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/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
47 #include <net/dcbnl.h>
49 #include <net/netprio_cgroup.h>
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
61 /* 802.15.4 specific */
64 void netdev_set_default_ethtool_ops(struct net_device
*dev
,
65 const struct ethtool_ops
*ops
);
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
72 * Transmit return codes: transmit return codes originate from three different
75 * - qdisc return codes
76 * - driver transmit return codes
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
105 __NETDEV_TX_MIN
= INT_MIN
, /* make sure enum is signed */
106 NETDEV_TX_OK
= 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY
= 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED
= 0x20, /* driver tx lock was already taken */
110 typedef enum netdev_tx netdev_tx_t
;
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 static inline bool dev_xmit_complete(int rc
)
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 if (likely(rc
< NET_XMIT_MASK
))
131 * Compute the worst case header length according to the protocols
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
139 # define LL_MAX_HEADER 96
142 # define LL_MAX_HEADER 32
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
157 struct net_device_stats
{
158 unsigned long rx_packets
;
159 unsigned long tx_packets
;
160 unsigned long rx_bytes
;
161 unsigned long tx_bytes
;
162 unsigned long rx_errors
;
163 unsigned long tx_errors
;
164 unsigned long rx_dropped
;
165 unsigned long tx_dropped
;
166 unsigned long multicast
;
167 unsigned long collisions
;
168 unsigned long rx_length_errors
;
169 unsigned long rx_over_errors
;
170 unsigned long rx_crc_errors
;
171 unsigned long rx_frame_errors
;
172 unsigned long rx_fifo_errors
;
173 unsigned long rx_missed_errors
;
174 unsigned long tx_aborted_errors
;
175 unsigned long tx_carrier_errors
;
176 unsigned long tx_fifo_errors
;
177 unsigned long tx_heartbeat_errors
;
178 unsigned long tx_window_errors
;
179 unsigned long rx_compressed
;
180 unsigned long tx_compressed
;
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed
;
196 struct netdev_hw_addr
{
197 struct list_head list
;
198 unsigned char addr
[MAX_ADDR_LEN
];
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
209 struct rcu_head rcu_head
;
212 struct netdev_hw_addr_list
{
213 struct list_head list
;
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data
[HH_DATA_ALIGN(LL_MAX_HEADER
) / sizeof(long)];
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 int (*create
) (struct sk_buff
*skb
, struct net_device
*dev
,
261 unsigned short type
, const void *daddr
,
262 const void *saddr
, unsigned int len
);
263 int (*parse
)(const struct sk_buff
*skb
, unsigned char *haddr
);
264 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
265 void (*cache_update
)(struct hh_cache
*hh
,
266 const struct net_device
*dev
,
267 const unsigned char *haddr
);
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
275 enum netdev_state_t
{
277 __LINK_STATE_PRESENT
,
278 __LINK_STATE_NOCARRIER
,
279 __LINK_STATE_LINKWATCH_PENDING
,
280 __LINK_STATE_DORMANT
,
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
288 struct netdev_boot_setup
{
292 #define NETDEV_BOOT_SETUP_MAX 8
294 int __init
netdev_boot_setup(char *str
);
297 * Structure for NAPI scheduling similar to tasklet but with weighting
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
306 struct list_head poll_list
;
310 unsigned int gro_count
;
311 int (*poll
)(struct napi_struct
*, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock
;
316 struct net_device
*dev
;
317 struct sk_buff
*gro_list
;
319 struct hrtimer timer
;
320 struct list_head dev_list
;
321 struct hlist_node napi_hash_node
;
322 unsigned int napi_id
;
326 NAPI_STATE_SCHED
, /* Poll is scheduled */
327 NAPI_STATE_DISABLE
, /* Disable pending */
328 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED
, /* In NAPI hash */
339 typedef enum gro_result gro_result_t
;
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
382 enum rx_handler_result
{
388 typedef enum rx_handler_result rx_handler_result_t
;
389 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
391 void __napi_schedule(struct napi_struct
*n
);
392 void __napi_schedule_irqoff(struct napi_struct
*n
);
394 static inline bool napi_disable_pending(struct napi_struct
*n
)
396 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
400 * napi_schedule_prep - check if napi can be scheduled
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
408 static inline bool napi_schedule_prep(struct napi_struct
*n
)
410 return !napi_disable_pending(n
) &&
411 !test_and_set_bit(NAPI_STATE_SCHED
, &n
->state
);
415 * napi_schedule - schedule NAPI poll
418 * Schedule NAPI poll routine to be called if it is not already
421 static inline void napi_schedule(struct napi_struct
*n
)
423 if (napi_schedule_prep(n
))
428 * napi_schedule_irqoff - schedule NAPI poll
431 * Variant of napi_schedule(), assuming hard irqs are masked.
433 static inline void napi_schedule_irqoff(struct napi_struct
*n
)
435 if (napi_schedule_prep(n
))
436 __napi_schedule_irqoff(n
);
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct
*napi
)
442 if (napi_schedule_prep(napi
)) {
443 __napi_schedule(napi
);
449 void __napi_complete(struct napi_struct
*n
);
450 void napi_complete_done(struct napi_struct
*n
, int work_done
);
452 * napi_complete - NAPI processing complete
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
458 static inline void napi_complete(struct napi_struct
*n
)
460 return napi_complete_done(n
, 0);
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
470 struct napi_struct
*napi_by_id(unsigned int napi_id
);
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
476 * generate a new napi_id and store a @napi under it in napi_hash
478 void napi_hash_add(struct napi_struct
*napi
);
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
487 void napi_hash_del(struct napi_struct
*napi
);
490 * napi_disable - prevent NAPI from scheduling
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
496 void napi_disable(struct napi_struct
*n
);
499 * napi_enable - enable NAPI scheduling
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
505 static inline void napi_enable(struct napi_struct
*n
)
507 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
514 * napi_synchronize - wait until NAPI is not running
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
521 static inline void napi_synchronize(const struct napi_struct
*n
)
523 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
527 # define napi_synchronize(n) barrier()
530 enum netdev_queue_state_t
{
531 __QUEUE_STATE_DRV_XOFF
,
532 __QUEUE_STATE_STACK_XOFF
,
533 __QUEUE_STATE_FROZEN
,
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
556 struct netdev_queue
{
560 struct net_device
*dev
;
561 struct Qdisc __rcu
*qdisc
;
562 struct Qdisc
*qdisc_sleeping
;
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
575 * please use this field instead of dev->trans_start
577 unsigned long trans_start
;
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
583 unsigned long trans_timeout
;
590 unsigned long tx_maxrate
;
591 } ____cacheline_aligned_in_smp
;
593 static inline int netdev_queue_numa_node_read(const struct netdev_queue
*q
)
595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
602 static inline void netdev_queue_numa_node_write(struct netdev_queue
*q
, int node
)
604 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
619 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
626 struct rps_dev_flow
{
629 unsigned int last_qtail
;
631 #define RPS_NO_FILTER 0xffff
634 * The rps_dev_flow_table structure contains a table of flow mappings.
636 struct rps_dev_flow_table
{
639 struct rps_dev_flow flows
[0];
641 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
642 ((_num) * sizeof(struct rps_dev_flow)))
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
654 struct rps_sock_flow_table
{
657 u32 ents
[0] ____cacheline_aligned_in_smp
;
659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
661 #define RPS_NO_CPU 0xffff
663 extern u32 rps_cpu_mask
;
664 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
666 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
670 unsigned int index
= hash
& table
->mask
;
671 u32 val
= hash
& ~rps_cpu_mask
;
673 /* We only give a hint, preemption can change cpu under us */
674 val
|= raw_smp_processor_id();
676 if (table
->ents
[index
] != val
)
677 table
->ents
[index
] = val
;
681 #ifdef CONFIG_RFS_ACCEL
682 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
685 #endif /* CONFIG_RPS */
687 /* This structure contains an instance of an RX queue. */
688 struct netdev_rx_queue
{
690 struct rps_map __rcu
*rps_map
;
691 struct rps_dev_flow_table __rcu
*rps_flow_table
;
694 struct net_device
*dev
;
695 } ____cacheline_aligned_in_smp
;
698 * RX queue sysfs structures and functions.
700 struct rx_queue_attribute
{
701 struct attribute attr
;
702 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
703 struct rx_queue_attribute
*attr
, char *buf
);
704 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
705 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
715 unsigned int alloc_len
;
719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 struct xps_dev_maps
{
728 struct xps_map __rcu
*cpu_map
[0];
730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732 #endif /* CONFIG_XPS */
734 #define TC_MAX_QUEUE 16
735 #define TC_BITMASK 15
736 /* HW offloaded queuing disciplines txq count and offset maps */
737 struct netdev_tc_txq
{
742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
747 struct netdev_fcoe_hbainfo
{
748 char manufacturer
[64];
749 char serial_number
[64];
750 char hardware_version
[64];
751 char driver_version
[64];
752 char optionrom_version
[64];
753 char firmware_version
[64];
755 char model_description
[256];
759 #define MAX_PHYS_ITEM_ID_LEN 32
761 /* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
764 struct netdev_phys_item_id
{
765 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
766 unsigned char id_len
;
769 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
770 struct sk_buff
*skb
);
773 * This structure defines the management hooks for network devices.
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
797 * Called when a packet needs to be transmitted.
798 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
799 * the queue before that can happen; it's for obsolete devices and weird
800 * corner cases, but the stack really does a non-trivial amount
801 * of useless work if you return NETDEV_TX_BUSY.
802 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
803 * Required can not be NULL.
805 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
806 * void *accel_priv, select_queue_fallback_t fallback);
807 * Called to decide which queue to when device supports multiple
810 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
811 * This function is called to allow device receiver to make
812 * changes to configuration when multicast or promiscious is enabled.
814 * void (*ndo_set_rx_mode)(struct net_device *dev);
815 * This function is called device changes address list filtering.
816 * If driver handles unicast address filtering, it should set
817 * IFF_UNICAST_FLT to its priv_flags.
819 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
820 * This function is called when the Media Access Control address
821 * needs to be changed. If this interface is not defined, the
822 * mac address can not be changed.
824 * int (*ndo_validate_addr)(struct net_device *dev);
825 * Test if Media Access Control address is valid for the device.
827 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
828 * Called when a user request an ioctl which can't be handled by
829 * the generic interface code. If not defined ioctl's return
830 * not supported error code.
832 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
833 * Used to set network devices bus interface parameters. This interface
834 * is retained for legacy reason, new devices should use the bus
835 * interface (PCI) for low level management.
837 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
838 * Called when a user wants to change the Maximum Transfer Unit
839 * of a device. If not defined, any request to change MTU will
840 * will return an error.
842 * void (*ndo_tx_timeout)(struct net_device *dev);
843 * Callback uses when the transmitter has not made any progress
844 * for dev->watchdog ticks.
846 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
847 * struct rtnl_link_stats64 *storage);
848 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
849 * Called when a user wants to get the network device usage
850 * statistics. Drivers must do one of the following:
851 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
852 * rtnl_link_stats64 structure passed by the caller.
853 * 2. Define @ndo_get_stats to update a net_device_stats structure
854 * (which should normally be dev->stats) and return a pointer to
855 * it. The structure may be changed asynchronously only if each
856 * field is written atomically.
857 * 3. Update dev->stats asynchronously and atomically, and define
860 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
861 * If device support VLAN filtering this function is called when a
862 * VLAN id is registered.
864 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
865 * If device support VLAN filtering this function is called when a
866 * VLAN id is unregistered.
868 * void (*ndo_poll_controller)(struct net_device *dev);
870 * SR-IOV management functions.
871 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
872 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
873 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
875 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
876 * int (*ndo_get_vf_config)(struct net_device *dev,
877 * int vf, struct ifla_vf_info *ivf);
878 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
879 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
880 * struct nlattr *port[]);
881 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
882 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
883 * Called to setup 'tc' number of traffic classes in the net device. This
884 * is always called from the stack with the rtnl lock held and netif tx
885 * queues stopped. This allows the netdevice to perform queue management
888 * Fiber Channel over Ethernet (FCoE) offload functions.
889 * int (*ndo_fcoe_enable)(struct net_device *dev);
890 * Called when the FCoE protocol stack wants to start using LLD for FCoE
891 * so the underlying device can perform whatever needed configuration or
892 * initialization to support acceleration of FCoE traffic.
894 * int (*ndo_fcoe_disable)(struct net_device *dev);
895 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
896 * so the underlying device can perform whatever needed clean-ups to
897 * stop supporting acceleration of FCoE traffic.
899 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
900 * struct scatterlist *sgl, unsigned int sgc);
901 * Called when the FCoE Initiator wants to initialize an I/O that
902 * is a possible candidate for Direct Data Placement (DDP). The LLD can
903 * perform necessary setup and returns 1 to indicate the device is set up
904 * successfully to perform DDP on this I/O, otherwise this returns 0.
906 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
907 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
908 * indicated by the FC exchange id 'xid', so the underlying device can
909 * clean up and reuse resources for later DDP requests.
911 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
912 * struct scatterlist *sgl, unsigned int sgc);
913 * Called when the FCoE Target wants to initialize an I/O that
914 * is a possible candidate for Direct Data Placement (DDP). The LLD can
915 * perform necessary setup and returns 1 to indicate the device is set up
916 * successfully to perform DDP on this I/O, otherwise this returns 0.
918 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
919 * struct netdev_fcoe_hbainfo *hbainfo);
920 * Called when the FCoE Protocol stack wants information on the underlying
921 * device. This information is utilized by the FCoE protocol stack to
922 * register attributes with Fiber Channel management service as per the
923 * FC-GS Fabric Device Management Information(FDMI) specification.
925 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
926 * Called when the underlying device wants to override default World Wide
927 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
928 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
929 * protocol stack to use.
932 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
933 * u16 rxq_index, u32 flow_id);
934 * Set hardware filter for RFS. rxq_index is the target queue index;
935 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
936 * Return the filter ID on success, or a negative error code.
938 * Slave management functions (for bridge, bonding, etc).
939 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
940 * Called to make another netdev an underling.
942 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
943 * Called to release previously enslaved netdev.
945 * Feature/offload setting functions.
946 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
947 * netdev_features_t features);
948 * Adjusts the requested feature flags according to device-specific
949 * constraints, and returns the resulting flags. Must not modify
952 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
953 * Called to update device configuration to new features. Passed
954 * feature set might be less than what was returned by ndo_fix_features()).
955 * Must return >0 or -errno if it changed dev->features itself.
957 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
958 * struct net_device *dev,
959 * const unsigned char *addr, u16 vid, u16 flags)
960 * Adds an FDB entry to dev for addr.
961 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
962 * struct net_device *dev,
963 * const unsigned char *addr, u16 vid)
964 * Deletes the FDB entry from dev coresponding to addr.
965 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
966 * struct net_device *dev, struct net_device *filter_dev,
968 * Used to add FDB entries to dump requests. Implementers should add
969 * entries to skb and update idx with the number of entries.
971 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
973 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
974 * struct net_device *dev, u32 filter_mask)
975 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
978 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
979 * Called to change device carrier. Soft-devices (like dummy, team, etc)
980 * which do not represent real hardware may define this to allow their
981 * userspace components to manage their virtual carrier state. Devices
982 * that determine carrier state from physical hardware properties (eg
983 * network cables) or protocol-dependent mechanisms (eg
984 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
986 * int (*ndo_get_phys_port_id)(struct net_device *dev,
987 * struct netdev_phys_item_id *ppid);
988 * Called to get ID of physical port of this device. If driver does
989 * not implement this, it is assumed that the hw is not able to have
990 * multiple net devices on single physical port.
992 * void (*ndo_add_vxlan_port)(struct net_device *dev,
993 * sa_family_t sa_family, __be16 port);
994 * Called by vxlan to notiy a driver about the UDP port and socket
995 * address family that vxlan is listnening to. It is called only when
996 * a new port starts listening. The operation is protected by the
997 * vxlan_net->sock_lock.
999 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1000 * sa_family_t sa_family, __be16 port);
1001 * Called by vxlan to notify the driver about a UDP port and socket
1002 * address family that vxlan is not listening to anymore. The operation
1003 * is protected by the vxlan_net->sock_lock.
1005 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1006 * struct net_device *dev)
1007 * Called by upper layer devices to accelerate switching or other
1008 * station functionality into hardware. 'pdev is the lowerdev
1009 * to use for the offload and 'dev' is the net device that will
1010 * back the offload. Returns a pointer to the private structure
1011 * the upper layer will maintain.
1012 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1013 * Called by upper layer device to delete the station created
1014 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1015 * the station and priv is the structure returned by the add
1017 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1018 * struct net_device *dev,
1020 * Callback to use for xmit over the accelerated station. This
1021 * is used in place of ndo_start_xmit on accelerated net
1023 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1024 * struct net_device *dev
1025 * netdev_features_t features);
1026 * Called by core transmit path to determine if device is capable of
1027 * performing offload operations on a given packet. This is to give
1028 * the device an opportunity to implement any restrictions that cannot
1029 * be otherwise expressed by feature flags. The check is called with
1030 * the set of features that the stack has calculated and it returns
1031 * those the driver believes to be appropriate.
1032 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1033 * int queue_index, u32 maxrate);
1034 * Called when a user wants to set a max-rate limitation of specific
1036 * int (*ndo_get_iflink)(const struct net_device *dev);
1037 * Called to get the iflink value of this device.
1039 struct net_device_ops
{
1040 int (*ndo_init
)(struct net_device
*dev
);
1041 void (*ndo_uninit
)(struct net_device
*dev
);
1042 int (*ndo_open
)(struct net_device
*dev
);
1043 int (*ndo_stop
)(struct net_device
*dev
);
1044 netdev_tx_t (*ndo_start_xmit
) (struct sk_buff
*skb
,
1045 struct net_device
*dev
);
1046 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1047 struct sk_buff
*skb
,
1049 select_queue_fallback_t fallback
);
1050 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1052 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1053 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1055 int (*ndo_validate_addr
)(struct net_device
*dev
);
1056 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1057 struct ifreq
*ifr
, int cmd
);
1058 int (*ndo_set_config
)(struct net_device
*dev
,
1060 int (*ndo_change_mtu
)(struct net_device
*dev
,
1062 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1063 struct neigh_parms
*);
1064 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1066 struct rtnl_link_stats64
* (*ndo_get_stats64
)(struct net_device
*dev
,
1067 struct rtnl_link_stats64
*storage
);
1068 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1070 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1071 __be16 proto
, u16 vid
);
1072 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1073 __be16 proto
, u16 vid
);
1074 #ifdef CONFIG_NET_POLL_CONTROLLER
1075 void (*ndo_poll_controller
)(struct net_device
*dev
);
1076 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1077 struct netpoll_info
*info
);
1078 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1080 #ifdef CONFIG_NET_RX_BUSY_POLL
1081 int (*ndo_busy_poll
)(struct napi_struct
*dev
);
1083 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1084 int queue
, u8
*mac
);
1085 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1086 int queue
, u16 vlan
, u8 qos
);
1087 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1088 int vf
, int min_tx_rate
,
1090 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1091 int vf
, bool setting
);
1092 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1094 struct ifla_vf_info
*ivf
);
1095 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1096 int vf
, int link_state
);
1097 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1099 struct nlattr
*port
[]);
1100 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1101 int vf
, struct sk_buff
*skb
);
1102 int (*ndo_setup_tc
)(struct net_device
*dev
, u8 tc
);
1103 #if IS_ENABLED(CONFIG_FCOE)
1104 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1105 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1106 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1108 struct scatterlist
*sgl
,
1110 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1112 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1114 struct scatterlist
*sgl
,
1116 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1117 struct netdev_fcoe_hbainfo
*hbainfo
);
1120 #if IS_ENABLED(CONFIG_LIBFCOE)
1121 #define NETDEV_FCOE_WWNN 0
1122 #define NETDEV_FCOE_WWPN 1
1123 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1124 u64
*wwn
, int type
);
1127 #ifdef CONFIG_RFS_ACCEL
1128 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1129 const struct sk_buff
*skb
,
1133 int (*ndo_add_slave
)(struct net_device
*dev
,
1134 struct net_device
*slave_dev
);
1135 int (*ndo_del_slave
)(struct net_device
*dev
,
1136 struct net_device
*slave_dev
);
1137 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1138 netdev_features_t features
);
1139 int (*ndo_set_features
)(struct net_device
*dev
,
1140 netdev_features_t features
);
1141 int (*ndo_neigh_construct
)(struct neighbour
*n
);
1142 void (*ndo_neigh_destroy
)(struct neighbour
*n
);
1144 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1145 struct nlattr
*tb
[],
1146 struct net_device
*dev
,
1147 const unsigned char *addr
,
1150 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1151 struct nlattr
*tb
[],
1152 struct net_device
*dev
,
1153 const unsigned char *addr
,
1155 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1156 struct netlink_callback
*cb
,
1157 struct net_device
*dev
,
1158 struct net_device
*filter_dev
,
1161 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1162 struct nlmsghdr
*nlh
,
1164 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1166 struct net_device
*dev
,
1168 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1169 struct nlmsghdr
*nlh
,
1171 int (*ndo_change_carrier
)(struct net_device
*dev
,
1173 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1174 struct netdev_phys_item_id
*ppid
);
1175 int (*ndo_get_phys_port_name
)(struct net_device
*dev
,
1176 char *name
, size_t len
);
1177 void (*ndo_add_vxlan_port
)(struct net_device
*dev
,
1178 sa_family_t sa_family
,
1180 void (*ndo_del_vxlan_port
)(struct net_device
*dev
,
1181 sa_family_t sa_family
,
1184 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1185 struct net_device
*dev
);
1186 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1189 netdev_tx_t (*ndo_dfwd_start_xmit
) (struct sk_buff
*skb
,
1190 struct net_device
*dev
,
1192 int (*ndo_get_lock_subclass
)(struct net_device
*dev
);
1193 netdev_features_t (*ndo_features_check
) (struct sk_buff
*skb
,
1194 struct net_device
*dev
,
1195 netdev_features_t features
);
1196 int (*ndo_set_tx_maxrate
)(struct net_device
*dev
,
1199 int (*ndo_get_iflink
)(const struct net_device
*dev
);
1203 * enum net_device_priv_flags - &struct net_device priv_flags
1205 * These are the &struct net_device, they are only set internally
1206 * by drivers and used in the kernel. These flags are invisible to
1207 * userspace, this means that the order of these flags can change
1208 * during any kernel release.
1210 * You should have a pretty good reason to be extending these flags.
1212 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1213 * @IFF_EBRIDGE: Ethernet bridging device
1214 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1215 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1216 * @IFF_MASTER_ALB: bonding master, balance-alb
1217 * @IFF_BONDING: bonding master or slave
1218 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1219 * @IFF_ISATAP: ISATAP interface (RFC4214)
1220 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1221 * @IFF_WAN_HDLC: WAN HDLC device
1222 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1224 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1225 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1226 * @IFF_MACVLAN_PORT: device used as macvlan port
1227 * @IFF_BRIDGE_PORT: device used as bridge port
1228 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1229 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1230 * @IFF_UNICAST_FLT: Supports unicast filtering
1231 * @IFF_TEAM_PORT: device used as team port
1232 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1233 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1234 * change when it's running
1235 * @IFF_MACVLAN: Macvlan device
1237 enum netdev_priv_flags
{
1238 IFF_802_1Q_VLAN
= 1<<0,
1240 IFF_SLAVE_INACTIVE
= 1<<2,
1241 IFF_MASTER_8023AD
= 1<<3,
1242 IFF_MASTER_ALB
= 1<<4,
1244 IFF_SLAVE_NEEDARP
= 1<<6,
1246 IFF_MASTER_ARPMON
= 1<<8,
1247 IFF_WAN_HDLC
= 1<<9,
1248 IFF_XMIT_DST_RELEASE
= 1<<10,
1249 IFF_DONT_BRIDGE
= 1<<11,
1250 IFF_DISABLE_NETPOLL
= 1<<12,
1251 IFF_MACVLAN_PORT
= 1<<13,
1252 IFF_BRIDGE_PORT
= 1<<14,
1253 IFF_OVS_DATAPATH
= 1<<15,
1254 IFF_TX_SKB_SHARING
= 1<<16,
1255 IFF_UNICAST_FLT
= 1<<17,
1256 IFF_TEAM_PORT
= 1<<18,
1257 IFF_SUPP_NOFCS
= 1<<19,
1258 IFF_LIVE_ADDR_CHANGE
= 1<<20,
1259 IFF_MACVLAN
= 1<<21,
1260 IFF_XMIT_DST_RELEASE_PERM
= 1<<22,
1261 IFF_IPVLAN_MASTER
= 1<<23,
1262 IFF_IPVLAN_SLAVE
= 1<<24,
1265 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1266 #define IFF_EBRIDGE IFF_EBRIDGE
1267 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1268 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1269 #define IFF_MASTER_ALB IFF_MASTER_ALB
1270 #define IFF_BONDING IFF_BONDING
1271 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1272 #define IFF_ISATAP IFF_ISATAP
1273 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1274 #define IFF_WAN_HDLC IFF_WAN_HDLC
1275 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1276 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1277 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1278 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1279 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1280 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1281 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1282 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1283 #define IFF_TEAM_PORT IFF_TEAM_PORT
1284 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1285 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1286 #define IFF_MACVLAN IFF_MACVLAN
1287 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1288 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1289 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1292 * struct net_device - The DEVICE structure.
1293 * Actually, this whole structure is a big mistake. It mixes I/O
1294 * data with strictly "high-level" data, and it has to know about
1295 * almost every data structure used in the INET module.
1297 * @name: This is the first field of the "visible" part of this structure
1298 * (i.e. as seen by users in the "Space.c" file). It is the name
1301 * @name_hlist: Device name hash chain, please keep it close to name[]
1302 * @ifalias: SNMP alias
1303 * @mem_end: Shared memory end
1304 * @mem_start: Shared memory start
1305 * @base_addr: Device I/O address
1306 * @irq: Device IRQ number
1308 * @state: Generic network queuing layer state, see netdev_state_t
1309 * @dev_list: The global list of network devices
1310 * @napi_list: List entry, that is used for polling napi devices
1311 * @unreg_list: List entry, that is used, when we are unregistering the
1312 * device, see the function unregister_netdev
1313 * @close_list: List entry, that is used, when we are closing the device
1315 * @adj_list: Directly linked devices, like slaves for bonding
1316 * @all_adj_list: All linked devices, *including* neighbours
1317 * @features: Currently active device features
1318 * @hw_features: User-changeable features
1320 * @wanted_features: User-requested features
1321 * @vlan_features: Mask of features inheritable by VLAN devices
1323 * @hw_enc_features: Mask of features inherited by encapsulating devices
1324 * This field indicates what encapsulation
1325 * offloads the hardware is capable of doing,
1326 * and drivers will need to set them appropriately.
1328 * @mpls_features: Mask of features inheritable by MPLS
1330 * @ifindex: interface index
1331 * @group: The group, that the device belongs to
1333 * @stats: Statistics struct, which was left as a legacy, use
1334 * rtnl_link_stats64 instead
1336 * @rx_dropped: Dropped packets by core network,
1337 * do not use this in drivers
1338 * @tx_dropped: Dropped packets by core network,
1339 * do not use this in drivers
1341 * @carrier_changes: Stats to monitor carrier on<->off transitions
1343 * @wireless_handlers: List of functions to handle Wireless Extensions,
1345 * see <net/iw_handler.h> for details.
1346 * @wireless_data: Instance data managed by the core of wireless extensions
1348 * @netdev_ops: Includes several pointers to callbacks,
1349 * if one wants to override the ndo_*() functions
1350 * @ethtool_ops: Management operations
1351 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1352 * of Layer 2 headers.
1354 * @flags: Interface flags (a la BSD)
1355 * @priv_flags: Like 'flags' but invisible to userspace,
1356 * see if.h for the definitions
1357 * @gflags: Global flags ( kept as legacy )
1358 * @padded: How much padding added by alloc_netdev()
1359 * @operstate: RFC2863 operstate
1360 * @link_mode: Mapping policy to operstate
1361 * @if_port: Selectable AUI, TP, ...
1363 * @mtu: Interface MTU value
1364 * @type: Interface hardware type
1365 * @hard_header_len: Hardware header length
1367 * @needed_headroom: Extra headroom the hardware may need, but not in all
1368 * cases can this be guaranteed
1369 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1370 * cases can this be guaranteed. Some cases also use
1371 * LL_MAX_HEADER instead to allocate the skb
1373 * interface address info:
1375 * @perm_addr: Permanent hw address
1376 * @addr_assign_type: Hw address assignment type
1377 * @addr_len: Hardware address length
1378 * @neigh_priv_len; Used in neigh_alloc(),
1379 * initialized only in atm/clip.c
1380 * @dev_id: Used to differentiate devices that share
1381 * the same link layer address
1382 * @dev_port: Used to differentiate devices that share
1384 * @addr_list_lock: XXX: need comments on this one
1385 * @uc: unicast mac addresses
1386 * @mc: multicast mac addresses
1387 * @dev_addrs: list of device hw addresses
1388 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1389 * @uc_promisc: Counter, that indicates, that promiscuous mode
1390 * has been enabled due to the need to listen to
1391 * additional unicast addresses in a device that
1392 * does not implement ndo_set_rx_mode()
1393 * @promiscuity: Number of times, the NIC is told to work in
1394 * Promiscuous mode, if it becomes 0 the NIC will
1395 * exit from working in Promiscuous mode
1396 * @allmulti: Counter, enables or disables allmulticast mode
1398 * @vlan_info: VLAN info
1399 * @dsa_ptr: dsa specific data
1400 * @tipc_ptr: TIPC specific data
1401 * @atalk_ptr: AppleTalk link
1402 * @ip_ptr: IPv4 specific data
1403 * @dn_ptr: DECnet specific data
1404 * @ip6_ptr: IPv6 specific data
1405 * @ax25_ptr: AX.25 specific data
1406 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1408 * @last_rx: Time of last Rx
1409 * @dev_addr: Hw address (before bcast,
1410 * because most packets are unicast)
1412 * @_rx: Array of RX queues
1413 * @num_rx_queues: Number of RX queues
1414 * allocated at register_netdev() time
1415 * @real_num_rx_queues: Number of RX queues currently active in device
1417 * @rx_handler: handler for received packets
1418 * @rx_handler_data: XXX: need comments on this one
1419 * @ingress_queue: XXX: need comments on this one
1420 * @broadcast: hw bcast address
1422 * @_tx: Array of TX queues
1423 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1424 * @real_num_tx_queues: Number of TX queues currently active in device
1425 * @qdisc: Root qdisc from userspace point of view
1426 * @tx_queue_len: Max frames per queue allowed
1427 * @tx_global_lock: XXX: need comments on this one
1429 * @xps_maps: XXX: need comments on this one
1431 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1432 * indexed by RX queue number. Assigned by driver.
1433 * This must only be set if the ndo_rx_flow_steer
1434 * operation is defined
1436 * @trans_start: Time (in jiffies) of last Tx
1437 * @watchdog_timeo: Represents the timeout that is used by
1438 * the watchdog ( see dev_watchdog() )
1439 * @watchdog_timer: List of timers
1441 * @pcpu_refcnt: Number of references to this device
1442 * @todo_list: Delayed register/unregister
1443 * @index_hlist: Device index hash chain
1444 * @link_watch_list: XXX: need comments on this one
1446 * @reg_state: Register/unregister state machine
1447 * @dismantle: Device is going to be freed
1448 * @rtnl_link_state: This enum represents the phases of creating
1451 * @destructor: Called from unregister,
1452 * can be used to call free_netdev
1453 * @npinfo: XXX: need comments on this one
1454 * @nd_net: Network namespace this network device is inside
1456 * @ml_priv: Mid-layer private
1457 * @lstats: Loopback statistics
1458 * @tstats: Tunnel statistics
1459 * @dstats: Dummy statistics
1460 * @vstats: Virtual ethernet statistics
1465 * @dev: Class/net/name entry
1466 * @sysfs_groups: Space for optional device, statistics and wireless
1469 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1470 * @rtnl_link_ops: Rtnl_link_ops
1472 * @gso_max_size: Maximum size of generic segmentation offload
1473 * @gso_max_segs: Maximum number of segments that can be passed to the
1475 * @gso_min_segs: Minimum number of segments that can be passed to the
1478 * @dcbnl_ops: Data Center Bridging netlink ops
1479 * @num_tc: Number of traffic classes in the net device
1480 * @tc_to_txq: XXX: need comments on this one
1481 * @prio_tc_map XXX: need comments on this one
1483 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1485 * @priomap: XXX: need comments on this one
1486 * @phydev: Physical device may attach itself
1487 * for hardware timestamping
1489 * @qdisc_tx_busylock: XXX: need comments on this one
1491 * @pm_qos_req: Power Management QoS object
1493 * FIXME: cleanup struct net_device such that network protocol info
1498 char name
[IFNAMSIZ
];
1499 struct hlist_node name_hlist
;
1502 * I/O specific fields
1503 * FIXME: Merge these and struct ifmap into one
1505 unsigned long mem_end
;
1506 unsigned long mem_start
;
1507 unsigned long base_addr
;
1511 * Some hardware also needs these fields (state,dev_list,
1512 * napi_list,unreg_list,close_list) but they are not
1513 * part of the usual set specified in Space.c.
1516 unsigned long state
;
1518 struct list_head dev_list
;
1519 struct list_head napi_list
;
1520 struct list_head unreg_list
;
1521 struct list_head close_list
;
1522 struct list_head ptype_all
;
1523 struct list_head ptype_specific
;
1526 struct list_head upper
;
1527 struct list_head lower
;
1531 struct list_head upper
;
1532 struct list_head lower
;
1535 netdev_features_t features
;
1536 netdev_features_t hw_features
;
1537 netdev_features_t wanted_features
;
1538 netdev_features_t vlan_features
;
1539 netdev_features_t hw_enc_features
;
1540 netdev_features_t mpls_features
;
1545 struct net_device_stats stats
;
1547 atomic_long_t rx_dropped
;
1548 atomic_long_t tx_dropped
;
1550 atomic_t carrier_changes
;
1552 #ifdef CONFIG_WIRELESS_EXT
1553 const struct iw_handler_def
* wireless_handlers
;
1554 struct iw_public_data
* wireless_data
;
1556 const struct net_device_ops
*netdev_ops
;
1557 const struct ethtool_ops
*ethtool_ops
;
1558 #ifdef CONFIG_NET_SWITCHDEV
1559 const struct swdev_ops
*swdev_ops
;
1562 const struct header_ops
*header_ops
;
1565 unsigned int priv_flags
;
1567 unsigned short gflags
;
1568 unsigned short padded
;
1570 unsigned char operstate
;
1571 unsigned char link_mode
;
1573 unsigned char if_port
;
1577 unsigned short type
;
1578 unsigned short hard_header_len
;
1580 unsigned short needed_headroom
;
1581 unsigned short needed_tailroom
;
1583 /* Interface address info. */
1584 unsigned char perm_addr
[MAX_ADDR_LEN
];
1585 unsigned char addr_assign_type
;
1586 unsigned char addr_len
;
1587 unsigned short neigh_priv_len
;
1588 unsigned short dev_id
;
1589 unsigned short dev_port
;
1590 spinlock_t addr_list_lock
;
1591 struct netdev_hw_addr_list uc
;
1592 struct netdev_hw_addr_list mc
;
1593 struct netdev_hw_addr_list dev_addrs
;
1596 struct kset
*queues_kset
;
1599 unsigned char name_assign_type
;
1602 unsigned int promiscuity
;
1603 unsigned int allmulti
;
1606 /* Protocol specific pointers */
1608 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1609 struct vlan_info __rcu
*vlan_info
;
1611 #if IS_ENABLED(CONFIG_NET_DSA)
1612 struct dsa_switch_tree
*dsa_ptr
;
1614 #if IS_ENABLED(CONFIG_TIPC)
1615 struct tipc_bearer __rcu
*tipc_ptr
;
1618 struct in_device __rcu
*ip_ptr
;
1619 struct dn_dev __rcu
*dn_ptr
;
1620 struct inet6_dev __rcu
*ip6_ptr
;
1622 struct wireless_dev
*ieee80211_ptr
;
1623 struct wpan_dev
*ieee802154_ptr
;
1626 * Cache lines mostly used on receive path (including eth_type_trans())
1628 unsigned long last_rx
;
1630 /* Interface address info used in eth_type_trans() */
1631 unsigned char *dev_addr
;
1635 struct netdev_rx_queue
*_rx
;
1637 unsigned int num_rx_queues
;
1638 unsigned int real_num_rx_queues
;
1642 unsigned long gro_flush_timeout
;
1643 rx_handler_func_t __rcu
*rx_handler
;
1644 void __rcu
*rx_handler_data
;
1646 struct netdev_queue __rcu
*ingress_queue
;
1647 unsigned char broadcast
[MAX_ADDR_LEN
];
1651 * Cache lines mostly used on transmit path
1653 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1654 unsigned int num_tx_queues
;
1655 unsigned int real_num_tx_queues
;
1656 struct Qdisc
*qdisc
;
1657 unsigned long tx_queue_len
;
1658 spinlock_t tx_global_lock
;
1661 struct xps_dev_maps __rcu
*xps_maps
;
1663 #ifdef CONFIG_RFS_ACCEL
1664 struct cpu_rmap
*rx_cpu_rmap
;
1667 /* These may be needed for future network-power-down code. */
1670 * trans_start here is expensive for high speed devices on SMP,
1671 * please use netdev_queue->trans_start instead.
1673 unsigned long trans_start
;
1676 struct timer_list watchdog_timer
;
1678 int __percpu
*pcpu_refcnt
;
1679 struct list_head todo_list
;
1681 struct hlist_node index_hlist
;
1682 struct list_head link_watch_list
;
1684 enum { NETREG_UNINITIALIZED
=0,
1685 NETREG_REGISTERED
, /* completed register_netdevice */
1686 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1687 NETREG_UNREGISTERED
, /* completed unregister todo */
1688 NETREG_RELEASED
, /* called free_netdev */
1689 NETREG_DUMMY
, /* dummy device for NAPI poll */
1695 RTNL_LINK_INITIALIZED
,
1696 RTNL_LINK_INITIALIZING
,
1697 } rtnl_link_state
:16;
1699 void (*destructor
)(struct net_device
*dev
);
1701 #ifdef CONFIG_NETPOLL
1702 struct netpoll_info __rcu
*npinfo
;
1705 possible_net_t nd_net
;
1707 /* mid-layer private */
1710 struct pcpu_lstats __percpu
*lstats
;
1711 struct pcpu_sw_netstats __percpu
*tstats
;
1712 struct pcpu_dstats __percpu
*dstats
;
1713 struct pcpu_vstats __percpu
*vstats
;
1716 struct garp_port __rcu
*garp_port
;
1717 struct mrp_port __rcu
*mrp_port
;
1720 const struct attribute_group
*sysfs_groups
[4];
1721 const struct attribute_group
*sysfs_rx_queue_group
;
1723 const struct rtnl_link_ops
*rtnl_link_ops
;
1725 /* for setting kernel sock attribute on TCP connection setup */
1726 #define GSO_MAX_SIZE 65536
1727 unsigned int gso_max_size
;
1728 #define GSO_MAX_SEGS 65535
1732 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1735 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1736 u8 prio_tc_map
[TC_BITMASK
+ 1];
1738 #if IS_ENABLED(CONFIG_FCOE)
1739 unsigned int fcoe_ddp_xid
;
1741 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1742 struct netprio_map __rcu
*priomap
;
1744 struct phy_device
*phydev
;
1745 struct lock_class_key
*qdisc_tx_busylock
;
1746 struct pm_qos_request pm_qos_req
;
1748 #define to_net_dev(d) container_of(d, struct net_device, dev)
1750 #define NETDEV_ALIGN 32
1753 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1755 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1759 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1761 if (tc
>= dev
->num_tc
)
1764 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1769 void netdev_reset_tc(struct net_device
*dev
)
1772 memset(dev
->tc_to_txq
, 0, sizeof(dev
->tc_to_txq
));
1773 memset(dev
->prio_tc_map
, 0, sizeof(dev
->prio_tc_map
));
1777 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
)
1779 if (tc
>= dev
->num_tc
)
1782 dev
->tc_to_txq
[tc
].count
= count
;
1783 dev
->tc_to_txq
[tc
].offset
= offset
;
1788 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
)
1790 if (num_tc
> TC_MAX_QUEUE
)
1793 dev
->num_tc
= num_tc
;
1798 int netdev_get_num_tc(struct net_device
*dev
)
1804 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1807 return &dev
->_tx
[index
];
1810 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1811 const struct sk_buff
*skb
)
1813 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1816 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1817 void (*f
)(struct net_device
*,
1818 struct netdev_queue
*,
1824 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1825 f(dev
, &dev
->_tx
[i
], arg
);
1828 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1829 struct sk_buff
*skb
,
1833 * Net namespace inlines
1836 struct net
*dev_net(const struct net_device
*dev
)
1838 return read_pnet(&dev
->nd_net
);
1842 void dev_net_set(struct net_device
*dev
, struct net
*net
)
1844 write_pnet(&dev
->nd_net
, net
);
1847 static inline bool netdev_uses_dsa(struct net_device
*dev
)
1849 #if IS_ENABLED(CONFIG_NET_DSA)
1850 if (dev
->dsa_ptr
!= NULL
)
1851 return dsa_uses_tagged_protocol(dev
->dsa_ptr
);
1857 * netdev_priv - access network device private data
1858 * @dev: network device
1860 * Get network device private data
1862 static inline void *netdev_priv(const struct net_device
*dev
)
1864 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
1867 /* Set the sysfs physical device reference for the network logical device
1868 * if set prior to registration will cause a symlink during initialization.
1870 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1872 /* Set the sysfs device type for the network logical device to allow
1873 * fine-grained identification of different network device types. For
1874 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1876 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1878 /* Default NAPI poll() weight
1879 * Device drivers are strongly advised to not use bigger value
1881 #define NAPI_POLL_WEIGHT 64
1884 * netif_napi_add - initialize a napi context
1885 * @dev: network device
1886 * @napi: napi context
1887 * @poll: polling function
1888 * @weight: default weight
1890 * netif_napi_add() must be used to initialize a napi context prior to calling
1891 * *any* of the other napi related functions.
1893 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
1894 int (*poll
)(struct napi_struct
*, int), int weight
);
1897 * netif_napi_del - remove a napi context
1898 * @napi: napi context
1900 * netif_napi_del() removes a napi context from the network device napi list
1902 void netif_napi_del(struct napi_struct
*napi
);
1904 struct napi_gro_cb
{
1905 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1908 /* Length of frag0. */
1909 unsigned int frag0_len
;
1911 /* This indicates where we are processing relative to skb->data. */
1914 /* This is non-zero if the packet cannot be merged with the new skb. */
1917 /* Save the IP ID here and check when we get to the transport layer */
1920 /* Number of segments aggregated. */
1923 /* Start offset for remote checksum offload */
1924 u16 gro_remcsum_start
;
1926 /* jiffies when first packet was created/queued */
1929 /* Used in ipv6_gro_receive() and foo-over-udp */
1932 /* This is non-zero if the packet may be of the same flow. */
1935 /* Used in udp_gro_receive */
1938 /* GRO checksum is valid */
1941 /* Number of checksums via CHECKSUM_UNNECESSARY */
1946 #define NAPI_GRO_FREE 1
1947 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1949 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1954 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1957 /* used in skb_gro_receive() slow path */
1958 struct sk_buff
*last
;
1961 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1963 struct packet_type
{
1964 __be16 type
; /* This is really htons(ether_type). */
1965 struct net_device
*dev
; /* NULL is wildcarded here */
1966 int (*func
) (struct sk_buff
*,
1967 struct net_device
*,
1968 struct packet_type
*,
1969 struct net_device
*);
1970 bool (*id_match
)(struct packet_type
*ptype
,
1972 void *af_packet_priv
;
1973 struct list_head list
;
1976 struct offload_callbacks
{
1977 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
1978 netdev_features_t features
);
1979 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1980 struct sk_buff
*skb
);
1981 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
1984 struct packet_offload
{
1985 __be16 type
; /* This is really htons(ether_type). */
1986 struct offload_callbacks callbacks
;
1987 struct list_head list
;
1992 struct udp_offload_callbacks
{
1993 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1994 struct sk_buff
*skb
,
1995 struct udp_offload
*uoff
);
1996 int (*gro_complete
)(struct sk_buff
*skb
,
1998 struct udp_offload
*uoff
);
2001 struct udp_offload
{
2004 struct udp_offload_callbacks callbacks
;
2007 /* often modified stats are per cpu, other are shared (netdev->stats) */
2008 struct pcpu_sw_netstats
{
2013 struct u64_stats_sync syncp
;
2016 #define netdev_alloc_pcpu_stats(type) \
2018 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2021 for_each_possible_cpu(i) { \
2022 typeof(type) *stat; \
2023 stat = per_cpu_ptr(pcpu_stats, i); \
2024 u64_stats_init(&stat->syncp); \
2030 #include <linux/notifier.h>
2032 /* netdevice notifier chain. Please remember to update the rtnetlink
2033 * notification exclusion list in rtnetlink_event() when adding new
2036 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2037 #define NETDEV_DOWN 0x0002
2038 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2039 detected a hardware crash and restarted
2040 - we can use this eg to kick tcp sessions
2042 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2043 #define NETDEV_REGISTER 0x0005
2044 #define NETDEV_UNREGISTER 0x0006
2045 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2046 #define NETDEV_CHANGEADDR 0x0008
2047 #define NETDEV_GOING_DOWN 0x0009
2048 #define NETDEV_CHANGENAME 0x000A
2049 #define NETDEV_FEAT_CHANGE 0x000B
2050 #define NETDEV_BONDING_FAILOVER 0x000C
2051 #define NETDEV_PRE_UP 0x000D
2052 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2053 #define NETDEV_POST_TYPE_CHANGE 0x000F
2054 #define NETDEV_POST_INIT 0x0010
2055 #define NETDEV_UNREGISTER_FINAL 0x0011
2056 #define NETDEV_RELEASE 0x0012
2057 #define NETDEV_NOTIFY_PEERS 0x0013
2058 #define NETDEV_JOIN 0x0014
2059 #define NETDEV_CHANGEUPPER 0x0015
2060 #define NETDEV_RESEND_IGMP 0x0016
2061 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2062 #define NETDEV_CHANGEINFODATA 0x0018
2063 #define NETDEV_BONDING_INFO 0x0019
2065 int register_netdevice_notifier(struct notifier_block
*nb
);
2066 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2068 struct netdev_notifier_info
{
2069 struct net_device
*dev
;
2072 struct netdev_notifier_change_info
{
2073 struct netdev_notifier_info info
; /* must be first */
2074 unsigned int flags_changed
;
2077 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2078 struct net_device
*dev
)
2083 static inline struct net_device
*
2084 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2089 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2092 extern rwlock_t dev_base_lock
; /* Device list lock */
2094 #define for_each_netdev(net, d) \
2095 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2096 #define for_each_netdev_reverse(net, d) \
2097 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2098 #define for_each_netdev_rcu(net, d) \
2099 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2100 #define for_each_netdev_safe(net, d, n) \
2101 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2102 #define for_each_netdev_continue(net, d) \
2103 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2104 #define for_each_netdev_continue_rcu(net, d) \
2105 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2106 #define for_each_netdev_in_bond_rcu(bond, slave) \
2107 for_each_netdev_rcu(&init_net, slave) \
2108 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2109 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2111 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2113 struct list_head
*lh
;
2117 lh
= dev
->dev_list
.next
;
2118 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2121 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2123 struct list_head
*lh
;
2127 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2128 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2131 static inline struct net_device
*first_net_device(struct net
*net
)
2133 return list_empty(&net
->dev_base_head
) ? NULL
:
2134 net_device_entry(net
->dev_base_head
.next
);
2137 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2139 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2141 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2144 int netdev_boot_setup_check(struct net_device
*dev
);
2145 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2146 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2147 const char *hwaddr
);
2148 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2149 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2150 void dev_add_pack(struct packet_type
*pt
);
2151 void dev_remove_pack(struct packet_type
*pt
);
2152 void __dev_remove_pack(struct packet_type
*pt
);
2153 void dev_add_offload(struct packet_offload
*po
);
2154 void dev_remove_offload(struct packet_offload
*po
);
2156 int dev_get_iflink(const struct net_device
*dev
);
2157 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2158 unsigned short mask
);
2159 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2160 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2161 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2162 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2163 int dev_open(struct net_device
*dev
);
2164 int dev_close(struct net_device
*dev
);
2165 int dev_close_many(struct list_head
*head
, bool unlink
);
2166 void dev_disable_lro(struct net_device
*dev
);
2167 int dev_loopback_xmit(struct sock
*sk
, struct sk_buff
*newskb
);
2168 int dev_queue_xmit_sk(struct sock
*sk
, struct sk_buff
*skb
);
2169 static inline int dev_queue_xmit(struct sk_buff
*skb
)
2171 return dev_queue_xmit_sk(skb
->sk
, skb
);
2173 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2174 int register_netdevice(struct net_device
*dev
);
2175 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2176 void unregister_netdevice_many(struct list_head
*head
);
2177 static inline void unregister_netdevice(struct net_device
*dev
)
2179 unregister_netdevice_queue(dev
, NULL
);
2182 int netdev_refcnt_read(const struct net_device
*dev
);
2183 void free_netdev(struct net_device
*dev
);
2184 void netdev_freemem(struct net_device
*dev
);
2185 void synchronize_net(void);
2186 int init_dummy_netdev(struct net_device
*dev
);
2188 DECLARE_PER_CPU(int, xmit_recursion
);
2189 static inline int dev_recursion_level(void)
2191 return this_cpu_read(xmit_recursion
);
2194 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2195 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2196 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2197 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2198 int dev_restart(struct net_device
*dev
);
2199 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2201 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2203 return NAPI_GRO_CB(skb
)->data_offset
;
2206 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2208 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2211 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2213 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2216 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2217 unsigned int offset
)
2219 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2222 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2224 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2227 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2228 unsigned int offset
)
2230 if (!pskb_may_pull(skb
, hlen
))
2233 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2234 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2235 return skb
->data
+ offset
;
2238 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2240 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2241 skb_network_offset(skb
);
2244 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2245 const void *start
, unsigned int len
)
2247 if (NAPI_GRO_CB(skb
)->csum_valid
)
2248 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2249 csum_partial(start
, len
, 0));
2252 /* GRO checksum functions. These are logical equivalents of the normal
2253 * checksum functions (in skbuff.h) except that they operate on the GRO
2254 * offsets and fields in sk_buff.
2257 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2259 static inline bool skb_at_gro_remcsum_start(struct sk_buff
*skb
)
2261 return (NAPI_GRO_CB(skb
)->gro_remcsum_start
- skb_headroom(skb
) ==
2262 skb_gro_offset(skb
));
2265 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2269 return ((skb
->ip_summed
!= CHECKSUM_PARTIAL
||
2270 skb_checksum_start_offset(skb
) <
2271 skb_gro_offset(skb
)) &&
2272 !skb_at_gro_remcsum_start(skb
) &&
2273 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2274 (!zero_okay
|| check
));
2277 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2280 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2281 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2284 NAPI_GRO_CB(skb
)->csum
= psum
;
2286 return __skb_gro_checksum_complete(skb
);
2289 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2291 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2292 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2293 NAPI_GRO_CB(skb
)->csum_cnt
--;
2295 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2296 * verified a new top level checksum or an encapsulated one
2297 * during GRO. This saves work if we fallback to normal path.
2299 __skb_incr_checksum_unnecessary(skb
);
2303 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2306 __sum16 __ret = 0; \
2307 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2308 __ret = __skb_gro_checksum_validate_complete(skb, \
2309 compute_pseudo(skb, proto)); \
2311 __skb_mark_checksum_bad(skb); \
2313 skb_gro_incr_csum_unnecessary(skb); \
2317 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2318 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2320 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2322 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2324 #define skb_gro_checksum_simple_validate(skb) \
2325 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2327 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2329 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2330 !NAPI_GRO_CB(skb
)->csum_valid
);
2333 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2334 __sum16 check
, __wsum pseudo
)
2336 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2337 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2340 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2342 if (__skb_gro_checksum_convert_check(skb)) \
2343 __skb_gro_checksum_convert(skb, check, \
2344 compute_pseudo(skb, proto)); \
2347 struct gro_remcsum
{
2352 static inline void skb_gro_remcsum_init(struct gro_remcsum
*grc
)
2358 static inline void skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2359 int start
, int offset
,
2360 struct gro_remcsum
*grc
,
2365 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2368 NAPI_GRO_CB(skb
)->gro_remcsum_start
=
2369 ((unsigned char *)ptr
+ start
) - skb
->head
;
2373 delta
= remcsum_adjust(ptr
, NAPI_GRO_CB(skb
)->csum
, start
, offset
);
2375 /* Adjust skb->csum since we changed the packet */
2376 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2378 grc
->offset
= (ptr
+ offset
) - (void *)skb
->head
;
2382 static inline void skb_gro_remcsum_cleanup(struct sk_buff
*skb
,
2383 struct gro_remcsum
*grc
)
2388 remcsum_unadjust((__sum16
*)(skb
->head
+ grc
->offset
), grc
->delta
);
2391 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2392 unsigned short type
,
2393 const void *daddr
, const void *saddr
,
2396 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2399 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2402 static inline int dev_parse_header(const struct sk_buff
*skb
,
2403 unsigned char *haddr
)
2405 const struct net_device
*dev
= skb
->dev
;
2407 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2409 return dev
->header_ops
->parse(skb
, haddr
);
2412 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2413 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2414 static inline int unregister_gifconf(unsigned int family
)
2416 return register_gifconf(family
, NULL
);
2419 #ifdef CONFIG_NET_FLOW_LIMIT
2420 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2421 struct sd_flow_limit
{
2423 unsigned int num_buckets
;
2424 unsigned int history_head
;
2425 u16 history
[FLOW_LIMIT_HISTORY
];
2429 extern int netdev_flow_limit_table_len
;
2430 #endif /* CONFIG_NET_FLOW_LIMIT */
2433 * Incoming packets are placed on per-cpu queues
2435 struct softnet_data
{
2436 struct list_head poll_list
;
2437 struct sk_buff_head process_queue
;
2440 unsigned int processed
;
2441 unsigned int time_squeeze
;
2442 unsigned int cpu_collision
;
2443 unsigned int received_rps
;
2445 struct softnet_data
*rps_ipi_list
;
2447 #ifdef CONFIG_NET_FLOW_LIMIT
2448 struct sd_flow_limit __rcu
*flow_limit
;
2450 struct Qdisc
*output_queue
;
2451 struct Qdisc
**output_queue_tailp
;
2452 struct sk_buff
*completion_queue
;
2455 /* Elements below can be accessed between CPUs for RPS */
2456 struct call_single_data csd ____cacheline_aligned_in_smp
;
2457 struct softnet_data
*rps_ipi_next
;
2459 unsigned int input_queue_head
;
2460 unsigned int input_queue_tail
;
2462 unsigned int dropped
;
2463 struct sk_buff_head input_pkt_queue
;
2464 struct napi_struct backlog
;
2468 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2471 sd
->input_queue_head
++;
2475 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2476 unsigned int *qtail
)
2479 *qtail
= ++sd
->input_queue_tail
;
2483 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2485 void __netif_schedule(struct Qdisc
*q
);
2486 void netif_schedule_queue(struct netdev_queue
*txq
);
2488 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2492 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2493 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2496 static inline void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2498 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2502 * netif_start_queue - allow transmit
2503 * @dev: network device
2505 * Allow upper layers to call the device hard_start_xmit routine.
2507 static inline void netif_start_queue(struct net_device
*dev
)
2509 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2512 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2516 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2517 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2518 netif_tx_start_queue(txq
);
2522 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2525 * netif_wake_queue - restart transmit
2526 * @dev: network device
2528 * Allow upper layers to call the device hard_start_xmit routine.
2529 * Used for flow control when transmit resources are available.
2531 static inline void netif_wake_queue(struct net_device
*dev
)
2533 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2536 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2540 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2541 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2542 netif_tx_wake_queue(txq
);
2546 static inline void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2548 if (WARN_ON(!dev_queue
)) {
2549 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2552 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2556 * netif_stop_queue - stop transmitted packets
2557 * @dev: network device
2559 * Stop upper layers calling the device hard_start_xmit routine.
2560 * Used for flow control when transmit resources are unavailable.
2562 static inline void netif_stop_queue(struct net_device
*dev
)
2564 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2567 static inline void netif_tx_stop_all_queues(struct net_device
*dev
)
2571 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2572 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2573 netif_tx_stop_queue(txq
);
2577 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2579 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2583 * netif_queue_stopped - test if transmit queue is flowblocked
2584 * @dev: network device
2586 * Test if transmit queue on device is currently unable to send.
2588 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2590 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2593 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2595 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2599 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2601 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2605 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2607 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2611 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2612 * @dev_queue: pointer to transmit queue
2614 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2615 * to give appropriate hint to the cpu.
2617 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2620 prefetchw(&dev_queue
->dql
.num_queued
);
2625 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2626 * @dev_queue: pointer to transmit queue
2628 * BQL enabled drivers might use this helper in their TX completion path,
2629 * to give appropriate hint to the cpu.
2631 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2634 prefetchw(&dev_queue
->dql
.limit
);
2638 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2642 dql_queued(&dev_queue
->dql
, bytes
);
2644 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2647 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2650 * The XOFF flag must be set before checking the dql_avail below,
2651 * because in netdev_tx_completed_queue we update the dql_completed
2652 * before checking the XOFF flag.
2656 /* check again in case another CPU has just made room avail */
2657 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2658 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2663 * netdev_sent_queue - report the number of bytes queued to hardware
2664 * @dev: network device
2665 * @bytes: number of bytes queued to the hardware device queue
2667 * Report the number of bytes queued for sending/completion to the network
2668 * device hardware queue. @bytes should be a good approximation and should
2669 * exactly match netdev_completed_queue() @bytes
2671 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2673 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2676 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2677 unsigned int pkts
, unsigned int bytes
)
2680 if (unlikely(!bytes
))
2683 dql_completed(&dev_queue
->dql
, bytes
);
2686 * Without the memory barrier there is a small possiblity that
2687 * netdev_tx_sent_queue will miss the update and cause the queue to
2688 * be stopped forever
2692 if (dql_avail(&dev_queue
->dql
) < 0)
2695 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2696 netif_schedule_queue(dev_queue
);
2701 * netdev_completed_queue - report bytes and packets completed by device
2702 * @dev: network device
2703 * @pkts: actual number of packets sent over the medium
2704 * @bytes: actual number of bytes sent over the medium
2706 * Report the number of bytes and packets transmitted by the network device
2707 * hardware queue over the physical medium, @bytes must exactly match the
2708 * @bytes amount passed to netdev_sent_queue()
2710 static inline void netdev_completed_queue(struct net_device
*dev
,
2711 unsigned int pkts
, unsigned int bytes
)
2713 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
2716 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
2719 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
2725 * netdev_reset_queue - reset the packets and bytes count of a network device
2726 * @dev_queue: network device
2728 * Reset the bytes and packet count of a network device and clear the
2729 * software flow control OFF bit for this network device
2731 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
2733 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
2737 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2738 * @dev: network device
2739 * @queue_index: given tx queue index
2741 * Returns 0 if given tx queue index >= number of device tx queues,
2742 * otherwise returns the originally passed tx queue index.
2744 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
2746 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
2747 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2748 dev
->name
, queue_index
,
2749 dev
->real_num_tx_queues
);
2757 * netif_running - test if up
2758 * @dev: network device
2760 * Test if the device has been brought up.
2762 static inline bool netif_running(const struct net_device
*dev
)
2764 return test_bit(__LINK_STATE_START
, &dev
->state
);
2768 * Routines to manage the subqueues on a device. We only need start
2769 * stop, and a check if it's stopped. All other device management is
2770 * done at the overall netdevice level.
2771 * Also test the device if we're multiqueue.
2775 * netif_start_subqueue - allow sending packets on subqueue
2776 * @dev: network device
2777 * @queue_index: sub queue index
2779 * Start individual transmit queue of a device with multiple transmit queues.
2781 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
2783 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2785 netif_tx_start_queue(txq
);
2789 * netif_stop_subqueue - stop sending packets on subqueue
2790 * @dev: network device
2791 * @queue_index: sub queue index
2793 * Stop individual transmit queue of a device with multiple transmit queues.
2795 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
2797 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2798 netif_tx_stop_queue(txq
);
2802 * netif_subqueue_stopped - test status of subqueue
2803 * @dev: network device
2804 * @queue_index: sub queue index
2806 * Check individual transmit queue of a device with multiple transmit queues.
2808 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
2811 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2813 return netif_tx_queue_stopped(txq
);
2816 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
2817 struct sk_buff
*skb
)
2819 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
2822 void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
);
2825 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
2828 static inline int netif_set_xps_queue(struct net_device
*dev
,
2829 const struct cpumask
*mask
,
2837 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2838 * as a distribution range limit for the returned value.
2840 static inline u16
skb_tx_hash(const struct net_device
*dev
,
2841 struct sk_buff
*skb
)
2843 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
2847 * netif_is_multiqueue - test if device has multiple transmit queues
2848 * @dev: network device
2850 * Check if device has multiple transmit queues
2852 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
2854 return dev
->num_tx_queues
> 1;
2857 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
2860 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
2862 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
2870 static inline unsigned int get_netdev_rx_queue_index(
2871 struct netdev_rx_queue
*queue
)
2873 struct net_device
*dev
= queue
->dev
;
2874 int index
= queue
- dev
->_rx
;
2876 BUG_ON(index
>= dev
->num_rx_queues
);
2881 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2882 int netif_get_num_default_rss_queues(void);
2884 enum skb_free_reason
{
2885 SKB_REASON_CONSUMED
,
2889 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
2890 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
2893 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2894 * interrupt context or with hardware interrupts being disabled.
2895 * (in_irq() || irqs_disabled())
2897 * We provide four helpers that can be used in following contexts :
2899 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2900 * replacing kfree_skb(skb)
2902 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2903 * Typically used in place of consume_skb(skb) in TX completion path
2905 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2906 * replacing kfree_skb(skb)
2908 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2909 * and consumed a packet. Used in place of consume_skb(skb)
2911 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
2913 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
2916 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
2918 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
2921 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
2923 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
2926 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
2928 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
2931 int netif_rx(struct sk_buff
*skb
);
2932 int netif_rx_ni(struct sk_buff
*skb
);
2933 int netif_receive_skb_sk(struct sock
*sk
, struct sk_buff
*skb
);
2934 static inline int netif_receive_skb(struct sk_buff
*skb
)
2936 return netif_receive_skb_sk(skb
->sk
, skb
);
2938 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
2939 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
2940 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
2941 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
2942 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
2943 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
2945 static inline void napi_free_frags(struct napi_struct
*napi
)
2947 kfree_skb(napi
->skb
);
2951 int netdev_rx_handler_register(struct net_device
*dev
,
2952 rx_handler_func_t
*rx_handler
,
2953 void *rx_handler_data
);
2954 void netdev_rx_handler_unregister(struct net_device
*dev
);
2956 bool dev_valid_name(const char *name
);
2957 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
2958 int dev_ethtool(struct net
*net
, struct ifreq
*);
2959 unsigned int dev_get_flags(const struct net_device
*);
2960 int __dev_change_flags(struct net_device
*, unsigned int flags
);
2961 int dev_change_flags(struct net_device
*, unsigned int);
2962 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
2963 unsigned int gchanges
);
2964 int dev_change_name(struct net_device
*, const char *);
2965 int dev_set_alias(struct net_device
*, const char *, size_t);
2966 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
2967 int dev_set_mtu(struct net_device
*, int);
2968 void dev_set_group(struct net_device
*, int);
2969 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
2970 int dev_change_carrier(struct net_device
*, bool new_carrier
);
2971 int dev_get_phys_port_id(struct net_device
*dev
,
2972 struct netdev_phys_item_id
*ppid
);
2973 int dev_get_phys_port_name(struct net_device
*dev
,
2974 char *name
, size_t len
);
2975 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
2976 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
2977 struct netdev_queue
*txq
, int *ret
);
2978 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2979 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2980 bool is_skb_forwardable(struct net_device
*dev
, struct sk_buff
*skb
);
2982 extern int netdev_budget
;
2984 /* Called by rtnetlink.c:rtnl_unlock() */
2985 void netdev_run_todo(void);
2988 * dev_put - release reference to device
2989 * @dev: network device
2991 * Release reference to device to allow it to be freed.
2993 static inline void dev_put(struct net_device
*dev
)
2995 this_cpu_dec(*dev
->pcpu_refcnt
);
2999 * dev_hold - get reference to device
3000 * @dev: network device
3002 * Hold reference to device to keep it from being freed.
3004 static inline void dev_hold(struct net_device
*dev
)
3006 this_cpu_inc(*dev
->pcpu_refcnt
);
3009 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3010 * and _off may be called from IRQ context, but it is caller
3011 * who is responsible for serialization of these calls.
3013 * The name carrier is inappropriate, these functions should really be
3014 * called netif_lowerlayer_*() because they represent the state of any
3015 * kind of lower layer not just hardware media.
3018 void linkwatch_init_dev(struct net_device
*dev
);
3019 void linkwatch_fire_event(struct net_device
*dev
);
3020 void linkwatch_forget_dev(struct net_device
*dev
);
3023 * netif_carrier_ok - test if carrier present
3024 * @dev: network device
3026 * Check if carrier is present on device
3028 static inline bool netif_carrier_ok(const struct net_device
*dev
)
3030 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
3033 unsigned long dev_trans_start(struct net_device
*dev
);
3035 void __netdev_watchdog_up(struct net_device
*dev
);
3037 void netif_carrier_on(struct net_device
*dev
);
3039 void netif_carrier_off(struct net_device
*dev
);
3042 * netif_dormant_on - mark device as dormant.
3043 * @dev: network device
3045 * Mark device as dormant (as per RFC2863).
3047 * The dormant state indicates that the relevant interface is not
3048 * actually in a condition to pass packets (i.e., it is not 'up') but is
3049 * in a "pending" state, waiting for some external event. For "on-
3050 * demand" interfaces, this new state identifies the situation where the
3051 * interface is waiting for events to place it in the up state.
3054 static inline void netif_dormant_on(struct net_device
*dev
)
3056 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3057 linkwatch_fire_event(dev
);
3061 * netif_dormant_off - set device as not dormant.
3062 * @dev: network device
3064 * Device is not in dormant state.
3066 static inline void netif_dormant_off(struct net_device
*dev
)
3068 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3069 linkwatch_fire_event(dev
);
3073 * netif_dormant - test if carrier present
3074 * @dev: network device
3076 * Check if carrier is present on device
3078 static inline bool netif_dormant(const struct net_device
*dev
)
3080 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3085 * netif_oper_up - test if device is operational
3086 * @dev: network device
3088 * Check if carrier is operational
3090 static inline bool netif_oper_up(const struct net_device
*dev
)
3092 return (dev
->operstate
== IF_OPER_UP
||
3093 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3097 * netif_device_present - is device available or removed
3098 * @dev: network device
3100 * Check if device has not been removed from system.
3102 static inline bool netif_device_present(struct net_device
*dev
)
3104 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3107 void netif_device_detach(struct net_device
*dev
);
3109 void netif_device_attach(struct net_device
*dev
);
3112 * Network interface message level settings
3116 NETIF_MSG_DRV
= 0x0001,
3117 NETIF_MSG_PROBE
= 0x0002,
3118 NETIF_MSG_LINK
= 0x0004,
3119 NETIF_MSG_TIMER
= 0x0008,
3120 NETIF_MSG_IFDOWN
= 0x0010,
3121 NETIF_MSG_IFUP
= 0x0020,
3122 NETIF_MSG_RX_ERR
= 0x0040,
3123 NETIF_MSG_TX_ERR
= 0x0080,
3124 NETIF_MSG_TX_QUEUED
= 0x0100,
3125 NETIF_MSG_INTR
= 0x0200,
3126 NETIF_MSG_TX_DONE
= 0x0400,
3127 NETIF_MSG_RX_STATUS
= 0x0800,
3128 NETIF_MSG_PKTDATA
= 0x1000,
3129 NETIF_MSG_HW
= 0x2000,
3130 NETIF_MSG_WOL
= 0x4000,
3133 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3134 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3135 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3136 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3137 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3138 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3139 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3140 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3141 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3142 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3143 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3144 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3145 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3146 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3147 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3149 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3152 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3153 return default_msg_enable_bits
;
3154 if (debug_value
== 0) /* no output */
3156 /* set low N bits */
3157 return (1 << debug_value
) - 1;
3160 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3162 spin_lock(&txq
->_xmit_lock
);
3163 txq
->xmit_lock_owner
= cpu
;
3166 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3168 spin_lock_bh(&txq
->_xmit_lock
);
3169 txq
->xmit_lock_owner
= smp_processor_id();
3172 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3174 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3176 txq
->xmit_lock_owner
= smp_processor_id();
3180 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3182 txq
->xmit_lock_owner
= -1;
3183 spin_unlock(&txq
->_xmit_lock
);
3186 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3188 txq
->xmit_lock_owner
= -1;
3189 spin_unlock_bh(&txq
->_xmit_lock
);
3192 static inline void txq_trans_update(struct netdev_queue
*txq
)
3194 if (txq
->xmit_lock_owner
!= -1)
3195 txq
->trans_start
= jiffies
;
3199 * netif_tx_lock - grab network device transmit lock
3200 * @dev: network device
3202 * Get network device transmit lock
3204 static inline void netif_tx_lock(struct net_device
*dev
)
3209 spin_lock(&dev
->tx_global_lock
);
3210 cpu
= smp_processor_id();
3211 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3212 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3214 /* We are the only thread of execution doing a
3215 * freeze, but we have to grab the _xmit_lock in
3216 * order to synchronize with threads which are in
3217 * the ->hard_start_xmit() handler and already
3218 * checked the frozen bit.
3220 __netif_tx_lock(txq
, cpu
);
3221 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3222 __netif_tx_unlock(txq
);
3226 static inline void netif_tx_lock_bh(struct net_device
*dev
)
3232 static inline void netif_tx_unlock(struct net_device
*dev
)
3236 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3237 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3239 /* No need to grab the _xmit_lock here. If the
3240 * queue is not stopped for another reason, we
3243 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3244 netif_schedule_queue(txq
);
3246 spin_unlock(&dev
->tx_global_lock
);
3249 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
3251 netif_tx_unlock(dev
);
3255 #define HARD_TX_LOCK(dev, txq, cpu) { \
3256 if ((dev->features & NETIF_F_LLTX) == 0) { \
3257 __netif_tx_lock(txq, cpu); \
3261 #define HARD_TX_TRYLOCK(dev, txq) \
3262 (((dev->features & NETIF_F_LLTX) == 0) ? \
3263 __netif_tx_trylock(txq) : \
3266 #define HARD_TX_UNLOCK(dev, txq) { \
3267 if ((dev->features & NETIF_F_LLTX) == 0) { \
3268 __netif_tx_unlock(txq); \
3272 static inline void netif_tx_disable(struct net_device
*dev
)
3278 cpu
= smp_processor_id();
3279 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3280 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3282 __netif_tx_lock(txq
, cpu
);
3283 netif_tx_stop_queue(txq
);
3284 __netif_tx_unlock(txq
);
3289 static inline void netif_addr_lock(struct net_device
*dev
)
3291 spin_lock(&dev
->addr_list_lock
);
3294 static inline void netif_addr_lock_nested(struct net_device
*dev
)
3296 int subclass
= SINGLE_DEPTH_NESTING
;
3298 if (dev
->netdev_ops
->ndo_get_lock_subclass
)
3299 subclass
= dev
->netdev_ops
->ndo_get_lock_subclass(dev
);
3301 spin_lock_nested(&dev
->addr_list_lock
, subclass
);
3304 static inline void netif_addr_lock_bh(struct net_device
*dev
)
3306 spin_lock_bh(&dev
->addr_list_lock
);
3309 static inline void netif_addr_unlock(struct net_device
*dev
)
3311 spin_unlock(&dev
->addr_list_lock
);
3314 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
3316 spin_unlock_bh(&dev
->addr_list_lock
);
3320 * dev_addrs walker. Should be used only for read access. Call with
3321 * rcu_read_lock held.
3323 #define for_each_dev_addr(dev, ha) \
3324 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3326 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3328 void ether_setup(struct net_device
*dev
);
3330 /* Support for loadable net-drivers */
3331 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
3332 unsigned char name_assign_type
,
3333 void (*setup
)(struct net_device
*),
3334 unsigned int txqs
, unsigned int rxqs
);
3335 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3336 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3338 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3339 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3342 int register_netdev(struct net_device
*dev
);
3343 void unregister_netdev(struct net_device
*dev
);
3345 /* General hardware address lists handling functions */
3346 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
3347 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3348 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
3349 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3350 int __hw_addr_sync_dev(struct netdev_hw_addr_list
*list
,
3351 struct net_device
*dev
,
3352 int (*sync
)(struct net_device
*, const unsigned char *),
3353 int (*unsync
)(struct net_device
*,
3354 const unsigned char *));
3355 void __hw_addr_unsync_dev(struct netdev_hw_addr_list
*list
,
3356 struct net_device
*dev
,
3357 int (*unsync
)(struct net_device
*,
3358 const unsigned char *));
3359 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
3361 /* Functions used for device addresses handling */
3362 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
3363 unsigned char addr_type
);
3364 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
3365 unsigned char addr_type
);
3366 void dev_addr_flush(struct net_device
*dev
);
3367 int dev_addr_init(struct net_device
*dev
);
3369 /* Functions used for unicast addresses handling */
3370 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
3371 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3372 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
3373 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
3374 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3375 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
3376 void dev_uc_flush(struct net_device
*dev
);
3377 void dev_uc_init(struct net_device
*dev
);
3380 * __dev_uc_sync - Synchonize device's unicast list
3381 * @dev: device to sync
3382 * @sync: function to call if address should be added
3383 * @unsync: function to call if address should be removed
3385 * Add newly added addresses to the interface, and release
3386 * addresses that have been deleted.
3388 static inline int __dev_uc_sync(struct net_device
*dev
,
3389 int (*sync
)(struct net_device
*,
3390 const unsigned char *),
3391 int (*unsync
)(struct net_device
*,
3392 const unsigned char *))
3394 return __hw_addr_sync_dev(&dev
->uc
, dev
, sync
, unsync
);
3398 * __dev_uc_unsync - Remove synchronized addresses from device
3399 * @dev: device to sync
3400 * @unsync: function to call if address should be removed
3402 * Remove all addresses that were added to the device by dev_uc_sync().
3404 static inline void __dev_uc_unsync(struct net_device
*dev
,
3405 int (*unsync
)(struct net_device
*,
3406 const unsigned char *))
3408 __hw_addr_unsync_dev(&dev
->uc
, dev
, unsync
);
3411 /* Functions used for multicast addresses handling */
3412 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
3413 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
3414 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3415 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
3416 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
3417 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
3418 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3419 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
3420 void dev_mc_flush(struct net_device
*dev
);
3421 void dev_mc_init(struct net_device
*dev
);
3424 * __dev_mc_sync - Synchonize device's multicast list
3425 * @dev: device to sync
3426 * @sync: function to call if address should be added
3427 * @unsync: function to call if address should be removed
3429 * Add newly added addresses to the interface, and release
3430 * addresses that have been deleted.
3432 static inline int __dev_mc_sync(struct net_device
*dev
,
3433 int (*sync
)(struct net_device
*,
3434 const unsigned char *),
3435 int (*unsync
)(struct net_device
*,
3436 const unsigned char *))
3438 return __hw_addr_sync_dev(&dev
->mc
, dev
, sync
, unsync
);
3442 * __dev_mc_unsync - Remove synchronized addresses from device
3443 * @dev: device to sync
3444 * @unsync: function to call if address should be removed
3446 * Remove all addresses that were added to the device by dev_mc_sync().
3448 static inline void __dev_mc_unsync(struct net_device
*dev
,
3449 int (*unsync
)(struct net_device
*,
3450 const unsigned char *))
3452 __hw_addr_unsync_dev(&dev
->mc
, dev
, unsync
);
3455 /* Functions used for secondary unicast and multicast support */
3456 void dev_set_rx_mode(struct net_device
*dev
);
3457 void __dev_set_rx_mode(struct net_device
*dev
);
3458 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
3459 int dev_set_allmulti(struct net_device
*dev
, int inc
);
3460 void netdev_state_change(struct net_device
*dev
);
3461 void netdev_notify_peers(struct net_device
*dev
);
3462 void netdev_features_change(struct net_device
*dev
);
3463 /* Load a device via the kmod */
3464 void dev_load(struct net
*net
, const char *name
);
3465 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
3466 struct rtnl_link_stats64
*storage
);
3467 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
3468 const struct net_device_stats
*netdev_stats
);
3470 extern int netdev_max_backlog
;
3471 extern int netdev_tstamp_prequeue
;
3472 extern int weight_p
;
3473 extern int bpf_jit_enable
;
3475 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3476 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3477 struct list_head
**iter
);
3478 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3479 struct list_head
**iter
);
3481 /* iterate through upper list, must be called under RCU read lock */
3482 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3483 for (iter = &(dev)->adj_list.upper, \
3484 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3486 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3488 /* iterate through upper list, must be called under RCU read lock */
3489 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3490 for (iter = &(dev)->all_adj_list.upper, \
3491 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3493 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3495 void *netdev_lower_get_next_private(struct net_device
*dev
,
3496 struct list_head
**iter
);
3497 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3498 struct list_head
**iter
);
3500 #define netdev_for_each_lower_private(dev, priv, iter) \
3501 for (iter = (dev)->adj_list.lower.next, \
3502 priv = netdev_lower_get_next_private(dev, &(iter)); \
3504 priv = netdev_lower_get_next_private(dev, &(iter)))
3506 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3507 for (iter = &(dev)->adj_list.lower, \
3508 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3510 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3512 void *netdev_lower_get_next(struct net_device
*dev
,
3513 struct list_head
**iter
);
3514 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3515 for (iter = &(dev)->adj_list.lower, \
3516 ldev = netdev_lower_get_next(dev, &(iter)); \
3518 ldev = netdev_lower_get_next(dev, &(iter)))
3520 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3521 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3522 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3523 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3524 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
3525 int netdev_master_upper_dev_link(struct net_device
*dev
,
3526 struct net_device
*upper_dev
);
3527 int netdev_master_upper_dev_link_private(struct net_device
*dev
,
3528 struct net_device
*upper_dev
,
3530 void netdev_upper_dev_unlink(struct net_device
*dev
,
3531 struct net_device
*upper_dev
);
3532 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3533 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3534 struct net_device
*lower_dev
);
3536 /* RSS keys are 40 or 52 bytes long */
3537 #define NETDEV_RSS_KEY_LEN 52
3538 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
];
3539 void netdev_rss_key_fill(void *buffer
, size_t len
);
3541 int dev_get_nest_level(struct net_device
*dev
,
3542 bool (*type_check
)(struct net_device
*dev
));
3543 int skb_checksum_help(struct sk_buff
*skb
);
3544 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3545 netdev_features_t features
, bool tx_path
);
3546 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3547 netdev_features_t features
);
3549 struct netdev_bonding_info
{
3554 struct netdev_notifier_bonding_info
{
3555 struct netdev_notifier_info info
; /* must be first */
3556 struct netdev_bonding_info bonding_info
;
3559 void netdev_bonding_info_change(struct net_device
*dev
,
3560 struct netdev_bonding_info
*bonding_info
);
3563 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3565 return __skb_gso_segment(skb
, features
, true);
3567 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3569 static inline bool can_checksum_protocol(netdev_features_t features
,
3572 return ((features
& NETIF_F_GEN_CSUM
) ||
3573 ((features
& NETIF_F_V4_CSUM
) &&
3574 protocol
== htons(ETH_P_IP
)) ||
3575 ((features
& NETIF_F_V6_CSUM
) &&
3576 protocol
== htons(ETH_P_IPV6
)) ||
3577 ((features
& NETIF_F_FCOE_CRC
) &&
3578 protocol
== htons(ETH_P_FCOE
)));
3582 void netdev_rx_csum_fault(struct net_device
*dev
);
3584 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3588 /* rx skb timestamps */
3589 void net_enable_timestamp(void);
3590 void net_disable_timestamp(void);
3592 #ifdef CONFIG_PROC_FS
3593 int __init
dev_proc_init(void);
3595 #define dev_proc_init() 0
3598 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
3599 struct sk_buff
*skb
, struct net_device
*dev
,
3602 skb
->xmit_more
= more
? 1 : 0;
3603 return ops
->ndo_start_xmit(skb
, dev
);
3606 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3607 struct netdev_queue
*txq
, bool more
)
3609 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3612 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
3613 if (rc
== NETDEV_TX_OK
)
3614 txq_trans_update(txq
);
3619 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3621 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3624 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3626 return netdev_class_create_file_ns(class_attr
, NULL
);
3629 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3631 netdev_class_remove_file_ns(class_attr
, NULL
);
3634 extern struct kobj_ns_type_operations net_ns_type_operations
;
3636 const char *netdev_drivername(const struct net_device
*dev
);
3638 void linkwatch_run_queue(void);
3640 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
3641 netdev_features_t f2
)
3643 if (f1
& NETIF_F_GEN_CSUM
)
3644 f1
|= (NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3645 if (f2
& NETIF_F_GEN_CSUM
)
3646 f2
|= (NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3648 if (f1
& NETIF_F_GEN_CSUM
)
3649 f1
&= ~(NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3654 static inline netdev_features_t
netdev_get_wanted_features(
3655 struct net_device
*dev
)
3657 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
3659 netdev_features_t
netdev_increment_features(netdev_features_t all
,
3660 netdev_features_t one
, netdev_features_t mask
);
3662 /* Allow TSO being used on stacked device :
3663 * Performing the GSO segmentation before last device
3664 * is a performance improvement.
3666 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
3667 netdev_features_t mask
)
3669 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
3672 int __netdev_update_features(struct net_device
*dev
);
3673 void netdev_update_features(struct net_device
*dev
);
3674 void netdev_change_features(struct net_device
*dev
);
3676 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
3677 struct net_device
*dev
);
3679 netdev_features_t
passthru_features_check(struct sk_buff
*skb
,
3680 struct net_device
*dev
,
3681 netdev_features_t features
);
3682 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
3684 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
3686 netdev_features_t feature
= gso_type
<< NETIF_F_GSO_SHIFT
;
3688 /* check flags correspondence */
3689 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
3690 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
3691 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
3692 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
3693 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
3694 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
3695 BUILD_BUG_ON(SKB_GSO_GRE
!= (NETIF_F_GSO_GRE
>> NETIF_F_GSO_SHIFT
));
3696 BUILD_BUG_ON(SKB_GSO_GRE_CSUM
!= (NETIF_F_GSO_GRE_CSUM
>> NETIF_F_GSO_SHIFT
));
3697 BUILD_BUG_ON(SKB_GSO_IPIP
!= (NETIF_F_GSO_IPIP
>> NETIF_F_GSO_SHIFT
));
3698 BUILD_BUG_ON(SKB_GSO_SIT
!= (NETIF_F_GSO_SIT
>> NETIF_F_GSO_SHIFT
));
3699 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL
!= (NETIF_F_GSO_UDP_TUNNEL
>> NETIF_F_GSO_SHIFT
));
3700 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM
!= (NETIF_F_GSO_UDP_TUNNEL_CSUM
>> NETIF_F_GSO_SHIFT
));
3701 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
3703 return (features
& feature
) == feature
;
3706 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
3708 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
3709 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
3712 static inline bool netif_needs_gso(struct net_device
*dev
, struct sk_buff
*skb
,
3713 netdev_features_t features
)
3715 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
3716 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
3717 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
3720 static inline void netif_set_gso_max_size(struct net_device
*dev
,
3723 dev
->gso_max_size
= size
;
3726 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
3727 int pulled_hlen
, u16 mac_offset
,
3730 skb
->protocol
= protocol
;
3731 skb
->encapsulation
= 1;
3732 skb_push(skb
, pulled_hlen
);
3733 skb_reset_transport_header(skb
);
3734 skb
->mac_header
= mac_offset
;
3735 skb
->network_header
= skb
->mac_header
+ mac_len
;
3736 skb
->mac_len
= mac_len
;
3739 static inline bool netif_is_macvlan(struct net_device
*dev
)
3741 return dev
->priv_flags
& IFF_MACVLAN
;
3744 static inline bool netif_is_macvlan_port(struct net_device
*dev
)
3746 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
3749 static inline bool netif_is_ipvlan(struct net_device
*dev
)
3751 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
3754 static inline bool netif_is_ipvlan_port(struct net_device
*dev
)
3756 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
3759 static inline bool netif_is_bond_master(struct net_device
*dev
)
3761 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
3764 static inline bool netif_is_bond_slave(struct net_device
*dev
)
3766 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
3769 static inline bool netif_supports_nofcs(struct net_device
*dev
)
3771 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
3774 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3775 static inline void netif_keep_dst(struct net_device
*dev
)
3777 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
3780 extern struct pernet_operations __net_initdata loopback_net_ops
;
3782 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3784 /* netdev_printk helpers, similar to dev_printk */
3786 static inline const char *netdev_name(const struct net_device
*dev
)
3788 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
3789 return "(unnamed net_device)";
3793 static inline const char *netdev_reg_state(const struct net_device
*dev
)
3795 switch (dev
->reg_state
) {
3796 case NETREG_UNINITIALIZED
: return " (uninitialized)";
3797 case NETREG_REGISTERED
: return "";
3798 case NETREG_UNREGISTERING
: return " (unregistering)";
3799 case NETREG_UNREGISTERED
: return " (unregistered)";
3800 case NETREG_RELEASED
: return " (released)";
3801 case NETREG_DUMMY
: return " (dummy)";
3804 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
3805 return " (unknown)";
3809 void netdev_printk(const char *level
, const struct net_device
*dev
,
3810 const char *format
, ...);
3812 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
3814 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
3816 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
3818 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
3820 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
3822 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
3824 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
3826 #define MODULE_ALIAS_NETDEV(device) \
3827 MODULE_ALIAS("netdev-" device)
3829 #if defined(CONFIG_DYNAMIC_DEBUG)
3830 #define netdev_dbg(__dev, format, args...) \
3832 dynamic_netdev_dbg(__dev, format, ##args); \
3834 #elif defined(DEBUG)
3835 #define netdev_dbg(__dev, format, args...) \
3836 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3838 #define netdev_dbg(__dev, format, args...) \
3841 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3845 #if defined(VERBOSE_DEBUG)
3846 #define netdev_vdbg netdev_dbg
3849 #define netdev_vdbg(dev, format, args...) \
3852 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3858 * netdev_WARN() acts like dev_printk(), but with the key difference
3859 * of using a WARN/WARN_ON to get the message out, including the
3860 * file/line information and a backtrace.
3862 #define netdev_WARN(dev, format, args...) \
3863 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3864 netdev_reg_state(dev), ##args)
3866 /* netif printk helpers, similar to netdev_printk */
3868 #define netif_printk(priv, type, level, dev, fmt, args...) \
3870 if (netif_msg_##type(priv)) \
3871 netdev_printk(level, (dev), fmt, ##args); \
3874 #define netif_level(level, priv, type, dev, fmt, args...) \
3876 if (netif_msg_##type(priv)) \
3877 netdev_##level(dev, fmt, ##args); \
3880 #define netif_emerg(priv, type, dev, fmt, args...) \
3881 netif_level(emerg, priv, type, dev, fmt, ##args)
3882 #define netif_alert(priv, type, dev, fmt, args...) \
3883 netif_level(alert, priv, type, dev, fmt, ##args)
3884 #define netif_crit(priv, type, dev, fmt, args...) \
3885 netif_level(crit, priv, type, dev, fmt, ##args)
3886 #define netif_err(priv, type, dev, fmt, args...) \
3887 netif_level(err, priv, type, dev, fmt, ##args)
3888 #define netif_warn(priv, type, dev, fmt, args...) \
3889 netif_level(warn, priv, type, dev, fmt, ##args)
3890 #define netif_notice(priv, type, dev, fmt, args...) \
3891 netif_level(notice, priv, type, dev, fmt, ##args)
3892 #define netif_info(priv, type, dev, fmt, args...) \
3893 netif_level(info, priv, type, dev, fmt, ##args)
3895 #if defined(CONFIG_DYNAMIC_DEBUG)
3896 #define netif_dbg(priv, type, netdev, format, args...) \
3898 if (netif_msg_##type(priv)) \
3899 dynamic_netdev_dbg(netdev, format, ##args); \
3901 #elif defined(DEBUG)
3902 #define netif_dbg(priv, type, dev, format, args...) \
3903 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3905 #define netif_dbg(priv, type, dev, format, args...) \
3908 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3913 #if defined(VERBOSE_DEBUG)
3914 #define netif_vdbg netif_dbg
3916 #define netif_vdbg(priv, type, dev, format, args...) \
3919 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3925 * The list of packet types we will receive (as opposed to discard)
3926 * and the routines to invoke.
3928 * Why 16. Because with 16 the only overlap we get on a hash of the
3929 * low nibble of the protocol value is RARP/SNAP/X.25.
3931 * NOTE: That is no longer true with the addition of VLAN tags. Not
3932 * sure which should go first, but I bet it won't make much
3933 * difference if we are running VLANs. The good news is that
3934 * this protocol won't be in the list unless compiled in, so
3935 * the average user (w/out VLANs) will not be adversely affected.
3951 #define PTYPE_HASH_SIZE (16)
3952 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3954 #endif /* _LINUX_NETDEVICE_H */