 unsigned char addr_type
);
3798 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
3799 unsigned char addr_type
);
3800 void dev_addr_flush(struct net_device
*dev
);
3801 int dev_addr_init(struct net_device
*dev
);
3803 /* Functions used for unicast addresses handling */
3804 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
3805 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3806 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
3807 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
3808 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3809 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
3810 void dev_uc_flush(struct net_device
*dev
);
3811 void dev_uc_init(struct net_device
*dev
);
3814 * __dev_uc_sync - Synchonize device's unicast list
3815 * @dev: device to sync
3816 * @sync: function to call if address should be added
3817 * @unsync: function to call if address should be removed
3819 * Add newly added addresses to the interface, and release
3820 * addresses that have been deleted.
3822 static inline int __dev_uc_sync(struct net_device
*dev
,
3823 int (*sync
)(struct net_device
*,
3824 const unsigned char *),
3825 int (*unsync
)(struct net_device
*,
3826 const unsigned char *))
3828 return __hw_addr_sync_dev(&dev
->uc
, dev
, sync
, unsync
);
3832 * __dev_uc_unsync - Remove synchronized addresses from device
3833 * @dev: device to sync
3834 * @unsync: function to call if address should be removed
3836 * Remove all addresses that were added to the device by dev_uc_sync().
3838 static inline void __dev_uc_unsync(struct net_device
*dev
,
3839 int (*unsync
)(struct net_device
*,
3840 const unsigned char *))
3842 __hw_addr_unsync_dev(&dev
->uc
, dev
, unsync
);
3845 /* Functions used for multicast addresses handling */
3846 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
3847 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
3848 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3849 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
3850 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
3851 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
3852 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3853 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
3854 void dev_mc_flush(struct net_device
*dev
);
3855 void dev_mc_init(struct net_device
*dev
);
3858 * __dev_mc_sync - Synchonize device's multicast list
3859 * @dev: device to sync
3860 * @sync: function to call if address should be added
3861 * @unsync: function to call if address should be removed
3863 * Add newly added addresses to the interface, and release
3864 * addresses that have been deleted.
3866 static inline int __dev_mc_sync(struct net_device
*dev
,
3867 int (*sync
)(struct net_device
*,
3868 const unsigned char *),
3869 int (*unsync
)(struct net_device
*,
3870 const unsigned char *))
3872 return __hw_addr_sync_dev(&dev
->mc
, dev
, sync
, unsync
);
3876 * __dev_mc_unsync - Remove synchronized addresses from device
3877 * @dev: device to sync
3878 * @unsync: function to call if address should be removed
3880 * Remove all addresses that were added to the device by dev_mc_sync().
3882 static inline void __dev_mc_unsync(struct net_device
*dev
,
3883 int (*unsync
)(struct net_device
*,
3884 const unsigned char *))
3886 __hw_addr_unsync_dev(&dev
->mc
, dev
, unsync
);
3889 /* Functions used for secondary unicast and multicast support */
3890 void dev_set_rx_mode(struct net_device
*dev
);
3891 void __dev_set_rx_mode(struct net_device
*dev
);
3892 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
3893 int dev_set_allmulti(struct net_device
*dev
, int inc
);
3894 void netdev_state_change(struct net_device
*dev
);
3895 void netdev_notify_peers(struct net_device
*dev
);
3896 void netdev_features_change(struct net_device
*dev
);
3897 /* Load a device via the kmod */
3898 void dev_load(struct net
*net
, const char *name
);
3899 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
3900 struct rtnl_link_stats64
*storage
);
3901 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
3902 const struct net_device_stats
*netdev_stats
);
3904 extern int netdev_max_backlog
;
3905 extern int netdev_tstamp_prequeue
;
3906 extern int weight_p
;
3907 extern int dev_weight_rx_bias
;
3908 extern int dev_weight_tx_bias
;
3909 extern int dev_rx_weight
;
3910 extern int dev_tx_weight
;
3912 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3913 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3914 struct list_head
**iter
);
3915 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3916 struct list_head
**iter
);
3918 /* iterate through upper list, must be called under RCU read lock */
3919 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3920 for (iter = &(dev)->adj_list.upper, \
3921 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3923 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3925 int netdev_walk_all_upper_dev_rcu(struct net_device
*dev
,
3926 int (*fn
)(struct net_device
*upper_dev
,
3930 bool netdev_has_upper_dev_all_rcu(struct net_device
*dev
,
3931 struct net_device
*upper_dev
);
3933 bool netdev_has_any_upper_dev(struct net_device
*dev
);
3935 void *netdev_lower_get_next_private(struct net_device
*dev
,
3936 struct list_head
**iter
);
3937 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3938 struct list_head
**iter
);
3940 #define netdev_for_each_lower_private(dev, priv, iter) \
3941 for (iter = (dev)->adj_list.lower.next, \
3942 priv = netdev_lower_get_next_private(dev, &(iter)); \
3944 priv = netdev_lower_get_next_private(dev, &(iter)))
3946 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3947 for (iter = &(dev)->adj_list.lower, \
3948 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3950 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3952 void *netdev_lower_get_next(struct net_device
*dev
,
3953 struct list_head
**iter
);
3955 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3956 for (iter = (dev)->adj_list.lower.next, \
3957 ldev = netdev_lower_get_next(dev, &(iter)); \
3959 ldev = netdev_lower_get_next(dev, &(iter)))
3961 struct net_device
*netdev_all_lower_get_next(struct net_device
*dev
,
3962 struct list_head
**iter
);
3963 struct net_device
*netdev_all_lower_get_next_rcu(struct net_device
*dev
,
3964 struct list_head
**iter
);
3966 int netdev_walk_all_lower_dev(struct net_device
*dev
,
3967 int (*fn
)(struct net_device
*lower_dev
,
3970 int netdev_walk_all_lower_dev_rcu(struct net_device
*dev
,
3971 int (*fn
)(struct net_device
*lower_dev
,
3975 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3976 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3977 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3978 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3979 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
,
3980 struct netlink_ext_ack
*extack
);
3981 int netdev_master_upper_dev_link(struct net_device
*dev
,
3982 struct net_device
*upper_dev
,
3983 void *upper_priv
, void *upper_info
,
3984 struct netlink_ext_ack
*extack
);
3985 void netdev_upper_dev_unlink(struct net_device
*dev
,
3986 struct net_device
*upper_dev
);
3987 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3988 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3989 struct net_device
*lower_dev
);
3990 void netdev_lower_state_changed(struct net_device
*lower_dev
,
3991 void *lower_state_info
);
3993 /* RSS keys are 40 or 52 bytes long */
3994 #define NETDEV_RSS_KEY_LEN 52
3995 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
] __read_mostly
;
3996 void netdev_rss_key_fill(void *buffer
, size_t len
);
3998 int dev_get_nest_level(struct net_device
*dev
);
3999 int skb_checksum_help(struct sk_buff
*skb
);
4000 int skb_crc32c_csum_help(struct sk_buff
*skb
);
4001 int skb_csum_hwoffload_help(struct sk_buff
*skb
,
4002 const netdev_features_t features
);
4004 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
4005 netdev_features_t features
, bool tx_path
);
4006 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
4007 netdev_features_t features
);
4009 struct netdev_bonding_info
{
4014 struct netdev_notifier_bonding_info
{
4015 struct netdev_notifier_info info
; /* must be first */
4016 struct netdev_bonding_info bonding_info
;
4019 void netdev_bonding_info_change(struct net_device
*dev
,
4020 struct netdev_bonding_info
*bonding_info
);
4023 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
4025 return __skb_gso_segment(skb
, features
, true);
4027 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
4029 static inline bool can_checksum_protocol(netdev_features_t features
,
4032 if (protocol
== htons(ETH_P_FCOE
))
4033 return !!(features
& NETIF_F_FCOE_CRC
);
4035 /* Assume this is an IP checksum (not SCTP CRC) */
4037 if (features
& NETIF_F_HW_CSUM
) {
4038 /* Can checksum everything */
4043 case htons(ETH_P_IP
):
4044 return !!(features
& NETIF_F_IP_CSUM
);
4045 case htons(ETH_P_IPV6
):
4046 return !!(features
& NETIF_F_IPV6_CSUM
);
4053 void netdev_rx_csum_fault(struct net_device
*dev
);
4055 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
4059 /* rx skb timestamps */
4060 void net_enable_timestamp(void);
4061 void net_disable_timestamp(void);
4063 #ifdef CONFIG_PROC_FS
4064 int __init
dev_proc_init(void);
4066 #define dev_proc_init() 0
4069 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
4070 struct sk_buff
*skb
, struct net_device
*dev
,
4073 skb
->xmit_more
= more
? 1 : 0;
4074 return ops
->ndo_start_xmit(skb
, dev
);
4077 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
4078 struct netdev_queue
*txq
, bool more
)
4080 const struct net_device_ops
*ops
= dev
->netdev_ops
;
4083 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
4084 if (rc
== NETDEV_TX_OK
)
4085 txq_trans_update(txq
);
4090 int netdev_class_create_file_ns(const struct class_attribute
*class_attr
,
4092 void netdev_class_remove_file_ns(const struct class_attribute
*class_attr
,
4095 static inline int netdev_class_create_file(const struct class_attribute
*class_attr
)
4097 return netdev_class_create_file_ns(class_attr
, NULL
);
4100 static inline void netdev_class_remove_file(const struct class_attribute
*class_attr
)
4102 netdev_class_remove_file_ns(class_attr
, NULL
);
4105 extern const struct kobj_ns_type_operations net_ns_type_operations
;
4107 const char *netdev_drivername(const struct net_device
*dev
);
4109 void linkwatch_run_queue(void);
4111 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
4112 netdev_features_t f2
)
4114 if ((f1
^ f2
) & NETIF_F_HW_CSUM
) {
4115 if (f1
& NETIF_F_HW_CSUM
)
4116 f1
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
4118 f2
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
4124 static inline netdev_features_t
netdev_get_wanted_features(
4125 struct net_device
*dev
)
4127 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
4129 netdev_features_t
netdev_increment_features(netdev_features_t all
,
4130 netdev_features_t one
, netdev_features_t mask
);
4132 /* Allow TSO being used on stacked device :
4133 * Performing the GSO segmentation before last device
4134 * is a performance improvement.
4136 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
4137 netdev_features_t mask
)
4139 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
4142 int __netdev_update_features(struct net_device
*dev
);
4143 void netdev_update_features(struct net_device
*dev
);
4144 void netdev_change_features(struct net_device
*dev
);
4146 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
4147 struct net_device
*dev
);
4149 netdev_features_t
passthru_features_check(struct sk_buff
*skb
,
4150 struct net_device
*dev
,
4151 netdev_features_t features
);
4152 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
4154 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
4156 netdev_features_t feature
= (netdev_features_t
)gso_type
<< NETIF_F_GSO_SHIFT
;
4158 /* check flags correspondence */
4159 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
4160 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
4161 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
4162 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID
!= (NETIF_F_TSO_MANGLEID
>> NETIF_F_GSO_SHIFT
));
4163 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
4164 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
4165 BUILD_BUG_ON(SKB_GSO_GRE
!= (NETIF_F_GSO_GRE
>> NETIF_F_GSO_SHIFT
));
4166 BUILD_BUG_ON(SKB_GSO_GRE_CSUM
!= (NETIF_F_GSO_GRE_CSUM
>> NETIF_F_GSO_SHIFT
));
4167 BUILD_BUG_ON(SKB_GSO_IPXIP4
!= (NETIF_F_GSO_IPXIP4
>> NETIF_F_GSO_SHIFT
));
4168 BUILD_BUG_ON(SKB_GSO_IPXIP6
!= (NETIF_F_GSO_IPXIP6
>> NETIF_F_GSO_SHIFT
));
4169 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL
!= (NETIF_F_GSO_UDP_TUNNEL
>> NETIF_F_GSO_SHIFT
));
4170 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM
!= (NETIF_F_GSO_UDP_TUNNEL_CSUM
>> NETIF_F_GSO_SHIFT
));
4171 BUILD_BUG_ON(SKB_GSO_PARTIAL
!= (NETIF_F_GSO_PARTIAL
>> NETIF_F_GSO_SHIFT
));
4172 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
4173 BUILD_BUG_ON(SKB_GSO_SCTP
!= (NETIF_F_GSO_SCTP
>> NETIF_F_GSO_SHIFT
));
4174 BUILD_BUG_ON(SKB_GSO_ESP
!= (NETIF_F_GSO_ESP
>> NETIF_F_GSO_SHIFT
));
4175 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_GSO_UDP
>> NETIF_F_GSO_SHIFT
));
4177 return (features
& feature
) == feature
;
4180 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
4182 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
4183 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
4186 static inline bool netif_needs_gso(struct sk_buff
*skb
,
4187 netdev_features_t features
)
4189 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
4190 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
4191 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
4194 static inline void netif_set_gso_max_size(struct net_device
*dev
,
4197 dev
->gso_max_size
= size
;
4200 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
4201 int pulled_hlen
, u16 mac_offset
,
4204 skb
->protocol
= protocol
;
4205 skb
->encapsulation
= 1;
4206 skb_push(skb
, pulled_hlen
);
4207 skb_reset_transport_header(skb
);
4208 skb
->mac_header
= mac_offset
;
4209 skb
->network_header
= skb
->mac_header
+ mac_len
;
4210 skb
->mac_len
= mac_len
;
4213 static inline bool netif_is_macsec(const struct net_device
*dev
)
4215 return dev
->priv_flags
& IFF_MACSEC
;
4218 static inline bool netif_is_macvlan(const struct net_device
*dev
)
4220 return dev
->priv_flags
& IFF_MACVLAN
;
4223 static inline bool netif_is_macvlan_port(const struct net_device
*dev
)
4225 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
4228 static inline bool netif_is_ipvlan(const struct net_device
*dev
)
4230 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
4233 static inline bool netif_is_ipvlan_port(const struct net_device
*dev
)
4235 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
4238 static inline bool netif_is_bond_master(const struct net_device
*dev
)
4240 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
4243 static inline bool netif_is_bond_slave(const struct net_device
*dev
)
4245 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
4248 static inline bool netif_supports_nofcs(struct net_device
*dev
)
4250 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
4253 static inline bool netif_has_l3_rx_handler(const struct net_device
*dev
)
4255 return dev
->priv_flags
& IFF_L3MDEV_RX_HANDLER
;
4258 static inline bool netif_is_l3_master(const struct net_device
*dev
)
4260 return dev
->priv_flags
& IFF_L3MDEV_MASTER
;
4263 static inline bool netif_is_l3_slave(const struct net_device
*dev
)
4265 return dev
->priv_flags
& IFF_L3MDEV_SLAVE
;
4268 static inline bool netif_is_bridge_master(const struct net_device
*dev
)
4270 return dev
->priv_flags
& IFF_EBRIDGE
;
4273 static inline bool netif_is_bridge_port(const struct net_device
*dev
)
4275 return dev
->priv_flags
& IFF_BRIDGE_PORT
;
4278 static inline bool netif_is_ovs_master(const struct net_device
*dev
)
4280 return dev
->priv_flags
& IFF_OPENVSWITCH
;
4283 static inline bool netif_is_ovs_port(const struct net_device
*dev
)
4285 return dev
->priv_flags
& IFF_OVS_DATAPATH
;
4288 static inline bool netif_is_team_master(const struct net_device
*dev
)
4290 return dev
->priv_flags
& IFF_TEAM
;
4293 static inline bool netif_is_team_port(const struct net_device
*dev
)
4295 return dev
->priv_flags
& IFF_TEAM_PORT
;
4298 static inline bool netif_is_lag_master(const struct net_device
*dev
)
4300 return netif_is_bond_master(dev
) || netif_is_team_master(dev
);
4303 static inline bool netif_is_lag_port(const struct net_device
*dev
)
4305 return netif_is_bond_slave(dev
) || netif_is_team_port(dev
);
4308 static inline bool netif_is_rxfh_configured(const struct net_device
*dev
)
4310 return dev
->priv_flags
& IFF_RXFH_CONFIGURED
;
4313 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4314 static inline void netif_keep_dst(struct net_device
*dev
)
4316 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
4319 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4320 static inline bool netif_reduces_vlan_mtu(struct net_device
*dev
)
4322 /* TODO: reserve and use an additional IFF bit, if we get more users */
4323 return dev
->priv_flags
& IFF_MACSEC
;
4326 extern struct pernet_operations __net_initdata loopback_net_ops
;
4328 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4330 /* netdev_printk helpers, similar to dev_printk */
4332 static inline const char *netdev_name(const struct net_device
*dev
)
4334 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
4335 return "(unnamed net_device)";
4339 static inline bool netdev_unregistering(const struct net_device
*dev
)
4341 return dev
->reg_state
== NETREG_UNREGISTERING
;
4344 static inline const char *netdev_reg_state(const struct net_device
*dev
)
4346 switch (dev
->reg_state
) {
4347 case NETREG_UNINITIALIZED
: return " (uninitialized)";
4348 case NETREG_REGISTERED
: return "";
4349 case NETREG_UNREGISTERING
: return " (unregistering)";
4350 case NETREG_UNREGISTERED
: return " (unregistered)";
4351 case NETREG_RELEASED
: return " (released)";
4352 case NETREG_DUMMY
: return " (dummy)";
4355 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
4356 return " (unknown)";
4360 void netdev_printk(const char *level
, const struct net_device
*dev
,
4361 const char *format
, ...);
4363 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
4365 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
4367 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
4369 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
4371 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
4373 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
4375 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
4377 #define netdev_level_once(level, dev, fmt, ...) \
4379 static bool __print_once __read_mostly; \
4381 if (!__print_once) { \
4382 __print_once = true; \
4383 netdev_printk(level, dev, fmt, ##__VA_ARGS__); \
4387 #define netdev_emerg_once(dev, fmt, ...) \
4388 netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4389 #define netdev_alert_once(dev, fmt, ...) \
4390 netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4391 #define netdev_crit_once(dev, fmt, ...) \
4392 netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4393 #define netdev_err_once(dev, fmt, ...) \
4394 netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4395 #define netdev_warn_once(dev, fmt, ...) \
4396 netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4397 #define netdev_notice_once(dev, fmt, ...) \
4398 netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4399 #define netdev_info_once(dev, fmt, ...) \
4400 netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4402 #define MODULE_ALIAS_NETDEV(device) \
4403 MODULE_ALIAS("netdev-" device)
4405 #if defined(CONFIG_DYNAMIC_DEBUG)
4406 #define netdev_dbg(__dev, format, args...) \
4408 dynamic_netdev_dbg(__dev, format, ##args); \
4410 #elif defined(DEBUG)
4411 #define netdev_dbg(__dev, format, args...) \
4412 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4414 #define netdev_dbg(__dev, format, args...) \
4417 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4421 #if defined(VERBOSE_DEBUG)
4422 #define netdev_vdbg netdev_dbg
4425 #define netdev_vdbg(dev, format, args...) \
4428 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4434 * netdev_WARN() acts like dev_printk(), but with the key difference
4435 * of using a WARN/WARN_ON to get the message out, including the
4436 * file/line information and a backtrace.
4438 #define netdev_WARN(dev, format, args...) \
4439 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4440 netdev_reg_state(dev), ##args)
4442 #define netdev_WARN_ONCE(dev, format, args...) \
4443 WARN_ONCE(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4444 netdev_reg_state(dev), ##args)
4446 /* netif printk helpers, similar to netdev_printk */
4448 #define netif_printk(priv, type, level, dev, fmt, args...) \
4450 if (netif_msg_##type(priv)) \
4451 netdev_printk(level, (dev), fmt, ##args); \
4454 #define netif_level(level, priv, type, dev, fmt, args...) \
4456 if (netif_msg_##type(priv)) \
4457 netdev_##level(dev, fmt, ##args); \
4460 #define netif_emerg(priv, type, dev, fmt, args...) \
4461 netif_level(emerg, priv, type, dev, fmt, ##args)
4462 #define netif_alert(priv, type, dev, fmt, args...) \
4463 netif_level(alert, priv, type, dev, fmt, ##args)
4464 #define netif_crit(priv, type, dev, fmt, args...) \
4465 netif_level(crit, priv, type, dev, fmt, ##args)
4466 #define netif_err(priv, type, dev, fmt, args...) \
4467 netif_level(err, priv, type, dev, fmt, ##args)
4468 #define netif_warn(priv, type, dev, fmt, args...) \
4469 netif_level(warn, priv, type, dev, fmt, ##args)
4470 #define netif_notice(priv, type, dev, fmt, args...) \
4471 netif_level(notice, priv, type, dev, fmt, ##args)
4472 #define netif_info(priv, type, dev, fmt, args...) \
4473 netif_level(info, priv, type, dev, fmt, ##args)
4475 #if defined(CONFIG_DYNAMIC_DEBUG)
4476 #define netif_dbg(priv, type, netdev, format, args...) \
4478 if (netif_msg_##type(priv)) \
4479 dynamic_netdev_dbg(netdev, format, ##args); \
4481 #elif defined(DEBUG)
4482 #define netif_dbg(priv, type, dev, format, args...) \
4483 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4485 #define netif_dbg(priv, type, dev, format, args...) \
4488 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4493 /* if @cond then downgrade to debug, else print at @level */
4494 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4497 netif_dbg(priv, type, netdev, fmt, ##args); \
4499 netif_ ## level(priv, type, netdev, fmt, ##args); \
4502 #if defined(VERBOSE_DEBUG)
4503 #define netif_vdbg netif_dbg
4505 #define netif_vdbg(priv, type, dev, format, args...) \
4508 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4514 * The list of packet types we will receive (as opposed to discard)
4515 * and the routines to invoke.
4517 * Why 16. Because with 16 the only overlap we get on a hash of the
4518 * low nibble of the protocol value is RARP/SNAP/X.25.
4532 #define PTYPE_HASH_SIZE (16)
4533 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4535 #endif /* _LINUX_NETDEVICE_H */