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 <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
59 /* source back-compat hooks */
60 #define SET_ETHTOOL_OPS(netdev,ops) \
61 ( (netdev)->ethtool_ops = (ops) )
63 void netdev_set_default_ethtool_ops(struct net_device
*dev
,
64 const struct ethtool_ops
*ops
);
66 /* hardware address assignment types */
67 #define NET_ADDR_PERM 0 /* address is permanent (default) */
68 #define NET_ADDR_RANDOM 1 /* address is generated randomly */
69 #define NET_ADDR_STOLEN 2 /* address is stolen from other device */
70 #define NET_ADDR_SET 3 /* address is set using
71 * dev_set_mac_address() */
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously, in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
99 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
101 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
104 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
105 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 /* Driver transmit return codes */
108 #define NETDEV_TX_MASK 0xf0
111 __NETDEV_TX_MIN
= INT_MIN
, /* make sure enum is signed */
112 NETDEV_TX_OK
= 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY
= 0x10, /* driver tx path was busy*/
114 NETDEV_TX_LOCKED
= 0x20, /* driver tx lock was already taken */
116 typedef enum netdev_tx netdev_tx_t
;
119 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
122 static inline bool dev_xmit_complete(int rc
)
125 * Positive cases with an skb consumed by a driver:
126 * - successful transmission (rc == NETDEV_TX_OK)
127 * - error while transmitting (rc < 0)
128 * - error while queueing to a different device (rc & NET_XMIT_MASK)
130 if (likely(rc
< NET_XMIT_MASK
))
137 * Compute the worst case header length according to the protocols
141 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats
{
164 unsigned long rx_packets
;
165 unsigned long tx_packets
;
166 unsigned long rx_bytes
;
167 unsigned long tx_bytes
;
168 unsigned long rx_errors
;
169 unsigned long tx_errors
;
170 unsigned long rx_dropped
;
171 unsigned long tx_dropped
;
172 unsigned long multicast
;
173 unsigned long collisions
;
174 unsigned long rx_length_errors
;
175 unsigned long rx_over_errors
;
176 unsigned long rx_crc_errors
;
177 unsigned long rx_frame_errors
;
178 unsigned long rx_fifo_errors
;
179 unsigned long rx_missed_errors
;
180 unsigned long tx_aborted_errors
;
181 unsigned long tx_carrier_errors
;
182 unsigned long tx_fifo_errors
;
183 unsigned long tx_heartbeat_errors
;
184 unsigned long tx_window_errors
;
185 unsigned long rx_compressed
;
186 unsigned long tx_compressed
;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed
;
202 struct netdev_hw_addr
{
203 struct list_head list
;
204 unsigned char addr
[MAX_ADDR_LEN
];
206 #define NETDEV_HW_ADDR_T_LAN 1
207 #define NETDEV_HW_ADDR_T_SAN 2
208 #define NETDEV_HW_ADDR_T_SLAVE 3
209 #define NETDEV_HW_ADDR_T_UNICAST 4
210 #define NETDEV_HW_ADDR_T_MULTICAST 5
215 struct rcu_head rcu_head
;
218 struct netdev_hw_addr_list
{
219 struct list_head list
;
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 list_for_each_entry(ha, &(l)->list, list)
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data
[HH_DATA_ALIGN(LL_MAX_HEADER
) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create
) (struct sk_buff
*skb
, struct net_device
*dev
,
267 unsigned short type
, const void *daddr
,
268 const void *saddr
, unsigned int len
);
269 int (*parse
)(const struct sk_buff
*skb
, unsigned char *haddr
);
270 int (*rebuild
)(struct sk_buff
*skb
);
271 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
272 void (*cache_update
)(struct hh_cache
*hh
,
273 const struct net_device
*dev
,
274 const unsigned char *haddr
);
277 /* These flag bits are private to the generic network queueing
278 * layer, they may not be explicitly referenced by any other
282 enum netdev_state_t
{
284 __LINK_STATE_PRESENT
,
285 __LINK_STATE_NOCARRIER
,
286 __LINK_STATE_LINKWATCH_PENDING
,
287 __LINK_STATE_DORMANT
,
292 * This structure holds at boot time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup
{
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init
netdev_boot_setup(char *str
);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-cpu poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list
;
317 unsigned int gro_count
;
318 int (*poll
)(struct napi_struct
*, int);
319 #ifdef CONFIG_NETPOLL
320 spinlock_t poll_lock
;
323 struct net_device
*dev
;
324 struct sk_buff
*gro_list
;
326 struct list_head dev_list
;
327 struct hlist_node napi_hash_node
;
328 unsigned int napi_id
;
332 NAPI_STATE_SCHED
, /* Poll is scheduled */
333 NAPI_STATE_DISABLE
, /* Disable pending */
334 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED
, /* In NAPI hash */
345 typedef enum gro_result gro_result_t
;
348 * enum rx_handler_result - Possible return values for rx_handlers.
349 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
351 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
352 * case skb->dev was changed by rx_handler.
353 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
354 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
356 * rx_handlers are functions called from inside __netif_receive_skb(), to do
357 * special processing of the skb, prior to delivery to protocol handlers.
359 * Currently, a net_device can only have a single rx_handler registered. Trying
360 * to register a second rx_handler will return -EBUSY.
362 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
363 * To unregister a rx_handler on a net_device, use
364 * netdev_rx_handler_unregister().
366 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
369 * If the rx_handler consumed to skb in some way, it should return
370 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
371 * the skb to be delivered in some other ways.
373 * If the rx_handler changed skb->dev, to divert the skb to another
374 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
375 * new device will be called if it exists.
377 * If the rx_handler consider the skb should be ignored, it should return
378 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
379 * are registered on exact device (ptype->dev == skb->dev).
381 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
382 * delivered, it should return RX_HANDLER_PASS.
384 * A device without a registered rx_handler will behave as if rx_handler
385 * returned RX_HANDLER_PASS.
388 enum rx_handler_result
{
394 typedef enum rx_handler_result rx_handler_result_t
;
395 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
397 void __napi_schedule(struct napi_struct
*n
);
399 static inline bool napi_disable_pending(struct napi_struct
*n
)
401 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
405 * napi_schedule_prep - check if napi can be scheduled
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
413 static inline bool napi_schedule_prep(struct napi_struct
*n
)
415 return !napi_disable_pending(n
) &&
416 !test_and_set_bit(NAPI_STATE_SCHED
, &n
->state
);
420 * napi_schedule - schedule NAPI poll
423 * Schedule NAPI poll routine to be called if it is not already
426 static inline void napi_schedule(struct napi_struct
*n
)
428 if (napi_schedule_prep(n
))
432 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
433 static inline bool napi_reschedule(struct napi_struct
*napi
)
435 if (napi_schedule_prep(napi
)) {
436 __napi_schedule(napi
);
443 * napi_complete - NAPI processing complete
446 * Mark NAPI processing as complete.
448 void __napi_complete(struct napi_struct
*n
);
449 void napi_complete(struct napi_struct
*n
);
452 * napi_by_id - lookup a NAPI by napi_id
453 * @napi_id: hashed napi_id
455 * lookup @napi_id in napi_hash table
456 * must be called under rcu_read_lock()
458 struct napi_struct
*napi_by_id(unsigned int napi_id
);
461 * napi_hash_add - add a NAPI to global hashtable
462 * @napi: napi context
464 * generate a new napi_id and store a @napi under it in napi_hash
466 void napi_hash_add(struct napi_struct
*napi
);
469 * napi_hash_del - remove a NAPI from global table
470 * @napi: napi context
472 * Warning: caller must observe rcu grace period
473 * before freeing memory containing @napi
475 void napi_hash_del(struct napi_struct
*napi
);
478 * napi_disable - prevent NAPI from scheduling
481 * Stop NAPI from being scheduled on this context.
482 * Waits till any outstanding processing completes.
484 static inline void napi_disable(struct napi_struct
*n
)
487 set_bit(NAPI_STATE_DISABLE
, &n
->state
);
488 while (test_and_set_bit(NAPI_STATE_SCHED
, &n
->state
))
490 clear_bit(NAPI_STATE_DISABLE
, &n
->state
);
494 * napi_enable - enable NAPI scheduling
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct
*n
)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
503 smp_mb__before_clear_bit();
504 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct
*n
)
518 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
522 # define napi_synchronize(n) barrier()
525 enum netdev_queue_state_t
{
526 __QUEUE_STATE_DRV_XOFF
,
527 __QUEUE_STATE_STACK_XOFF
,
528 __QUEUE_STATE_FROZEN
,
529 #define QUEUE_STATE_ANY_XOFF ((1 << __QUEUE_STATE_DRV_XOFF) | \
530 (1 << __QUEUE_STATE_STACK_XOFF))
531 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
532 (1 << __QUEUE_STATE_FROZEN))
535 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
536 * netif_tx_* functions below are used to manipulate this flag. The
537 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
538 * queue independently. The netif_xmit_*stopped functions below are called
539 * to check if the queue has been stopped by the driver or stack (either
540 * of the XOFF bits are set in the state). Drivers should not need to call
541 * netif_xmit*stopped functions, they should only be using netif_tx_*.
544 struct netdev_queue
{
548 struct net_device
*dev
;
550 struct Qdisc
*qdisc_sleeping
;
554 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
560 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
563 * please use this field instead of dev->trans_start
565 unsigned long trans_start
;
568 * Number of TX timeouts for this queue
569 * (/sys/class/net/DEV/Q/trans_timeout)
571 unsigned long trans_timeout
;
578 } ____cacheline_aligned_in_smp
;
580 static inline int netdev_queue_numa_node_read(const struct netdev_queue
*q
)
582 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 static inline void netdev_queue_numa_node_write(struct netdev_queue
*q
, int node
)
591 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 * This structure holds an RPS map which can be of variable length. The
599 * map is an array of CPUs.
606 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
609 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
610 * tail pointer for that CPU's input queue at the time of last enqueue, and
611 * a hardware filter index.
613 struct rps_dev_flow
{
616 unsigned int last_qtail
;
618 #define RPS_NO_FILTER 0xffff
621 * The rps_dev_flow_table structure contains a table of flow mappings.
623 struct rps_dev_flow_table
{
626 struct rps_dev_flow flows
[0];
628 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
629 ((_num) * sizeof(struct rps_dev_flow)))
632 * The rps_sock_flow_table contains mappings of flows to the last CPU
633 * on which they were processed by the application (set in recvmsg).
635 struct rps_sock_flow_table
{
639 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
640 ((_num) * sizeof(u16)))
642 #define RPS_NO_CPU 0xffff
644 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
648 unsigned int cpu
, index
= hash
& table
->mask
;
650 /* We only give a hint, preemption can change cpu under us */
651 cpu
= raw_smp_processor_id();
653 if (table
->ents
[index
] != cpu
)
654 table
->ents
[index
] = cpu
;
658 static inline void rps_reset_sock_flow(struct rps_sock_flow_table
*table
,
662 table
->ents
[hash
& table
->mask
] = RPS_NO_CPU
;
665 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
667 #ifdef CONFIG_RFS_ACCEL
668 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
671 #endif /* CONFIG_RPS */
673 /* This structure contains an instance of an RX queue. */
674 struct netdev_rx_queue
{
676 struct rps_map __rcu
*rps_map
;
677 struct rps_dev_flow_table __rcu
*rps_flow_table
;
680 struct net_device
*dev
;
681 } ____cacheline_aligned_in_smp
;
684 * RX queue sysfs structures and functions.
686 struct rx_queue_attribute
{
687 struct attribute attr
;
688 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
689 struct rx_queue_attribute
*attr
, char *buf
);
690 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
691 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
696 * This structure holds an XPS map which can be of variable length. The
697 * map is an array of queues.
701 unsigned int alloc_len
;
705 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
706 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
710 * This structure holds all XPS maps for device. Maps are indexed by CPU.
712 struct xps_dev_maps
{
714 struct xps_map __rcu
*cpu_map
[0];
716 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
717 (nr_cpu_ids * sizeof(struct xps_map *)))
718 #endif /* CONFIG_XPS */
720 #define TC_MAX_QUEUE 16
721 #define TC_BITMASK 15
722 /* HW offloaded queuing disciplines txq count and offset maps */
723 struct netdev_tc_txq
{
728 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
730 * This structure is to hold information about the device
731 * configured to run FCoE protocol stack.
733 struct netdev_fcoe_hbainfo
{
734 char manufacturer
[64];
735 char serial_number
[64];
736 char hardware_version
[64];
737 char driver_version
[64];
738 char optionrom_version
[64];
739 char firmware_version
[64];
741 char model_description
[256];
745 #define MAX_PHYS_PORT_ID_LEN 32
747 /* This structure holds a unique identifier to identify the
748 * physical port used by a netdevice.
750 struct netdev_phys_port_id
{
751 unsigned char id
[MAX_PHYS_PORT_ID_LEN
];
752 unsigned char id_len
;
755 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
756 struct sk_buff
*skb
);
759 * This structure defines the management hooks for network devices.
760 * The following hooks can be defined; unless noted otherwise, they are
761 * optional and can be filled with a null pointer.
763 * int (*ndo_init)(struct net_device *dev);
764 * This function is called once when network device is registered.
765 * The network device can use this to any late stage initializaton
766 * or semantic validattion. It can fail with an error code which will
767 * be propogated back to register_netdev
769 * void (*ndo_uninit)(struct net_device *dev);
770 * This function is called when device is unregistered or when registration
771 * fails. It is not called if init fails.
773 * int (*ndo_open)(struct net_device *dev);
774 * This function is called when network device transistions to the up
777 * int (*ndo_stop)(struct net_device *dev);
778 * This function is called when network device transistions to the down
781 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
782 * struct net_device *dev);
783 * Called when a packet needs to be transmitted.
784 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
785 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
786 * Required can not be NULL.
788 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
789 * void *accel_priv, select_queue_fallback_t fallback);
790 * Called to decide which queue to when device supports multiple
793 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
794 * This function is called to allow device receiver to make
795 * changes to configuration when multicast or promiscious is enabled.
797 * void (*ndo_set_rx_mode)(struct net_device *dev);
798 * This function is called device changes address list filtering.
799 * If driver handles unicast address filtering, it should set
800 * IFF_UNICAST_FLT to its priv_flags.
802 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
803 * This function is called when the Media Access Control address
804 * needs to be changed. If this interface is not defined, the
805 * mac address can not be changed.
807 * int (*ndo_validate_addr)(struct net_device *dev);
808 * Test if Media Access Control address is valid for the device.
810 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
811 * Called when a user request an ioctl which can't be handled by
812 * the generic interface code. If not defined ioctl's return
813 * not supported error code.
815 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
816 * Used to set network devices bus interface parameters. This interface
817 * is retained for legacy reason, new devices should use the bus
818 * interface (PCI) for low level management.
820 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
821 * Called when a user wants to change the Maximum Transfer Unit
822 * of a device. If not defined, any request to change MTU will
823 * will return an error.
825 * void (*ndo_tx_timeout)(struct net_device *dev);
826 * Callback uses when the transmitter has not made any progress
827 * for dev->watchdog ticks.
829 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
830 * struct rtnl_link_stats64 *storage);
831 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
832 * Called when a user wants to get the network device usage
833 * statistics. Drivers must do one of the following:
834 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
835 * rtnl_link_stats64 structure passed by the caller.
836 * 2. Define @ndo_get_stats to update a net_device_stats structure
837 * (which should normally be dev->stats) and return a pointer to
838 * it. The structure may be changed asynchronously only if each
839 * field is written atomically.
840 * 3. Update dev->stats asynchronously and atomically, and define
843 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
844 * If device support VLAN filtering this function is called when a
845 * VLAN id is registered.
847 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
848 * If device support VLAN filtering this function is called when a
849 * VLAN id is unregistered.
851 * void (*ndo_poll_controller)(struct net_device *dev);
853 * SR-IOV management functions.
854 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
855 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
856 * int (*ndo_set_vf_tx_rate)(struct net_device *dev, int vf, int rate);
857 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
858 * int (*ndo_get_vf_config)(struct net_device *dev,
859 * int vf, struct ifla_vf_info *ivf);
860 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
861 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
862 * struct nlattr *port[]);
863 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
864 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
865 * Called to setup 'tc' number of traffic classes in the net device. This
866 * is always called from the stack with the rtnl lock held and netif tx
867 * queues stopped. This allows the netdevice to perform queue management
870 * Fiber Channel over Ethernet (FCoE) offload functions.
871 * int (*ndo_fcoe_enable)(struct net_device *dev);
872 * Called when the FCoE protocol stack wants to start using LLD for FCoE
873 * so the underlying device can perform whatever needed configuration or
874 * initialization to support acceleration of FCoE traffic.
876 * int (*ndo_fcoe_disable)(struct net_device *dev);
877 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
878 * so the underlying device can perform whatever needed clean-ups to
879 * stop supporting acceleration of FCoE traffic.
881 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
882 * struct scatterlist *sgl, unsigned int sgc);
883 * Called when the FCoE Initiator wants to initialize an I/O that
884 * is a possible candidate for Direct Data Placement (DDP). The LLD can
885 * perform necessary setup and returns 1 to indicate the device is set up
886 * successfully to perform DDP on this I/O, otherwise this returns 0.
888 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
889 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
890 * indicated by the FC exchange id 'xid', so the underlying device can
891 * clean up and reuse resources for later DDP requests.
893 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
894 * struct scatterlist *sgl, unsigned int sgc);
895 * Called when the FCoE Target wants to initialize an I/O that
896 * is a possible candidate for Direct Data Placement (DDP). The LLD can
897 * perform necessary setup and returns 1 to indicate the device is set up
898 * successfully to perform DDP on this I/O, otherwise this returns 0.
900 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
901 * struct netdev_fcoe_hbainfo *hbainfo);
902 * Called when the FCoE Protocol stack wants information on the underlying
903 * device. This information is utilized by the FCoE protocol stack to
904 * register attributes with Fiber Channel management service as per the
905 * FC-GS Fabric Device Management Information(FDMI) specification.
907 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
908 * Called when the underlying device wants to override default World Wide
909 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
910 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
911 * protocol stack to use.
914 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
915 * u16 rxq_index, u32 flow_id);
916 * Set hardware filter for RFS. rxq_index is the target queue index;
917 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
918 * Return the filter ID on success, or a negative error code.
920 * Slave management functions (for bridge, bonding, etc).
921 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
922 * Called to make another netdev an underling.
924 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
925 * Called to release previously enslaved netdev.
927 * Feature/offload setting functions.
928 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
929 * netdev_features_t features);
930 * Adjusts the requested feature flags according to device-specific
931 * constraints, and returns the resulting flags. Must not modify
934 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
935 * Called to update device configuration to new features. Passed
936 * feature set might be less than what was returned by ndo_fix_features()).
937 * Must return >0 or -errno if it changed dev->features itself.
939 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
940 * struct net_device *dev,
941 * const unsigned char *addr, u16 flags)
942 * Adds an FDB entry to dev for addr.
943 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
944 * struct net_device *dev,
945 * const unsigned char *addr)
946 * Deletes the FDB entry from dev coresponding to addr.
947 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
948 * struct net_device *dev, int idx)
949 * Used to add FDB entries to dump requests. Implementers should add
950 * entries to skb and update idx with the number of entries.
952 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
953 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
954 * struct net_device *dev, u32 filter_mask)
956 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
957 * Called to change device carrier. Soft-devices (like dummy, team, etc)
958 * which do not represent real hardware may define this to allow their
959 * userspace components to manage their virtual carrier state. Devices
960 * that determine carrier state from physical hardware properties (eg
961 * network cables) or protocol-dependent mechanisms (eg
962 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
964 * int (*ndo_get_phys_port_id)(struct net_device *dev,
965 * struct netdev_phys_port_id *ppid);
966 * Called to get ID of physical port of this device. If driver does
967 * not implement this, it is assumed that the hw is not able to have
968 * multiple net devices on single physical port.
970 * void (*ndo_add_vxlan_port)(struct net_device *dev,
971 * sa_family_t sa_family, __be16 port);
972 * Called by vxlan to notiy a driver about the UDP port and socket
973 * address family that vxlan is listnening to. It is called only when
974 * a new port starts listening. The operation is protected by the
975 * vxlan_net->sock_lock.
977 * void (*ndo_del_vxlan_port)(struct net_device *dev,
978 * sa_family_t sa_family, __be16 port);
979 * Called by vxlan to notify the driver about a UDP port and socket
980 * address family that vxlan is not listening to anymore. The operation
981 * is protected by the vxlan_net->sock_lock.
983 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
984 * struct net_device *dev)
985 * Called by upper layer devices to accelerate switching or other
986 * station functionality into hardware. 'pdev is the lowerdev
987 * to use for the offload and 'dev' is the net device that will
988 * back the offload. Returns a pointer to the private structure
989 * the upper layer will maintain.
990 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
991 * Called by upper layer device to delete the station created
992 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
993 * the station and priv is the structure returned by the add
995 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
996 * struct net_device *dev,
998 * Callback to use for xmit over the accelerated station. This
999 * is used in place of ndo_start_xmit on accelerated net
1002 struct net_device_ops
{
1003 int (*ndo_init
)(struct net_device
*dev
);
1004 void (*ndo_uninit
)(struct net_device
*dev
);
1005 int (*ndo_open
)(struct net_device
*dev
);
1006 int (*ndo_stop
)(struct net_device
*dev
);
1007 netdev_tx_t (*ndo_start_xmit
) (struct sk_buff
*skb
,
1008 struct net_device
*dev
);
1009 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1010 struct sk_buff
*skb
,
1012 select_queue_fallback_t fallback
);
1013 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1015 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1016 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1018 int (*ndo_validate_addr
)(struct net_device
*dev
);
1019 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1020 struct ifreq
*ifr
, int cmd
);
1021 int (*ndo_set_config
)(struct net_device
*dev
,
1023 int (*ndo_change_mtu
)(struct net_device
*dev
,
1025 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1026 struct neigh_parms
*);
1027 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1029 struct rtnl_link_stats64
* (*ndo_get_stats64
)(struct net_device
*dev
,
1030 struct rtnl_link_stats64
*storage
);
1031 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1033 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1034 __be16 proto
, u16 vid
);
1035 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1036 __be16 proto
, u16 vid
);
1037 #ifdef CONFIG_NET_POLL_CONTROLLER
1038 void (*ndo_poll_controller
)(struct net_device
*dev
);
1039 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1040 struct netpoll_info
*info
,
1042 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1044 #ifdef CONFIG_NET_RX_BUSY_POLL
1045 int (*ndo_busy_poll
)(struct napi_struct
*dev
);
1047 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1048 int queue
, u8
*mac
);
1049 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1050 int queue
, u16 vlan
, u8 qos
);
1051 int (*ndo_set_vf_tx_rate
)(struct net_device
*dev
,
1053 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1054 int vf
, bool setting
);
1055 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1057 struct ifla_vf_info
*ivf
);
1058 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1059 int vf
, int link_state
);
1060 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1062 struct nlattr
*port
[]);
1063 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1064 int vf
, struct sk_buff
*skb
);
1065 int (*ndo_setup_tc
)(struct net_device
*dev
, u8 tc
);
1066 #if IS_ENABLED(CONFIG_FCOE)
1067 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1068 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1069 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1071 struct scatterlist
*sgl
,
1073 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1075 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1077 struct scatterlist
*sgl
,
1079 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1080 struct netdev_fcoe_hbainfo
*hbainfo
);
1083 #if IS_ENABLED(CONFIG_LIBFCOE)
1084 #define NETDEV_FCOE_WWNN 0
1085 #define NETDEV_FCOE_WWPN 1
1086 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1087 u64
*wwn
, int type
);
1090 #ifdef CONFIG_RFS_ACCEL
1091 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1092 const struct sk_buff
*skb
,
1096 int (*ndo_add_slave
)(struct net_device
*dev
,
1097 struct net_device
*slave_dev
);
1098 int (*ndo_del_slave
)(struct net_device
*dev
,
1099 struct net_device
*slave_dev
);
1100 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1101 netdev_features_t features
);
1102 int (*ndo_set_features
)(struct net_device
*dev
,
1103 netdev_features_t features
);
1104 int (*ndo_neigh_construct
)(struct neighbour
*n
);
1105 void (*ndo_neigh_destroy
)(struct neighbour
*n
);
1107 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1108 struct nlattr
*tb
[],
1109 struct net_device
*dev
,
1110 const unsigned char *addr
,
1112 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1113 struct nlattr
*tb
[],
1114 struct net_device
*dev
,
1115 const unsigned char *addr
);
1116 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1117 struct netlink_callback
*cb
,
1118 struct net_device
*dev
,
1121 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1122 struct nlmsghdr
*nlh
);
1123 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1125 struct net_device
*dev
,
1127 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1128 struct nlmsghdr
*nlh
);
1129 int (*ndo_change_carrier
)(struct net_device
*dev
,
1131 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1132 struct netdev_phys_port_id
*ppid
);
1133 void (*ndo_add_vxlan_port
)(struct net_device
*dev
,
1134 sa_family_t sa_family
,
1136 void (*ndo_del_vxlan_port
)(struct net_device
*dev
,
1137 sa_family_t sa_family
,
1140 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1141 struct net_device
*dev
);
1142 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1145 netdev_tx_t (*ndo_dfwd_start_xmit
) (struct sk_buff
*skb
,
1146 struct net_device
*dev
,
1151 * The DEVICE structure.
1152 * Actually, this whole structure is a big mistake. It mixes I/O
1153 * data with strictly "high-level" data, and it has to know about
1154 * almost every data structure used in the INET module.
1156 * FIXME: cleanup struct net_device such that network protocol info
1163 * This is the first field of the "visible" part of this structure
1164 * (i.e. as seen by users in the "Space.c" file). It is the name
1167 char name
[IFNAMSIZ
];
1169 /* device name hash chain, please keep it close to name[] */
1170 struct hlist_node name_hlist
;
1176 * I/O specific fields
1177 * FIXME: Merge these and struct ifmap into one
1179 unsigned long mem_end
; /* shared mem end */
1180 unsigned long mem_start
; /* shared mem start */
1181 unsigned long base_addr
; /* device I/O address */
1182 int irq
; /* device IRQ number */
1185 * Some hardware also needs these fields, but they are not
1186 * part of the usual set specified in Space.c.
1189 unsigned long state
;
1191 struct list_head dev_list
;
1192 struct list_head napi_list
;
1193 struct list_head unreg_list
;
1194 struct list_head close_list
;
1196 /* directly linked devices, like slaves for bonding */
1198 struct list_head upper
;
1199 struct list_head lower
;
1202 /* all linked devices, *including* neighbours */
1204 struct list_head upper
;
1205 struct list_head lower
;
1209 /* currently active device features */
1210 netdev_features_t features
;
1211 /* user-changeable features */
1212 netdev_features_t hw_features
;
1213 /* user-requested features */
1214 netdev_features_t wanted_features
;
1215 /* mask of features inheritable by VLAN devices */
1216 netdev_features_t vlan_features
;
1217 /* mask of features inherited by encapsulating devices
1218 * This field indicates what encapsulation offloads
1219 * the hardware is capable of doing, and drivers will
1220 * need to set them appropriately.
1222 netdev_features_t hw_enc_features
;
1223 /* mask of fetures inheritable by MPLS */
1224 netdev_features_t mpls_features
;
1226 /* Interface index. Unique device identifier */
1230 struct net_device_stats stats
;
1231 atomic_long_t rx_dropped
; /* dropped packets by core network
1232 * Do not use this in drivers.
1235 #ifdef CONFIG_WIRELESS_EXT
1236 /* List of functions to handle Wireless Extensions (instead of ioctl).
1237 * See <net/iw_handler.h> for details. Jean II */
1238 const struct iw_handler_def
* wireless_handlers
;
1239 /* Instance data managed by the core of Wireless Extensions. */
1240 struct iw_public_data
* wireless_data
;
1242 /* Management operations */
1243 const struct net_device_ops
*netdev_ops
;
1244 const struct ethtool_ops
*ethtool_ops
;
1245 const struct forwarding_accel_ops
*fwd_ops
;
1247 /* Hardware header description */
1248 const struct header_ops
*header_ops
;
1250 unsigned int flags
; /* interface flags (a la BSD) */
1251 unsigned int priv_flags
; /* Like 'flags' but invisible to userspace.
1252 * See if.h for definitions. */
1253 unsigned short gflags
;
1254 unsigned short padded
; /* How much padding added by alloc_netdev() */
1256 unsigned char operstate
; /* RFC2863 operstate */
1257 unsigned char link_mode
; /* mapping policy to operstate */
1259 unsigned char if_port
; /* Selectable AUI, TP,..*/
1260 unsigned char dma
; /* DMA channel */
1262 unsigned int mtu
; /* interface MTU value */
1263 unsigned short type
; /* interface hardware type */
1264 unsigned short hard_header_len
; /* hardware hdr length */
1266 /* extra head- and tailroom the hardware may need, but not in all cases
1267 * can this be guaranteed, especially tailroom. Some cases also use
1268 * LL_MAX_HEADER instead to allocate the skb.
1270 unsigned short needed_headroom
;
1271 unsigned short needed_tailroom
;
1273 /* Interface address info. */
1274 unsigned char perm_addr
[MAX_ADDR_LEN
]; /* permanent hw address */
1275 unsigned char addr_assign_type
; /* hw address assignment type */
1276 unsigned char addr_len
; /* hardware address length */
1277 unsigned short neigh_priv_len
;
1278 unsigned short dev_id
; /* Used to differentiate devices
1279 * that share the same link
1282 spinlock_t addr_list_lock
;
1283 struct netdev_hw_addr_list uc
; /* Unicast mac addresses */
1284 struct netdev_hw_addr_list mc
; /* Multicast mac addresses */
1285 struct netdev_hw_addr_list dev_addrs
; /* list of device
1289 struct kset
*queues_kset
;
1293 unsigned int promiscuity
;
1294 unsigned int allmulti
;
1297 /* Protocol specific pointers */
1299 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1300 struct vlan_info __rcu
*vlan_info
; /* VLAN info */
1302 #if IS_ENABLED(CONFIG_NET_DSA)
1303 struct dsa_switch_tree
*dsa_ptr
; /* dsa specific data */
1305 #if IS_ENABLED(CONFIG_TIPC)
1306 struct tipc_bearer __rcu
*tipc_ptr
; /* TIPC specific data */
1308 void *atalk_ptr
; /* AppleTalk link */
1309 struct in_device __rcu
*ip_ptr
; /* IPv4 specific data */
1310 struct dn_dev __rcu
*dn_ptr
; /* DECnet specific data */
1311 struct inet6_dev __rcu
*ip6_ptr
; /* IPv6 specific data */
1312 void *ax25_ptr
; /* AX.25 specific data */
1313 struct wireless_dev
*ieee80211_ptr
; /* IEEE 802.11 specific data,
1314 assign before registering */
1317 * Cache lines mostly used on receive path (including eth_type_trans())
1319 unsigned long last_rx
; /* Time of last Rx
1320 * This should not be set in
1321 * drivers, unless really needed,
1322 * because network stack (bonding)
1323 * use it if/when necessary, to
1324 * avoid dirtying this cache line.
1327 /* Interface address info used in eth_type_trans() */
1328 unsigned char *dev_addr
; /* hw address, (before bcast
1329 because most packets are
1334 struct netdev_rx_queue
*_rx
;
1336 /* Number of RX queues allocated at register_netdev() time */
1337 unsigned int num_rx_queues
;
1339 /* Number of RX queues currently active in device */
1340 unsigned int real_num_rx_queues
;
1344 rx_handler_func_t __rcu
*rx_handler
;
1345 void __rcu
*rx_handler_data
;
1347 struct netdev_queue __rcu
*ingress_queue
;
1348 unsigned char broadcast
[MAX_ADDR_LEN
]; /* hw bcast add */
1352 * Cache lines mostly used on transmit path
1354 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1356 /* Number of TX queues allocated at alloc_netdev_mq() time */
1357 unsigned int num_tx_queues
;
1359 /* Number of TX queues currently active in device */
1360 unsigned int real_num_tx_queues
;
1362 /* root qdisc from userspace point of view */
1363 struct Qdisc
*qdisc
;
1365 unsigned long tx_queue_len
; /* Max frames per queue allowed */
1366 spinlock_t tx_global_lock
;
1369 struct xps_dev_maps __rcu
*xps_maps
;
1371 #ifdef CONFIG_RFS_ACCEL
1372 /* CPU reverse-mapping for RX completion interrupts, indexed
1373 * by RX queue number. Assigned by driver. This must only be
1374 * set if the ndo_rx_flow_steer operation is defined. */
1375 struct cpu_rmap
*rx_cpu_rmap
;
1378 /* These may be needed for future network-power-down code. */
1381 * trans_start here is expensive for high speed devices on SMP,
1382 * please use netdev_queue->trans_start instead.
1384 unsigned long trans_start
; /* Time (in jiffies) of last Tx */
1386 int watchdog_timeo
; /* used by dev_watchdog() */
1387 struct timer_list watchdog_timer
;
1389 /* Number of references to this device */
1390 int __percpu
*pcpu_refcnt
;
1392 /* delayed register/unregister */
1393 struct list_head todo_list
;
1394 /* device index hash chain */
1395 struct hlist_node index_hlist
;
1397 struct list_head link_watch_list
;
1399 /* register/unregister state machine */
1400 enum { NETREG_UNINITIALIZED
=0,
1401 NETREG_REGISTERED
, /* completed register_netdevice */
1402 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1403 NETREG_UNREGISTERED
, /* completed unregister todo */
1404 NETREG_RELEASED
, /* called free_netdev */
1405 NETREG_DUMMY
, /* dummy device for NAPI poll */
1408 bool dismantle
; /* device is going do be freed */
1411 RTNL_LINK_INITIALIZED
,
1412 RTNL_LINK_INITIALIZING
,
1413 } rtnl_link_state
:16;
1415 /* Called from unregister, can be used to call free_netdev */
1416 void (*destructor
)(struct net_device
*dev
);
1418 #ifdef CONFIG_NETPOLL
1419 struct netpoll_info __rcu
*npinfo
;
1422 #ifdef CONFIG_NET_NS
1423 /* Network namespace this network device is inside */
1427 /* mid-layer private */
1430 struct pcpu_lstats __percpu
*lstats
; /* loopback stats */
1431 struct pcpu_sw_netstats __percpu
*tstats
;
1432 struct pcpu_dstats __percpu
*dstats
; /* dummy stats */
1433 struct pcpu_vstats __percpu
*vstats
; /* veth stats */
1436 struct garp_port __rcu
*garp_port
;
1438 struct mrp_port __rcu
*mrp_port
;
1440 /* class/net/name entry */
1442 /* space for optional device, statistics, and wireless sysfs groups */
1443 const struct attribute_group
*sysfs_groups
[4];
1444 /* space for optional per-rx queue attributes */
1445 const struct attribute_group
*sysfs_rx_queue_group
;
1447 /* rtnetlink link ops */
1448 const struct rtnl_link_ops
*rtnl_link_ops
;
1450 /* for setting kernel sock attribute on TCP connection setup */
1451 #define GSO_MAX_SIZE 65536
1452 unsigned int gso_max_size
;
1453 #define GSO_MAX_SEGS 65535
1457 /* Data Center Bridging netlink ops */
1458 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1461 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1462 u8 prio_tc_map
[TC_BITMASK
+ 1];
1464 #if IS_ENABLED(CONFIG_FCOE)
1465 /* max exchange id for FCoE LRO by ddp */
1466 unsigned int fcoe_ddp_xid
;
1468 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1469 struct netprio_map __rcu
*priomap
;
1471 /* phy device may attach itself for hardware timestamping */
1472 struct phy_device
*phydev
;
1474 struct lock_class_key
*qdisc_tx_busylock
;
1476 /* group the device belongs to */
1479 struct pm_qos_request pm_qos_req
;
1481 #define to_net_dev(d) container_of(d, struct net_device, dev)
1483 #define NETDEV_ALIGN 32
1486 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1488 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1492 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1494 if (tc
>= dev
->num_tc
)
1497 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1502 void netdev_reset_tc(struct net_device
*dev
)
1505 memset(dev
->tc_to_txq
, 0, sizeof(dev
->tc_to_txq
));
1506 memset(dev
->prio_tc_map
, 0, sizeof(dev
->prio_tc_map
));
1510 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
)
1512 if (tc
>= dev
->num_tc
)
1515 dev
->tc_to_txq
[tc
].count
= count
;
1516 dev
->tc_to_txq
[tc
].offset
= offset
;
1521 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
)
1523 if (num_tc
> TC_MAX_QUEUE
)
1526 dev
->num_tc
= num_tc
;
1531 int netdev_get_num_tc(struct net_device
*dev
)
1537 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1540 return &dev
->_tx
[index
];
1543 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1544 void (*f
)(struct net_device
*,
1545 struct netdev_queue
*,
1551 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1552 f(dev
, &dev
->_tx
[i
], arg
);
1555 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1556 struct sk_buff
*skb
,
1560 * Net namespace inlines
1563 struct net
*dev_net(const struct net_device
*dev
)
1565 return read_pnet(&dev
->nd_net
);
1569 void dev_net_set(struct net_device
*dev
, struct net
*net
)
1571 #ifdef CONFIG_NET_NS
1572 release_net(dev
->nd_net
);
1573 dev
->nd_net
= hold_net(net
);
1577 static inline bool netdev_uses_dsa_tags(struct net_device
*dev
)
1579 #ifdef CONFIG_NET_DSA_TAG_DSA
1580 if (dev
->dsa_ptr
!= NULL
)
1581 return dsa_uses_dsa_tags(dev
->dsa_ptr
);
1587 static inline bool netdev_uses_trailer_tags(struct net_device
*dev
)
1589 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1590 if (dev
->dsa_ptr
!= NULL
)
1591 return dsa_uses_trailer_tags(dev
->dsa_ptr
);
1598 * netdev_priv - access network device private data
1599 * @dev: network device
1601 * Get network device private data
1603 static inline void *netdev_priv(const struct net_device
*dev
)
1605 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
1608 /* Set the sysfs physical device reference for the network logical device
1609 * if set prior to registration will cause a symlink during initialization.
1611 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1613 /* Set the sysfs device type for the network logical device to allow
1614 * fine-grained identification of different network device types. For
1615 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1617 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1619 /* Default NAPI poll() weight
1620 * Device drivers are strongly advised to not use bigger value
1622 #define NAPI_POLL_WEIGHT 64
1625 * netif_napi_add - initialize a napi context
1626 * @dev: network device
1627 * @napi: napi context
1628 * @poll: polling function
1629 * @weight: default weight
1631 * netif_napi_add() must be used to initialize a napi context prior to calling
1632 * *any* of the other napi related functions.
1634 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
1635 int (*poll
)(struct napi_struct
*, int), int weight
);
1638 * netif_napi_del - remove a napi context
1639 * @napi: napi context
1641 * netif_napi_del() removes a napi context from the network device napi list
1643 void netif_napi_del(struct napi_struct
*napi
);
1645 struct napi_gro_cb
{
1646 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1649 /* Length of frag0. */
1650 unsigned int frag0_len
;
1652 /* This indicates where we are processing relative to skb->data. */
1655 /* This is non-zero if the packet cannot be merged with the new skb. */
1658 /* Save the IP ID here and check when we get to the transport layer */
1661 /* Number of segments aggregated. */
1664 /* This is non-zero if the packet may be of the same flow. */
1669 #define NAPI_GRO_FREE 1
1670 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1672 /* jiffies when first packet was created/queued */
1675 /* Used in ipv6_gro_receive() */
1678 /* Used in udp_gro_receive */
1681 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1684 /* used in skb_gro_receive() slow path */
1685 struct sk_buff
*last
;
1688 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1690 struct packet_type
{
1691 __be16 type
; /* This is really htons(ether_type). */
1692 struct net_device
*dev
; /* NULL is wildcarded here */
1693 int (*func
) (struct sk_buff
*,
1694 struct net_device
*,
1695 struct packet_type
*,
1696 struct net_device
*);
1697 bool (*id_match
)(struct packet_type
*ptype
,
1699 void *af_packet_priv
;
1700 struct list_head list
;
1703 struct offload_callbacks
{
1704 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
1705 netdev_features_t features
);
1706 int (*gso_send_check
)(struct sk_buff
*skb
);
1707 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
1708 struct sk_buff
*skb
);
1709 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
1712 struct packet_offload
{
1713 __be16 type
; /* This is really htons(ether_type). */
1714 struct offload_callbacks callbacks
;
1715 struct list_head list
;
1718 struct udp_offload
{
1720 struct offload_callbacks callbacks
;
1723 /* often modified stats are per cpu, other are shared (netdev->stats) */
1724 struct pcpu_sw_netstats
{
1729 struct u64_stats_sync syncp
;
1732 #include <linux/notifier.h>
1734 /* netdevice notifier chain. Please remember to update the rtnetlink
1735 * notification exclusion list in rtnetlink_event() when adding new
1738 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1739 #define NETDEV_DOWN 0x0002
1740 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1741 detected a hardware crash and restarted
1742 - we can use this eg to kick tcp sessions
1744 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1745 #define NETDEV_REGISTER 0x0005
1746 #define NETDEV_UNREGISTER 0x0006
1747 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1748 #define NETDEV_CHANGEADDR 0x0008
1749 #define NETDEV_GOING_DOWN 0x0009
1750 #define NETDEV_CHANGENAME 0x000A
1751 #define NETDEV_FEAT_CHANGE 0x000B
1752 #define NETDEV_BONDING_FAILOVER 0x000C
1753 #define NETDEV_PRE_UP 0x000D
1754 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1755 #define NETDEV_POST_TYPE_CHANGE 0x000F
1756 #define NETDEV_POST_INIT 0x0010
1757 #define NETDEV_UNREGISTER_FINAL 0x0011
1758 #define NETDEV_RELEASE 0x0012
1759 #define NETDEV_NOTIFY_PEERS 0x0013
1760 #define NETDEV_JOIN 0x0014
1761 #define NETDEV_CHANGEUPPER 0x0015
1762 #define NETDEV_RESEND_IGMP 0x0016
1763 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1765 int register_netdevice_notifier(struct notifier_block
*nb
);
1766 int unregister_netdevice_notifier(struct notifier_block
*nb
);
1768 struct netdev_notifier_info
{
1769 struct net_device
*dev
;
1772 struct netdev_notifier_change_info
{
1773 struct netdev_notifier_info info
; /* must be first */
1774 unsigned int flags_changed
;
1777 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
1778 struct net_device
*dev
)
1783 static inline struct net_device
*
1784 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
1789 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
1792 extern rwlock_t dev_base_lock
; /* Device list lock */
1794 #define for_each_netdev(net, d) \
1795 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1796 #define for_each_netdev_reverse(net, d) \
1797 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1798 #define for_each_netdev_rcu(net, d) \
1799 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1800 #define for_each_netdev_safe(net, d, n) \
1801 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1802 #define for_each_netdev_continue(net, d) \
1803 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1804 #define for_each_netdev_continue_rcu(net, d) \
1805 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1806 #define for_each_netdev_in_bond_rcu(bond, slave) \
1807 for_each_netdev_rcu(&init_net, slave) \
1808 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1809 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1811 static inline struct net_device
*next_net_device(struct net_device
*dev
)
1813 struct list_head
*lh
;
1817 lh
= dev
->dev_list
.next
;
1818 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
1821 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
1823 struct list_head
*lh
;
1827 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
1828 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
1831 static inline struct net_device
*first_net_device(struct net
*net
)
1833 return list_empty(&net
->dev_base_head
) ? NULL
:
1834 net_device_entry(net
->dev_base_head
.next
);
1837 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
1839 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
1841 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
1844 int netdev_boot_setup_check(struct net_device
*dev
);
1845 unsigned long netdev_boot_base(const char *prefix
, int unit
);
1846 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
1847 const char *hwaddr
);
1848 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
1849 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
1850 void dev_add_pack(struct packet_type
*pt
);
1851 void dev_remove_pack(struct packet_type
*pt
);
1852 void __dev_remove_pack(struct packet_type
*pt
);
1853 void dev_add_offload(struct packet_offload
*po
);
1854 void dev_remove_offload(struct packet_offload
*po
);
1856 struct net_device
*dev_get_by_flags_rcu(struct net
*net
, unsigned short flags
,
1857 unsigned short mask
);
1858 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
1859 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
1860 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
1861 int dev_alloc_name(struct net_device
*dev
, const char *name
);
1862 int dev_open(struct net_device
*dev
);
1863 int dev_close(struct net_device
*dev
);
1864 void dev_disable_lro(struct net_device
*dev
);
1865 int dev_loopback_xmit(struct sk_buff
*newskb
);
1866 int dev_queue_xmit(struct sk_buff
*skb
);
1867 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
1868 int register_netdevice(struct net_device
*dev
);
1869 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
1870 void unregister_netdevice_many(struct list_head
*head
);
1871 static inline void unregister_netdevice(struct net_device
*dev
)
1873 unregister_netdevice_queue(dev
, NULL
);
1876 int netdev_refcnt_read(const struct net_device
*dev
);
1877 void free_netdev(struct net_device
*dev
);
1878 void netdev_freemem(struct net_device
*dev
);
1879 void synchronize_net(void);
1880 int init_dummy_netdev(struct net_device
*dev
);
1882 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
1883 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
1884 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
1885 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
1886 int dev_restart(struct net_device
*dev
);
1887 #ifdef CONFIG_NETPOLL_TRAP
1888 int netpoll_trap(void);
1890 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
1892 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
1894 return NAPI_GRO_CB(skb
)->data_offset
;
1897 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
1899 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
1902 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
1904 NAPI_GRO_CB(skb
)->data_offset
+= len
;
1907 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
1908 unsigned int offset
)
1910 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
1913 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
1915 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
1918 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
1919 unsigned int offset
)
1921 if (!pskb_may_pull(skb
, hlen
))
1924 NAPI_GRO_CB(skb
)->frag0
= NULL
;
1925 NAPI_GRO_CB(skb
)->frag0_len
= 0;
1926 return skb
->data
+ offset
;
1929 static inline void *skb_gro_mac_header(struct sk_buff
*skb
)
1931 return NAPI_GRO_CB(skb
)->frag0
?: skb_mac_header(skb
);
1934 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
1936 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
1937 skb_network_offset(skb
);
1940 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
1941 const void *start
, unsigned int len
)
1943 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1944 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
1945 csum_partial(start
, len
, 0));
1948 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
1949 unsigned short type
,
1950 const void *daddr
, const void *saddr
,
1953 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
1956 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
1959 static inline int dev_parse_header(const struct sk_buff
*skb
,
1960 unsigned char *haddr
)
1962 const struct net_device
*dev
= skb
->dev
;
1964 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
1966 return dev
->header_ops
->parse(skb
, haddr
);
1969 static inline int dev_rebuild_header(struct sk_buff
*skb
)
1971 const struct net_device
*dev
= skb
->dev
;
1973 if (!dev
->header_ops
|| !dev
->header_ops
->rebuild
)
1975 return dev
->header_ops
->rebuild(skb
);
1978 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
1979 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
1980 static inline int unregister_gifconf(unsigned int family
)
1982 return register_gifconf(family
, NULL
);
1985 #ifdef CONFIG_NET_FLOW_LIMIT
1986 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
1987 struct sd_flow_limit
{
1989 unsigned int num_buckets
;
1990 unsigned int history_head
;
1991 u16 history
[FLOW_LIMIT_HISTORY
];
1995 extern int netdev_flow_limit_table_len
;
1996 #endif /* CONFIG_NET_FLOW_LIMIT */
1999 * Incoming packets are placed on per-cpu queues
2001 struct softnet_data
{
2002 struct Qdisc
*output_queue
;
2003 struct Qdisc
**output_queue_tailp
;
2004 struct list_head poll_list
;
2005 struct sk_buff
*completion_queue
;
2006 struct sk_buff_head process_queue
;
2009 unsigned int processed
;
2010 unsigned int time_squeeze
;
2011 unsigned int cpu_collision
;
2012 unsigned int received_rps
;
2015 struct softnet_data
*rps_ipi_list
;
2017 /* Elements below can be accessed between CPUs for RPS */
2018 struct call_single_data csd ____cacheline_aligned_in_smp
;
2019 struct softnet_data
*rps_ipi_next
;
2021 unsigned int input_queue_head
;
2022 unsigned int input_queue_tail
;
2024 unsigned int dropped
;
2025 struct sk_buff_head input_pkt_queue
;
2026 struct napi_struct backlog
;
2028 #ifdef CONFIG_NET_FLOW_LIMIT
2029 struct sd_flow_limit __rcu
*flow_limit
;
2033 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2036 sd
->input_queue_head
++;
2040 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2041 unsigned int *qtail
)
2044 *qtail
= ++sd
->input_queue_tail
;
2048 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2050 void __netif_schedule(struct Qdisc
*q
);
2052 static inline void netif_schedule_queue(struct netdev_queue
*txq
)
2054 if (!(txq
->state
& QUEUE_STATE_ANY_XOFF
))
2055 __netif_schedule(txq
->qdisc
);
2058 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2062 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2063 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2066 static inline void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2068 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2072 * netif_start_queue - allow transmit
2073 * @dev: network device
2075 * Allow upper layers to call the device hard_start_xmit routine.
2077 static inline void netif_start_queue(struct net_device
*dev
)
2079 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2082 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2086 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2087 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2088 netif_tx_start_queue(txq
);
2092 static inline void netif_tx_wake_queue(struct netdev_queue
*dev_queue
)
2094 #ifdef CONFIG_NETPOLL_TRAP
2095 if (netpoll_trap()) {
2096 netif_tx_start_queue(dev_queue
);
2100 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
))
2101 __netif_schedule(dev_queue
->qdisc
);
2105 * netif_wake_queue - restart transmit
2106 * @dev: network device
2108 * Allow upper layers to call the device hard_start_xmit routine.
2109 * Used for flow control when transmit resources are available.
2111 static inline void netif_wake_queue(struct net_device
*dev
)
2113 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2116 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2120 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2121 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2122 netif_tx_wake_queue(txq
);
2126 static inline void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2128 if (WARN_ON(!dev_queue
)) {
2129 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2132 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2136 * netif_stop_queue - stop transmitted packets
2137 * @dev: network device
2139 * Stop upper layers calling the device hard_start_xmit routine.
2140 * Used for flow control when transmit resources are unavailable.
2142 static inline void netif_stop_queue(struct net_device
*dev
)
2144 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2147 static inline void netif_tx_stop_all_queues(struct net_device
*dev
)
2151 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2152 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2153 netif_tx_stop_queue(txq
);
2157 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2159 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2163 * netif_queue_stopped - test if transmit queue is flowblocked
2164 * @dev: network device
2166 * Test if transmit queue on device is currently unable to send.
2168 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2170 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2173 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2175 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2178 static inline bool netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2180 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2183 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2187 dql_queued(&dev_queue
->dql
, bytes
);
2189 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2192 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2195 * The XOFF flag must be set before checking the dql_avail below,
2196 * because in netdev_tx_completed_queue we update the dql_completed
2197 * before checking the XOFF flag.
2201 /* check again in case another CPU has just made room avail */
2202 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2203 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2208 * netdev_sent_queue - report the number of bytes queued to hardware
2209 * @dev: network device
2210 * @bytes: number of bytes queued to the hardware device queue
2212 * Report the number of bytes queued for sending/completion to the network
2213 * device hardware queue. @bytes should be a good approximation and should
2214 * exactly match netdev_completed_queue() @bytes
2216 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2218 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2221 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2222 unsigned int pkts
, unsigned int bytes
)
2225 if (unlikely(!bytes
))
2228 dql_completed(&dev_queue
->dql
, bytes
);
2231 * Without the memory barrier there is a small possiblity that
2232 * netdev_tx_sent_queue will miss the update and cause the queue to
2233 * be stopped forever
2237 if (dql_avail(&dev_queue
->dql
) < 0)
2240 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2241 netif_schedule_queue(dev_queue
);
2246 * netdev_completed_queue - report bytes and packets completed by device
2247 * @dev: network device
2248 * @pkts: actual number of packets sent over the medium
2249 * @bytes: actual number of bytes sent over the medium
2251 * Report the number of bytes and packets transmitted by the network device
2252 * hardware queue over the physical medium, @bytes must exactly match the
2253 * @bytes amount passed to netdev_sent_queue()
2255 static inline void netdev_completed_queue(struct net_device
*dev
,
2256 unsigned int pkts
, unsigned int bytes
)
2258 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
2261 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
2264 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
2270 * netdev_reset_queue - reset the packets and bytes count of a network device
2271 * @dev_queue: network device
2273 * Reset the bytes and packet count of a network device and clear the
2274 * software flow control OFF bit for this network device
2276 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
2278 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
2282 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2283 * @dev: network device
2284 * @queue_index: given tx queue index
2286 * Returns 0 if given tx queue index >= number of device tx queues,
2287 * otherwise returns the originally passed tx queue index.
2289 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
2291 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
2292 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2293 dev
->name
, queue_index
,
2294 dev
->real_num_tx_queues
);
2302 * netif_running - test if up
2303 * @dev: network device
2305 * Test if the device has been brought up.
2307 static inline bool netif_running(const struct net_device
*dev
)
2309 return test_bit(__LINK_STATE_START
, &dev
->state
);
2313 * Routines to manage the subqueues on a device. We only need start
2314 * stop, and a check if it's stopped. All other device management is
2315 * done at the overall netdevice level.
2316 * Also test the device if we're multiqueue.
2320 * netif_start_subqueue - allow sending packets on subqueue
2321 * @dev: network device
2322 * @queue_index: sub queue index
2324 * Start individual transmit queue of a device with multiple transmit queues.
2326 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
2328 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2330 netif_tx_start_queue(txq
);
2334 * netif_stop_subqueue - stop sending packets on subqueue
2335 * @dev: network device
2336 * @queue_index: sub queue index
2338 * Stop individual transmit queue of a device with multiple transmit queues.
2340 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
2342 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2343 #ifdef CONFIG_NETPOLL_TRAP
2347 netif_tx_stop_queue(txq
);
2351 * netif_subqueue_stopped - test status of subqueue
2352 * @dev: network device
2353 * @queue_index: sub queue index
2355 * Check individual transmit queue of a device with multiple transmit queues.
2357 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
2360 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2362 return netif_tx_queue_stopped(txq
);
2365 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
2366 struct sk_buff
*skb
)
2368 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
2372 * netif_wake_subqueue - allow sending packets on subqueue
2373 * @dev: network device
2374 * @queue_index: sub queue index
2376 * Resume individual transmit queue of a device with multiple transmit queues.
2378 static inline void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
)
2380 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
2381 #ifdef CONFIG_NETPOLL_TRAP
2385 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF
, &txq
->state
))
2386 __netif_schedule(txq
->qdisc
);
2390 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
2393 static inline int netif_set_xps_queue(struct net_device
*dev
,
2394 const struct cpumask
*mask
,
2402 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2403 * as a distribution range limit for the returned value.
2405 static inline u16
skb_tx_hash(const struct net_device
*dev
,
2406 const struct sk_buff
*skb
)
2408 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
2412 * netif_is_multiqueue - test if device has multiple transmit queues
2413 * @dev: network device
2415 * Check if device has multiple transmit queues
2417 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
2419 return dev
->num_tx_queues
> 1;
2422 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
2425 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
2427 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
2434 static inline int netif_copy_real_num_queues(struct net_device
*to_dev
,
2435 const struct net_device
*from_dev
)
2439 err
= netif_set_real_num_tx_queues(to_dev
,
2440 from_dev
->real_num_tx_queues
);
2444 return netif_set_real_num_rx_queues(to_dev
,
2445 from_dev
->real_num_rx_queues
);
2452 static inline unsigned int get_netdev_rx_queue_index(
2453 struct netdev_rx_queue
*queue
)
2455 struct net_device
*dev
= queue
->dev
;
2456 int index
= queue
- dev
->_rx
;
2458 BUG_ON(index
>= dev
->num_rx_queues
);
2463 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2464 int netif_get_num_default_rss_queues(void);
2466 enum skb_free_reason
{
2467 SKB_REASON_CONSUMED
,
2471 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
2472 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
2475 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2476 * interrupt context or with hardware interrupts being disabled.
2477 * (in_irq() || irqs_disabled())
2479 * We provide four helpers that can be used in following contexts :
2481 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2482 * replacing kfree_skb(skb)
2484 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2485 * Typically used in place of consume_skb(skb) in TX completion path
2487 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2488 * replacing kfree_skb(skb)
2490 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2491 * and consumed a packet. Used in place of consume_skb(skb)
2493 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
2495 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
2498 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
2500 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
2503 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
2505 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
2508 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
2510 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
2513 int netif_rx(struct sk_buff
*skb
);
2514 int netif_rx_ni(struct sk_buff
*skb
);
2515 int netif_receive_skb(struct sk_buff
*skb
);
2516 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
2517 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
2518 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
2519 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
2520 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
2521 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
2523 static inline void napi_free_frags(struct napi_struct
*napi
)
2525 kfree_skb(napi
->skb
);
2529 int netdev_rx_handler_register(struct net_device
*dev
,
2530 rx_handler_func_t
*rx_handler
,
2531 void *rx_handler_data
);
2532 void netdev_rx_handler_unregister(struct net_device
*dev
);
2534 bool dev_valid_name(const char *name
);
2535 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
2536 int dev_ethtool(struct net
*net
, struct ifreq
*);
2537 unsigned int dev_get_flags(const struct net_device
*);
2538 int __dev_change_flags(struct net_device
*, unsigned int flags
);
2539 int dev_change_flags(struct net_device
*, unsigned int);
2540 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
2541 unsigned int gchanges
);
2542 int dev_change_name(struct net_device
*, const char *);
2543 int dev_set_alias(struct net_device
*, const char *, size_t);
2544 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
2545 int dev_set_mtu(struct net_device
*, int);
2546 void dev_set_group(struct net_device
*, int);
2547 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
2548 int dev_change_carrier(struct net_device
*, bool new_carrier
);
2549 int dev_get_phys_port_id(struct net_device
*dev
,
2550 struct netdev_phys_port_id
*ppid
);
2551 int dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
2552 struct netdev_queue
*txq
);
2553 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
2555 extern int netdev_budget
;
2557 /* Called by rtnetlink.c:rtnl_unlock() */
2558 void netdev_run_todo(void);
2561 * dev_put - release reference to device
2562 * @dev: network device
2564 * Release reference to device to allow it to be freed.
2566 static inline void dev_put(struct net_device
*dev
)
2568 this_cpu_dec(*dev
->pcpu_refcnt
);
2572 * dev_hold - get reference to device
2573 * @dev: network device
2575 * Hold reference to device to keep it from being freed.
2577 static inline void dev_hold(struct net_device
*dev
)
2579 this_cpu_inc(*dev
->pcpu_refcnt
);
2582 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2583 * and _off may be called from IRQ context, but it is caller
2584 * who is responsible for serialization of these calls.
2586 * The name carrier is inappropriate, these functions should really be
2587 * called netif_lowerlayer_*() because they represent the state of any
2588 * kind of lower layer not just hardware media.
2591 void linkwatch_init_dev(struct net_device
*dev
);
2592 void linkwatch_fire_event(struct net_device
*dev
);
2593 void linkwatch_forget_dev(struct net_device
*dev
);
2596 * netif_carrier_ok - test if carrier present
2597 * @dev: network device
2599 * Check if carrier is present on device
2601 static inline bool netif_carrier_ok(const struct net_device
*dev
)
2603 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
2606 unsigned long dev_trans_start(struct net_device
*dev
);
2608 void __netdev_watchdog_up(struct net_device
*dev
);
2610 void netif_carrier_on(struct net_device
*dev
);
2612 void netif_carrier_off(struct net_device
*dev
);
2615 * netif_dormant_on - mark device as dormant.
2616 * @dev: network device
2618 * Mark device as dormant (as per RFC2863).
2620 * The dormant state indicates that the relevant interface is not
2621 * actually in a condition to pass packets (i.e., it is not 'up') but is
2622 * in a "pending" state, waiting for some external event. For "on-
2623 * demand" interfaces, this new state identifies the situation where the
2624 * interface is waiting for events to place it in the up state.
2627 static inline void netif_dormant_on(struct net_device
*dev
)
2629 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
2630 linkwatch_fire_event(dev
);
2634 * netif_dormant_off - set device as not dormant.
2635 * @dev: network device
2637 * Device is not in dormant state.
2639 static inline void netif_dormant_off(struct net_device
*dev
)
2641 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
2642 linkwatch_fire_event(dev
);
2646 * netif_dormant - test if carrier present
2647 * @dev: network device
2649 * Check if carrier is present on device
2651 static inline bool netif_dormant(const struct net_device
*dev
)
2653 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
2658 * netif_oper_up - test if device is operational
2659 * @dev: network device
2661 * Check if carrier is operational
2663 static inline bool netif_oper_up(const struct net_device
*dev
)
2665 return (dev
->operstate
== IF_OPER_UP
||
2666 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
2670 * netif_device_present - is device available or removed
2671 * @dev: network device
2673 * Check if device has not been removed from system.
2675 static inline bool netif_device_present(struct net_device
*dev
)
2677 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
2680 void netif_device_detach(struct net_device
*dev
);
2682 void netif_device_attach(struct net_device
*dev
);
2685 * Network interface message level settings
2689 NETIF_MSG_DRV
= 0x0001,
2690 NETIF_MSG_PROBE
= 0x0002,
2691 NETIF_MSG_LINK
= 0x0004,
2692 NETIF_MSG_TIMER
= 0x0008,
2693 NETIF_MSG_IFDOWN
= 0x0010,
2694 NETIF_MSG_IFUP
= 0x0020,
2695 NETIF_MSG_RX_ERR
= 0x0040,
2696 NETIF_MSG_TX_ERR
= 0x0080,
2697 NETIF_MSG_TX_QUEUED
= 0x0100,
2698 NETIF_MSG_INTR
= 0x0200,
2699 NETIF_MSG_TX_DONE
= 0x0400,
2700 NETIF_MSG_RX_STATUS
= 0x0800,
2701 NETIF_MSG_PKTDATA
= 0x1000,
2702 NETIF_MSG_HW
= 0x2000,
2703 NETIF_MSG_WOL
= 0x4000,
2706 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2707 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2708 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2709 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2710 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2711 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2712 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2713 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2714 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2715 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2716 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2717 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2718 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2719 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2720 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2722 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
2725 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
2726 return default_msg_enable_bits
;
2727 if (debug_value
== 0) /* no output */
2729 /* set low N bits */
2730 return (1 << debug_value
) - 1;
2733 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
2735 spin_lock(&txq
->_xmit_lock
);
2736 txq
->xmit_lock_owner
= cpu
;
2739 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
2741 spin_lock_bh(&txq
->_xmit_lock
);
2742 txq
->xmit_lock_owner
= smp_processor_id();
2745 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
2747 bool ok
= spin_trylock(&txq
->_xmit_lock
);
2749 txq
->xmit_lock_owner
= smp_processor_id();
2753 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
2755 txq
->xmit_lock_owner
= -1;
2756 spin_unlock(&txq
->_xmit_lock
);
2759 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
2761 txq
->xmit_lock_owner
= -1;
2762 spin_unlock_bh(&txq
->_xmit_lock
);
2765 static inline void txq_trans_update(struct netdev_queue
*txq
)
2767 if (txq
->xmit_lock_owner
!= -1)
2768 txq
->trans_start
= jiffies
;
2772 * netif_tx_lock - grab network device transmit lock
2773 * @dev: network device
2775 * Get network device transmit lock
2777 static inline void netif_tx_lock(struct net_device
*dev
)
2782 spin_lock(&dev
->tx_global_lock
);
2783 cpu
= smp_processor_id();
2784 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2785 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2787 /* We are the only thread of execution doing a
2788 * freeze, but we have to grab the _xmit_lock in
2789 * order to synchronize with threads which are in
2790 * the ->hard_start_xmit() handler and already
2791 * checked the frozen bit.
2793 __netif_tx_lock(txq
, cpu
);
2794 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
2795 __netif_tx_unlock(txq
);
2799 static inline void netif_tx_lock_bh(struct net_device
*dev
)
2805 static inline void netif_tx_unlock(struct net_device
*dev
)
2809 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2810 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2812 /* No need to grab the _xmit_lock here. If the
2813 * queue is not stopped for another reason, we
2816 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
2817 netif_schedule_queue(txq
);
2819 spin_unlock(&dev
->tx_global_lock
);
2822 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
2824 netif_tx_unlock(dev
);
2828 #define HARD_TX_LOCK(dev, txq, cpu) { \
2829 if ((dev->features & NETIF_F_LLTX) == 0) { \
2830 __netif_tx_lock(txq, cpu); \
2834 #define HARD_TX_UNLOCK(dev, txq) { \
2835 if ((dev->features & NETIF_F_LLTX) == 0) { \
2836 __netif_tx_unlock(txq); \
2840 static inline void netif_tx_disable(struct net_device
*dev
)
2846 cpu
= smp_processor_id();
2847 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2848 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2850 __netif_tx_lock(txq
, cpu
);
2851 netif_tx_stop_queue(txq
);
2852 __netif_tx_unlock(txq
);
2857 static inline void netif_addr_lock(struct net_device
*dev
)
2859 spin_lock(&dev
->addr_list_lock
);
2862 static inline void netif_addr_lock_nested(struct net_device
*dev
)
2864 spin_lock_nested(&dev
->addr_list_lock
, SINGLE_DEPTH_NESTING
);
2867 static inline void netif_addr_lock_bh(struct net_device
*dev
)
2869 spin_lock_bh(&dev
->addr_list_lock
);
2872 static inline void netif_addr_unlock(struct net_device
*dev
)
2874 spin_unlock(&dev
->addr_list_lock
);
2877 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
2879 spin_unlock_bh(&dev
->addr_list_lock
);
2883 * dev_addrs walker. Should be used only for read access. Call with
2884 * rcu_read_lock held.
2886 #define for_each_dev_addr(dev, ha) \
2887 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2889 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2891 void ether_setup(struct net_device
*dev
);
2893 /* Support for loadable net-drivers */
2894 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
2895 void (*setup
)(struct net_device
*),
2896 unsigned int txqs
, unsigned int rxqs
);
2897 #define alloc_netdev(sizeof_priv, name, setup) \
2898 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2900 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2901 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2903 int register_netdev(struct net_device
*dev
);
2904 void unregister_netdev(struct net_device
*dev
);
2906 /* General hardware address lists handling functions */
2907 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
2908 struct netdev_hw_addr_list
*from_list
, int addr_len
);
2909 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
2910 struct netdev_hw_addr_list
*from_list
, int addr_len
);
2911 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
2913 /* Functions used for device addresses handling */
2914 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
2915 unsigned char addr_type
);
2916 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
2917 unsigned char addr_type
);
2918 void dev_addr_flush(struct net_device
*dev
);
2919 int dev_addr_init(struct net_device
*dev
);
2921 /* Functions used for unicast addresses handling */
2922 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
2923 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
2924 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
2925 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
2926 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
2927 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
2928 void dev_uc_flush(struct net_device
*dev
);
2929 void dev_uc_init(struct net_device
*dev
);
2931 /* Functions used for multicast addresses handling */
2932 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
2933 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
2934 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
2935 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
2936 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
2937 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
2938 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
2939 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
2940 void dev_mc_flush(struct net_device
*dev
);
2941 void dev_mc_init(struct net_device
*dev
);
2943 /* Functions used for secondary unicast and multicast support */
2944 void dev_set_rx_mode(struct net_device
*dev
);
2945 void __dev_set_rx_mode(struct net_device
*dev
);
2946 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
2947 int dev_set_allmulti(struct net_device
*dev
, int inc
);
2948 void netdev_state_change(struct net_device
*dev
);
2949 void netdev_notify_peers(struct net_device
*dev
);
2950 void netdev_features_change(struct net_device
*dev
);
2951 /* Load a device via the kmod */
2952 void dev_load(struct net
*net
, const char *name
);
2953 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
2954 struct rtnl_link_stats64
*storage
);
2955 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
2956 const struct net_device_stats
*netdev_stats
);
2958 extern int netdev_max_backlog
;
2959 extern int netdev_tstamp_prequeue
;
2960 extern int weight_p
;
2961 extern int bpf_jit_enable
;
2963 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
2964 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
2965 struct list_head
**iter
);
2967 /* iterate through upper list, must be called under RCU read lock */
2968 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
2969 for (iter = &(dev)->all_adj_list.upper, \
2970 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
2972 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
2974 void *netdev_lower_get_next_private(struct net_device
*dev
,
2975 struct list_head
**iter
);
2976 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
2977 struct list_head
**iter
);
2979 #define netdev_for_each_lower_private(dev, priv, iter) \
2980 for (iter = (dev)->adj_list.lower.next, \
2981 priv = netdev_lower_get_next_private(dev, &(iter)); \
2983 priv = netdev_lower_get_next_private(dev, &(iter)))
2985 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
2986 for (iter = &(dev)->adj_list.lower, \
2987 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
2989 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
2991 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
2992 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
2993 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
2994 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
2995 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
2996 int netdev_master_upper_dev_link(struct net_device
*dev
,
2997 struct net_device
*upper_dev
);
2998 int netdev_master_upper_dev_link_private(struct net_device
*dev
,
2999 struct net_device
*upper_dev
,
3001 void netdev_upper_dev_unlink(struct net_device
*dev
,
3002 struct net_device
*upper_dev
);
3003 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3004 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3005 struct net_device
*lower_dev
);
3006 int skb_checksum_help(struct sk_buff
*skb
);
3007 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3008 netdev_features_t features
, bool tx_path
);
3009 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3010 netdev_features_t features
);
3013 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3015 return __skb_gso_segment(skb
, features
, true);
3017 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3019 static inline bool can_checksum_protocol(netdev_features_t features
,
3022 return ((features
& NETIF_F_GEN_CSUM
) ||
3023 ((features
& NETIF_F_V4_CSUM
) &&
3024 protocol
== htons(ETH_P_IP
)) ||
3025 ((features
& NETIF_F_V6_CSUM
) &&
3026 protocol
== htons(ETH_P_IPV6
)) ||
3027 ((features
& NETIF_F_FCOE_CRC
) &&
3028 protocol
== htons(ETH_P_FCOE
)));
3032 void netdev_rx_csum_fault(struct net_device
*dev
);
3034 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3038 /* rx skb timestamps */
3039 void net_enable_timestamp(void);
3040 void net_disable_timestamp(void);
3042 #ifdef CONFIG_PROC_FS
3043 int __init
dev_proc_init(void);
3045 #define dev_proc_init() 0
3048 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3050 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3053 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3055 return netdev_class_create_file_ns(class_attr
, NULL
);
3058 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3060 netdev_class_remove_file_ns(class_attr
, NULL
);
3063 extern struct kobj_ns_type_operations net_ns_type_operations
;
3065 const char *netdev_drivername(const struct net_device
*dev
);
3067 void linkwatch_run_queue(void);
3069 static inline netdev_features_t
netdev_get_wanted_features(
3070 struct net_device
*dev
)
3072 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
3074 netdev_features_t
netdev_increment_features(netdev_features_t all
,
3075 netdev_features_t one
, netdev_features_t mask
);
3077 /* Allow TSO being used on stacked device :
3078 * Performing the GSO segmentation before last device
3079 * is a performance improvement.
3081 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
3082 netdev_features_t mask
)
3084 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
3087 int __netdev_update_features(struct net_device
*dev
);
3088 void netdev_update_features(struct net_device
*dev
);
3089 void netdev_change_features(struct net_device
*dev
);
3091 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
3092 struct net_device
*dev
);
3094 netdev_features_t
netif_skb_dev_features(struct sk_buff
*skb
,
3095 const struct net_device
*dev
);
3096 static inline netdev_features_t
netif_skb_features(struct sk_buff
*skb
)
3098 return netif_skb_dev_features(skb
, skb
->dev
);
3101 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
3103 netdev_features_t feature
= gso_type
<< NETIF_F_GSO_SHIFT
;
3105 /* check flags correspondence */
3106 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
3107 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
3108 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
3109 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
3110 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
3111 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
3113 return (features
& feature
) == feature
;
3116 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
3118 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
3119 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
3122 static inline bool netif_needs_gso(struct sk_buff
*skb
,
3123 netdev_features_t features
)
3125 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
3126 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
3127 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
3130 static inline void netif_set_gso_max_size(struct net_device
*dev
,
3133 dev
->gso_max_size
= size
;
3136 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
3137 int pulled_hlen
, u16 mac_offset
,
3140 skb
->protocol
= protocol
;
3141 skb
->encapsulation
= 1;
3142 skb_push(skb
, pulled_hlen
);
3143 skb_reset_transport_header(skb
);
3144 skb
->mac_header
= mac_offset
;
3145 skb
->network_header
= skb
->mac_header
+ mac_len
;
3146 skb
->mac_len
= mac_len
;
3149 static inline bool netif_is_macvlan(struct net_device
*dev
)
3151 return dev
->priv_flags
& IFF_MACVLAN
;
3154 static inline bool netif_is_bond_master(struct net_device
*dev
)
3156 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
3159 static inline bool netif_is_bond_slave(struct net_device
*dev
)
3161 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
3164 static inline bool netif_supports_nofcs(struct net_device
*dev
)
3166 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
3169 extern struct pernet_operations __net_initdata loopback_net_ops
;
3171 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3173 /* netdev_printk helpers, similar to dev_printk */
3175 static inline const char *netdev_name(const struct net_device
*dev
)
3177 if (dev
->reg_state
!= NETREG_REGISTERED
)
3178 return "(unregistered net_device)";
3183 int netdev_printk(const char *level
, const struct net_device
*dev
,
3184 const char *format
, ...);
3186 int netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
3188 int netdev_alert(const struct net_device
*dev
, const char *format
, ...);
3190 int netdev_crit(const struct net_device
*dev
, const char *format
, ...);
3192 int netdev_err(const struct net_device
*dev
, const char *format
, ...);
3194 int netdev_warn(const struct net_device
*dev
, const char *format
, ...);
3196 int netdev_notice(const struct net_device
*dev
, const char *format
, ...);
3198 int netdev_info(const struct net_device
*dev
, const char *format
, ...);
3200 #define MODULE_ALIAS_NETDEV(device) \
3201 MODULE_ALIAS("netdev-" device)
3203 #if defined(CONFIG_DYNAMIC_DEBUG)
3204 #define netdev_dbg(__dev, format, args...) \
3206 dynamic_netdev_dbg(__dev, format, ##args); \
3208 #elif defined(DEBUG)
3209 #define netdev_dbg(__dev, format, args...) \
3210 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3212 #define netdev_dbg(__dev, format, args...) \
3215 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3220 #if defined(VERBOSE_DEBUG)
3221 #define netdev_vdbg netdev_dbg
3224 #define netdev_vdbg(dev, format, args...) \
3227 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3233 * netdev_WARN() acts like dev_printk(), but with the key difference
3234 * of using a WARN/WARN_ON to get the message out, including the
3235 * file/line information and a backtrace.
3237 #define netdev_WARN(dev, format, args...) \
3238 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3240 /* netif printk helpers, similar to netdev_printk */
3242 #define netif_printk(priv, type, level, dev, fmt, args...) \
3244 if (netif_msg_##type(priv)) \
3245 netdev_printk(level, (dev), fmt, ##args); \
3248 #define netif_level(level, priv, type, dev, fmt, args...) \
3250 if (netif_msg_##type(priv)) \
3251 netdev_##level(dev, fmt, ##args); \
3254 #define netif_emerg(priv, type, dev, fmt, args...) \
3255 netif_level(emerg, priv, type, dev, fmt, ##args)
3256 #define netif_alert(priv, type, dev, fmt, args...) \
3257 netif_level(alert, priv, type, dev, fmt, ##args)
3258 #define netif_crit(priv, type, dev, fmt, args...) \
3259 netif_level(crit, priv, type, dev, fmt, ##args)
3260 #define netif_err(priv, type, dev, fmt, args...) \
3261 netif_level(err, priv, type, dev, fmt, ##args)
3262 #define netif_warn(priv, type, dev, fmt, args...) \
3263 netif_level(warn, priv, type, dev, fmt, ##args)
3264 #define netif_notice(priv, type, dev, fmt, args...) \
3265 netif_level(notice, priv, type, dev, fmt, ##args)
3266 #define netif_info(priv, type, dev, fmt, args...) \
3267 netif_level(info, priv, type, dev, fmt, ##args)
3269 #if defined(CONFIG_DYNAMIC_DEBUG)
3270 #define netif_dbg(priv, type, netdev, format, args...) \
3272 if (netif_msg_##type(priv)) \
3273 dynamic_netdev_dbg(netdev, format, ##args); \
3275 #elif defined(DEBUG)
3276 #define netif_dbg(priv, type, dev, format, args...) \
3277 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3279 #define netif_dbg(priv, type, dev, format, args...) \
3282 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3287 #if defined(VERBOSE_DEBUG)
3288 #define netif_vdbg netif_dbg
3290 #define netif_vdbg(priv, type, dev, format, args...) \
3293 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3299 * The list of packet types we will receive (as opposed to discard)
3300 * and the routines to invoke.
3302 * Why 16. Because with 16 the only overlap we get on a hash of the
3303 * low nibble of the protocol value is RARP/SNAP/X.25.
3305 * NOTE: That is no longer true with the addition of VLAN tags. Not
3306 * sure which should go first, but I bet it won't make much
3307 * difference if we are running VLANs. The good news is that
3308 * this protocol won't be in the list unless compiled in, so
3309 * the average user (w/out VLANs) will not be adversely affected.
3325 #define PTYPE_HASH_SIZE (16)
3326 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3328 #endif /* _LINUX_NETDEVICE_H */