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).
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
654 struct rps_sock_flow_table
{
657 u32 ents
[0] ____cacheline_aligned_in_smp
;
659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
661 #define RPS_NO_CPU 0xffff
663 extern u32 rps_cpu_mask
;
664 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
666 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
670 unsigned int index
= hash
& table
->mask
;
671 u32 val
= hash
& ~rps_cpu_mask
;
673 /* We only give a hint, preemption can change cpu under us */
674 val
|= raw_smp_processor_id();
676 if (table
->ents
[index
] != val
)
677 table
->ents
[index
] = val
;
681 #ifdef CONFIG_RFS_ACCEL
682 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
685 #endif /* CONFIG_RPS */
687 /* This structure contains an instance of an RX queue. */
688 struct netdev_rx_queue
{
690 struct rps_map __rcu
*rps_map
;
691 struct rps_dev_flow_table __rcu
*rps_flow_table
;
694 struct net_device
*dev
;
695 } ____cacheline_aligned_in_smp
;
698 * RX queue sysfs structures and functions.
700 struct rx_queue_attribute
{
701 struct attribute attr
;
702 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
703 struct rx_queue_attribute
*attr
, char *buf
);
704 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
705 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
715 unsigned int alloc_len
;
719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
726 struct xps_dev_maps
{
728 struct xps_map __rcu
*cpu_map
[0];
730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732 #endif /* CONFIG_XPS */
734 #define TC_MAX_QUEUE 16
735 #define TC_BITMASK 15
736 /* HW offloaded queuing disciplines txq count and offset maps */
737 struct netdev_tc_txq
{
742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
747 struct netdev_fcoe_hbainfo
{
748 char manufacturer
[64];
749 char serial_number
[64];
750 char hardware_version
[64];
751 char driver_version
[64];
752 char optionrom_version
[64];
753 char firmware_version
[64];
755 char model_description
[256];
759 #define MAX_PHYS_ITEM_ID_LEN 32
761 /* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
764 struct netdev_phys_item_id
{
765 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
766 unsigned char id_len
;
769 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
770 struct sk_buff
*skb
);
773 * This structure defines the management hooks for network devices.
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
797 * Called when a packet needs to be transmitted.
798 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
799 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
800 * Required can not be NULL.
802 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
803 * void *accel_priv, select_queue_fallback_t fallback);
804 * Called to decide which queue to when device supports multiple
807 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
808 * This function is called to allow device receiver to make
809 * changes to configuration when multicast or promiscious is enabled.
811 * void (*ndo_set_rx_mode)(struct net_device *dev);
812 * This function is called device changes address list filtering.
813 * If driver handles unicast address filtering, it should set
814 * IFF_UNICAST_FLT to its priv_flags.
816 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
817 * This function is called when the Media Access Control address
818 * needs to be changed. If this interface is not defined, the
819 * mac address can not be changed.
821 * int (*ndo_validate_addr)(struct net_device *dev);
822 * Test if Media Access Control address is valid for the device.
824 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
825 * Called when a user request an ioctl which can't be handled by
826 * the generic interface code. If not defined ioctl's return
827 * not supported error code.
829 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
830 * Used to set network devices bus interface parameters. This interface
831 * is retained for legacy reason, new devices should use the bus
832 * interface (PCI) for low level management.
834 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
835 * Called when a user wants to change the Maximum Transfer Unit
836 * of a device. If not defined, any request to change MTU will
837 * will return an error.
839 * void (*ndo_tx_timeout)(struct net_device *dev);
840 * Callback uses when the transmitter has not made any progress
841 * for dev->watchdog ticks.
843 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
844 * struct rtnl_link_stats64 *storage);
845 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
846 * Called when a user wants to get the network device usage
847 * statistics. Drivers must do one of the following:
848 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
849 * rtnl_link_stats64 structure passed by the caller.
850 * 2. Define @ndo_get_stats to update a net_device_stats structure
851 * (which should normally be dev->stats) and return a pointer to
852 * it. The structure may be changed asynchronously only if each
853 * field is written atomically.
854 * 3. Update dev->stats asynchronously and atomically, and define
857 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
858 * If device support VLAN filtering this function is called when a
859 * VLAN id is registered.
861 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
862 * If device support VLAN filtering this function is called when a
863 * VLAN id is unregistered.
865 * void (*ndo_poll_controller)(struct net_device *dev);
867 * SR-IOV management functions.
868 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
869 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
870 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
872 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
873 * int (*ndo_get_vf_config)(struct net_device *dev,
874 * int vf, struct ifla_vf_info *ivf);
875 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
876 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
877 * struct nlattr *port[]);
878 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
879 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
880 * Called to setup 'tc' number of traffic classes in the net device. This
881 * is always called from the stack with the rtnl lock held and netif tx
882 * queues stopped. This allows the netdevice to perform queue management
885 * Fiber Channel over Ethernet (FCoE) offload functions.
886 * int (*ndo_fcoe_enable)(struct net_device *dev);
887 * Called when the FCoE protocol stack wants to start using LLD for FCoE
888 * so the underlying device can perform whatever needed configuration or
889 * initialization to support acceleration of FCoE traffic.
891 * int (*ndo_fcoe_disable)(struct net_device *dev);
892 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
893 * so the underlying device can perform whatever needed clean-ups to
894 * stop supporting acceleration of FCoE traffic.
896 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
897 * struct scatterlist *sgl, unsigned int sgc);
898 * Called when the FCoE Initiator wants to initialize an I/O that
899 * is a possible candidate for Direct Data Placement (DDP). The LLD can
900 * perform necessary setup and returns 1 to indicate the device is set up
901 * successfully to perform DDP on this I/O, otherwise this returns 0.
903 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
904 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
905 * indicated by the FC exchange id 'xid', so the underlying device can
906 * clean up and reuse resources for later DDP requests.
908 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
909 * struct scatterlist *sgl, unsigned int sgc);
910 * Called when the FCoE Target wants to initialize an I/O that
911 * is a possible candidate for Direct Data Placement (DDP). The LLD can
912 * perform necessary setup and returns 1 to indicate the device is set up
913 * successfully to perform DDP on this I/O, otherwise this returns 0.
915 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
916 * struct netdev_fcoe_hbainfo *hbainfo);
917 * Called when the FCoE Protocol stack wants information on the underlying
918 * device. This information is utilized by the FCoE protocol stack to
919 * register attributes with Fiber Channel management service as per the
920 * FC-GS Fabric Device Management Information(FDMI) specification.
922 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
923 * Called when the underlying device wants to override default World Wide
924 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
925 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
926 * protocol stack to use.
929 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
930 * u16 rxq_index, u32 flow_id);
931 * Set hardware filter for RFS. rxq_index is the target queue index;
932 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
933 * Return the filter ID on success, or a negative error code.
935 * Slave management functions (for bridge, bonding, etc).
936 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
937 * Called to make another netdev an underling.
939 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
940 * Called to release previously enslaved netdev.
942 * Feature/offload setting functions.
943 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
944 * netdev_features_t features);
945 * Adjusts the requested feature flags according to device-specific
946 * constraints, and returns the resulting flags. Must not modify
949 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
950 * Called to update device configuration to new features. Passed
951 * feature set might be less than what was returned by ndo_fix_features()).
952 * Must return >0 or -errno if it changed dev->features itself.
954 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
955 * struct net_device *dev,
956 * const unsigned char *addr, u16 vid, u16 flags)
957 * Adds an FDB entry to dev for addr.
958 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
959 * struct net_device *dev,
960 * const unsigned char *addr, u16 vid)
961 * Deletes the FDB entry from dev coresponding to addr.
962 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
963 * struct net_device *dev, struct net_device *filter_dev,
965 * Used to add FDB entries to dump requests. Implementers should add
966 * entries to skb and update idx with the number of entries.
968 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
969 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
970 * struct net_device *dev, u32 filter_mask)
972 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
973 * Called to change device carrier. Soft-devices (like dummy, team, etc)
974 * which do not represent real hardware may define this to allow their
975 * userspace components to manage their virtual carrier state. Devices
976 * that determine carrier state from physical hardware properties (eg
977 * network cables) or protocol-dependent mechanisms (eg
978 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
980 * int (*ndo_get_phys_port_id)(struct net_device *dev,
981 * struct netdev_phys_item_id *ppid);
982 * Called to get ID of physical port of this device. If driver does
983 * not implement this, it is assumed that the hw is not able to have
984 * multiple net devices on single physical port.
986 * void (*ndo_add_vxlan_port)(struct net_device *dev,
987 * sa_family_t sa_family, __be16 port);
988 * Called by vxlan to notiy a driver about the UDP port and socket
989 * address family that vxlan is listnening to. It is called only when
990 * a new port starts listening. The operation is protected by the
991 * vxlan_net->sock_lock.
993 * void (*ndo_del_vxlan_port)(struct net_device *dev,
994 * sa_family_t sa_family, __be16 port);
995 * Called by vxlan to notify the driver about a UDP port and socket
996 * address family that vxlan is not listening to anymore. The operation
997 * is protected by the vxlan_net->sock_lock.
999 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1000 * struct net_device *dev)
1001 * Called by upper layer devices to accelerate switching or other
1002 * station functionality into hardware. 'pdev is the lowerdev
1003 * to use for the offload and 'dev' is the net device that will
1004 * back the offload. Returns a pointer to the private structure
1005 * the upper layer will maintain.
1006 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1007 * Called by upper layer device to delete the station created
1008 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1009 * the station and priv is the structure returned by the add
1011 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1012 * struct net_device *dev,
1014 * Callback to use for xmit over the accelerated station. This
1015 * is used in place of ndo_start_xmit on accelerated net
1017 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1018 * struct net_device *dev
1019 * netdev_features_t features);
1020 * Called by core transmit path to determine if device is capable of
1021 * performing offload operations on a given packet. This is to give
1022 * the device an opportunity to implement any restrictions that cannot
1023 * be otherwise expressed by feature flags. The check is called with
1024 * the set of features that the stack has calculated and it returns
1025 * those the driver believes to be appropriate.
1027 * int (*ndo_switch_parent_id_get)(struct net_device *dev,
1028 * struct netdev_phys_item_id *psid);
1029 * Called to get an ID of the switch chip this port is part of.
1030 * If driver implements this, it indicates that it represents a port
1032 * int (*ndo_switch_port_stp_update)(struct net_device *dev, u8 state);
1033 * Called to notify switch device port of bridge port STP
1036 struct net_device_ops
{
1037 int (*ndo_init
)(struct net_device
*dev
);
1038 void (*ndo_uninit
)(struct net_device
*dev
);
1039 int (*ndo_open
)(struct net_device
*dev
);
1040 int (*ndo_stop
)(struct net_device
*dev
);
1041 netdev_tx_t (*ndo_start_xmit
) (struct sk_buff
*skb
,
1042 struct net_device
*dev
);
1043 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1044 struct sk_buff
*skb
,
1046 select_queue_fallback_t fallback
);
1047 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1049 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1050 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1052 int (*ndo_validate_addr
)(struct net_device
*dev
);
1053 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1054 struct ifreq
*ifr
, int cmd
);
1055 int (*ndo_set_config
)(struct net_device
*dev
,
1057 int (*ndo_change_mtu
)(struct net_device
*dev
,
1059 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1060 struct neigh_parms
*);
1061 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1063 struct rtnl_link_stats64
* (*ndo_get_stats64
)(struct net_device
*dev
,
1064 struct rtnl_link_stats64
*storage
);
1065 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1067 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1068 __be16 proto
, u16 vid
);
1069 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1070 __be16 proto
, u16 vid
);
1071 #ifdef CONFIG_NET_POLL_CONTROLLER
1072 void (*ndo_poll_controller
)(struct net_device
*dev
);
1073 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1074 struct netpoll_info
*info
);
1075 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1077 #ifdef CONFIG_NET_RX_BUSY_POLL
1078 int (*ndo_busy_poll
)(struct napi_struct
*dev
);
1080 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1081 int queue
, u8
*mac
);
1082 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1083 int queue
, u16 vlan
, u8 qos
);
1084 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1085 int vf
, int min_tx_rate
,
1087 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1088 int vf
, bool setting
);
1089 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1091 struct ifla_vf_info
*ivf
);
1092 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1093 int vf
, int link_state
);
1094 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1096 struct nlattr
*port
[]);
1097 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1098 int vf
, struct sk_buff
*skb
);
1099 int (*ndo_setup_tc
)(struct net_device
*dev
, u8 tc
);
1100 #if IS_ENABLED(CONFIG_FCOE)
1101 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1102 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1103 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1105 struct scatterlist
*sgl
,
1107 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1109 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1111 struct scatterlist
*sgl
,
1113 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1114 struct netdev_fcoe_hbainfo
*hbainfo
);
1117 #if IS_ENABLED(CONFIG_LIBFCOE)
1118 #define NETDEV_FCOE_WWNN 0
1119 #define NETDEV_FCOE_WWPN 1
1120 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1121 u64
*wwn
, int type
);
1124 #ifdef CONFIG_RFS_ACCEL
1125 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1126 const struct sk_buff
*skb
,
1130 int (*ndo_add_slave
)(struct net_device
*dev
,
1131 struct net_device
*slave_dev
);
1132 int (*ndo_del_slave
)(struct net_device
*dev
,
1133 struct net_device
*slave_dev
);
1134 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1135 netdev_features_t features
);
1136 int (*ndo_set_features
)(struct net_device
*dev
,
1137 netdev_features_t features
);
1138 int (*ndo_neigh_construct
)(struct neighbour
*n
);
1139 void (*ndo_neigh_destroy
)(struct neighbour
*n
);
1141 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1142 struct nlattr
*tb
[],
1143 struct net_device
*dev
,
1144 const unsigned char *addr
,
1147 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1148 struct nlattr
*tb
[],
1149 struct net_device
*dev
,
1150 const unsigned char *addr
,
1152 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1153 struct netlink_callback
*cb
,
1154 struct net_device
*dev
,
1155 struct net_device
*filter_dev
,
1158 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1159 struct nlmsghdr
*nlh
,
1161 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1163 struct net_device
*dev
,
1165 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1166 struct nlmsghdr
*nlh
,
1168 int (*ndo_change_carrier
)(struct net_device
*dev
,
1170 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1171 struct netdev_phys_item_id
*ppid
);
1172 void (*ndo_add_vxlan_port
)(struct net_device
*dev
,
1173 sa_family_t sa_family
,
1175 void (*ndo_del_vxlan_port
)(struct net_device
*dev
,
1176 sa_family_t sa_family
,
1179 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1180 struct net_device
*dev
);
1181 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1184 netdev_tx_t (*ndo_dfwd_start_xmit
) (struct sk_buff
*skb
,
1185 struct net_device
*dev
,
1187 int (*ndo_get_lock_subclass
)(struct net_device
*dev
);
1188 netdev_features_t (*ndo_features_check
) (struct sk_buff
*skb
,
1189 struct net_device
*dev
,
1190 netdev_features_t features
);
1191 #ifdef CONFIG_NET_SWITCHDEV
1192 int (*ndo_switch_parent_id_get
)(struct net_device
*dev
,
1193 struct netdev_phys_item_id
*psid
);
1194 int (*ndo_switch_port_stp_update
)(struct net_device
*dev
,
1200 * enum net_device_priv_flags - &struct net_device priv_flags
1202 * These are the &struct net_device, they are only set internally
1203 * by drivers and used in the kernel. These flags are invisible to
1204 * userspace, this means that the order of these flags can change
1205 * during any kernel release.
1207 * You should have a pretty good reason to be extending these flags.
1209 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1210 * @IFF_EBRIDGE: Ethernet bridging device
1211 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1212 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1213 * @IFF_MASTER_ALB: bonding master, balance-alb
1214 * @IFF_BONDING: bonding master or slave
1215 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1216 * @IFF_ISATAP: ISATAP interface (RFC4214)
1217 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1218 * @IFF_WAN_HDLC: WAN HDLC device
1219 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1221 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1222 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1223 * @IFF_MACVLAN_PORT: device used as macvlan port
1224 * @IFF_BRIDGE_PORT: device used as bridge port
1225 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1226 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1227 * @IFF_UNICAST_FLT: Supports unicast filtering
1228 * @IFF_TEAM_PORT: device used as team port
1229 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1230 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1231 * change when it's running
1232 * @IFF_MACVLAN: Macvlan device
1234 enum netdev_priv_flags
{
1235 IFF_802_1Q_VLAN
= 1<<0,
1237 IFF_SLAVE_INACTIVE
= 1<<2,
1238 IFF_MASTER_8023AD
= 1<<3,
1239 IFF_MASTER_ALB
= 1<<4,
1241 IFF_SLAVE_NEEDARP
= 1<<6,
1243 IFF_MASTER_ARPMON
= 1<<8,
1244 IFF_WAN_HDLC
= 1<<9,
1245 IFF_XMIT_DST_RELEASE
= 1<<10,
1246 IFF_DONT_BRIDGE
= 1<<11,
1247 IFF_DISABLE_NETPOLL
= 1<<12,
1248 IFF_MACVLAN_PORT
= 1<<13,
1249 IFF_BRIDGE_PORT
= 1<<14,
1250 IFF_OVS_DATAPATH
= 1<<15,
1251 IFF_TX_SKB_SHARING
= 1<<16,
1252 IFF_UNICAST_FLT
= 1<<17,
1253 IFF_TEAM_PORT
= 1<<18,
1254 IFF_SUPP_NOFCS
= 1<<19,
1255 IFF_LIVE_ADDR_CHANGE
= 1<<20,
1256 IFF_MACVLAN
= 1<<21,
1257 IFF_XMIT_DST_RELEASE_PERM
= 1<<22,
1258 IFF_IPVLAN_MASTER
= 1<<23,
1259 IFF_IPVLAN_SLAVE
= 1<<24,
1262 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1263 #define IFF_EBRIDGE IFF_EBRIDGE
1264 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1265 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1266 #define IFF_MASTER_ALB IFF_MASTER_ALB
1267 #define IFF_BONDING IFF_BONDING
1268 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1269 #define IFF_ISATAP IFF_ISATAP
1270 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1271 #define IFF_WAN_HDLC IFF_WAN_HDLC
1272 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1273 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1274 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1275 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1276 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1277 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1278 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1279 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1280 #define IFF_TEAM_PORT IFF_TEAM_PORT
1281 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1282 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1283 #define IFF_MACVLAN IFF_MACVLAN
1284 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1285 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1286 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1289 * struct net_device - The DEVICE structure.
1290 * Actually, this whole structure is a big mistake. It mixes I/O
1291 * data with strictly "high-level" data, and it has to know about
1292 * almost every data structure used in the INET module.
1294 * @name: This is the first field of the "visible" part of this structure
1295 * (i.e. as seen by users in the "Space.c" file). It is the name
1298 * @name_hlist: Device name hash chain, please keep it close to name[]
1299 * @ifalias: SNMP alias
1300 * @mem_end: Shared memory end
1301 * @mem_start: Shared memory start
1302 * @base_addr: Device I/O address
1303 * @irq: Device IRQ number
1305 * @state: Generic network queuing layer state, see netdev_state_t
1306 * @dev_list: The global list of network devices
1307 * @napi_list: List entry, that is used for polling napi devices
1308 * @unreg_list: List entry, that is used, when we are unregistering the
1309 * device, see the function unregister_netdev
1310 * @close_list: List entry, that is used, when we are closing the device
1312 * @adj_list: Directly linked devices, like slaves for bonding
1313 * @all_adj_list: All linked devices, *including* neighbours
1314 * @features: Currently active device features
1315 * @hw_features: User-changeable features
1317 * @wanted_features: User-requested features
1318 * @vlan_features: Mask of features inheritable by VLAN devices
1320 * @hw_enc_features: Mask of features inherited by encapsulating devices
1321 * This field indicates what encapsulation
1322 * offloads the hardware is capable of doing,
1323 * and drivers will need to set them appropriately.
1325 * @mpls_features: Mask of features inheritable by MPLS
1327 * @ifindex: interface index
1328 * @iflink: unique device identifier
1330 * @stats: Statistics struct, which was left as a legacy, use
1331 * rtnl_link_stats64 instead
1333 * @rx_dropped: Dropped packets by core network,
1334 * do not use this in drivers
1335 * @tx_dropped: Dropped packets by core network,
1336 * do not use this in drivers
1338 * @carrier_changes: Stats to monitor carrier on<->off transitions
1340 * @wireless_handlers: List of functions to handle Wireless Extensions,
1342 * see <net/iw_handler.h> for details.
1343 * @wireless_data: Instance data managed by the core of wireless extensions
1345 * @netdev_ops: Includes several pointers to callbacks,
1346 * if one wants to override the ndo_*() functions
1347 * @ethtool_ops: Management operations
1348 * @fwd_ops: Management operations
1349 * @header_ops: Includes callbacks for creating,parsing,rebuilding,etc
1350 * of Layer 2 headers.
1352 * @flags: Interface flags (a la BSD)
1353 * @priv_flags: Like 'flags' but invisible to userspace,
1354 * see if.h for the definitions
1355 * @gflags: Global flags ( kept as legacy )
1356 * @padded: How much padding added by alloc_netdev()
1357 * @operstate: RFC2863 operstate
1358 * @link_mode: Mapping policy to operstate
1359 * @if_port: Selectable AUI, TP, ...
1361 * @mtu: Interface MTU value
1362 * @type: Interface hardware type
1363 * @hard_header_len: Hardware header length
1365 * @needed_headroom: Extra headroom the hardware may need, but not in all
1366 * cases can this be guaranteed
1367 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1368 * cases can this be guaranteed. Some cases also use
1369 * LL_MAX_HEADER instead to allocate the skb
1371 * interface address info:
1373 * @perm_addr: Permanent hw address
1374 * @addr_assign_type: Hw address assignment type
1375 * @addr_len: Hardware address length
1376 * @neigh_priv_len; Used in neigh_alloc(),
1377 * initialized only in atm/clip.c
1378 * @dev_id: Used to differentiate devices that share
1379 * the same link layer address
1380 * @dev_port: Used to differentiate devices that share
1382 * @addr_list_lock: XXX: need comments on this one
1383 * @uc: unicast mac addresses
1384 * @mc: multicast mac addresses
1385 * @dev_addrs: list of device hw addresses
1386 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1387 * @uc_promisc: Counter, that indicates, that promiscuous mode
1388 * has been enabled due to the need to listen to
1389 * additional unicast addresses in a device that
1390 * does not implement ndo_set_rx_mode()
1391 * @promiscuity: Number of times, the NIC is told to work in
1392 * Promiscuous mode, if it becomes 0 the NIC will
1393 * exit from working in Promiscuous mode
1394 * @allmulti: Counter, enables or disables allmulticast mode
1396 * @vlan_info: VLAN info
1397 * @dsa_ptr: dsa specific data
1398 * @tipc_ptr: TIPC specific data
1399 * @atalk_ptr: AppleTalk link
1400 * @ip_ptr: IPv4 specific data
1401 * @dn_ptr: DECnet specific data
1402 * @ip6_ptr: IPv6 specific data
1403 * @ax25_ptr: AX.25 specific data
1404 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1406 * @last_rx: Time of last Rx
1407 * @dev_addr: Hw address (before bcast,
1408 * because most packets are unicast)
1410 * @_rx: Array of RX queues
1411 * @num_rx_queues: Number of RX queues
1412 * allocated at register_netdev() time
1413 * @real_num_rx_queues: Number of RX queues currently active in device
1415 * @rx_handler: handler for received packets
1416 * @rx_handler_data: XXX: need comments on this one
1417 * @ingress_queue: XXX: need comments on this one
1418 * @broadcast: hw bcast address
1420 * @_tx: Array of TX queues
1421 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1422 * @real_num_tx_queues: Number of TX queues currently active in device
1423 * @qdisc: Root qdisc from userspace point of view
1424 * @tx_queue_len: Max frames per queue allowed
1425 * @tx_global_lock: XXX: need comments on this one
1427 * @xps_maps: XXX: need comments on this one
1429 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1430 * indexed by RX queue number. Assigned by driver.
1431 * This must only be set if the ndo_rx_flow_steer
1432 * operation is defined
1434 * @trans_start: Time (in jiffies) of last Tx
1435 * @watchdog_timeo: Represents the timeout that is used by
1436 * the watchdog ( see dev_watchdog() )
1437 * @watchdog_timer: List of timers
1439 * @pcpu_refcnt: Number of references to this device
1440 * @todo_list: Delayed register/unregister
1441 * @index_hlist: Device index hash chain
1442 * @link_watch_list: XXX: need comments on this one
1444 * @reg_state: Register/unregister state machine
1445 * @dismantle: Device is going to be freed
1446 * @rtnl_link_state: This enum represents the phases of creating
1449 * @destructor: Called from unregister,
1450 * can be used to call free_netdev
1451 * @npinfo: XXX: need comments on this one
1452 * @nd_net: Network namespace this network device is inside
1454 * @ml_priv: Mid-layer private
1455 * @lstats: Loopback statistics
1456 * @tstats: Tunnel statistics
1457 * @dstats: Dummy statistics
1458 * @vstats: Virtual ethernet statistics
1463 * @dev: Class/net/name entry
1464 * @sysfs_groups: Space for optional device, statistics and wireless
1467 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1468 * @rtnl_link_ops: Rtnl_link_ops
1470 * @gso_max_size: Maximum size of generic segmentation offload
1471 * @gso_max_segs: Maximum number of segments that can be passed to the
1473 * @gso_min_segs: Minimum number of segments that can be passed to the
1476 * @dcbnl_ops: Data Center Bridging netlink ops
1477 * @num_tc: Number of traffic classes in the net device
1478 * @tc_to_txq: XXX: need comments on this one
1479 * @prio_tc_map XXX: need comments on this one
1481 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1483 * @priomap: XXX: need comments on this one
1484 * @phydev: Physical device may attach itself
1485 * for hardware timestamping
1487 * @qdisc_tx_busylock: XXX: need comments on this one
1489 * @group: The group, that the device belongs to
1490 * @pm_qos_req: Power Management QoS object
1492 * FIXME: cleanup struct net_device such that network protocol info
1497 char name
[IFNAMSIZ
];
1498 struct hlist_node name_hlist
;
1501 * I/O specific fields
1502 * FIXME: Merge these and struct ifmap into one
1504 unsigned long mem_end
;
1505 unsigned long mem_start
;
1506 unsigned long base_addr
;
1510 * Some hardware also needs these fields (state,dev_list,
1511 * napi_list,unreg_list,close_list) but they are not
1512 * part of the usual set specified in Space.c.
1515 unsigned long state
;
1517 struct list_head dev_list
;
1518 struct list_head napi_list
;
1519 struct list_head unreg_list
;
1520 struct list_head close_list
;
1521 struct list_head ptype_all
;
1522 struct list_head ptype_specific
;
1525 struct list_head upper
;
1526 struct list_head lower
;
1530 struct list_head upper
;
1531 struct list_head lower
;
1534 netdev_features_t features
;
1535 netdev_features_t hw_features
;
1536 netdev_features_t wanted_features
;
1537 netdev_features_t vlan_features
;
1538 netdev_features_t hw_enc_features
;
1539 netdev_features_t mpls_features
;
1544 struct net_device_stats stats
;
1546 atomic_long_t rx_dropped
;
1547 atomic_long_t tx_dropped
;
1549 atomic_t carrier_changes
;
1551 #ifdef CONFIG_WIRELESS_EXT
1552 const struct iw_handler_def
* wireless_handlers
;
1553 struct iw_public_data
* wireless_data
;
1555 const struct net_device_ops
*netdev_ops
;
1556 const struct ethtool_ops
*ethtool_ops
;
1557 const struct forwarding_accel_ops
*fwd_ops
;
1559 const struct header_ops
*header_ops
;
1562 unsigned int priv_flags
;
1564 unsigned short gflags
;
1565 unsigned short padded
;
1567 unsigned char operstate
;
1568 unsigned char link_mode
;
1570 unsigned char if_port
;
1574 unsigned short type
;
1575 unsigned short hard_header_len
;
1577 unsigned short needed_headroom
;
1578 unsigned short needed_tailroom
;
1580 /* Interface address info. */
1581 unsigned char perm_addr
[MAX_ADDR_LEN
];
1582 unsigned char addr_assign_type
;
1583 unsigned char addr_len
;
1584 unsigned short neigh_priv_len
;
1585 unsigned short dev_id
;
1586 unsigned short dev_port
;
1587 spinlock_t addr_list_lock
;
1588 struct netdev_hw_addr_list uc
;
1589 struct netdev_hw_addr_list mc
;
1590 struct netdev_hw_addr_list dev_addrs
;
1593 struct kset
*queues_kset
;
1596 unsigned char name_assign_type
;
1599 unsigned int promiscuity
;
1600 unsigned int allmulti
;
1603 /* Protocol specific pointers */
1605 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1606 struct vlan_info __rcu
*vlan_info
;
1608 #if IS_ENABLED(CONFIG_NET_DSA)
1609 struct dsa_switch_tree
*dsa_ptr
;
1611 #if IS_ENABLED(CONFIG_TIPC)
1612 struct tipc_bearer __rcu
*tipc_ptr
;
1615 struct in_device __rcu
*ip_ptr
;
1616 struct dn_dev __rcu
*dn_ptr
;
1617 struct inet6_dev __rcu
*ip6_ptr
;
1619 struct wireless_dev
*ieee80211_ptr
;
1620 struct wpan_dev
*ieee802154_ptr
;
1623 * Cache lines mostly used on receive path (including eth_type_trans())
1625 unsigned long last_rx
;
1627 /* Interface address info used in eth_type_trans() */
1628 unsigned char *dev_addr
;
1632 struct netdev_rx_queue
*_rx
;
1634 unsigned int num_rx_queues
;
1635 unsigned int real_num_rx_queues
;
1639 unsigned long gro_flush_timeout
;
1640 rx_handler_func_t __rcu
*rx_handler
;
1641 void __rcu
*rx_handler_data
;
1643 struct netdev_queue __rcu
*ingress_queue
;
1644 unsigned char broadcast
[MAX_ADDR_LEN
];
1648 * Cache lines mostly used on transmit path
1650 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1651 unsigned int num_tx_queues
;
1652 unsigned int real_num_tx_queues
;
1653 struct Qdisc
*qdisc
;
1654 unsigned long tx_queue_len
;
1655 spinlock_t tx_global_lock
;
1658 struct xps_dev_maps __rcu
*xps_maps
;
1660 #ifdef CONFIG_RFS_ACCEL
1661 struct cpu_rmap
*rx_cpu_rmap
;
1664 /* These may be needed for future network-power-down code. */
1667 * trans_start here is expensive for high speed devices on SMP,
1668 * please use netdev_queue->trans_start instead.
1670 unsigned long trans_start
;
1673 struct timer_list watchdog_timer
;
1675 int __percpu
*pcpu_refcnt
;
1676 struct list_head todo_list
;
1678 struct hlist_node index_hlist
;
1679 struct list_head link_watch_list
;
1681 enum { NETREG_UNINITIALIZED
=0,
1682 NETREG_REGISTERED
, /* completed register_netdevice */
1683 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1684 NETREG_UNREGISTERED
, /* completed unregister todo */
1685 NETREG_RELEASED
, /* called free_netdev */
1686 NETREG_DUMMY
, /* dummy device for NAPI poll */
1692 RTNL_LINK_INITIALIZED
,
1693 RTNL_LINK_INITIALIZING
,
1694 } rtnl_link_state
:16;
1696 void (*destructor
)(struct net_device
*dev
);
1698 #ifdef CONFIG_NETPOLL
1699 struct netpoll_info __rcu
*npinfo
;
1702 #ifdef CONFIG_NET_NS
1706 /* mid-layer private */
1709 struct pcpu_lstats __percpu
*lstats
;
1710 struct pcpu_sw_netstats __percpu
*tstats
;
1711 struct pcpu_dstats __percpu
*dstats
;
1712 struct pcpu_vstats __percpu
*vstats
;
1715 struct garp_port __rcu
*garp_port
;
1716 struct mrp_port __rcu
*mrp_port
;
1719 const struct attribute_group
*sysfs_groups
[4];
1720 const struct attribute_group
*sysfs_rx_queue_group
;
1722 const struct rtnl_link_ops
*rtnl_link_ops
;
1724 /* for setting kernel sock attribute on TCP connection setup */
1725 #define GSO_MAX_SIZE 65536
1726 unsigned int gso_max_size
;
1727 #define GSO_MAX_SEGS 65535
1731 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1734 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1735 u8 prio_tc_map
[TC_BITMASK
+ 1];
1737 #if IS_ENABLED(CONFIG_FCOE)
1738 unsigned int fcoe_ddp_xid
;
1740 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1741 struct netprio_map __rcu
*priomap
;
1743 struct phy_device
*phydev
;
1744 struct lock_class_key
*qdisc_tx_busylock
;
1746 struct pm_qos_request pm_qos_req
;
1748 #define to_net_dev(d) container_of(d, struct net_device, dev)
1750 #define NETDEV_ALIGN 32
1753 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1755 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1759 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1761 if (tc
>= dev
->num_tc
)
1764 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1769 void netdev_reset_tc(struct net_device
*dev
)
1772 memset(dev
->tc_to_txq
, 0, sizeof(dev
->tc_to_txq
));
1773 memset(dev
->prio_tc_map
, 0, sizeof(dev
->prio_tc_map
));
1777 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
)
1779 if (tc
>= dev
->num_tc
)
1782 dev
->tc_to_txq
[tc
].count
= count
;
1783 dev
->tc_to_txq
[tc
].offset
= offset
;
1788 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
)
1790 if (num_tc
> TC_MAX_QUEUE
)
1793 dev
->num_tc
= num_tc
;
1798 int netdev_get_num_tc(struct net_device
*dev
)
1804 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1807 return &dev
->_tx
[index
];
1810 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1811 const struct sk_buff
*skb
)
1813 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1816 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1817 void (*f
)(struct net_device
*,
1818 struct netdev_queue
*,
1824 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1825 f(dev
, &dev
->_tx
[i
], arg
);
1828 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1829 struct sk_buff
*skb
,
1833 * Net namespace inlines
1836 struct net
*dev_net(const struct net_device
*dev
)
1838 return read_pnet(&dev
->nd_net
);
1842 void dev_net_set(struct net_device
*dev
, struct net
*net
)
1844 #ifdef CONFIG_NET_NS
1845 release_net(dev
->nd_net
);
1846 dev
->nd_net
= hold_net(net
);
1850 static inline bool netdev_uses_dsa(struct net_device
*dev
)
1852 #if IS_ENABLED(CONFIG_NET_DSA)
1853 if (dev
->dsa_ptr
!= NULL
)
1854 return dsa_uses_tagged_protocol(dev
->dsa_ptr
);
1860 * netdev_priv - access network device private data
1861 * @dev: network device
1863 * Get network device private data
1865 static inline void *netdev_priv(const struct net_device
*dev
)
1867 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
1870 /* Set the sysfs physical device reference for the network logical device
1871 * if set prior to registration will cause a symlink during initialization.
1873 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1875 /* Set the sysfs device type for the network logical device to allow
1876 * fine-grained identification of different network device types. For
1877 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1879 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1881 /* Default NAPI poll() weight
1882 * Device drivers are strongly advised to not use bigger value
1884 #define NAPI_POLL_WEIGHT 64
1887 * netif_napi_add - initialize a napi context
1888 * @dev: network device
1889 * @napi: napi context
1890 * @poll: polling function
1891 * @weight: default weight
1893 * netif_napi_add() must be used to initialize a napi context prior to calling
1894 * *any* of the other napi related functions.
1896 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
1897 int (*poll
)(struct napi_struct
*, int), int weight
);
1900 * netif_napi_del - remove a napi context
1901 * @napi: napi context
1903 * netif_napi_del() removes a napi context from the network device napi list
1905 void netif_napi_del(struct napi_struct
*napi
);
1907 struct napi_gro_cb
{
1908 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1911 /* Length of frag0. */
1912 unsigned int frag0_len
;
1914 /* This indicates where we are processing relative to skb->data. */
1917 /* This is non-zero if the packet cannot be merged with the new skb. */
1920 /* Save the IP ID here and check when we get to the transport layer */
1923 /* Number of segments aggregated. */
1926 /* This is non-zero if the packet may be of the same flow. */
1931 #define NAPI_GRO_FREE 1
1932 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1934 /* jiffies when first packet was created/queued */
1937 /* Used in ipv6_gro_receive() and foo-over-udp */
1940 /* Used in udp_gro_receive */
1943 /* GRO checksum is valid */
1946 /* Number of checksums via CHECKSUM_UNNECESSARY */
1949 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1952 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1955 /* used in skb_gro_receive() slow path */
1956 struct sk_buff
*last
;
1959 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1961 struct packet_type
{
1962 __be16 type
; /* This is really htons(ether_type). */
1963 struct net_device
*dev
; /* NULL is wildcarded here */
1964 int (*func
) (struct sk_buff
*,
1965 struct net_device
*,
1966 struct packet_type
*,
1967 struct net_device
*);
1968 bool (*id_match
)(struct packet_type
*ptype
,
1970 void *af_packet_priv
;
1971 struct list_head list
;
1974 struct offload_callbacks
{
1975 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
1976 netdev_features_t features
);
1977 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1978 struct sk_buff
*skb
);
1979 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
1982 struct packet_offload
{
1983 __be16 type
; /* This is really htons(ether_type). */
1984 struct offload_callbacks callbacks
;
1985 struct list_head list
;
1990 struct udp_offload_callbacks
{
1991 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1992 struct sk_buff
*skb
,
1993 struct udp_offload
*uoff
);
1994 int (*gro_complete
)(struct sk_buff
*skb
,
1996 struct udp_offload
*uoff
);
1999 struct udp_offload
{
2002 struct udp_offload_callbacks callbacks
;
2005 /* often modified stats are per cpu, other are shared (netdev->stats) */
2006 struct pcpu_sw_netstats
{
2011 struct u64_stats_sync syncp
;
2014 #define netdev_alloc_pcpu_stats(type) \
2016 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2019 for_each_possible_cpu(i) { \
2020 typeof(type) *stat; \
2021 stat = per_cpu_ptr(pcpu_stats, i); \
2022 u64_stats_init(&stat->syncp); \
2028 #include <linux/notifier.h>
2030 /* netdevice notifier chain. Please remember to update the rtnetlink
2031 * notification exclusion list in rtnetlink_event() when adding new
2034 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2035 #define NETDEV_DOWN 0x0002
2036 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2037 detected a hardware crash and restarted
2038 - we can use this eg to kick tcp sessions
2040 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2041 #define NETDEV_REGISTER 0x0005
2042 #define NETDEV_UNREGISTER 0x0006
2043 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2044 #define NETDEV_CHANGEADDR 0x0008
2045 #define NETDEV_GOING_DOWN 0x0009
2046 #define NETDEV_CHANGENAME 0x000A
2047 #define NETDEV_FEAT_CHANGE 0x000B
2048 #define NETDEV_BONDING_FAILOVER 0x000C
2049 #define NETDEV_PRE_UP 0x000D
2050 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2051 #define NETDEV_POST_TYPE_CHANGE 0x000F
2052 #define NETDEV_POST_INIT 0x0010
2053 #define NETDEV_UNREGISTER_FINAL 0x0011
2054 #define NETDEV_RELEASE 0x0012
2055 #define NETDEV_NOTIFY_PEERS 0x0013
2056 #define NETDEV_JOIN 0x0014
2057 #define NETDEV_CHANGEUPPER 0x0015
2058 #define NETDEV_RESEND_IGMP 0x0016
2059 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2060 #define NETDEV_CHANGEINFODATA 0x0018
2061 #define NETDEV_BONDING_INFO 0x0019
2063 int register_netdevice_notifier(struct notifier_block
*nb
);
2064 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2066 struct netdev_notifier_info
{
2067 struct net_device
*dev
;
2070 struct netdev_notifier_change_info
{
2071 struct netdev_notifier_info info
; /* must be first */
2072 unsigned int flags_changed
;
2075 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2076 struct net_device
*dev
)
2081 static inline struct net_device
*
2082 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2087 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2090 extern rwlock_t dev_base_lock
; /* Device list lock */
2092 #define for_each_netdev(net, d) \
2093 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2094 #define for_each_netdev_reverse(net, d) \
2095 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2096 #define for_each_netdev_rcu(net, d) \
2097 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2098 #define for_each_netdev_safe(net, d, n) \
2099 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2100 #define for_each_netdev_continue(net, d) \
2101 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2102 #define for_each_netdev_continue_rcu(net, d) \
2103 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2104 #define for_each_netdev_in_bond_rcu(bond, slave) \
2105 for_each_netdev_rcu(&init_net, slave) \
2106 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2107 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2109 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2111 struct list_head
*lh
;
2115 lh
= dev
->dev_list
.next
;
2116 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2119 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2121 struct list_head
*lh
;
2125 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2126 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2129 static inline struct net_device
*first_net_device(struct net
*net
)
2131 return list_empty(&net
->dev_base_head
) ? NULL
:
2132 net_device_entry(net
->dev_base_head
.next
);
2135 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2137 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2139 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2142 int netdev_boot_setup_check(struct net_device
*dev
);
2143 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2144 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2145 const char *hwaddr
);
2146 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2147 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2148 void dev_add_pack(struct packet_type
*pt
);
2149 void dev_remove_pack(struct packet_type
*pt
);
2150 void __dev_remove_pack(struct packet_type
*pt
);
2151 void dev_add_offload(struct packet_offload
*po
);
2152 void dev_remove_offload(struct packet_offload
*po
);
2154 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2155 unsigned short mask
);
2156 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2157 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2158 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2159 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2160 int dev_open(struct net_device
*dev
);
2161 int dev_close(struct net_device
*dev
);
2162 void dev_disable_lro(struct net_device
*dev
);
2163 int dev_loopback_xmit(struct sk_buff
*newskb
);
2164 int dev_queue_xmit(struct sk_buff
*skb
);
2165 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2166 int register_netdevice(struct net_device
*dev
);
2167 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2168 void unregister_netdevice_many(struct list_head
*head
);
2169 static inline void unregister_netdevice(struct net_device
*dev
)
2171 unregister_netdevice_queue(dev
, NULL
);
2174 int netdev_refcnt_read(const struct net_device
*dev
);
2175 void free_netdev(struct net_device
*dev
);
2176 void netdev_freemem(struct net_device
*dev
);
2177 void synchronize_net(void);
2178 int init_dummy_netdev(struct net_device
*dev
);
2180 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2181 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2182 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2183 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2184 int dev_restart(struct net_device
*dev
);
2185 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2187 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2189 return NAPI_GRO_CB(skb
)->data_offset
;
2192 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2194 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2197 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2199 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2202 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2203 unsigned int offset
)
2205 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2208 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2210 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2213 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2214 unsigned int offset
)
2216 if (!pskb_may_pull(skb
, hlen
))
2219 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2220 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2221 return skb
->data
+ offset
;
2224 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2226 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2227 skb_network_offset(skb
);
2230 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2231 const void *start
, unsigned int len
)
2233 if (NAPI_GRO_CB(skb
)->csum_valid
)
2234 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2235 csum_partial(start
, len
, 0));
2238 /* GRO checksum functions. These are logical equivalents of the normal
2239 * checksum functions (in skbuff.h) except that they operate on the GRO
2240 * offsets and fields in sk_buff.
2243 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2245 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2249 return ((skb
->ip_summed
!= CHECKSUM_PARTIAL
||
2250 skb_checksum_start_offset(skb
) <
2251 skb_gro_offset(skb
)) &&
2252 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2253 (!zero_okay
|| check
));
2256 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2259 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2260 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2263 NAPI_GRO_CB(skb
)->csum
= psum
;
2265 return __skb_gro_checksum_complete(skb
);
2268 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2270 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2271 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2272 NAPI_GRO_CB(skb
)->csum_cnt
--;
2274 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2275 * verified a new top level checksum or an encapsulated one
2276 * during GRO. This saves work if we fallback to normal path.
2278 __skb_incr_checksum_unnecessary(skb
);
2282 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2285 __sum16 __ret = 0; \
2286 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2287 __ret = __skb_gro_checksum_validate_complete(skb, \
2288 compute_pseudo(skb, proto)); \
2290 __skb_mark_checksum_bad(skb); \
2292 skb_gro_incr_csum_unnecessary(skb); \
2296 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2297 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2299 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2301 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2303 #define skb_gro_checksum_simple_validate(skb) \
2304 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2306 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2308 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2309 !NAPI_GRO_CB(skb
)->csum_valid
);
2312 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2313 __sum16 check
, __wsum pseudo
)
2315 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2316 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2319 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2321 if (__skb_gro_checksum_convert_check(skb)) \
2322 __skb_gro_checksum_convert(skb, check, \
2323 compute_pseudo(skb, proto)); \
2326 struct gro_remcsum
{
2331 static inline void skb_gro_remcsum_init(struct gro_remcsum
*grc
)
2336 static inline void skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2337 int start
, int offset
,
2338 struct gro_remcsum
*grc
)
2342 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2344 delta
= remcsum_adjust(ptr
, NAPI_GRO_CB(skb
)->csum
, start
, offset
);
2346 /* Adjust skb->csum since we changed the packet */
2347 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2349 grc
->offset
= (ptr
+ offset
) - (void *)skb
->head
;
2353 static inline void skb_gro_remcsum_cleanup(struct sk_buff
*skb
,
2354 struct gro_remcsum
*grc
)
2359 remcsum_unadjust((__sum16
*)(skb
->head
+ grc
->offset
), grc
->delta
);
2362 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2363 unsigned short type
,
2364 const void *daddr
, const void *saddr
,
2367 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2370 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2373 static inline int dev_parse_header(const struct sk_buff
*skb
,
2374 unsigned char *haddr
)
2376 const struct net_device
*dev
= skb
->dev
;
2378 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2380 return dev
->header_ops
->parse(skb
, haddr
);
2383 static inline int dev_rebuild_header(struct sk_buff
*skb
)
2385 const struct net_device
*dev
= skb
->dev
;
2387 if (!dev
->header_ops
|| !dev
->header_ops
->rebuild
)
2389 return dev
->header_ops
->rebuild(skb
);
2392 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2393 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2394 static inline int unregister_gifconf(unsigned int family
)
2396 return register_gifconf(family
, NULL
);
2399 #ifdef CONFIG_NET_FLOW_LIMIT
2400 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2401 struct sd_flow_limit
{
2403 unsigned int num_buckets
;
2404 unsigned int history_head
;
2405 u16 history
[FLOW_LIMIT_HISTORY
];
2409 extern int netdev_flow_limit_table_len
;
2410 #endif /* CONFIG_NET_FLOW_LIMIT */
2413 * Incoming packets are placed on per-cpu queues
2415 struct softnet_data
{
2416 struct list_head poll_list
;
2417 struct sk_buff_head process_queue
;
2420 unsigned int processed
;
2421 unsigned int time_squeeze
;
2422 unsigned int cpu_collision
;
2423 unsigned int received_rps
;
2425 struct softnet_data
*rps_ipi_list
;
2427 #ifdef CONFIG_NET_FLOW_LIMIT
2428 struct sd_flow_limit __rcu
*flow_limit
;
2430 struct Qdisc
*output_queue
;
2431 struct Qdisc
**output_queue_tailp
;
2432 struct sk_buff
*completion_queue
;
2435 /* Elements below can be accessed between CPUs for RPS */
2436 struct call_single_data csd ____cacheline_aligned_in_smp
;
2437 struct softnet_data
*rps_ipi_next
;
2439 unsigned int input_queue_head
;
2440 unsigned int input_queue_tail
;
2442 unsigned int dropped
;
2443 struct sk_buff_head input_pkt_queue
;
2444 struct napi_struct backlog
;
2448 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2451 sd
->input_queue_head
++;
2455 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2456 unsigned int *qtail
)
2459 *qtail
= ++sd
->input_queue_tail
;
2463 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2465 void __netif_schedule(struct Qdisc
*q
);
2466 void netif_schedule_queue(struct netdev_queue
*txq
);
2468 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2472 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2473 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2476 static inline void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2478 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2482 * netif_start_queue - allow transmit
2483 * @dev: network device
2485 * Allow upper layers to call the device hard_start_xmit routine.
2487 static inline void netif_start_queue(struct net_device
*dev
)
2489 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2492 static inline void netif_tx_start_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_start_queue(txq
);
2502 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2505 * netif_wake_queue - restart transmit
2506 * @dev: network device
2508 * Allow upper layers to call the device hard_start_xmit routine.
2509 * Used for flow control when transmit resources are available.
2511 static inline void netif_wake_queue(struct net_device
*dev
)
2513 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2516 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2520 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2521 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2522 netif_tx_wake_queue(txq
);
2526 static inline void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2528 if (WARN_ON(!dev_queue
)) {
2529 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2532 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2536 * netif_stop_queue - stop transmitted packets
2537 * @dev: network device
2539 * Stop upper layers calling the device hard_start_xmit routine.
2540 * Used for flow control when transmit resources are unavailable.
2542 static inline void netif_stop_queue(struct net_device
*dev
)
2544 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2547 static inline void netif_tx_stop_all_queues(struct net_device
*dev
)
2551 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2552 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2553 netif_tx_stop_queue(txq
);
2557 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2559 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2563 * netif_queue_stopped - test if transmit queue is flowblocked
2564 * @dev: network device
2566 * Test if transmit queue on device is currently unable to send.
2568 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2570 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2573 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2575 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2579 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2581 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2585 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2587 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2591 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2592 * @dev_queue: pointer to transmit queue
2594 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2595 * to give appropriate hint to the cpu.
2597 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2600 prefetchw(&dev_queue
->dql
.num_queued
);
2605 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2606 * @dev_queue: pointer to transmit queue
2608 * BQL enabled drivers might use this helper in their TX completion path,
2609 * to give appropriate hint to the cpu.
2611 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2614 prefetchw(&dev_queue
->dql
.limit
);
2618 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2622 dql_queued(&dev_queue
->dql
, bytes
);
2624 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2627 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2630 * The XOFF flag must be set before checking the dql_avail below,
2631 * because in netdev_tx_completed_queue we update the dql_completed
2632 * before checking the XOFF flag.
2636 /* check again in case another CPU has just made room avail */
2637 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2638 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2643 * netdev_sent_queue - report the number of bytes queued to hardware
2644 * @dev: network device
2645 * @bytes: number of bytes queued to the hardware device queue
2647 * Report the number of bytes queued for sending/completion to the network
2648 * device hardware queue. @bytes should be a good approximation and should
2649 * exactly match netdev_completed_queue() @bytes
2651 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2653 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2656 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2657 unsigned int pkts
, unsigned int bytes
)
2660 if (unlikely(!bytes
))
2663 dql_completed(&dev_queue
->dql
, bytes
);
2666 * Without the memory barrier there is a small possiblity that
2667 * netdev_tx_sent_queue will miss the update and cause the queue to
2668 * be stopped forever
2672 if (dql_avail(&dev_queue
->dql
) < 0)
2675 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2676 netif_schedule_queue(dev_queue
);
2681 * netdev_completed_queue - report bytes and packets completed by device
2682 * @dev: network device
2683 * @pkts: actual number of packets sent over the medium
2684 * @bytes: actual number of bytes sent over the medium
2686 * Report the number of bytes and packets transmitted by the network device
2687 * hardware queue over the physical medium, @bytes must exactly match the
2688 * @bytes amount passed to netdev_sent_queue()
2690 static inline void netdev_completed_queue(struct net_device
*dev
,
2691 unsigned int pkts
, unsigned int bytes
)
2693 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
2696 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
2699 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
2705 * netdev_reset_queue - reset the packets and bytes count of a network device
2706 * @dev_queue: network device
2708 * Reset the bytes and packet count of a network device and clear the
2709 * software flow control OFF bit for this network device
2711 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
2713 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
2717 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2718 * @dev: network device
2719 * @queue_index: given tx queue index
2721 * Returns 0 if given tx queue index >= number of device tx queues,
2722 * otherwise returns the originally passed tx queue index.
2724 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
2726 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
2727 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2728 dev
->name
, queue_index
,
2729 dev
->real_num_tx_queues
);
2737 * netif_running - test if up
2738 * @dev: network device
2740 * Test if the device has been brought up.
2742 static inline bool netif_running(const struct net_device
*dev
)
2744 return test_bit(__LINK_STATE_START
, &dev
->state
);
2748 * Routines to manage the subqueues on a device. We only need start
2749 * stop, and a check if it's stopped. All other device management is
2750 * done at the overall netdevice level.
2751 * Also test the device if we're multiqueue.
2755 * netif_start_subqueue - allow sending packets on subqueue
2756 * @dev: network device
2757 * @queue_index: sub queue index
2759 * Start individual transmit queue of a device with multiple transmit queues.
2761 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
2763 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2765 netif_tx_start_queue(txq
);
2769 * netif_stop_subqueue - stop sending packets on subqueue
2770 * @dev: network device
2771 * @queue_index: sub queue index
2773 * Stop individual transmit queue of a device with multiple transmit queues.
2775 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
2777 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2778 netif_tx_stop_queue(txq
);
2782 * netif_subqueue_stopped - test status of subqueue
2783 * @dev: network device
2784 * @queue_index: sub queue index
2786 * Check individual transmit queue of a device with multiple transmit queues.
2788 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
2791 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2793 return netif_tx_queue_stopped(txq
);
2796 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
2797 struct sk_buff
*skb
)
2799 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
2802 void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
);
2805 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
2808 static inline int netif_set_xps_queue(struct net_device
*dev
,
2809 const struct cpumask
*mask
,
2817 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2818 * as a distribution range limit for the returned value.
2820 static inline u16
skb_tx_hash(const struct net_device
*dev
,
2821 struct sk_buff
*skb
)
2823 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
2827 * netif_is_multiqueue - test if device has multiple transmit queues
2828 * @dev: network device
2830 * Check if device has multiple transmit queues
2832 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
2834 return dev
->num_tx_queues
> 1;
2837 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
2840 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
2842 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
2850 static inline unsigned int get_netdev_rx_queue_index(
2851 struct netdev_rx_queue
*queue
)
2853 struct net_device
*dev
= queue
->dev
;
2854 int index
= queue
- dev
->_rx
;
2856 BUG_ON(index
>= dev
->num_rx_queues
);
2861 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2862 int netif_get_num_default_rss_queues(void);
2864 enum skb_free_reason
{
2865 SKB_REASON_CONSUMED
,
2869 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
2870 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
2873 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2874 * interrupt context or with hardware interrupts being disabled.
2875 * (in_irq() || irqs_disabled())
2877 * We provide four helpers that can be used in following contexts :
2879 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2880 * replacing kfree_skb(skb)
2882 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2883 * Typically used in place of consume_skb(skb) in TX completion path
2885 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2886 * replacing kfree_skb(skb)
2888 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2889 * and consumed a packet. Used in place of consume_skb(skb)
2891 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
2893 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
2896 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
2898 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
2901 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
2903 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
2906 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
2908 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
2911 int netif_rx(struct sk_buff
*skb
);
2912 int netif_rx_ni(struct sk_buff
*skb
);
2913 int netif_receive_skb(struct sk_buff
*skb
);
2914 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
2915 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
2916 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
2917 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
2918 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
2919 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
2921 static inline void napi_free_frags(struct napi_struct
*napi
)
2923 kfree_skb(napi
->skb
);
2927 int netdev_rx_handler_register(struct net_device
*dev
,
2928 rx_handler_func_t
*rx_handler
,
2929 void *rx_handler_data
);
2930 void netdev_rx_handler_unregister(struct net_device
*dev
);
2932 bool dev_valid_name(const char *name
);
2933 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
2934 int dev_ethtool(struct net
*net
, struct ifreq
*);
2935 unsigned int dev_get_flags(const struct net_device
*);
2936 int __dev_change_flags(struct net_device
*, unsigned int flags
);
2937 int dev_change_flags(struct net_device
*, unsigned int);
2938 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
2939 unsigned int gchanges
);
2940 int dev_change_name(struct net_device
*, const char *);
2941 int dev_set_alias(struct net_device
*, const char *, size_t);
2942 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
2943 int dev_set_mtu(struct net_device
*, int);
2944 void dev_set_group(struct net_device
*, int);
2945 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
2946 int dev_change_carrier(struct net_device
*, bool new_carrier
);
2947 int dev_get_phys_port_id(struct net_device
*dev
,
2948 struct netdev_phys_item_id
*ppid
);
2949 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
2950 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
2951 struct netdev_queue
*txq
, int *ret
);
2952 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2953 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2954 bool is_skb_forwardable(struct net_device
*dev
, struct sk_buff
*skb
);
2956 extern int netdev_budget
;
2958 /* Called by rtnetlink.c:rtnl_unlock() */
2959 void netdev_run_todo(void);
2962 * dev_put - release reference to device
2963 * @dev: network device
2965 * Release reference to device to allow it to be freed.
2967 static inline void dev_put(struct net_device
*dev
)
2969 this_cpu_dec(*dev
->pcpu_refcnt
);
2973 * dev_hold - get reference to device
2974 * @dev: network device
2976 * Hold reference to device to keep it from being freed.
2978 static inline void dev_hold(struct net_device
*dev
)
2980 this_cpu_inc(*dev
->pcpu_refcnt
);
2983 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2984 * and _off may be called from IRQ context, but it is caller
2985 * who is responsible for serialization of these calls.
2987 * The name carrier is inappropriate, these functions should really be
2988 * called netif_lowerlayer_*() because they represent the state of any
2989 * kind of lower layer not just hardware media.
2992 void linkwatch_init_dev(struct net_device
*dev
);
2993 void linkwatch_fire_event(struct net_device
*dev
);
2994 void linkwatch_forget_dev(struct net_device
*dev
);
2997 * netif_carrier_ok - test if carrier present
2998 * @dev: network device
3000 * Check if carrier is present on device
3002 static inline bool netif_carrier_ok(const struct net_device
*dev
)
3004 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
3007 unsigned long dev_trans_start(struct net_device
*dev
);
3009 void __netdev_watchdog_up(struct net_device
*dev
);
3011 void netif_carrier_on(struct net_device
*dev
);
3013 void netif_carrier_off(struct net_device
*dev
);
3016 * netif_dormant_on - mark device as dormant.
3017 * @dev: network device
3019 * Mark device as dormant (as per RFC2863).
3021 * The dormant state indicates that the relevant interface is not
3022 * actually in a condition to pass packets (i.e., it is not 'up') but is
3023 * in a "pending" state, waiting for some external event. For "on-
3024 * demand" interfaces, this new state identifies the situation where the
3025 * interface is waiting for events to place it in the up state.
3028 static inline void netif_dormant_on(struct net_device
*dev
)
3030 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3031 linkwatch_fire_event(dev
);
3035 * netif_dormant_off - set device as not dormant.
3036 * @dev: network device
3038 * Device is not in dormant state.
3040 static inline void netif_dormant_off(struct net_device
*dev
)
3042 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3043 linkwatch_fire_event(dev
);
3047 * netif_dormant - test if carrier present
3048 * @dev: network device
3050 * Check if carrier is present on device
3052 static inline bool netif_dormant(const struct net_device
*dev
)
3054 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3059 * netif_oper_up - test if device is operational
3060 * @dev: network device
3062 * Check if carrier is operational
3064 static inline bool netif_oper_up(const struct net_device
*dev
)
3066 return (dev
->operstate
== IF_OPER_UP
||
3067 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3071 * netif_device_present - is device available or removed
3072 * @dev: network device
3074 * Check if device has not been removed from system.
3076 static inline bool netif_device_present(struct net_device
*dev
)
3078 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3081 void netif_device_detach(struct net_device
*dev
);
3083 void netif_device_attach(struct net_device
*dev
);
3086 * Network interface message level settings
3090 NETIF_MSG_DRV
= 0x0001,
3091 NETIF_MSG_PROBE
= 0x0002,
3092 NETIF_MSG_LINK
= 0x0004,
3093 NETIF_MSG_TIMER
= 0x0008,
3094 NETIF_MSG_IFDOWN
= 0x0010,
3095 NETIF_MSG_IFUP
= 0x0020,
3096 NETIF_MSG_RX_ERR
= 0x0040,
3097 NETIF_MSG_TX_ERR
= 0x0080,
3098 NETIF_MSG_TX_QUEUED
= 0x0100,
3099 NETIF_MSG_INTR
= 0x0200,
3100 NETIF_MSG_TX_DONE
= 0x0400,
3101 NETIF_MSG_RX_STATUS
= 0x0800,
3102 NETIF_MSG_PKTDATA
= 0x1000,
3103 NETIF_MSG_HW
= 0x2000,
3104 NETIF_MSG_WOL
= 0x4000,
3107 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3108 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3109 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3110 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3111 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3112 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3113 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3114 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3115 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3116 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3117 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3118 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3119 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3120 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3121 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3123 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3126 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3127 return default_msg_enable_bits
;
3128 if (debug_value
== 0) /* no output */
3130 /* set low N bits */
3131 return (1 << debug_value
) - 1;
3134 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3136 spin_lock(&txq
->_xmit_lock
);
3137 txq
->xmit_lock_owner
= cpu
;
3140 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3142 spin_lock_bh(&txq
->_xmit_lock
);
3143 txq
->xmit_lock_owner
= smp_processor_id();
3146 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3148 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3150 txq
->xmit_lock_owner
= smp_processor_id();
3154 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3156 txq
->xmit_lock_owner
= -1;
3157 spin_unlock(&txq
->_xmit_lock
);
3160 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3162 txq
->xmit_lock_owner
= -1;
3163 spin_unlock_bh(&txq
->_xmit_lock
);
3166 static inline void txq_trans_update(struct netdev_queue
*txq
)
3168 if (txq
->xmit_lock_owner
!= -1)
3169 txq
->trans_start
= jiffies
;
3173 * netif_tx_lock - grab network device transmit lock
3174 * @dev: network device
3176 * Get network device transmit lock
3178 static inline void netif_tx_lock(struct net_device
*dev
)
3183 spin_lock(&dev
->tx_global_lock
);
3184 cpu
= smp_processor_id();
3185 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3186 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3188 /* We are the only thread of execution doing a
3189 * freeze, but we have to grab the _xmit_lock in
3190 * order to synchronize with threads which are in
3191 * the ->hard_start_xmit() handler and already
3192 * checked the frozen bit.
3194 __netif_tx_lock(txq
, cpu
);
3195 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3196 __netif_tx_unlock(txq
);
3200 static inline void netif_tx_lock_bh(struct net_device
*dev
)
3206 static inline void netif_tx_unlock(struct net_device
*dev
)
3210 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3211 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3213 /* No need to grab the _xmit_lock here. If the
3214 * queue is not stopped for another reason, we
3217 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3218 netif_schedule_queue(txq
);
3220 spin_unlock(&dev
->tx_global_lock
);
3223 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
3225 netif_tx_unlock(dev
);
3229 #define HARD_TX_LOCK(dev, txq, cpu) { \
3230 if ((dev->features & NETIF_F_LLTX) == 0) { \
3231 __netif_tx_lock(txq, cpu); \
3235 #define HARD_TX_TRYLOCK(dev, txq) \
3236 (((dev->features & NETIF_F_LLTX) == 0) ? \
3237 __netif_tx_trylock(txq) : \
3240 #define HARD_TX_UNLOCK(dev, txq) { \
3241 if ((dev->features & NETIF_F_LLTX) == 0) { \
3242 __netif_tx_unlock(txq); \
3246 static inline void netif_tx_disable(struct net_device
*dev
)
3252 cpu
= smp_processor_id();
3253 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3254 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3256 __netif_tx_lock(txq
, cpu
);
3257 netif_tx_stop_queue(txq
);
3258 __netif_tx_unlock(txq
);
3263 static inline void netif_addr_lock(struct net_device
*dev
)
3265 spin_lock(&dev
->addr_list_lock
);
3268 static inline void netif_addr_lock_nested(struct net_device
*dev
)
3270 int subclass
= SINGLE_DEPTH_NESTING
;
3272 if (dev
->netdev_ops
->ndo_get_lock_subclass
)
3273 subclass
= dev
->netdev_ops
->ndo_get_lock_subclass(dev
);
3275 spin_lock_nested(&dev
->addr_list_lock
, subclass
);
3278 static inline void netif_addr_lock_bh(struct net_device
*dev
)
3280 spin_lock_bh(&dev
->addr_list_lock
);
3283 static inline void netif_addr_unlock(struct net_device
*dev
)
3285 spin_unlock(&dev
->addr_list_lock
);
3288 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
3290 spin_unlock_bh(&dev
->addr_list_lock
);
3294 * dev_addrs walker. Should be used only for read access. Call with
3295 * rcu_read_lock held.
3297 #define for_each_dev_addr(dev, ha) \
3298 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3300 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3302 void ether_setup(struct net_device
*dev
);
3304 /* Support for loadable net-drivers */
3305 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
3306 unsigned char name_assign_type
,
3307 void (*setup
)(struct net_device
*),
3308 unsigned int txqs
, unsigned int rxqs
);
3309 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3310 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3312 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3313 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3316 int register_netdev(struct net_device
*dev
);
3317 void unregister_netdev(struct net_device
*dev
);
3319 /* General hardware address lists handling functions */
3320 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
3321 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3322 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
3323 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3324 int __hw_addr_sync_dev(struct netdev_hw_addr_list
*list
,
3325 struct net_device
*dev
,
3326 int (*sync
)(struct net_device
*, const unsigned char *),
3327 int (*unsync
)(struct net_device
*,
3328 const unsigned char *));
3329 void __hw_addr_unsync_dev(struct netdev_hw_addr_list
*list
,
3330 struct net_device
*dev
,
3331 int (*unsync
)(struct net_device
*,
3332 const unsigned char *));
3333 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
3335 /* Functions used for device addresses handling */
3336 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
3337 unsigned char addr_type
);
3338 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
3339 unsigned char addr_type
);
3340 void dev_addr_flush(struct net_device
*dev
);
3341 int dev_addr_init(struct net_device
*dev
);
3343 /* Functions used for unicast addresses handling */
3344 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
3345 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3346 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
3347 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
3348 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3349 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
3350 void dev_uc_flush(struct net_device
*dev
);
3351 void dev_uc_init(struct net_device
*dev
);
3354 * __dev_uc_sync - Synchonize device's unicast list
3355 * @dev: device to sync
3356 * @sync: function to call if address should be added
3357 * @unsync: function to call if address should be removed
3359 * Add newly added addresses to the interface, and release
3360 * addresses that have been deleted.
3362 static inline int __dev_uc_sync(struct net_device
*dev
,
3363 int (*sync
)(struct net_device
*,
3364 const unsigned char *),
3365 int (*unsync
)(struct net_device
*,
3366 const unsigned char *))
3368 return __hw_addr_sync_dev(&dev
->uc
, dev
, sync
, unsync
);
3372 * __dev_uc_unsync - Remove synchronized addresses from device
3373 * @dev: device to sync
3374 * @unsync: function to call if address should be removed
3376 * Remove all addresses that were added to the device by dev_uc_sync().
3378 static inline void __dev_uc_unsync(struct net_device
*dev
,
3379 int (*unsync
)(struct net_device
*,
3380 const unsigned char *))
3382 __hw_addr_unsync_dev(&dev
->uc
, dev
, unsync
);
3385 /* Functions used for multicast addresses handling */
3386 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
3387 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
3388 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3389 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
3390 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
3391 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
3392 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3393 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
3394 void dev_mc_flush(struct net_device
*dev
);
3395 void dev_mc_init(struct net_device
*dev
);
3398 * __dev_mc_sync - Synchonize device's multicast list
3399 * @dev: device to sync
3400 * @sync: function to call if address should be added
3401 * @unsync: function to call if address should be removed
3403 * Add newly added addresses to the interface, and release
3404 * addresses that have been deleted.
3406 static inline int __dev_mc_sync(struct net_device
*dev
,
3407 int (*sync
)(struct net_device
*,
3408 const unsigned char *),
3409 int (*unsync
)(struct net_device
*,
3410 const unsigned char *))
3412 return __hw_addr_sync_dev(&dev
->mc
, dev
, sync
, unsync
);
3416 * __dev_mc_unsync - Remove synchronized addresses from device
3417 * @dev: device to sync
3418 * @unsync: function to call if address should be removed
3420 * Remove all addresses that were added to the device by dev_mc_sync().
3422 static inline void __dev_mc_unsync(struct net_device
*dev
,
3423 int (*unsync
)(struct net_device
*,
3424 const unsigned char *))
3426 __hw_addr_unsync_dev(&dev
->mc
, dev
, unsync
);
3429 /* Functions used for secondary unicast and multicast support */
3430 void dev_set_rx_mode(struct net_device
*dev
);
3431 void __dev_set_rx_mode(struct net_device
*dev
);
3432 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
3433 int dev_set_allmulti(struct net_device
*dev
, int inc
);
3434 void netdev_state_change(struct net_device
*dev
);
3435 void netdev_notify_peers(struct net_device
*dev
);
3436 void netdev_features_change(struct net_device
*dev
);
3437 /* Load a device via the kmod */
3438 void dev_load(struct net
*net
, const char *name
);
3439 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
3440 struct rtnl_link_stats64
*storage
);
3441 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
3442 const struct net_device_stats
*netdev_stats
);
3444 extern int netdev_max_backlog
;
3445 extern int netdev_tstamp_prequeue
;
3446 extern int weight_p
;
3447 extern int bpf_jit_enable
;
3449 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3450 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3451 struct list_head
**iter
);
3452 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3453 struct list_head
**iter
);
3455 /* iterate through upper list, must be called under RCU read lock */
3456 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3457 for (iter = &(dev)->adj_list.upper, \
3458 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3460 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3462 /* iterate through upper list, must be called under RCU read lock */
3463 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3464 for (iter = &(dev)->all_adj_list.upper, \
3465 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3467 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3469 void *netdev_lower_get_next_private(struct net_device
*dev
,
3470 struct list_head
**iter
);
3471 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3472 struct list_head
**iter
);
3474 #define netdev_for_each_lower_private(dev, priv, iter) \
3475 for (iter = (dev)->adj_list.lower.next, \
3476 priv = netdev_lower_get_next_private(dev, &(iter)); \
3478 priv = netdev_lower_get_next_private(dev, &(iter)))
3480 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3481 for (iter = &(dev)->adj_list.lower, \
3482 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3484 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3486 void *netdev_lower_get_next(struct net_device
*dev
,
3487 struct list_head
**iter
);
3488 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3489 for (iter = &(dev)->adj_list.lower, \
3490 ldev = netdev_lower_get_next(dev, &(iter)); \
3492 ldev = netdev_lower_get_next(dev, &(iter)))
3494 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3495 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3496 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3497 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3498 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
3499 int netdev_master_upper_dev_link(struct net_device
*dev
,
3500 struct net_device
*upper_dev
);
3501 int netdev_master_upper_dev_link_private(struct net_device
*dev
,
3502 struct net_device
*upper_dev
,
3504 void netdev_upper_dev_unlink(struct net_device
*dev
,
3505 struct net_device
*upper_dev
);
3506 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3507 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3508 struct net_device
*lower_dev
);
3510 /* RSS keys are 40 or 52 bytes long */
3511 #define NETDEV_RSS_KEY_LEN 52
3512 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
];
3513 void netdev_rss_key_fill(void *buffer
, size_t len
);
3515 int dev_get_nest_level(struct net_device
*dev
,
3516 bool (*type_check
)(struct net_device
*dev
));
3517 int skb_checksum_help(struct sk_buff
*skb
);
3518 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3519 netdev_features_t features
, bool tx_path
);
3520 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3521 netdev_features_t features
);
3523 struct netdev_bonding_info
{
3528 struct netdev_notifier_bonding_info
{
3529 struct netdev_notifier_info info
; /* must be first */
3530 struct netdev_bonding_info bonding_info
;
3533 void netdev_bonding_info_change(struct net_device
*dev
,
3534 struct netdev_bonding_info
*bonding_info
);
3537 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3539 return __skb_gso_segment(skb
, features
, true);
3541 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3543 static inline bool can_checksum_protocol(netdev_features_t features
,
3546 return ((features
& NETIF_F_GEN_CSUM
) ||
3547 ((features
& NETIF_F_V4_CSUM
) &&
3548 protocol
== htons(ETH_P_IP
)) ||
3549 ((features
& NETIF_F_V6_CSUM
) &&
3550 protocol
== htons(ETH_P_IPV6
)) ||
3551 ((features
& NETIF_F_FCOE_CRC
) &&
3552 protocol
== htons(ETH_P_FCOE
)));
3556 void netdev_rx_csum_fault(struct net_device
*dev
);
3558 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3562 /* rx skb timestamps */
3563 void net_enable_timestamp(void);
3564 void net_disable_timestamp(void);
3566 #ifdef CONFIG_PROC_FS
3567 int __init
dev_proc_init(void);
3569 #define dev_proc_init() 0
3572 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
3573 struct sk_buff
*skb
, struct net_device
*dev
,
3576 skb
->xmit_more
= more
? 1 : 0;
3577 return ops
->ndo_start_xmit(skb
, dev
);
3580 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3581 struct netdev_queue
*txq
, bool more
)
3583 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3586 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
3587 if (rc
== NETDEV_TX_OK
)
3588 txq_trans_update(txq
);
3593 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3595 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3598 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3600 return netdev_class_create_file_ns(class_attr
, NULL
);
3603 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3605 netdev_class_remove_file_ns(class_attr
, NULL
);
3608 extern struct kobj_ns_type_operations net_ns_type_operations
;
3610 const char *netdev_drivername(const struct net_device
*dev
);
3612 void linkwatch_run_queue(void);
3614 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
3615 netdev_features_t f2
)
3617 if (f1
& NETIF_F_GEN_CSUM
)
3618 f1
|= (NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3619 if (f2
& NETIF_F_GEN_CSUM
)
3620 f2
|= (NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3622 if (f1
& NETIF_F_GEN_CSUM
)
3623 f1
&= ~(NETIF_F_ALL_CSUM
& ~NETIF_F_GEN_CSUM
);
3628 static inline netdev_features_t
netdev_get_wanted_features(
3629 struct net_device
*dev
)
3631 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
3633 netdev_features_t
netdev_increment_features(netdev_features_t all
,
3634 netdev_features_t one
, netdev_features_t mask
);
3636 /* Allow TSO being used on stacked device :
3637 * Performing the GSO segmentation before last device
3638 * is a performance improvement.
3640 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
3641 netdev_features_t mask
)
3643 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
3646 int __netdev_update_features(struct net_device
*dev
);
3647 void netdev_update_features(struct net_device
*dev
);
3648 void netdev_change_features(struct net_device
*dev
);
3650 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
3651 struct net_device
*dev
);
3653 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
3655 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
3657 netdev_features_t feature
= gso_type
<< NETIF_F_GSO_SHIFT
;
3659 /* check flags correspondence */
3660 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
3661 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
3662 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
3663 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
3664 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
3665 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
3666 BUILD_BUG_ON(SKB_GSO_GRE
!= (NETIF_F_GSO_GRE
>> NETIF_F_GSO_SHIFT
));
3667 BUILD_BUG_ON(SKB_GSO_GRE_CSUM
!= (NETIF_F_GSO_GRE_CSUM
>> NETIF_F_GSO_SHIFT
));
3668 BUILD_BUG_ON(SKB_GSO_IPIP
!= (NETIF_F_GSO_IPIP
>> NETIF_F_GSO_SHIFT
));
3669 BUILD_BUG_ON(SKB_GSO_SIT
!= (NETIF_F_GSO_SIT
>> NETIF_F_GSO_SHIFT
));
3670 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL
!= (NETIF_F_GSO_UDP_TUNNEL
>> NETIF_F_GSO_SHIFT
));
3671 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM
!= (NETIF_F_GSO_UDP_TUNNEL_CSUM
>> NETIF_F_GSO_SHIFT
));
3672 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
3674 return (features
& feature
) == feature
;
3677 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
3679 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
3680 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
3683 static inline bool netif_needs_gso(struct net_device
*dev
, struct sk_buff
*skb
,
3684 netdev_features_t features
)
3686 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
3687 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
3688 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
3691 static inline void netif_set_gso_max_size(struct net_device
*dev
,
3694 dev
->gso_max_size
= size
;
3697 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
3698 int pulled_hlen
, u16 mac_offset
,
3701 skb
->protocol
= protocol
;
3702 skb
->encapsulation
= 1;
3703 skb_push(skb
, pulled_hlen
);
3704 skb_reset_transport_header(skb
);
3705 skb
->mac_header
= mac_offset
;
3706 skb
->network_header
= skb
->mac_header
+ mac_len
;
3707 skb
->mac_len
= mac_len
;
3710 static inline bool netif_is_macvlan(struct net_device
*dev
)
3712 return dev
->priv_flags
& IFF_MACVLAN
;
3715 static inline bool netif_is_macvlan_port(struct net_device
*dev
)
3717 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
3720 static inline bool netif_is_ipvlan(struct net_device
*dev
)
3722 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
3725 static inline bool netif_is_ipvlan_port(struct net_device
*dev
)
3727 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
3730 static inline bool netif_is_bond_master(struct net_device
*dev
)
3732 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
3735 static inline bool netif_is_bond_slave(struct net_device
*dev
)
3737 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
3740 static inline bool netif_supports_nofcs(struct net_device
*dev
)
3742 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
3745 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3746 static inline void netif_keep_dst(struct net_device
*dev
)
3748 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
3751 extern struct pernet_operations __net_initdata loopback_net_ops
;
3753 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3755 /* netdev_printk helpers, similar to dev_printk */
3757 static inline const char *netdev_name(const struct net_device
*dev
)
3759 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
3760 return "(unnamed net_device)";
3764 static inline const char *netdev_reg_state(const struct net_device
*dev
)
3766 switch (dev
->reg_state
) {
3767 case NETREG_UNINITIALIZED
: return " (uninitialized)";
3768 case NETREG_REGISTERED
: return "";
3769 case NETREG_UNREGISTERING
: return " (unregistering)";
3770 case NETREG_UNREGISTERED
: return " (unregistered)";
3771 case NETREG_RELEASED
: return " (released)";
3772 case NETREG_DUMMY
: return " (dummy)";
3775 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
3776 return " (unknown)";
3780 void netdev_printk(const char *level
, const struct net_device
*dev
,
3781 const char *format
, ...);
3783 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
3785 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
3787 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
3789 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
3791 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
3793 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
3795 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
3797 #define MODULE_ALIAS_NETDEV(device) \
3798 MODULE_ALIAS("netdev-" device)
3800 #if defined(CONFIG_DYNAMIC_DEBUG)
3801 #define netdev_dbg(__dev, format, args...) \
3803 dynamic_netdev_dbg(__dev, format, ##args); \
3805 #elif defined(DEBUG)
3806 #define netdev_dbg(__dev, format, args...) \
3807 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3809 #define netdev_dbg(__dev, format, args...) \
3812 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3816 #if defined(VERBOSE_DEBUG)
3817 #define netdev_vdbg netdev_dbg
3820 #define netdev_vdbg(dev, format, args...) \
3823 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3829 * netdev_WARN() acts like dev_printk(), but with the key difference
3830 * of using a WARN/WARN_ON to get the message out, including the
3831 * file/line information and a backtrace.
3833 #define netdev_WARN(dev, format, args...) \
3834 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3835 netdev_reg_state(dev), ##args)
3837 /* netif printk helpers, similar to netdev_printk */
3839 #define netif_printk(priv, type, level, dev, fmt, args...) \
3841 if (netif_msg_##type(priv)) \
3842 netdev_printk(level, (dev), fmt, ##args); \
3845 #define netif_level(level, priv, type, dev, fmt, args...) \
3847 if (netif_msg_##type(priv)) \
3848 netdev_##level(dev, fmt, ##args); \
3851 #define netif_emerg(priv, type, dev, fmt, args...) \
3852 netif_level(emerg, priv, type, dev, fmt, ##args)
3853 #define netif_alert(priv, type, dev, fmt, args...) \
3854 netif_level(alert, priv, type, dev, fmt, ##args)
3855 #define netif_crit(priv, type, dev, fmt, args...) \
3856 netif_level(crit, priv, type, dev, fmt, ##args)
3857 #define netif_err(priv, type, dev, fmt, args...) \
3858 netif_level(err, priv, type, dev, fmt, ##args)
3859 #define netif_warn(priv, type, dev, fmt, args...) \
3860 netif_level(warn, priv, type, dev, fmt, ##args)
3861 #define netif_notice(priv, type, dev, fmt, args...) \
3862 netif_level(notice, priv, type, dev, fmt, ##args)
3863 #define netif_info(priv, type, dev, fmt, args...) \
3864 netif_level(info, priv, type, dev, fmt, ##args)
3866 #if defined(CONFIG_DYNAMIC_DEBUG)
3867 #define netif_dbg(priv, type, netdev, format, args...) \
3869 if (netif_msg_##type(priv)) \
3870 dynamic_netdev_dbg(netdev, format, ##args); \
3872 #elif defined(DEBUG)
3873 #define netif_dbg(priv, type, dev, format, args...) \
3874 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3876 #define netif_dbg(priv, type, dev, format, args...) \
3879 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3884 #if defined(VERBOSE_DEBUG)
3885 #define netif_vdbg netif_dbg
3887 #define netif_vdbg(priv, type, dev, format, args...) \
3890 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3896 * The list of packet types we will receive (as opposed to discard)
3897 * and the routines to invoke.
3899 * Why 16. Because with 16 the only overlap we get on a hash of the
3900 * low nibble of the protocol value is RARP/SNAP/X.25.
3902 * NOTE: That is no longer true with the addition of VLAN tags. Not
3903 * sure which should go first, but I bet it won't make much
3904 * difference if we are running VLANs. The good news is that
3905 * this protocol won't be in the list unless compiled in, so
3906 * the average user (w/out VLANs) will not be adversely affected.
3922 #define PTYPE_HASH_SIZE (16)
3923 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3925 #endif /* _LINUX_NETDEVICE_H */