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 (*rebuild
)(struct sk_buff
*skb
);
265 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
266 void (*cache_update
)(struct hh_cache
*hh
,
267 const struct net_device
*dev
,
268 const unsigned char *haddr
);
271 /* These flag bits are private to the generic network queueing
272 * layer, they may not be explicitly referenced by any other
276 enum netdev_state_t
{
278 __LINK_STATE_PRESENT
,
279 __LINK_STATE_NOCARRIER
,
280 __LINK_STATE_LINKWATCH_PENDING
,
281 __LINK_STATE_DORMANT
,
286 * This structure holds at boot time configured netdevice settings. They
287 * are then used in the device probing.
289 struct netdev_boot_setup
{
293 #define NETDEV_BOOT_SETUP_MAX 8
295 int __init
netdev_boot_setup(char *str
);
298 * Structure for NAPI scheduling similar to tasklet but with weighting
301 /* The poll_list must only be managed by the entity which
302 * changes the state of the NAPI_STATE_SCHED bit. This means
303 * whoever atomically sets that bit can add this napi_struct
304 * to the per-cpu poll_list, and whoever clears that bit
305 * can remove from the list right before clearing the bit.
307 struct list_head poll_list
;
311 unsigned int gro_count
;
312 int (*poll
)(struct napi_struct
*, int);
313 #ifdef CONFIG_NETPOLL
314 spinlock_t poll_lock
;
317 struct net_device
*dev
;
318 struct sk_buff
*gro_list
;
320 struct hrtimer timer
;
321 struct list_head dev_list
;
322 struct hlist_node napi_hash_node
;
323 unsigned int napi_id
;
327 NAPI_STATE_SCHED
, /* Poll is scheduled */
328 NAPI_STATE_DISABLE
, /* Disable pending */
329 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
330 NAPI_STATE_HASHED
, /* In NAPI hash */
340 typedef enum gro_result gro_result_t
;
343 * enum rx_handler_result - Possible return values for rx_handlers.
344 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
346 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
347 * case skb->dev was changed by rx_handler.
348 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
349 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
351 * rx_handlers are functions called from inside __netif_receive_skb(), to do
352 * special processing of the skb, prior to delivery to protocol handlers.
354 * Currently, a net_device can only have a single rx_handler registered. Trying
355 * to register a second rx_handler will return -EBUSY.
357 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
358 * To unregister a rx_handler on a net_device, use
359 * netdev_rx_handler_unregister().
361 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
364 * If the rx_handler consumed to skb in some way, it should return
365 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
366 * the skb to be delivered in some other ways.
368 * If the rx_handler changed skb->dev, to divert the skb to another
369 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
370 * new device will be called if it exists.
372 * If the rx_handler consider the skb should be ignored, it should return
373 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
374 * are registered on exact device (ptype->dev == skb->dev).
376 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
377 * delivered, it should return RX_HANDLER_PASS.
379 * A device without a registered rx_handler will behave as if rx_handler
380 * returned RX_HANDLER_PASS.
383 enum rx_handler_result
{
389 typedef enum rx_handler_result rx_handler_result_t
;
390 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
392 void __napi_schedule(struct napi_struct
*n
);
393 void __napi_schedule_irqoff(struct napi_struct
*n
);
395 static inline bool napi_disable_pending(struct napi_struct
*n
)
397 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
401 * napi_schedule_prep - check if napi can be scheduled
404 * Test if NAPI routine is already running, and if not mark
405 * it as running. This is used as a condition variable
406 * insure only one NAPI poll instance runs. We also make
407 * sure there is no pending NAPI disable.
409 static inline bool napi_schedule_prep(struct napi_struct
*n
)
411 return !napi_disable_pending(n
) &&
412 !test_and_set_bit(NAPI_STATE_SCHED
, &n
->state
);
416 * napi_schedule - schedule NAPI poll
419 * Schedule NAPI poll routine to be called if it is not already
422 static inline void napi_schedule(struct napi_struct
*n
)
424 if (napi_schedule_prep(n
))
429 * napi_schedule_irqoff - schedule NAPI poll
432 * Variant of napi_schedule(), assuming hard irqs are masked.
434 static inline void napi_schedule_irqoff(struct napi_struct
*n
)
436 if (napi_schedule_prep(n
))
437 __napi_schedule_irqoff(n
);
440 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
441 static inline bool napi_reschedule(struct napi_struct
*napi
)
443 if (napi_schedule_prep(napi
)) {
444 __napi_schedule(napi
);
450 void __napi_complete(struct napi_struct
*n
);
451 void napi_complete_done(struct napi_struct
*n
, int work_done
);
453 * napi_complete - NAPI processing complete
456 * Mark NAPI processing as complete.
457 * Consider using napi_complete_done() instead.
459 static inline void napi_complete(struct napi_struct
*n
)
461 return napi_complete_done(n
, 0);
465 * napi_by_id - lookup a NAPI by napi_id
466 * @napi_id: hashed napi_id
468 * lookup @napi_id in napi_hash table
469 * must be called under rcu_read_lock()
471 struct napi_struct
*napi_by_id(unsigned int napi_id
);
474 * napi_hash_add - add a NAPI to global hashtable
475 * @napi: napi context
477 * generate a new napi_id and store a @napi under it in napi_hash
479 void napi_hash_add(struct napi_struct
*napi
);
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: napi context
485 * Warning: caller must observe rcu grace period
486 * before freeing memory containing @napi
488 void napi_hash_del(struct napi_struct
*napi
);
491 * napi_disable - prevent NAPI from scheduling
494 * Stop NAPI from being scheduled on this context.
495 * Waits till any outstanding processing completes.
497 void napi_disable(struct napi_struct
*n
);
500 * napi_enable - enable NAPI scheduling
503 * Resume NAPI from being scheduled on this context.
504 * Must be paired with napi_disable.
506 static inline void napi_enable(struct napi_struct
*n
)
508 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
509 smp_mb__before_atomic();
510 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
515 * napi_synchronize - wait until NAPI is not running
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
522 static inline void napi_synchronize(const struct napi_struct
*n
)
524 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
528 # define napi_synchronize(n) barrier()
531 enum netdev_queue_state_t
{
532 __QUEUE_STATE_DRV_XOFF
,
533 __QUEUE_STATE_STACK_XOFF
,
534 __QUEUE_STATE_FROZEN
,
537 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
541 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
544 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
557 struct netdev_queue
{
561 struct net_device
*dev
;
562 struct Qdisc __rcu
*qdisc
;
563 struct Qdisc
*qdisc_sleeping
;
567 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
576 * please use this field instead of dev->trans_start
578 unsigned long trans_start
;
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
584 unsigned long trans_timeout
;
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).
648 struct rps_sock_flow_table
{
652 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
653 ((_num) * sizeof(u16)))
655 #define RPS_NO_CPU 0xffff
657 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
661 unsigned int cpu
, index
= hash
& table
->mask
;
663 /* We only give a hint, preemption can change cpu under us */
664 cpu
= raw_smp_processor_id();
666 if (table
->ents
[index
] != cpu
)
667 table
->ents
[index
] = cpu
;
671 static inline void rps_reset_sock_flow(struct rps_sock_flow_table
*table
,
675 table
->ents
[hash
& table
->mask
] = RPS_NO_CPU
;
678 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
680 #ifdef CONFIG_RFS_ACCEL
681 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
684 #endif /* CONFIG_RPS */
686 /* This structure contains an instance of an RX queue. */
687 struct netdev_rx_queue
{
689 struct rps_map __rcu
*rps_map
;
690 struct rps_dev_flow_table __rcu
*rps_flow_table
;
693 struct net_device
*dev
;
694 } ____cacheline_aligned_in_smp
;
697 * RX queue sysfs structures and functions.
699 struct rx_queue_attribute
{
700 struct attribute attr
;
701 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
702 struct rx_queue_attribute
*attr
, char *buf
);
703 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
704 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
709 * This structure holds an XPS map which can be of variable length. The
710 * map is an array of queues.
714 unsigned int alloc_len
;
718 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
719 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
723 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 struct xps_dev_maps
{
727 struct xps_map __rcu
*cpu_map
[0];
729 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
730 (nr_cpu_ids * sizeof(struct xps_map *)))
731 #endif /* CONFIG_XPS */
733 #define TC_MAX_QUEUE 16
734 #define TC_BITMASK 15
735 /* HW offloaded queuing disciplines txq count and offset maps */
736 struct netdev_tc_txq
{
741 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 * This structure is to hold information about the device
744 * configured to run FCoE protocol stack.
746 struct netdev_fcoe_hbainfo
{
747 char manufacturer
[64];
748 char serial_number
[64];
749 char hardware_version
[64];
750 char driver_version
[64];
751 char optionrom_version
[64];
752 char firmware_version
[64];
754 char model_description
[256];
758 #define MAX_PHYS_ITEM_ID_LEN 32
760 /* This structure holds a unique identifier to identify some
761 * physical item (port for example) used by a netdevice.
763 struct netdev_phys_item_id
{
764 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
765 unsigned char id_len
;
768 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
769 struct sk_buff
*skb
);
772 * This structure defines the management hooks for network devices.
773 * The following hooks can be defined; unless noted otherwise, they are
774 * optional and can be filled with a null pointer.
776 * int (*ndo_init)(struct net_device *dev);
777 * This function is called once when network device is registered.
778 * The network device can use this to any late stage initializaton
779 * or semantic validattion. It can fail with an error code which will
780 * be propogated back to register_netdev
782 * void (*ndo_uninit)(struct net_device *dev);
783 * This function is called when device is unregistered or when registration
784 * fails. It is not called if init fails.
786 * int (*ndo_open)(struct net_device *dev);
787 * This function is called when network device transistions to the up
790 * int (*ndo_stop)(struct net_device *dev);
791 * This function is called when network device transistions to the down
794 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
795 * struct net_device *dev);
796 * Called when a packet needs to be transmitted.
797 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
798 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
799 * Required can not be NULL.
801 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
802 * void *accel_priv, select_queue_fallback_t fallback);
803 * Called to decide which queue to when device supports multiple
806 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
807 * This function is called to allow device receiver to make
808 * changes to configuration when multicast or promiscious is enabled.
810 * void (*ndo_set_rx_mode)(struct net_device *dev);
811 * This function is called device changes address list filtering.
812 * If driver handles unicast address filtering, it should set
813 * IFF_UNICAST_FLT to its priv_flags.
815 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
816 * This function is called when the Media Access Control address
817 * needs to be changed. If this interface is not defined, the
818 * mac address can not be changed.
820 * int (*ndo_validate_addr)(struct net_device *dev);
821 * Test if Media Access Control address is valid for the device.
823 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
824 * Called when a user request an ioctl which can't be handled by
825 * the generic interface code. If not defined ioctl's return
826 * not supported error code.
828 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
829 * Used to set network devices bus interface parameters. This interface
830 * is retained for legacy reason, new devices should use the bus
831 * interface (PCI) for low level management.
833 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
834 * Called when a user wants to change the Maximum Transfer Unit
835 * of a device. If not defined, any request to change MTU will
836 * will return an error.
838 * void (*ndo_tx_timeout)(struct net_device *dev);
839 * Callback uses when the transmitter has not made any progress
840 * for dev->watchdog ticks.
842 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
843 * struct rtnl_link_stats64 *storage);
844 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
845 * Called when a user wants to get the network device usage
846 * statistics. Drivers must do one of the following:
847 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
848 * rtnl_link_stats64 structure passed by the caller.
849 * 2. Define @ndo_get_stats to update a net_device_stats structure
850 * (which should normally be dev->stats) and return a pointer to
851 * it. The structure may be changed asynchronously only if each
852 * field is written atomically.
853 * 3. Update dev->stats asynchronously and atomically, and define
856 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
857 * If device support VLAN filtering this function is called when a
858 * VLAN id is registered.
860 * int (*ndo_vlan_rx_kill_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 unregistered.
864 * void (*ndo_poll_controller)(struct net_device *dev);
866 * SR-IOV management functions.
867 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
868 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
869 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
871 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
872 * int (*ndo_get_vf_config)(struct net_device *dev,
873 * int vf, struct ifla_vf_info *ivf);
874 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
875 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
876 * struct nlattr *port[]);
877 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
878 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
879 * Called to setup 'tc' number of traffic classes in the net device. This
880 * is always called from the stack with the rtnl lock held and netif tx
881 * queues stopped. This allows the netdevice to perform queue management
884 * Fiber Channel over Ethernet (FCoE) offload functions.
885 * int (*ndo_fcoe_enable)(struct net_device *dev);
886 * Called when the FCoE protocol stack wants to start using LLD for FCoE
887 * so the underlying device can perform whatever needed configuration or
888 * initialization to support acceleration of FCoE traffic.
890 * int (*ndo_fcoe_disable)(struct net_device *dev);
891 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
892 * so the underlying device can perform whatever needed clean-ups to
893 * stop supporting acceleration of FCoE traffic.
895 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
896 * struct scatterlist *sgl, unsigned int sgc);
897 * Called when the FCoE Initiator wants to initialize an I/O that
898 * is a possible candidate for Direct Data Placement (DDP). The LLD can
899 * perform necessary setup and returns 1 to indicate the device is set up
900 * successfully to perform DDP on this I/O, otherwise this returns 0.
902 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
903 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
904 * indicated by the FC exchange id 'xid', so the underlying device can
905 * clean up and reuse resources for later DDP requests.
907 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
908 * struct scatterlist *sgl, unsigned int sgc);
909 * Called when the FCoE Target wants to initialize an I/O that
910 * is a possible candidate for Direct Data Placement (DDP). The LLD can
911 * perform necessary setup and returns 1 to indicate the device is set up
912 * successfully to perform DDP on this I/O, otherwise this returns 0.
914 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
915 * struct netdev_fcoe_hbainfo *hbainfo);
916 * Called when the FCoE Protocol stack wants information on the underlying
917 * device. This information is utilized by the FCoE protocol stack to
918 * register attributes with Fiber Channel management service as per the
919 * FC-GS Fabric Device Management Information(FDMI) specification.
921 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
922 * Called when the underlying device wants to override default World Wide
923 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
924 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
925 * protocol stack to use.
928 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
929 * u16 rxq_index, u32 flow_id);
930 * Set hardware filter for RFS. rxq_index is the target queue index;
931 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
932 * Return the filter ID on success, or a negative error code.
934 * Slave management functions (for bridge, bonding, etc).
935 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
936 * Called to make another netdev an underling.
938 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
939 * Called to release previously enslaved netdev.
941 * Feature/offload setting functions.
942 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
943 * netdev_features_t features);
944 * Adjusts the requested feature flags according to device-specific
945 * constraints, and returns the resulting flags. Must not modify
948 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
949 * Called to update device configuration to new features. Passed
950 * feature set might be less than what was returned by ndo_fix_features()).
951 * Must return >0 or -errno if it changed dev->features itself.
953 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
954 * struct net_device *dev,
955 * const unsigned char *addr, u16 vid, u16 flags)
956 * Adds an FDB entry to dev for addr.
957 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
958 * struct net_device *dev,
959 * const unsigned char *addr, u16 vid)
960 * Deletes the FDB entry from dev coresponding to addr.
961 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
962 * struct net_device *dev, struct net_device *filter_dev,
964 * Used to add FDB entries to dump requests. Implementers should add
965 * entries to skb and update idx with the number of entries.
967 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
968 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
969 * struct net_device *dev, u32 filter_mask)
971 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
972 * Called to change device carrier. Soft-devices (like dummy, team, etc)
973 * which do not represent real hardware may define this to allow their
974 * userspace components to manage their virtual carrier state. Devices
975 * that determine carrier state from physical hardware properties (eg
976 * network cables) or protocol-dependent mechanisms (eg
977 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
979 * int (*ndo_get_phys_port_id)(struct net_device *dev,
980 * struct netdev_phys_item_id *ppid);
981 * Called to get ID of physical port of this device. If driver does
982 * not implement this, it is assumed that the hw is not able to have
983 * multiple net devices on single physical port.
985 * void (*ndo_add_vxlan_port)(struct net_device *dev,
986 * sa_family_t sa_family, __be16 port);
987 * Called by vxlan to notiy a driver about the UDP port and socket
988 * address family that vxlan is listnening to. It is called only when
989 * a new port starts listening. The operation is protected by the
990 * vxlan_net->sock_lock.
992 * void (*ndo_del_vxlan_port)(struct net_device *dev,
993 * sa_family_t sa_family, __be16 port);
994 * Called by vxlan to notify the driver about a UDP port and socket
995 * address family that vxlan is not listening to anymore. The operation
996 * is protected by the vxlan_net->sock_lock.
998 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
999 * struct net_device *dev)
1000 * Called by upper layer devices to accelerate switching or other
1001 * station functionality into hardware. 'pdev is the lowerdev
1002 * to use for the offload and 'dev' is the net device that will
1003 * back the offload. Returns a pointer to the private structure
1004 * the upper layer will maintain.
1005 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1006 * Called by upper layer device to delete the station created
1007 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1008 * the station and priv is the structure returned by the add
1010 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1011 * struct net_device *dev,
1013 * Callback to use for xmit over the accelerated station. This
1014 * is used in place of ndo_start_xmit on accelerated net
1016 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1017 * struct net_device *dev
1018 * netdev_features_t features);
1019 * Called by core transmit path to determine if device is capable of
1020 * performing offload operations on a given packet. This is to give
1021 * the device an opportunity to implement any restrictions that cannot
1022 * be otherwise expressed by feature flags. The check is called with
1023 * the set of features that the stack has calculated and it returns
1024 * those the driver believes to be appropriate.
1026 * int (*ndo_switch_parent_id_get)(struct net_device *dev,
1027 * struct netdev_phys_item_id *psid);
1028 * Called to get an ID of the switch chip this port is part of.
1029 * If driver implements this, it indicates that it represents a port
1031 * int (*ndo_switch_port_stp_update)(struct net_device *dev, u8 state);
1032 * Called to notify switch device port of bridge port STP
1035 struct net_device_ops
{
1036 int (*ndo_init
)(struct net_device
*dev
);
1037 void (*ndo_uninit
)(struct net_device
*dev
);
1038 int (*ndo_open
)(struct net_device
*dev
);
1039 int (*ndo_stop
)(struct net_device
*dev
);
1040 netdev_tx_t (*ndo_start_xmit
) (struct sk_buff
*skb
,
1041 struct net_device
*dev
);
1042 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1043 struct sk_buff
*skb
,
1045 select_queue_fallback_t fallback
);
1046 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1048 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1049 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1051 int (*ndo_validate_addr
)(struct net_device
*dev
);
1052 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1053 struct ifreq
*ifr
, int cmd
);
1054 int (*ndo_set_config
)(struct net_device
*dev
,
1056 int (*ndo_change_mtu
)(struct net_device
*dev
,
1058 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1059 struct neigh_parms
*);
1060 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1062 struct rtnl_link_stats64
* (*ndo_get_stats64
)(struct net_device
*dev
,
1063 struct rtnl_link_stats64
*storage
);
1064 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1066 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1067 __be16 proto
, u16 vid
);
1068 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1069 __be16 proto
, u16 vid
);
1070 #ifdef CONFIG_NET_POLL_CONTROLLER
1071 void (*ndo_poll_controller
)(struct net_device
*dev
);
1072 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1073 struct netpoll_info
*info
);
1074 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1076 #ifdef CONFIG_NET_RX_BUSY_POLL
1077 int (*ndo_busy_poll
)(struct napi_struct
*dev
);
1079 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1080 int queue
, u8
*mac
);
1081 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1082 int queue
, u16 vlan
, u8 qos
);
1083 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1084 int vf
, int min_tx_rate
,
1086 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1087 int vf
, bool setting
);
1088 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1090 struct ifla_vf_info
*ivf
);
1091 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1092 int vf
, int link_state
);
1093 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1095 struct nlattr
*port
[]);
1096 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1097 int vf
, struct sk_buff
*skb
);
1098 int (*ndo_setup_tc
)(struct net_device
*dev
, u8 tc
);
1099 #if IS_ENABLED(CONFIG_FCOE)
1100 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1101 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1102 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1104 struct scatterlist
*sgl
,
1106 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1108 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1110 struct scatterlist
*sgl
,
1112 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1113 struct netdev_fcoe_hbainfo
*hbainfo
);
1116 #if IS_ENABLED(CONFIG_LIBFCOE)
1117 #define NETDEV_FCOE_WWNN 0
1118 #define NETDEV_FCOE_WWPN 1
1119 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1120 u64
*wwn
, int type
);
1123 #ifdef CONFIG_RFS_ACCEL
1124 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1125 const struct sk_buff
*skb
,
1129 int (*ndo_add_slave
)(struct net_device
*dev
,
1130 struct net_device
*slave_dev
);
1131 int (*ndo_del_slave
)(struct net_device
*dev
,
1132 struct net_device
*slave_dev
);
1133 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1134 netdev_features_t features
);
1135 int (*ndo_set_features
)(struct net_device
*dev
,
1136 netdev_features_t features
);
1137 int (*ndo_neigh_construct
)(struct neighbour
*n
);
1138 void (*ndo_neigh_destroy
)(struct neighbour
*n
);
1140 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1141 struct nlattr
*tb
[],
1142 struct net_device
*dev
,
1143 const unsigned char *addr
,
1146 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1147 struct nlattr
*tb
[],
1148 struct net_device
*dev
,
1149 const unsigned char *addr
,
1151 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1152 struct netlink_callback
*cb
,
1153 struct net_device
*dev
,
1154 struct net_device
*filter_dev
,
1157 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1158 struct nlmsghdr
*nlh
,
1160 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1162 struct net_device
*dev
,
1164 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1165 struct nlmsghdr
*nlh
,
1167 int (*ndo_change_carrier
)(struct net_device
*dev
,
1169 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1170 struct netdev_phys_item_id
*ppid
);
1171 void (*ndo_add_vxlan_port
)(struct net_device
*dev
,
1172 sa_family_t sa_family
,
1174 void (*ndo_del_vxlan_port
)(struct net_device
*dev
,
1175 sa_family_t sa_family
,
1178 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1179 struct net_device
*dev
);
1180 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1183 netdev_tx_t (*ndo_dfwd_start_xmit
) (struct sk_buff
*skb
,
1184 struct net_device
*dev
,
1186 int (*ndo_get_lock_subclass
)(struct net_device
*dev
);
1187 netdev_features_t (*ndo_features_check
) (struct sk_buff
*skb
,
1188 struct net_device
*dev
,
1189 netdev_features_t features
);
1190 #ifdef CONFIG_NET_SWITCHDEV
1191 int (*ndo_switch_parent_id_get
)(struct net_device
*dev
,
1192 struct netdev_phys_item_id
*psid
);
1193 int (*ndo_switch_port_stp_update
)(struct net_device
*dev
,
1199 * enum net_device_priv_flags - &struct net_device priv_flags
1201 * These are the &struct net_device, they are only set internally
1202 * by drivers and used in the kernel. These flags are invisible to
1203 * userspace, this means that the order of these flags can change
1204 * during any kernel release.
1206 * You should have a pretty good reason to be extending these flags.
1208 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1209 * @IFF_EBRIDGE: Ethernet bridging device
1210 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1211 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1212 * @IFF_MASTER_ALB: bonding master, balance-alb
1213 * @IFF_BONDING: bonding master or slave
1214 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1215 * @IFF_ISATAP: ISATAP interface (RFC4214)
1216 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1217 * @IFF_WAN_HDLC: WAN HDLC device
1218 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1220 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1221 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1222 * @IFF_MACVLAN_PORT: device used as macvlan port
1223 * @IFF_BRIDGE_PORT: device used as bridge port
1224 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1225 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1226 * @IFF_UNICAST_FLT: Supports unicast filtering
1227 * @IFF_TEAM_PORT: device used as team port
1228 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1229 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1230 * change when it's running
1231 * @IFF_MACVLAN: Macvlan device
1233 enum netdev_priv_flags
{
1234 IFF_802_1Q_VLAN
= 1<<0,
1236 IFF_SLAVE_INACTIVE
= 1<<2,
1237 IFF_MASTER_8023AD
= 1<<3,
1238 IFF_MASTER_ALB
= 1<<4,
1240 IFF_SLAVE_NEEDARP
= 1<<6,
1242 IFF_MASTER_ARPMON
= 1<<8,
1243 IFF_WAN_HDLC
= 1<<9,
1244 IFF_XMIT_DST_RELEASE
= 1<<10,
1245 IFF_DONT_BRIDGE
= 1<<11,
1246 IFF_DISABLE_NETPOLL
= 1<<12,
1247 IFF_MACVLAN_PORT
= 1<<13,
1248 IFF_BRIDGE_PORT
= 1<<14,
1249 IFF_OVS_DATAPATH
= 1<<15,
1250 IFF_TX_SKB_SHARING
= 1<<16,
1251 IFF_UNICAST_FLT
= 1<<17,
1252 IFF_TEAM_PORT
= 1<<18,
1253 IFF_SUPP_NOFCS
= 1<<19,
1254 IFF_LIVE_ADDR_CHANGE
= 1<<20,
1255 IFF_MACVLAN
= 1<<21,
1256 IFF_XMIT_DST_RELEASE_PERM
= 1<<22,
1257 IFF_IPVLAN_MASTER
= 1<<23,
1258 IFF_IPVLAN_SLAVE
= 1<<24,
1261 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1262 #define IFF_EBRIDGE IFF_EBRIDGE
1263 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1264 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1265 #define IFF_MASTER_ALB IFF_MASTER_ALB
1266 #define IFF_BONDING IFF_BONDING
1267 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1268 #define IFF_ISATAP IFF_ISATAP
1269 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1270 #define IFF_WAN_HDLC IFF_WAN_HDLC
1271 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1272 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1273 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1274 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1275 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1276 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1277 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1278 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1279 #define IFF_TEAM_PORT IFF_TEAM_PORT
1280 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1281 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1282 #define IFF_MACVLAN IFF_MACVLAN
1283 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1284 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1285 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1288 * struct net_device - The DEVICE structure.
1289 * Actually, this whole structure is a big mistake. It mixes I/O
1290 * data with strictly "high-level" data, and it has to know about
1291 * almost every data structure used in the INET module.
1293 * @name: This is the first field of the "visible" part of this structure
1294 * (i.e. as seen by users in the "Space.c" file). It is the name
1297 * @name_hlist: Device name hash chain, please keep it close to name[]
1298 * @ifalias: SNMP alias
1299 * @mem_end: Shared memory end
1300 * @mem_start: Shared memory start
1301 * @base_addr: Device I/O address
1302 * @irq: Device IRQ number
1304 * @state: Generic network queuing layer state, see netdev_state_t
1305 * @dev_list: The global list of network devices
1306 * @napi_list: List entry, that is used for polling napi devices
1307 * @unreg_list: List entry, that is used, when we are unregistering the
1308 * device, see the function unregister_netdev
1309 * @close_list: List entry, that is used, when we are closing the device
1311 * @adj_list: Directly linked devices, like slaves for bonding
1312 * @all_adj_list: All linked devices, *including* neighbours
1313 * @features: Currently active device features
1314 * @hw_features: User-changeable features
1316 * @wanted_features: User-requested features
1317 * @vlan_features: Mask of features inheritable by VLAN devices
1319 * @hw_enc_features: Mask of features inherited by encapsulating devices
1320 * This field indicates what encapsulation
1321 * offloads the hardware is capable of doing,
1322 * and drivers will need to set them appropriately.
1324 * @mpls_features: Mask of features inheritable by MPLS
1326 * @ifindex: interface index
1327 * @iflink: unique device identifier
1329 * @stats: Statistics struct, which was left as a legacy, use
1330 * rtnl_link_stats64 instead
1332 * @rx_dropped: Dropped packets by core network,
1333 * do not use this in drivers
1334 * @tx_dropped: Dropped packets by core network,
1335 * do not use this in drivers
1337 * @carrier_changes: Stats to monitor carrier on<->off transitions
1339 * @wireless_handlers: List of functions to handle Wireless Extensions,
1341 * see <net/iw_handler.h> for details.
1342 * @wireless_data: Instance data managed by the core of wireless extensions
1344 * @netdev_ops: Includes several pointers to callbacks,
1345 * if one wants to override the ndo_*() functions
1346 * @ethtool_ops: Management operations
1347 * @fwd_ops: Management operations
1348 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1349 * of Layer 2 headers.
1351 * @flags: Interface flags (a la BSD)
1352 * @priv_flags: Like 'flags' but invisible to userspace,
1353 * see if.h for the definitions
1354 * @gflags: Global flags ( kept as legacy )
1355 * @padded: How much padding added by alloc_netdev()
1356 * @operstate: RFC2863 operstate
1357 * @link_mode: Mapping policy to operstate
1358 * @if_port: Selectable AUI, TP, ...
1360 * @mtu: Interface MTU value
1361 * @type: Interface hardware type
1362 * @hard_header_len: Hardware header length
1364 * @needed_headroom: Extra headroom the hardware may need, but not in all
1365 * cases can this be guaranteed
1366 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1367 * cases can this be guaranteed. Some cases also use
1368 * LL_MAX_HEADER instead to allocate the skb
1370 * interface address info:
1372 * @perm_addr: Permanent hw address
1373 * @addr_assign_type: Hw address assignment type
1374 * @addr_len: Hardware address length
1375 * @neigh_priv_len; Used in neigh_alloc(),
1376 * initialized only in atm/clip.c
1377 * @dev_id: Used to differentiate devices that share
1378 * the same link layer address
1379 * @dev_port: Used to differentiate devices that share
1381 * @addr_list_lock: XXX: need comments on this one
1382 * @uc: unicast mac addresses
1383 * @mc: multicast mac addresses
1384 * @dev_addrs: list of device hw addresses
1385 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1386 * @uc_promisc: Counter, that indicates, that promiscuous mode
1387 * has been enabled due to the need to listen to
1388 * additional unicast addresses in a device that
1389 * does not implement ndo_set_rx_mode()
1390 * @promiscuity: Number of times, the NIC is told to work in
1391 * Promiscuous mode, if it becomes 0 the NIC will
1392 * exit from working in Promiscuous mode
1393 * @allmulti: Counter, enables or disables allmulticast mode
1395 * @vlan_info: VLAN info
1396 * @dsa_ptr: dsa specific data
1397 * @tipc_ptr: TIPC specific data
1398 * @atalk_ptr: AppleTalk link
1399 * @ip_ptr: IPv4 specific data
1400 * @dn_ptr: DECnet specific data
1401 * @ip6_ptr: IPv6 specific data
1402 * @ax25_ptr: AX.25 specific data
1403 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1405 * @last_rx: Time of last Rx
1406 * @dev_addr: Hw address (before bcast,
1407 * because most packets are unicast)
1409 * @_rx: Array of RX queues
1410 * @num_rx_queues: Number of RX queues
1411 * allocated at register_netdev() time
1412 * @real_num_rx_queues: Number of RX queues currently active in device
1414 * @rx_handler: handler for received packets
1415 * @rx_handler_data: XXX: need comments on this one
1416 * @ingress_queue: XXX: need comments on this one
1417 * @broadcast: hw bcast address
1419 * @_tx: Array of TX queues
1420 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1421 * @real_num_tx_queues: Number of TX queues currently active in device
1422 * @qdisc: Root qdisc from userspace point of view
1423 * @tx_queue_len: Max frames per queue allowed
1424 * @tx_global_lock: XXX: need comments on this one
1426 * @xps_maps: XXX: need comments on this one
1428 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1429 * indexed by RX queue number. Assigned by driver.
1430 * This must only be set if the ndo_rx_flow_steer
1431 * operation is defined
1433 * @trans_start: Time (in jiffies) of last Tx
1434 * @watchdog_timeo: Represents the timeout that is used by
1435 * the watchdog ( see dev_watchdog() )
1436 * @watchdog_timer: List of timers
1438 * @pcpu_refcnt: Number of references to this device
1439 * @todo_list: Delayed register/unregister
1440 * @index_hlist: Device index hash chain
1441 * @link_watch_list: XXX: need comments on this one
1443 * @reg_state: Register/unregister state machine
1444 * @dismantle: Device is going to be freed
1445 * @rtnl_link_state: This enum represents the phases of creating
1448 * @destructor: Called from unregister,
1449 * can be used to call free_netdev
1450 * @npinfo: XXX: need comments on this one
1451 * @nd_net: Network namespace this network device is inside
1453 * @ml_priv: Mid-layer private
1454 * @lstats: Loopback statistics
1455 * @tstats: Tunnel statistics
1456 * @dstats: Dummy statistics
1457 * @vstats: Virtual ethernet statistics
1462 * @dev: Class/net/name entry
1463 * @sysfs_groups: Space for optional device, statistics and wireless
1466 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1467 * @rtnl_link_ops: Rtnl_link_ops
1469 * @gso_max_size: Maximum size of generic segmentation offload
1470 * @gso_max_segs: Maximum number of segments that can be passed to the
1472 * @gso_min_segs: Minimum number of segments that can be passed to the
1475 * @dcbnl_ops: Data Center Bridging netlink ops
1476 * @num_tc: Number of traffic classes in the net device
1477 * @tc_to_txq: XXX: need comments on this one
1478 * @prio_tc_map XXX: need comments on this one
1480 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1482 * @priomap: XXX: need comments on this one
1483 * @phydev: Physical device may attach itself
1484 * for hardware timestamping
1486 * @qdisc_tx_busylock: XXX: need comments on this one
1488 * @group: The group, that the device belongs to
1489 * @pm_qos_req: Power Management QoS object
1491 * FIXME: cleanup struct net_device such that network protocol info
1496 char name
[IFNAMSIZ
];
1497 struct hlist_node name_hlist
;
1500 * I/O specific fields
1501 * FIXME: Merge these and struct ifmap into one
1503 unsigned long mem_end
;
1504 unsigned long mem_start
;
1505 unsigned long base_addr
;
1509 * Some hardware also needs these fields (state,dev_list,
1510 * napi_list,unreg_list,close_list) but they are not
1511 * part of the usual set specified in Space.c.
1514 unsigned long state
;
1516 struct list_head dev_list
;
1517 struct list_head napi_list
;
1518 struct list_head unreg_list
;
1519 struct list_head close_list
;
1520 struct list_head ptype_all
;
1521 struct list_head ptype_specific
;
1524 struct list_head upper
;
1525 struct list_head lower
;
1529 struct list_head upper
;
1530 struct list_head lower
;
1533 netdev_features_t features
;
1534 netdev_features_t hw_features
;
1535 netdev_features_t wanted_features
;
1536 netdev_features_t vlan_features
;
1537 netdev_features_t hw_enc_features
;
1538 netdev_features_t mpls_features
;
1543 struct net_device_stats stats
;
1545 atomic_long_t rx_dropped
;
1546 atomic_long_t tx_dropped
;
1548 atomic_t carrier_changes
;
1550 #ifdef CONFIG_WIRELESS_EXT
1551 const struct iw_handler_def
* wireless_handlers
;
1552 struct iw_public_data
* wireless_data
;
1554 const struct net_device_ops
*netdev_ops
;
1555 const struct ethtool_ops
*ethtool_ops
;
1556 const struct forwarding_accel_ops
*fwd_ops
;
1558 const struct header_ops
*header_ops
;
1561 unsigned int priv_flags
;
1563 unsigned short gflags
;
1564 unsigned short padded
;
1566 unsigned char operstate
;
1567 unsigned char link_mode
;
1569 unsigned char if_port
;
1573 unsigned short type
;
1574 unsigned short hard_header_len
;
1576 unsigned short needed_headroom
;
1577 unsigned short needed_tailroom
;
1579 /* Interface address info. */
1580 unsigned char perm_addr
[MAX_ADDR_LEN
];
1581 unsigned char addr_assign_type
;
1582 unsigned char addr_len
;
1583 unsigned short neigh_priv_len
;
1584 unsigned short dev_id
;
1585 unsigned short dev_port
;
1586 spinlock_t addr_list_lock
;
1587 struct netdev_hw_addr_list uc
;
1588 struct netdev_hw_addr_list mc
;
1589 struct netdev_hw_addr_list dev_addrs
;
1592 struct kset
*queues_kset
;
1595 unsigned char name_assign_type
;
1598 unsigned int promiscuity
;
1599 unsigned int allmulti
;
1602 /* Protocol specific pointers */
1604 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1605 struct vlan_info __rcu
*vlan_info
;
1607 #if IS_ENABLED(CONFIG_NET_DSA)
1608 struct dsa_switch_tree
*dsa_ptr
;
1610 #if IS_ENABLED(CONFIG_TIPC)
1611 struct tipc_bearer __rcu
*tipc_ptr
;
1614 struct in_device __rcu
*ip_ptr
;
1615 struct dn_dev __rcu
*dn_ptr
;
1616 struct inet6_dev __rcu
*ip6_ptr
;
1618 struct wireless_dev
*ieee80211_ptr
;
1619 struct wpan_dev
*ieee802154_ptr
;
1622 * Cache lines mostly used on receive path (including eth_type_trans())
1624 unsigned long last_rx
;
1626 /* Interface address info used in eth_type_trans() */
1627 unsigned char *dev_addr
;
1631 struct netdev_rx_queue
*_rx
;
1633 unsigned int num_rx_queues
;
1634 unsigned int real_num_rx_queues
;
1638 unsigned long gro_flush_timeout
;
1639 rx_handler_func_t __rcu
*rx_handler
;
1640 void __rcu
*rx_handler_data
;
1642 struct netdev_queue __rcu
*ingress_queue
;
1643 unsigned char broadcast
[MAX_ADDR_LEN
];
1647 * Cache lines mostly used on transmit path
1649 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1650 unsigned int num_tx_queues
;
1651 unsigned int real_num_tx_queues
;
1652 struct Qdisc
*qdisc
;
1653 unsigned long tx_queue_len
;
1654 spinlock_t tx_global_lock
;
1657 struct xps_dev_maps __rcu
*xps_maps
;
1659 #ifdef CONFIG_RFS_ACCEL
1660 struct cpu_rmap
*rx_cpu_rmap
;
1663 /* These may be needed for future network-power-down code. */
1666 * trans_start here is expensive for high speed devices on SMP,
1667 * please use netdev_queue->trans_start instead.
1669 unsigned long trans_start
;
1672 struct timer_list watchdog_timer
;
1674 int __percpu
*pcpu_refcnt
;
1675 struct list_head todo_list
;
1677 struct hlist_node index_hlist
;
1678 struct list_head link_watch_list
;
1680 enum { NETREG_UNINITIALIZED
=0,
1681 NETREG_REGISTERED
, /* completed register_netdevice */
1682 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1683 NETREG_UNREGISTERED
, /* completed unregister todo */
1684 NETREG_RELEASED
, /* called free_netdev */
1685 NETREG_DUMMY
, /* dummy device for NAPI poll */
1691 RTNL_LINK_INITIALIZED
,
1692 RTNL_LINK_INITIALIZING
,
1693 } rtnl_link_state
:16;
1695 void (*destructor
)(struct net_device
*dev
);
1697 #ifdef CONFIG_NETPOLL
1698 struct netpoll_info __rcu
*npinfo
;
1701 #ifdef CONFIG_NET_NS
1705 /* mid-layer private */
1708 struct pcpu_lstats __percpu
*lstats
;
1709 struct pcpu_sw_netstats __percpu
*tstats
;
1710 struct pcpu_dstats __percpu
*dstats
;
1711 struct pcpu_vstats __percpu
*vstats
;
1714 struct garp_port __rcu
*garp_port
;
1715 struct mrp_port __rcu
*mrp_port
;
1718 const struct attribute_group
*sysfs_groups
[4];
1719 const struct attribute_group
*sysfs_rx_queue_group
;
1721 const struct rtnl_link_ops
*rtnl_link_ops
;
1723 /* for setting kernel sock attribute on TCP connection setup */
1724 #define GSO_MAX_SIZE 65536
1725 unsigned int gso_max_size
;
1726 #define GSO_MAX_SEGS 65535
1730 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1733 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1734 u8 prio_tc_map
[TC_BITMASK
+ 1];
1736 #if IS_ENABLED(CONFIG_FCOE)
1737 unsigned int fcoe_ddp_xid
;
1739 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1740 struct netprio_map __rcu
*priomap
;
1742 struct phy_device
*phydev
;
1743 struct lock_class_key
*qdisc_tx_busylock
;
1745 struct pm_qos_request pm_qos_req
;
1747 #define to_net_dev(d) container_of(d, struct net_device, dev)
1749 #define NETDEV_ALIGN 32
1752 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1754 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1758 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1760 if (tc
>= dev
->num_tc
)
1763 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1768 void netdev_reset_tc(struct net_device
*dev
)
1771 memset(dev
->tc_to_txq
, 0, sizeof(dev
->tc_to_txq
));
1772 memset(dev
->prio_tc_map
, 0, sizeof(dev
->prio_tc_map
));
1776 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
)
1778 if (tc
>= dev
->num_tc
)
1781 dev
->tc_to_txq
[tc
].count
= count
;
1782 dev
->tc_to_txq
[tc
].offset
= offset
;
1787 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
)
1789 if (num_tc
> TC_MAX_QUEUE
)
1792 dev
->num_tc
= num_tc
;
1797 int netdev_get_num_tc(struct net_device
*dev
)
1803 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1806 return &dev
->_tx
[index
];
1809 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1810 const struct sk_buff
*skb
)
1812 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1815 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1816 void (*f
)(struct net_device
*,
1817 struct netdev_queue
*,
1823 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1824 f(dev
, &dev
->_tx
[i
], arg
);
1827 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1828 struct sk_buff
*skb
,
1832 * Net namespace inlines
1835 struct net
*dev_net(const struct net_device
*dev
)
1837 return read_pnet(&dev
->nd_net
);
1841 void dev_net_set(struct net_device
*dev
, struct net
*net
)
1843 #ifdef CONFIG_NET_NS
1844 release_net(dev
->nd_net
);
1845 dev
->nd_net
= hold_net(net
);
1849 static inline bool netdev_uses_dsa(struct net_device
*dev
)
1851 #if IS_ENABLED(CONFIG_NET_DSA)
1852 if (dev
->dsa_ptr
!= NULL
)
1853 return dsa_uses_tagged_protocol(dev
->dsa_ptr
);
1859 * netdev_priv - access network device private data
1860 * @dev: network device
1862 * Get network device private data
1864 static inline void *netdev_priv(const struct net_device
*dev
)
1866 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
1869 /* Set the sysfs physical device reference for the network logical device
1870 * if set prior to registration will cause a symlink during initialization.
1872 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1874 /* Set the sysfs device type for the network logical device to allow
1875 * fine-grained identification of different network device types. For
1876 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1878 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1880 /* Default NAPI poll() weight
1881 * Device drivers are strongly advised to not use bigger value
1883 #define NAPI_POLL_WEIGHT 64
1886 * netif_napi_add - initialize a napi context
1887 * @dev: network device
1888 * @napi: napi context
1889 * @poll: polling function
1890 * @weight: default weight
1892 * netif_napi_add() must be used to initialize a napi context prior to calling
1893 * *any* of the other napi related functions.
1895 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
1896 int (*poll
)(struct napi_struct
*, int), int weight
);
1899 * netif_napi_del - remove a napi context
1900 * @napi: napi context
1902 * netif_napi_del() removes a napi context from the network device napi list
1904 void netif_napi_del(struct napi_struct
*napi
);
1906 struct napi_gro_cb
{
1907 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1910 /* Length of frag0. */
1911 unsigned int frag0_len
;
1913 /* This indicates where we are processing relative to skb->data. */
1916 /* This is non-zero if the packet cannot be merged with the new skb. */
1919 /* Save the IP ID here and check when we get to the transport layer */
1922 /* Number of segments aggregated. */
1925 /* This is non-zero if the packet may be of the same flow. */
1930 #define NAPI_GRO_FREE 1
1931 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1933 /* jiffies when first packet was created/queued */
1936 /* Used in ipv6_gro_receive() and foo-over-udp */
1939 /* Used in udp_gro_receive */
1942 /* GRO checksum is valid */
1945 /* Number of checksums via CHECKSUM_UNNECESSARY */
1948 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1951 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1954 /* used in skb_gro_receive() slow path */
1955 struct sk_buff
*last
;
1958 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1960 struct packet_type
{
1961 __be16 type
; /* This is really htons(ether_type). */
1962 struct net_device
*dev
; /* NULL is wildcarded here */
1963 int (*func
) (struct sk_buff
*,
1964 struct net_device
*,
1965 struct packet_type
*,
1966 struct net_device
*);
1967 bool (*id_match
)(struct packet_type
*ptype
,
1969 void *af_packet_priv
;
1970 struct list_head list
;
1973 struct offload_callbacks
{
1974 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
1975 netdev_features_t features
);
1976 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1977 struct sk_buff
*skb
);
1978 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
1981 struct packet_offload
{
1982 __be16 type
; /* This is really htons(ether_type). */
1983 struct offload_callbacks callbacks
;
1984 struct list_head list
;
1989 struct udp_offload_callbacks
{
1990 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1991 struct sk_buff
*skb
,
1992 struct udp_offload
*uoff
);
1993 int (*gro_complete
)(struct sk_buff
*skb
,
1995 struct udp_offload
*uoff
);
1998 struct udp_offload
{
2001 struct udp_offload_callbacks callbacks
;
2004 /* often modified stats are per cpu, other are shared (netdev->stats) */
2005 struct pcpu_sw_netstats
{
2010 struct u64_stats_sync syncp
;
2013 #define netdev_alloc_pcpu_stats(type) \
2015 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2018 for_each_possible_cpu(i) { \
2019 typeof(type) *stat; \
2020 stat = per_cpu_ptr(pcpu_stats, i); \
2021 u64_stats_init(&stat->syncp); \
2027 #include <linux/notifier.h>
2029 /* netdevice notifier chain. Please remember to update the rtnetlink
2030 * notification exclusion list in rtnetlink_event() when adding new
2033 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2034 #define NETDEV_DOWN 0x0002
2035 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2036 detected a hardware crash and restarted
2037 - we can use this eg to kick tcp sessions
2039 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2040 #define NETDEV_REGISTER 0x0005
2041 #define NETDEV_UNREGISTER 0x0006
2042 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2043 #define NETDEV_CHANGEADDR 0x0008
2044 #define NETDEV_GOING_DOWN 0x0009
2045 #define NETDEV_CHANGENAME 0x000A
2046 #define NETDEV_FEAT_CHANGE 0x000B
2047 #define NETDEV_BONDING_FAILOVER 0x000C
2048 #define NETDEV_PRE_UP 0x000D
2049 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2050 #define NETDEV_POST_TYPE_CHANGE 0x000F
2051 #define NETDEV_POST_INIT 0x0010
2052 #define NETDEV_UNREGISTER_FINAL 0x0011
2053 #define NETDEV_RELEASE 0x0012
2054 #define NETDEV_NOTIFY_PEERS 0x0013
2055 #define NETDEV_JOIN 0x0014
2056 #define NETDEV_CHANGEUPPER 0x0015
2057 #define NETDEV_RESEND_IGMP 0x0016
2058 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2059 #define NETDEV_CHANGEINFODATA 0x0018
2060 #define NETDEV_BONDING_INFO 0x0019
2062 int register_netdevice_notifier(struct notifier_block
*nb
);
2063 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2065 struct netdev_notifier_info
{
2066 struct net_device
*dev
;
2069 struct netdev_notifier_change_info
{
2070 struct netdev_notifier_info info
; /* must be first */
2071 unsigned int flags_changed
;
2074 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2075 struct net_device
*dev
)
2080 static inline struct net_device
*
2081 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2086 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2089 extern rwlock_t dev_base_lock
; /* Device list lock */
2091 #define for_each_netdev(net, d) \
2092 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2093 #define for_each_netdev_reverse(net, d) \
2094 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2095 #define for_each_netdev_rcu(net, d) \
2096 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2097 #define for_each_netdev_safe(net, d, n) \
2098 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2099 #define for_each_netdev_continue(net, d) \
2100 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2101 #define for_each_netdev_continue_rcu(net, d) \
2102 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2103 #define for_each_netdev_in_bond_rcu(bond, slave) \
2104 for_each_netdev_rcu(&init_net, slave) \
2105 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2106 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2108 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2110 struct list_head
*lh
;
2114 lh
= dev
->dev_list
.next
;
2115 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2118 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2120 struct list_head
*lh
;
2124 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2125 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2128 static inline struct net_device
*first_net_device(struct net
*net
)
2130 return list_empty(&net
->dev_base_head
) ? NULL
:
2131 net_device_entry(net
->dev_base_head
.next
);
2134 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2136 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2138 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2141 int netdev_boot_setup_check(struct net_device
*dev
);
2142 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2143 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2144 const char *hwaddr
);
2145 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2146 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2147 void dev_add_pack(struct packet_type
*pt
);
2148 void dev_remove_pack(struct packet_type
*pt
);
2149 void __dev_remove_pack(struct packet_type
*pt
);
2150 void dev_add_offload(struct packet_offload
*po
);
2151 void dev_remove_offload(struct packet_offload
*po
);
2153 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2154 unsigned short mask
);
2155 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2156 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2157 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2158 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2159 int dev_open(struct net_device
*dev
);
2160 int dev_close(struct net_device
*dev
);
2161 void dev_disable_lro(struct net_device
*dev
);
2162 int dev_loopback_xmit(struct sk_buff
*newskb
);
2163 int dev_queue_xmit(struct sk_buff
*skb
);
2164 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2165 int register_netdevice(struct net_device
*dev
);
2166 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2167 void unregister_netdevice_many(struct list_head
*head
);
2168 static inline void unregister_netdevice(struct net_device
*dev
)
2170 unregister_netdevice_queue(dev
, NULL
);
2173 int netdev_refcnt_read(const struct net_device
*dev
);
2174 void free_netdev(struct net_device
*dev
);
2175 void netdev_freemem(struct net_device
*dev
);
2176 void synchronize_net(void);
2177 int init_dummy_netdev(struct net_device
*dev
);
2179 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2180 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2181 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2182 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2183 int dev_restart(struct net_device
*dev
);
2184 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2186 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2188 return NAPI_GRO_CB(skb
)->data_offset
;
2191 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2193 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2196 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2198 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2201 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2202 unsigned int offset
)
2204 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2207 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2209 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2212 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2213 unsigned int offset
)
2215 if (!pskb_may_pull(skb
, hlen
))
2218 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2219 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2220 return skb
->data
+ offset
;
2223 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2225 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2226 skb_network_offset(skb
);
2229 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2230 const void *start
, unsigned int len
)
2232 if (NAPI_GRO_CB(skb
)->csum_valid
)
2233 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2234 csum_partial(start
, len
, 0));
2237 /* GRO checksum functions. These are logical equivalents of the normal
2238 * checksum functions (in skbuff.h) except that they operate on the GRO
2239 * offsets and fields in sk_buff.
2242 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2244 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2248 return (skb
->ip_summed
!= CHECKSUM_PARTIAL
&&
2249 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2250 (!zero_okay
|| check
));
2253 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2256 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2257 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2260 NAPI_GRO_CB(skb
)->csum
= psum
;
2262 return __skb_gro_checksum_complete(skb
);
2265 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2267 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2268 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2269 NAPI_GRO_CB(skb
)->csum_cnt
--;
2271 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2272 * verified a new top level checksum or an encapsulated one
2273 * during GRO. This saves work if we fallback to normal path.
2275 __skb_incr_checksum_unnecessary(skb
);
2279 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2282 __sum16 __ret = 0; \
2283 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2284 __ret = __skb_gro_checksum_validate_complete(skb, \
2285 compute_pseudo(skb, proto)); \
2287 __skb_mark_checksum_bad(skb); \
2289 skb_gro_incr_csum_unnecessary(skb); \
2293 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2294 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2296 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2298 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2300 #define skb_gro_checksum_simple_validate(skb) \
2301 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2303 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2305 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2306 !NAPI_GRO_CB(skb
)->csum_valid
);
2309 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2310 __sum16 check
, __wsum pseudo
)
2312 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2313 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2316 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2318 if (__skb_gro_checksum_convert_check(skb)) \
2319 __skb_gro_checksum_convert(skb, check, \
2320 compute_pseudo(skb, proto)); \
2323 static inline void skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2324 int start
, int offset
)
2328 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2330 delta
= remcsum_adjust(ptr
, NAPI_GRO_CB(skb
)->csum
, start
, offset
);
2332 /* Adjust skb->csum since we changed the packet */
2333 skb
->csum
= csum_add(skb
->csum
, delta
);
2334 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2338 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2339 unsigned short type
,
2340 const void *daddr
, const void *saddr
,
2343 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2346 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2349 static inline int dev_parse_header(const struct sk_buff
*skb
,
2350 unsigned char *haddr
)
2352 const struct net_device
*dev
= skb
->dev
;
2354 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2356 return dev
->header_ops
->parse(skb
, haddr
);
2359 static inline int dev_rebuild_header(struct sk_buff
*skb
)
2361 const struct net_device
*dev
= skb
->dev
;
2363 if (!dev
->header_ops
|| !dev
->header_ops
->rebuild
)
2365 return dev
->header_ops
->rebuild(skb
);
2368 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2369 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2370 static inline int unregister_gifconf(unsigned int family
)
2372 return register_gifconf(family
, NULL
);
2375 #ifdef CONFIG_NET_FLOW_LIMIT
2376 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2377 struct sd_flow_limit
{
2379 unsigned int num_buckets
;
2380 unsigned int history_head
;
2381 u16 history
[FLOW_LIMIT_HISTORY
];
2385 extern int netdev_flow_limit_table_len
;
2386 #endif /* CONFIG_NET_FLOW_LIMIT */
2389 * Incoming packets are placed on per-cpu queues
2391 struct softnet_data
{
2392 struct list_head poll_list
;
2393 struct sk_buff_head process_queue
;
2396 unsigned int processed
;
2397 unsigned int time_squeeze
;
2398 unsigned int cpu_collision
;
2399 unsigned int received_rps
;
2401 struct softnet_data
*rps_ipi_list
;
2403 #ifdef CONFIG_NET_FLOW_LIMIT
2404 struct sd_flow_limit __rcu
*flow_limit
;
2406 struct Qdisc
*output_queue
;
2407 struct Qdisc
**output_queue_tailp
;
2408 struct sk_buff
*completion_queue
;
2411 /* Elements below can be accessed between CPUs for RPS */
2412 struct call_single_data csd ____cacheline_aligned_in_smp
;
2413 struct softnet_data
*rps_ipi_next
;
2415 unsigned int input_queue_head
;
2416 unsigned int input_queue_tail
;
2418 unsigned int dropped
;
2419 struct sk_buff_head input_pkt_queue
;
2420 struct napi_struct backlog
;
2424 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2427 sd
->input_queue_head
++;
2431 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2432 unsigned int *qtail
)
2435 *qtail
= ++sd
->input_queue_tail
;
2439 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2441 void __netif_schedule(struct Qdisc
*q
);
2442 void netif_schedule_queue(struct netdev_queue
*txq
);
2444 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2448 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2449 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2452 static inline void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2454 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2458 * netif_start_queue - allow transmit
2459 * @dev: network device
2461 * Allow upper layers to call the device hard_start_xmit routine.
2463 static inline void netif_start_queue(struct net_device
*dev
)
2465 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2468 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2472 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2473 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2474 netif_tx_start_queue(txq
);
2478 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2481 * netif_wake_queue - restart transmit
2482 * @dev: network device
2484 * Allow upper layers to call the device hard_start_xmit routine.
2485 * Used for flow control when transmit resources are available.
2487 static inline void netif_wake_queue(struct net_device
*dev
)
2489 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2492 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2496 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2497 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2498 netif_tx_wake_queue(txq
);
2502 static inline void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2504 if (WARN_ON(!dev_queue
)) {
2505 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2508 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2512 * netif_stop_queue - stop transmitted packets
2513 * @dev: network device
2515 * Stop upper layers calling the device hard_start_xmit routine.
2516 * Used for flow control when transmit resources are unavailable.
2518 static inline void netif_stop_queue(struct net_device
*dev
)
2520 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2523 static inline void netif_tx_stop_all_queues(struct net_device
*dev
)
2527 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2528 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2529 netif_tx_stop_queue(txq
);
2533 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2535 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2539 * netif_queue_stopped - test if transmit queue is flowblocked
2540 * @dev: network device
2542 * Test if transmit queue on device is currently unable to send.
2544 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2546 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2549 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2551 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2555 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2557 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2561 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2563 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2567 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2568 * @dev_queue: pointer to transmit queue
2570 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2571 * to give appropriate hint to the cpu.
2573 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2576 prefetchw(&dev_queue
->dql
.num_queued
);
2581 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2582 * @dev_queue: pointer to transmit queue
2584 * BQL enabled drivers might use this helper in their TX completion path,
2585 * to give appropriate hint to the cpu.
2587 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2590 prefetchw(&dev_queue
->dql
.limit
);
2594 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2598 dql_queued(&dev_queue
->dql
, bytes
);
2600 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2603 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2606 * The XOFF flag must be set before checking the dql_avail below,
2607 * because in netdev_tx_completed_queue we update the dql_completed
2608 * before checking the XOFF flag.
2612 /* check again in case another CPU has just made room avail */
2613 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2614 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2619 * netdev_sent_queue - report the number of bytes queued to hardware
2620 * @dev: network device
2621 * @bytes: number of bytes queued to the hardware device queue
2623 * Report the number of bytes queued for sending/completion to the network
2624 * device hardware queue. @bytes should be a good approximation and should
2625 * exactly match netdev_completed_queue() @bytes
2627 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2629 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2632 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2633 unsigned int pkts
, unsigned int bytes
)
2636 if (unlikely(!bytes
))
2639 dql_completed(&dev_queue
->dql
, bytes
);
2642 * Without the memory barrier there is a small possiblity that
2643 * netdev_tx_sent_queue will miss the update and cause the queue to
2644 * be stopped forever
2648 if (dql_avail(&dev_queue
->dql
) < 0)
2651 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2652 netif_schedule_queue(dev_queue
);
2657 * netdev_completed_queue - report bytes and packets completed by device
2658 * @dev: network device
2659 * @pkts: actual number of packets sent over the medium
2660 * @bytes: actual number of bytes sent over the medium
2662 * Report the number of bytes and packets transmitted by the network device
2663 * hardware queue over the physical medium, @bytes must exactly match the
2664 * @bytes amount passed to netdev_sent_queue()
2666 static inline void netdev_completed_queue(struct net_device
*dev
,
2667 unsigned int pkts
, unsigned int bytes
)
2669 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
2672 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
2675 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
2681 * netdev_reset_queue - reset the packets and bytes count of a network device
2682 * @dev_queue: network device
2684 * Reset the bytes and packet count of a network device and clear the
2685 * software flow control OFF bit for this network device
2687 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
2689 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
2693 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2694 * @dev: network device
2695 * @queue_index: given tx queue index
2697 * Returns 0 if given tx queue index >= number of device tx queues,
2698 * otherwise returns the originally passed tx queue index.
2700 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
2702 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
2703 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2704 dev
->name
, queue_index
,
2705 dev
->real_num_tx_queues
);
2713 * netif_running - test if up
2714 * @dev: network device
2716 * Test if the device has been brought up.
2718 static inline bool netif_running(const struct net_device
*dev
)
2720 return test_bit(__LINK_STATE_START
, &dev
->state
);
2724 * Routines to manage the subqueues on a device. We only need start
2725 * stop, and a check if it's stopped. All other device management is
2726 * done at the overall netdevice level.
2727 * Also test the device if we're multiqueue.
2731 * netif_start_subqueue - allow sending packets on subqueue
2732 * @dev: network device
2733 * @queue_index: sub queue index
2735 * Start individual transmit queue of a device with multiple transmit queues.
2737 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
2739 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2741 netif_tx_start_queue(txq
);
2745 * netif_stop_subqueue - stop sending packets on subqueue
2746 * @dev: network device
2747 * @queue_index: sub queue index
2749 * Stop individual transmit queue of a device with multiple transmit queues.
2751 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
2753 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2754 netif_tx_stop_queue(txq
);
2758 * netif_subqueue_stopped - test status of subqueue
2759 * @dev: network device
2760 * @queue_index: sub queue index
2762 * Check individual transmit queue of a device with multiple transmit queues.
2764 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
2767 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2769 return netif_tx_queue_stopped(txq
);
2772 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
2773 struct sk_buff
*skb
)
2775 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
2778 void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
);
2781 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
2784 static inline int netif_set_xps_queue(struct net_device
*dev
,
2785 const struct cpumask
*mask
,
2793 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2794 * as a distribution range limit for the returned value.
2796 static inline u16
skb_tx_hash(const struct net_device
*dev
,
2797 struct sk_buff
*skb
)
2799 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
2803 * netif_is_multiqueue - test if device has multiple transmit queues
2804 * @dev: network device
2806 * Check if device has multiple transmit queues
2808 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
2810 return dev
->num_tx_queues
> 1;
2813 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
2816 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
2818 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
2826 static inline unsigned int get_netdev_rx_queue_index(
2827 struct netdev_rx_queue
*queue
)
2829 struct net_device
*dev
= queue
->dev
;
2830 int index
= queue
- dev
->_rx
;
2832 BUG_ON(index
>= dev
->num_rx_queues
);
2837 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2838 int netif_get_num_default_rss_queues(void);
2840 enum skb_free_reason
{
2841 SKB_REASON_CONSUMED
,
2845 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
2846 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
2849 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2850 * interrupt context or with hardware interrupts being disabled.
2851 * (in_irq() || irqs_disabled())
2853 * We provide four helpers that can be used in following contexts :
2855 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2856 * replacing kfree_skb(skb)
2858 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2859 * Typically used in place of consume_skb(skb) in TX completion path
2861 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2862 * replacing kfree_skb(skb)
2864 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2865 * and consumed a packet. Used in place of consume_skb(skb)
2867 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
2869 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
2872 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
2874 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
2877 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
2879 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
2882 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
2884 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
2887 int netif_rx(struct sk_buff
*skb
);
2888 int netif_rx_ni(struct sk_buff
*skb
);
2889 int netif_receive_skb(struct sk_buff
*skb
);
2890 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
2891 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
2892 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
2893 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
2894 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
2895 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
2897 static inline void napi_free_frags(struct napi_struct
*napi
)
2899 kfree_skb(napi
->skb
);
2903 int netdev_rx_handler_register(struct net_device
*dev
,
2904 rx_handler_func_t
*rx_handler
,
2905 void *rx_handler_data
);
2906 void netdev_rx_handler_unregister(struct net_device
*dev
);
2908 bool dev_valid_name(const char *name
);
2909 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
2910 int dev_ethtool(struct net
*net
, struct ifreq
*);
2911 unsigned int dev_get_flags(const struct net_device
*);
2912 int __dev_change_flags(struct net_device
*, unsigned int flags
);
2913 int dev_change_flags(struct net_device
*, unsigned int);
2914 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
2915 unsigned int gchanges
);
2916 int dev_change_name(struct net_device
*, const char *);
2917 int dev_set_alias(struct net_device
*, const char *, size_t);
2918 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
2919 int dev_set_mtu(struct net_device
*, int);
2920 void dev_set_group(struct net_device
*, int);
2921 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
2922 int dev_change_carrier(struct net_device
*, bool new_carrier
);
2923 int dev_get_phys_port_id(struct net_device
*dev
,
2924 struct netdev_phys_item_id
*ppid
);
2925 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
2926 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
2927 struct netdev_queue
*txq
, int *ret
);
2928 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2929 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2930 bool is_skb_forwardable(struct net_device
*dev
, struct sk_buff
*skb
);
2932 extern int netdev_budget
;
2934 /* Called by rtnetlink.c:rtnl_unlock() */
2935 void netdev_run_todo(void);
2938 * dev_put - release reference to device
2939 * @dev: network device
2941 * Release reference to device to allow it to be freed.
2943 static inline void dev_put(struct net_device
*dev
)
2945 this_cpu_dec(*dev
->pcpu_refcnt
);
2949 * dev_hold - get reference to device
2950 * @dev: network device
2952 * Hold reference to device to keep it from being freed.
2954 static inline void dev_hold(struct net_device
*dev
)
2956 this_cpu_inc(*dev
->pcpu_refcnt
);
2959 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2960 * and _off may be called from IRQ context, but it is caller
2961 * who is responsible for serialization of these calls.
2963 * The name carrier is inappropriate, these functions should really be
2964 * called netif_lowerlayer_*() because they represent the state of any
2965 * kind of lower layer not just hardware media.
2968 void linkwatch_init_dev(struct net_device
*dev
);
2969 void linkwatch_fire_event(struct net_device
*dev
);
2970 void linkwatch_forget_dev(struct net_device
*dev
);
2973 * netif_carrier_ok - test if carrier present
2974 * @dev: network device
2976 * Check if carrier is present on device
2978 static inline bool netif_carrier_ok(const struct net_device
*dev
)
2980 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
2983 unsigned long dev_trans_start(struct net_device
*dev
);
2985 void __netdev_watchdog_up(struct net_device
*dev
);
2987 void netif_carrier_on(struct net_device
*dev
);
2989 void netif_carrier_off(struct net_device
*dev
);
2992 * netif_dormant_on - mark device as dormant.
2993 * @dev: network device
2995 * Mark device as dormant (as per RFC2863).
2997 * The dormant state indicates that the relevant interface is not
2998 * actually in a condition to pass packets (i.e., it is not 'up') but is
2999 * in a "pending" state, waiting for some external event. For "on-
3000 * demand" interfaces, this new state identifies the situation where the
3001 * interface is waiting for events to place it in the up state.
3004 static inline void netif_dormant_on(struct net_device
*dev
)
3006 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3007 linkwatch_fire_event(dev
);
3011 * netif_dormant_off - set device as not dormant.
3012 * @dev: network device
3014 * Device is not in dormant state.
3016 static inline void netif_dormant_off(struct net_device
*dev
)
3018 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3019 linkwatch_fire_event(dev
);
3023 * netif_dormant - test if carrier present
3024 * @dev: network device
3026 * Check if carrier is present on device
3028 static inline bool netif_dormant(const struct net_device
*dev
)
3030 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3035 * netif_oper_up - test if device is operational
3036 * @dev: network device
3038 * Check if carrier is operational
3040 static inline bool netif_oper_up(const struct net_device
*dev
)
3042 return (dev
->operstate
== IF_OPER_UP
||
3043 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3047 * netif_device_present - is device available or removed
3048 * @dev: network device
3050 * Check if device has not been removed from system.
3052 static inline bool netif_device_present(struct net_device
*dev
)
3054 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3057 void netif_device_detach(struct net_device
*dev
);
3059 void netif_device_attach(struct net_device
*dev
);
3062 * Network interface message level settings
3066 NETIF_MSG_DRV
= 0x0001,
3067 NETIF_MSG_PROBE
= 0x0002,
3068 NETIF_MSG_LINK
= 0x0004,
3069 NETIF_MSG_TIMER
= 0x0008,
3070 NETIF_MSG_IFDOWN
= 0x0010,
3071 NETIF_MSG_IFUP
= 0x0020,
3072 NETIF_MSG_RX_ERR
= 0x0040,
3073 NETIF_MSG_TX_ERR
= 0x0080,
3074 NETIF_MSG_TX_QUEUED
= 0x0100,
3075 NETIF_MSG_INTR
= 0x0200,
3076 NETIF_MSG_TX_DONE
= 0x0400,
3077 NETIF_MSG_RX_STATUS
= 0x0800,
3078 NETIF_MSG_PKTDATA
= 0x1000,
3079 NETIF_MSG_HW
= 0x2000,
3080 NETIF_MSG_WOL
= 0x4000,
3083 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3084 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3085 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3086 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3087 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3088 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3089 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3090 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3091 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3092 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3093 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3094 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3095 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3096 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3097 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3099 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3102 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3103 return default_msg_enable_bits
;
3104 if (debug_value
== 0) /* no output */
3106 /* set low N bits */
3107 return (1 << debug_value
) - 1;
3110 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3112 spin_lock(&txq
->_xmit_lock
);
3113 txq
->xmit_lock_owner
= cpu
;
3116 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3118 spin_lock_bh(&txq
->_xmit_lock
);
3119 txq
->xmit_lock_owner
= smp_processor_id();
3122 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3124 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3126 txq
->xmit_lock_owner
= smp_processor_id();
3130 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3132 txq
->xmit_lock_owner
= -1;
3133 spin_unlock(&txq
->_xmit_lock
);
3136 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3138 txq
->xmit_lock_owner
= -1;
3139 spin_unlock_bh(&txq
->_xmit_lock
);
3142 static inline void txq_trans_update(struct netdev_queue
*txq
)
3144 if (txq
->xmit_lock_owner
!= -1)
3145 txq
->trans_start
= jiffies
;
3149 * netif_tx_lock - grab network device transmit lock
3150 * @dev: network device
3152 * Get network device transmit lock
3154 static inline void netif_tx_lock(struct net_device
*dev
)
3159 spin_lock(&dev
->tx_global_lock
);
3160 cpu
= smp_processor_id();
3161 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3162 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3164 /* We are the only thread of execution doing a
3165 * freeze, but we have to grab the _xmit_lock in
3166 * order to synchronize with threads which are in
3167 * the ->hard_start_xmit() handler and already
3168 * checked the frozen bit.
3170 __netif_tx_lock(txq
, cpu
);
3171 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3172 __netif_tx_unlock(txq
);
3176 static inline void netif_tx_lock_bh(struct net_device
*dev
)
3182 static inline void netif_tx_unlock(struct net_device
*dev
)
3186 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3187 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3189 /* No need to grab the _xmit_lock here. If the
3190 * queue is not stopped for another reason, we
3193 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3194 netif_schedule_queue(txq
);
3196 spin_unlock(&dev
->tx_global_lock
);
3199 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
3201 netif_tx_unlock(dev
);
3205 #define HARD_TX_LOCK(dev, txq, cpu) { \
3206 if ((dev->features & NETIF_F_LLTX) == 0) { \
3207 __netif_tx_lock(txq, cpu); \
3211 #define HARD_TX_TRYLOCK(dev, txq) \
3212 (((dev->features & NETIF_F_LLTX) == 0) ? \
3213 __netif_tx_trylock(txq) : \
3216 #define HARD_TX_UNLOCK(dev, txq) { \
3217 if ((dev->features & NETIF_F_LLTX) == 0) { \
3218 __netif_tx_unlock(txq); \
3222 static inline void netif_tx_disable(struct net_device
*dev
)
3228 cpu
= smp_processor_id();
3229 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3230 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3232 __netif_tx_lock(txq
, cpu
);
3233 netif_tx_stop_queue(txq
);
3234 __netif_tx_unlock(txq
);
3239 static inline void netif_addr_lock(struct net_device
*dev
)
3241 spin_lock(&dev
->addr_list_lock
);
3244 static inline void netif_addr_lock_nested(struct net_device
*dev
)
3246 int subclass
= SINGLE_DEPTH_NESTING
;
3248 if (dev
->netdev_ops
->ndo_get_lock_subclass
)
3249 subclass
= dev
->netdev_ops
->ndo_get_lock_subclass(dev
);
3251 spin_lock_nested(&dev
->addr_list_lock
, subclass
);
3254 static inline void netif_addr_lock_bh(struct net_device
*dev
)
3256 spin_lock_bh(&dev
->addr_list_lock
);
3259 static inline void netif_addr_unlock(struct net_device
*dev
)
3261 spin_unlock(&dev
->addr_list_lock
);
3264 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
3266 spin_unlock_bh(&dev
->addr_list_lock
);
3270 * dev_addrs walker. Should be used only for read access. Call with
3271 * rcu_read_lock held.
3273 #define for_each_dev_addr(dev, ha) \
3274 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3276 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3278 void ether_setup(struct net_device
*dev
);
3280 /* Support for loadable net-drivers */
3281 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
3282 unsigned char name_assign_type
,
3283 void (*setup
)(struct net_device
*),
3284 unsigned int txqs
, unsigned int rxqs
);
3285 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3286 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3288 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3289 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3292 int register_netdev(struct net_device
*dev
);
3293 void unregister_netdev(struct net_device
*dev
);
3295 /* General hardware address lists handling functions */
3296 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
3297 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3298 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
3299 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3300 int __hw_addr_sync_dev(struct netdev_hw_addr_list
*list
,
3301 struct net_device
*dev
,
3302 int (*sync
)(struct net_device
*, const unsigned char *),
3303 int (*unsync
)(struct net_device
*,
3304 const unsigned char *));
3305 void __hw_addr_unsync_dev(struct netdev_hw_addr_list
*list
,
3306 struct net_device
*dev
,
3307 int (*unsync
)(struct net_device
*,
3308 const unsigned char *));
3309 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
3311 /* Functions used for device addresses handling */
3312 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
3313 unsigned char addr_type
);
3314 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
3315 unsigned char addr_type
);
3316 void dev_addr_flush(struct net_device
*dev
);
3317 int dev_addr_init(struct net_device
*dev
);
3319 /* Functions used for unicast addresses handling */
3320 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
3321 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3322 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
3323 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
3324 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3325 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
3326 void dev_uc_flush(struct net_device
*dev
);
3327 void dev_uc_init(struct net_device
*dev
);
3330 * __dev_uc_sync - Synchonize device's unicast list
3331 * @dev: device to sync
3332 * @sync: function to call if address should be added
3333 * @unsync: function to call if address should be removed
3335 * Add newly added addresses to the interface, and release
3336 * addresses that have been deleted.
3338 static inline int __dev_uc_sync(struct net_device
*dev
,
3339 int (*sync
)(struct net_device
*,
3340 const unsigned char *),
3341 int (*unsync
)(struct net_device
*,
3342 const unsigned char *))
3344 return __hw_addr_sync_dev(&dev
->uc
, dev
, sync
, unsync
);
3348 * __dev_uc_unsync - Remove synchronized addresses from device
3349 * @dev: device to sync
3350 * @unsync: function to call if address should be removed
3352 * Remove all addresses that were added to the device by dev_uc_sync().
3354 static inline void __dev_uc_unsync(struct net_device
*dev
,
3355 int (*unsync
)(struct net_device
*,
3356 const unsigned char *))
3358 __hw_addr_unsync_dev(&dev
->uc
, dev
, unsync
);
3361 /* Functions used for multicast addresses handling */
3362 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
3363 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
3364 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3365 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
3366 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
3367 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
3368 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3369 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
3370 void dev_mc_flush(struct net_device
*dev
);
3371 void dev_mc_init(struct net_device
*dev
);
3374 * __dev_mc_sync - Synchonize device's multicast list
3375 * @dev: device to sync
3376 * @sync: function to call if address should be added
3377 * @unsync: function to call if address should be removed
3379 * Add newly added addresses to the interface, and release
3380 * addresses that have been deleted.
3382 static inline int __dev_mc_sync(struct net_device
*dev
,
3383 int (*sync
)(struct net_device
*,
3384 const unsigned char *),
3385 int (*unsync
)(struct net_device
*,
3386 const unsigned char *))
3388 return __hw_addr_sync_dev(&dev
->mc
, dev
, sync
, unsync
);
3392 * __dev_mc_unsync - Remove synchronized addresses from device
3393 * @dev: device to sync
3394 * @unsync: function to call if address should be removed
3396 * Remove all addresses that were added to the device by dev_mc_sync().
3398 static inline void __dev_mc_unsync(struct net_device
*dev
,
3399 int (*unsync
)(struct net_device
*,
3400 const unsigned char *))
3402 __hw_addr_unsync_dev(&dev
->mc
, dev
, unsync
);
3405 /* Functions used for secondary unicast and multicast support */
3406 void dev_set_rx_mode(struct net_device
*dev
);
3407 void __dev_set_rx_mode(struct net_device
*dev
);
3408 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
3409 int dev_set_allmulti(struct net_device
*dev
, int inc
);
3410 void netdev_state_change(struct net_device
*dev
);
3411 void netdev_notify_peers(struct net_device
*dev
);
3412 void netdev_features_change(struct net_device
*dev
);
3413 /* Load a device via the kmod */
3414 void dev_load(struct net
*net
, const char *name
);
3415 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
3416 struct rtnl_link_stats64
*storage
);
3417 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
3418 const struct net_device_stats
*netdev_stats
);
3420 extern int netdev_max_backlog
;
3421 extern int netdev_tstamp_prequeue
;
3422 extern int weight_p
;
3423 extern int bpf_jit_enable
;
3425 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3426 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3427 struct list_head
**iter
);
3428 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3429 struct list_head
**iter
);
3431 /* iterate through upper list, must be called under RCU read lock */
3432 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3433 for (iter = &(dev)->adj_list.upper, \
3434 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3436 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3438 /* iterate through upper list, must be called under RCU read lock */
3439 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3440 for (iter = &(dev)->all_adj_list.upper, \
3441 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3443 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3445 void *netdev_lower_get_next_private(struct net_device
*dev
,
3446 struct list_head
**iter
);
3447 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3448 struct list_head
**iter
);
3450 #define netdev_for_each_lower_private(dev, priv, iter) \
3451 for (iter = (dev)->adj_list.lower.next, \
3452 priv = netdev_lower_get_next_private(dev, &(iter)); \
3454 priv = netdev_lower_get_next_private(dev, &(iter)))
3456 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3457 for (iter = &(dev)->adj_list.lower, \
3458 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3460 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3462 void *netdev_lower_get_next(struct net_device
*dev
,
3463 struct list_head
**iter
);
3464 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3465 for (iter = &(dev)->adj_list.lower, \
3466 ldev = netdev_lower_get_next(dev, &(iter)); \
3468 ldev = netdev_lower_get_next(dev, &(iter)))
3470 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3471 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3472 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3473 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3474 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
3475 int netdev_master_upper_dev_link(struct net_device
*dev
,
3476 struct net_device
*upper_dev
);
3477 int netdev_master_upper_dev_link_private(struct net_device
*dev
,
3478 struct net_device
*upper_dev
,
3480 void netdev_upper_dev_unlink(struct net_device
*dev
,
3481 struct net_device
*upper_dev
);
3482 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3483 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3484 struct net_device
*lower_dev
);
3486 /* RSS keys are 40 or 52 bytes long */
3487 #define NETDEV_RSS_KEY_LEN 52
3488 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
];
3489 void netdev_rss_key_fill(void *buffer
, size_t len
);
3491 int dev_get_nest_level(struct net_device
*dev
,
3492 bool (*type_check
)(struct net_device
*dev
));
3493 int skb_checksum_help(struct sk_buff
*skb
);
3494 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3495 netdev_features_t features
, bool tx_path
);
3496 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3497 netdev_features_t features
);
3499 struct netdev_bonding_info
{
3504 struct netdev_notifier_bonding_info
{
3505 struct netdev_notifier_info info
; /* must be first */
3506 struct netdev_bonding_info bonding_info
;
3509 void netdev_bonding_info_change(struct net_device
*dev
,
3510 struct netdev_bonding_info
*bonding_info
);
3513 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3515 return __skb_gso_segment(skb
, features
, true);
3517 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3519 static inline bool can_checksum_protocol(netdev_features_t features
,
3522 return ((features
& NETIF_F_GEN_CSUM
) ||
3523 ((features
& NETIF_F_V4_CSUM
) &&
3524 protocol
== htons(ETH_P_IP
)) ||
3525 ((features
& NETIF_F_V6_CSUM
) &&
3526 protocol
== htons(ETH_P_IPV6
)) ||
3527 ((features
& NETIF_F_FCOE_CRC
) &&
3528 protocol
== htons(ETH_P_FCOE
)));
3532 void netdev_rx_csum_fault(struct net_device
*dev
);
3534 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3538 /* rx skb timestamps */
3539 void net_enable_timestamp(void);
3540 void net_disable_timestamp(void);
3542 #ifdef CONFIG_PROC_FS
3543 int __init
dev_proc_init(void);
3545 #define dev_proc_init() 0
3548 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
3549 struct sk_buff
*skb
, struct net_device
*dev
,
3552 skb
->xmit_more
= more
? 1 : 0;
3553 return ops
->ndo_start_xmit(skb
, dev
);
3556 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3557 struct netdev_queue
*txq
, bool more
)
3559 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3562 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
3563 if (rc
== NETDEV_TX_OK
)
3564 txq_trans_update(txq
);
3569 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3571 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3574 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3576 return netdev_class_create_file_ns(class_attr
, NULL
);
3579 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3581 netdev_class_remove_file_ns(class_attr
, NULL
);
3584 extern struct kobj_ns_type_operations net_ns_type_operations
;
3586 const char *netdev_drivername(const struct net_device
*dev
);
3588 void linkwatch_run_queue(void);
3590 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
3591 netdev_features_t f2
)
3593 if (f1
& NETIF_F_GEN_CSUM
)
3594 f1
|= (NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3595 if (f2
& NETIF_F_GEN_CSUM
)
3596 f2
|= (NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3598 if (f1
& NETIF_F_GEN_CSUM
)
3599 f1
&= ~(NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3604 static inline netdev_features_t
netdev_get_wanted_features(
3605 struct net_device
*dev
)
3607 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
3609 netdev_features_t
netdev_increment_features(netdev_features_t all
,
3610 netdev_features_t one
, netdev_features_t mask
);
3612 /* Allow TSO being used on stacked device :
3613 * Performing the GSO segmentation before last device
3614 * is a performance improvement.
3616 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
3617 netdev_features_t mask
)
3619 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
3622 int __netdev_update_features(struct net_device
*dev
);
3623 void netdev_update_features(struct net_device
*dev
);
3624 void netdev_change_features(struct net_device
*dev
);
3626 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
3627 struct net_device
*dev
);
3629 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
3631 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
3633 netdev_features_t feature
= gso_type
<< NETIF_F_GSO_SHIFT
;
3635 /* check flags correspondence */
3636 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
3637 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
3638 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
3639 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
3640 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
3641 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
3642 BUILD_BUG_ON(SKB_GSO_GRE
!= (NETIF_F_GSO_GRE
>> NETIF_F_GSO_SHIFT
));
3643 BUILD_BUG_ON(SKB_GSO_GRE_CSUM
!= (NETIF_F_GSO_GRE_CSUM
>> NETIF_F_GSO_SHIFT
));
3644 BUILD_BUG_ON(SKB_GSO_IPIP
!= (NETIF_F_GSO_IPIP
>> NETIF_F_GSO_SHIFT
));
3645 BUILD_BUG_ON(SKB_GSO_SIT
!= (NETIF_F_GSO_SIT
>> NETIF_F_GSO_SHIFT
));
3646 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL
!= (NETIF_F_GSO_UDP_TUNNEL
>> NETIF_F_GSO_SHIFT
));
3647 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM
!= (NETIF_F_GSO_UDP_TUNNEL_CSUM
>> NETIF_F_GSO_SHIFT
));
3648 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
3650 return (features
& feature
) == feature
;
3653 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
3655 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
3656 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
3659 static inline bool netif_needs_gso(struct net_device
*dev
, struct sk_buff
*skb
,
3660 netdev_features_t features
)
3662 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
3663 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
3664 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
3667 static inline void netif_set_gso_max_size(struct net_device
*dev
,
3670 dev
->gso_max_size
= size
;
3673 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
3674 int pulled_hlen
, u16 mac_offset
,
3677 skb
->protocol
= protocol
;
3678 skb
->encapsulation
= 1;
3679 skb_push(skb
, pulled_hlen
);
3680 skb_reset_transport_header(skb
);
3681 skb
->mac_header
= mac_offset
;
3682 skb
->network_header
= skb
->mac_header
+ mac_len
;
3683 skb
->mac_len
= mac_len
;
3686 static inline bool netif_is_macvlan(struct net_device
*dev
)
3688 return dev
->priv_flags
& IFF_MACVLAN
;
3691 static inline bool netif_is_macvlan_port(struct net_device
*dev
)
3693 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
3696 static inline bool netif_is_ipvlan(struct net_device
*dev
)
3698 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
3701 static inline bool netif_is_ipvlan_port(struct net_device
*dev
)
3703 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
3706 static inline bool netif_is_bond_master(struct net_device
*dev
)
3708 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
3711 static inline bool netif_is_bond_slave(struct net_device
*dev
)
3713 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
3716 static inline bool netif_supports_nofcs(struct net_device
*dev
)
3718 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
3721 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3722 static inline void netif_keep_dst(struct net_device
*dev
)
3724 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
3727 extern struct pernet_operations __net_initdata loopback_net_ops
;
3729 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3731 /* netdev_printk helpers, similar to dev_printk */
3733 static inline const char *netdev_name(const struct net_device
*dev
)
3735 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
3736 return "(unnamed net_device)";
3740 static inline const char *netdev_reg_state(const struct net_device
*dev
)
3742 switch (dev
->reg_state
) {
3743 case NETREG_UNINITIALIZED
: return " (uninitialized)";
3744 case NETREG_REGISTERED
: return "";
3745 case NETREG_UNREGISTERING
: return " (unregistering)";
3746 case NETREG_UNREGISTERED
: return " (unregistered)";
3747 case NETREG_RELEASED
: return " (released)";
3748 case NETREG_DUMMY
: return " (dummy)";
3751 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
3752 return " (unknown)";
3756 void netdev_printk(const char *level
, const struct net_device
*dev
,
3757 const char *format
, ...);
3759 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
3761 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
3763 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
3765 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
3767 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
3769 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
3771 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
3773 #define MODULE_ALIAS_NETDEV(device) \
3774 MODULE_ALIAS("netdev-" device)
3776 #if defined(CONFIG_DYNAMIC_DEBUG)
3777 #define netdev_dbg(__dev, format, args...) \
3779 dynamic_netdev_dbg(__dev, format, ##args); \
3781 #elif defined(DEBUG)
3782 #define netdev_dbg(__dev, format, args...) \
3783 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3785 #define netdev_dbg(__dev, format, args...) \
3788 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3792 #if defined(VERBOSE_DEBUG)
3793 #define netdev_vdbg netdev_dbg
3796 #define netdev_vdbg(dev, format, args...) \
3799 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3805 * netdev_WARN() acts like dev_printk(), but with the key difference
3806 * of using a WARN/WARN_ON to get the message out, including the
3807 * file/line information and a backtrace.
3809 #define netdev_WARN(dev, format, args...) \
3810 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3811 netdev_reg_state(dev), ##args)
3813 /* netif printk helpers, similar to netdev_printk */
3815 #define netif_printk(priv, type, level, dev, fmt, args...) \
3817 if (netif_msg_##type(priv)) \
3818 netdev_printk(level, (dev), fmt, ##args); \
3821 #define netif_level(level, priv, type, dev, fmt, args...) \
3823 if (netif_msg_##type(priv)) \
3824 netdev_##level(dev, fmt, ##args); \
3827 #define netif_emerg(priv, type, dev, fmt, args...) \
3828 netif_level(emerg, priv, type, dev, fmt, ##args)
3829 #define netif_alert(priv, type, dev, fmt, args...) \
3830 netif_level(alert, priv, type, dev, fmt, ##args)
3831 #define netif_crit(priv, type, dev, fmt, args...) \
3832 netif_level(crit, priv, type, dev, fmt, ##args)
3833 #define netif_err(priv, type, dev, fmt, args...) \
3834 netif_level(err, priv, type, dev, fmt, ##args)
3835 #define netif_warn(priv, type, dev, fmt, args...) \
3836 netif_level(warn, priv, type, dev, fmt, ##args)
3837 #define netif_notice(priv, type, dev, fmt, args...) \
3838 netif_level(notice, priv, type, dev, fmt, ##args)
3839 #define netif_info(priv, type, dev, fmt, args...) \
3840 netif_level(info, priv, type, dev, fmt, ##args)
3842 #if defined(CONFIG_DYNAMIC_DEBUG)
3843 #define netif_dbg(priv, type, netdev, format, args...) \
3845 if (netif_msg_##type(priv)) \
3846 dynamic_netdev_dbg(netdev, format, ##args); \
3848 #elif defined(DEBUG)
3849 #define netif_dbg(priv, type, dev, format, args...) \
3850 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3852 #define netif_dbg(priv, type, dev, format, args...) \
3855 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3860 #if defined(VERBOSE_DEBUG)
3861 #define netif_vdbg netif_dbg
3863 #define netif_vdbg(priv, type, dev, format, args...) \
3866 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3872 * The list of packet types we will receive (as opposed to discard)
3873 * and the routines to invoke.
3875 * Why 16. Because with 16 the only overlap we get on a hash of the
3876 * low nibble of the protocol value is RARP/SNAP/X.25.
3878 * NOTE: That is no longer true with the addition of VLAN tags. Not
3879 * sure which should go first, but I bet it won't make much
3880 * difference if we are running VLANs. The good news is that
3881 * this protocol won't be in the list unless compiled in, so
3882 * the average user (w/out VLANs) will not be adversely affected.
3898 #define PTYPE_HASH_SIZE (16)
3899 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3901 #endif /* _LINUX_NETDEVICE_H */