2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/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>
45 #include <net/dcbnl.h>
47 #include <net/netprio_cgroup.h>
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
59 struct dsa_switch_tree
;
63 /* 802.15.4 specific */
66 /* UDP Tunnel offloads */
67 struct udp_tunnel_info
;
70 void netdev_set_default_ethtool_ops(struct net_device
*dev
,
71 const struct ethtool_ops
*ops
);
73 /* Backlog congestion levels */
74 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
75 #define NET_RX_DROP 1 /* packet dropped */
78 * Transmit return codes: transmit return codes originate from three different
81 * - qdisc return codes
82 * - driver transmit return codes
85 * Drivers are allowed to return any one of those in their hard_start_xmit()
86 * function. Real network devices commonly used with qdiscs should only return
87 * the driver transmit return codes though - when qdiscs are used, the actual
88 * transmission happens asynchronously, so the value is not propagated to
89 * higher layers. Virtual network devices transmit synchronously; in this case
90 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
91 * others are propagated to higher layers.
94 /* qdisc ->enqueue() return codes. */
95 #define NET_XMIT_SUCCESS 0x00
96 #define NET_XMIT_DROP 0x01 /* skb dropped */
97 #define NET_XMIT_CN 0x02 /* congestion notification */
98 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
100 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
101 * indicates that the device will soon be dropping packets, or already drops
102 * some packets of the same priority; prompting us to send less aggressively. */
103 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
104 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106 /* Driver transmit return codes */
107 #define NETDEV_TX_MASK 0xf0
110 __NETDEV_TX_MIN
= INT_MIN
, /* make sure enum is signed */
111 NETDEV_TX_OK
= 0x00, /* driver took care of packet */
112 NETDEV_TX_BUSY
= 0x10, /* driver tx path was busy*/
114 typedef enum netdev_tx netdev_tx_t
;
117 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
118 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 static inline bool dev_xmit_complete(int rc
)
123 * Positive cases with an skb consumed by a driver:
124 * - successful transmission (rc == NETDEV_TX_OK)
125 * - error while transmitting (rc < 0)
126 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 if (likely(rc
< NET_XMIT_MASK
))
135 * Compute the worst-case header length according to the protocols
139 #if defined(CONFIG_HYPERV_NET)
140 # define LL_MAX_HEADER 128
141 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
142 # if defined(CONFIG_MAC80211_MESH)
143 # define LL_MAX_HEADER 128
145 # define LL_MAX_HEADER 96
148 # define LL_MAX_HEADER 32
151 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
152 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
153 #define MAX_HEADER LL_MAX_HEADER
155 #define MAX_HEADER (LL_MAX_HEADER + 48)
159 * Old network device statistics. Fields are native words
160 * (unsigned long) so they can be read and written atomically.
163 struct net_device_stats
{
164 unsigned long rx_packets
;
165 unsigned long tx_packets
;
166 unsigned long rx_bytes
;
167 unsigned long tx_bytes
;
168 unsigned long rx_errors
;
169 unsigned long tx_errors
;
170 unsigned long rx_dropped
;
171 unsigned long tx_dropped
;
172 unsigned long multicast
;
173 unsigned long collisions
;
174 unsigned long rx_length_errors
;
175 unsigned long rx_over_errors
;
176 unsigned long rx_crc_errors
;
177 unsigned long rx_frame_errors
;
178 unsigned long rx_fifo_errors
;
179 unsigned long rx_missed_errors
;
180 unsigned long tx_aborted_errors
;
181 unsigned long tx_carrier_errors
;
182 unsigned long tx_fifo_errors
;
183 unsigned long tx_heartbeat_errors
;
184 unsigned long tx_window_errors
;
185 unsigned long rx_compressed
;
186 unsigned long tx_compressed
;
190 #include <linux/cache.h>
191 #include <linux/skbuff.h>
194 #include <linux/static_key.h>
195 extern struct static_key rps_needed
;
196 extern struct static_key rfs_needed
;
203 struct netdev_hw_addr
{
204 struct list_head list
;
205 unsigned char addr
[MAX_ADDR_LEN
];
207 #define NETDEV_HW_ADDR_T_LAN 1
208 #define NETDEV_HW_ADDR_T_SAN 2
209 #define NETDEV_HW_ADDR_T_SLAVE 3
210 #define NETDEV_HW_ADDR_T_UNICAST 4
211 #define NETDEV_HW_ADDR_T_MULTICAST 5
216 struct rcu_head rcu_head
;
219 struct netdev_hw_addr_list
{
220 struct list_head list
;
224 #define netdev_hw_addr_list_count(l) ((l)->count)
225 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
226 #define netdev_hw_addr_list_for_each(ha, l) \
227 list_for_each_entry(ha, &(l)->list, list)
229 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
230 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
231 #define netdev_for_each_uc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
234 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
235 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
236 #define netdev_for_each_mc_addr(ha, dev) \
237 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
243 /* cached hardware header; allow for machine alignment needs. */
244 #define HH_DATA_MOD 16
245 #define HH_DATA_OFF(__len) \
246 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 unsigned long hh_data
[HH_DATA_ALIGN(LL_MAX_HEADER
) / sizeof(long)];
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
254 * dev->hard_header_len ? (dev->hard_header_len +
255 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
257 * We could use other alignment values, but we must maintain the
258 * relationship HH alignment <= LL alignment.
260 #define LL_RESERVED_SPACE(dev) \
261 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 int (*create
) (struct sk_buff
*skb
, struct net_device
*dev
,
267 unsigned short type
, const void *daddr
,
268 const void *saddr
, unsigned int len
);
269 int (*parse
)(const struct sk_buff
*skb
, unsigned char *haddr
);
270 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
271 void (*cache_update
)(struct hh_cache
*hh
,
272 const struct net_device
*dev
,
273 const unsigned char *haddr
);
274 bool (*validate
)(const char *ll_header
, unsigned int len
);
277 /* These flag bits are private to the generic network queueing
278 * layer; they may not be explicitly referenced by any other
282 enum netdev_state_t
{
284 __LINK_STATE_PRESENT
,
285 __LINK_STATE_NOCARRIER
,
286 __LINK_STATE_LINKWATCH_PENDING
,
287 __LINK_STATE_DORMANT
,
292 * This structure holds boot-time configured netdevice settings. They
293 * are then used in the device probing.
295 struct netdev_boot_setup
{
299 #define NETDEV_BOOT_SETUP_MAX 8
301 int __init
netdev_boot_setup(char *str
);
304 * Structure for NAPI scheduling similar to tasklet but with weighting
307 /* The poll_list must only be managed by the entity which
308 * changes the state of the NAPI_STATE_SCHED bit. This means
309 * whoever atomically sets that bit can add this napi_struct
310 * to the per-CPU poll_list, and whoever clears that bit
311 * can remove from the list right before clearing the bit.
313 struct list_head poll_list
;
317 unsigned int gro_count
;
318 int (*poll
)(struct napi_struct
*, int);
319 #ifdef CONFIG_NETPOLL
322 struct net_device
*dev
;
323 struct sk_buff
*gro_list
;
325 struct hrtimer timer
;
326 struct list_head dev_list
;
327 struct hlist_node napi_hash_node
;
328 unsigned int napi_id
;
332 NAPI_STATE_SCHED
, /* Poll is scheduled */
333 NAPI_STATE_MISSED
, /* reschedule a napi */
334 NAPI_STATE_DISABLE
, /* Disable pending */
335 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
336 NAPI_STATE_HASHED
, /* In NAPI hash (busy polling possible) */
337 NAPI_STATE_NO_BUSY_POLL
,/* Do not add in napi_hash, no busy polling */
338 NAPI_STATE_IN_BUSY_POLL
,/* sk_busy_loop() owns this NAPI */
342 NAPIF_STATE_SCHED
= BIT(NAPI_STATE_SCHED
),
343 NAPIF_STATE_MISSED
= BIT(NAPI_STATE_MISSED
),
344 NAPIF_STATE_DISABLE
= BIT(NAPI_STATE_DISABLE
),
345 NAPIF_STATE_NPSVC
= BIT(NAPI_STATE_NPSVC
),
346 NAPIF_STATE_HASHED
= BIT(NAPI_STATE_HASHED
),
347 NAPIF_STATE_NO_BUSY_POLL
= BIT(NAPI_STATE_NO_BUSY_POLL
),
348 NAPIF_STATE_IN_BUSY_POLL
= BIT(NAPI_STATE_IN_BUSY_POLL
),
359 typedef enum gro_result gro_result_t
;
362 * enum rx_handler_result - Possible return values for rx_handlers.
363 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
365 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
366 * case skb->dev was changed by rx_handler.
367 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
368 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
370 * rx_handlers are functions called from inside __netif_receive_skb(), to do
371 * special processing of the skb, prior to delivery to protocol handlers.
373 * Currently, a net_device can only have a single rx_handler registered. Trying
374 * to register a second rx_handler will return -EBUSY.
376 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
377 * To unregister a rx_handler on a net_device, use
378 * netdev_rx_handler_unregister().
380 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
383 * If the rx_handler consumed the skb in some way, it should return
384 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
385 * the skb to be delivered in some other way.
387 * If the rx_handler changed skb->dev, to divert the skb to another
388 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
389 * new device will be called if it exists.
391 * If the rx_handler decides the skb should be ignored, it should return
392 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
393 * are registered on exact device (ptype->dev == skb->dev).
395 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
396 * delivered, it should return RX_HANDLER_PASS.
398 * A device without a registered rx_handler will behave as if rx_handler
399 * returned RX_HANDLER_PASS.
402 enum rx_handler_result
{
408 typedef enum rx_handler_result rx_handler_result_t
;
409 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
411 void __napi_schedule(struct napi_struct
*n
);
412 void __napi_schedule_irqoff(struct napi_struct
*n
);
414 static inline bool napi_disable_pending(struct napi_struct
*n
)
416 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
419 bool napi_schedule_prep(struct napi_struct
*n
);
422 * napi_schedule - schedule NAPI poll
425 * Schedule NAPI poll routine to be called if it is not already
428 static inline void napi_schedule(struct napi_struct
*n
)
430 if (napi_schedule_prep(n
))
435 * napi_schedule_irqoff - schedule NAPI poll
438 * Variant of napi_schedule(), assuming hard irqs are masked.
440 static inline void napi_schedule_irqoff(struct napi_struct
*n
)
442 if (napi_schedule_prep(n
))
443 __napi_schedule_irqoff(n
);
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct
*napi
)
449 if (napi_schedule_prep(napi
)) {
450 __napi_schedule(napi
);
456 bool napi_complete_done(struct napi_struct
*n
, int work_done
);
458 * napi_complete - NAPI processing complete
461 * Mark NAPI processing as complete.
462 * Consider using napi_complete_done() instead.
463 * Return false if device should avoid rearming interrupts.
465 static inline bool napi_complete(struct napi_struct
*n
)
467 return napi_complete_done(n
, 0);
471 * napi_hash_del - remove a NAPI from global table
472 * @napi: NAPI context
474 * Warning: caller must observe RCU grace period
475 * before freeing memory containing @napi, if
476 * this function returns true.
477 * Note: core networking stack automatically calls it
478 * from netif_napi_del().
479 * Drivers might want to call this helper to combine all
480 * the needed RCU grace periods into a single one.
482 bool napi_hash_del(struct napi_struct
*napi
);
485 * napi_disable - prevent NAPI from scheduling
488 * Stop NAPI from being scheduled on this context.
489 * Waits till any outstanding processing completes.
491 void napi_disable(struct napi_struct
*n
);
494 * napi_enable - enable NAPI scheduling
497 * Resume NAPI from being scheduled on this context.
498 * Must be paired with napi_disable.
500 static inline void napi_enable(struct napi_struct
*n
)
502 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
503 smp_mb__before_atomic();
504 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
505 clear_bit(NAPI_STATE_NPSVC
, &n
->state
);
509 * napi_synchronize - wait until NAPI is not running
512 * Wait until NAPI is done being scheduled on this context.
513 * Waits till any outstanding processing completes but
514 * does not disable future activations.
516 static inline void napi_synchronize(const struct napi_struct
*n
)
518 if (IS_ENABLED(CONFIG_SMP
))
519 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
525 enum netdev_queue_state_t
{
526 __QUEUE_STATE_DRV_XOFF
,
527 __QUEUE_STATE_STACK_XOFF
,
528 __QUEUE_STATE_FROZEN
,
531 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
532 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
533 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
535 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
536 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
538 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
542 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
543 * netif_tx_* functions below are used to manipulate this flag. The
544 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
545 * queue independently. The netif_xmit_*stopped functions below are called
546 * to check if the queue has been stopped by the driver or stack (either
547 * of the XOFF bits are set in the state). Drivers should not need to call
548 * netif_xmit*stopped functions, they should only be using netif_tx_*.
551 struct netdev_queue
{
555 struct net_device
*dev
;
556 struct Qdisc __rcu
*qdisc
;
557 struct Qdisc
*qdisc_sleeping
;
561 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
564 unsigned long tx_maxrate
;
566 * Number of TX timeouts for this queue
567 * (/sys/class/net/DEV/Q/trans_timeout)
569 unsigned long trans_timeout
;
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
576 * Time (in jiffies) of last Tx
578 unsigned long trans_start
;
585 } ____cacheline_aligned_in_smp
;
587 static inline int netdev_queue_numa_node_read(const struct netdev_queue
*q
)
589 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 static inline void netdev_queue_numa_node_write(struct netdev_queue
*q
, int node
)
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 * This structure holds an RPS map which can be of variable length. The
606 * map is an array of CPUs.
613 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
616 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
617 * tail pointer for that CPU's input queue at the time of last enqueue, and
618 * a hardware filter index.
620 struct rps_dev_flow
{
623 unsigned int last_qtail
;
625 #define RPS_NO_FILTER 0xffff
628 * The rps_dev_flow_table structure contains a table of flow mappings.
630 struct rps_dev_flow_table
{
633 struct rps_dev_flow flows
[0];
635 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
636 ((_num) * sizeof(struct rps_dev_flow)))
639 * The rps_sock_flow_table contains mappings of flows to the last CPU
640 * on which they were processed by the application (set in recvmsg).
641 * Each entry is a 32bit value. Upper part is the high-order bits
642 * of flow hash, lower part is CPU number.
643 * rps_cpu_mask is used to partition the space, depending on number of
644 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
645 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
646 * meaning we use 32-6=26 bits for the hash.
648 struct rps_sock_flow_table
{
651 u32 ents
[0] ____cacheline_aligned_in_smp
;
653 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
655 #define RPS_NO_CPU 0xffff
657 extern u32 rps_cpu_mask
;
658 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
660 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
664 unsigned int index
= hash
& table
->mask
;
665 u32 val
= hash
& ~rps_cpu_mask
;
667 /* We only give a hint, preemption can change CPU under us */
668 val
|= raw_smp_processor_id();
670 if (table
->ents
[index
] != val
)
671 table
->ents
[index
] = val
;
675 #ifdef CONFIG_RFS_ACCEL
676 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
679 #endif /* CONFIG_RPS */
681 /* This structure contains an instance of an RX queue. */
682 struct netdev_rx_queue
{
684 struct rps_map __rcu
*rps_map
;
685 struct rps_dev_flow_table __rcu
*rps_flow_table
;
688 struct net_device
*dev
;
689 } ____cacheline_aligned_in_smp
;
692 * RX queue sysfs structures and functions.
694 struct rx_queue_attribute
{
695 struct attribute attr
;
696 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
697 struct rx_queue_attribute
*attr
, char *buf
);
698 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
699 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
704 * This structure holds an XPS map which can be of variable length. The
705 * map is an array of queues.
709 unsigned int alloc_len
;
713 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
714 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
715 - sizeof(struct xps_map)) / sizeof(u16))
718 * This structure holds all XPS maps for device. Maps are indexed by CPU.
720 struct xps_dev_maps
{
722 struct xps_map __rcu
*cpu_map
[0];
724 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) + \
725 (nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
726 #endif /* CONFIG_XPS */
728 #define TC_MAX_QUEUE 16
729 #define TC_BITMASK 15
730 /* HW offloaded queuing disciplines txq count and offset maps */
731 struct netdev_tc_txq
{
736 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
738 * This structure is to hold information about the device
739 * configured to run FCoE protocol stack.
741 struct netdev_fcoe_hbainfo
{
742 char manufacturer
[64];
743 char serial_number
[64];
744 char hardware_version
[64];
745 char driver_version
[64];
746 char optionrom_version
[64];
747 char firmware_version
[64];
749 char model_description
[256];
753 #define MAX_PHYS_ITEM_ID_LEN 32
755 /* This structure holds a unique identifier to identify some
756 * physical item (port for example) used by a netdevice.
758 struct netdev_phys_item_id
{
759 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
760 unsigned char id_len
;
763 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id
*a
,
764 struct netdev_phys_item_id
*b
)
766 return a
->id_len
== b
->id_len
&&
767 memcmp(a
->id
, b
->id
, a
->id_len
) == 0;
770 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
771 struct sk_buff
*skb
);
773 /* These structures hold the attributes of qdisc and classifiers
774 * that are being passed to the netdevice through the setup_tc op.
784 struct tc_cls_u32_offload
;
786 struct tc_to_netdev
{
789 struct tc_cls_u32_offload
*cls_u32
;
790 struct tc_cls_flower_offload
*cls_flower
;
791 struct tc_cls_matchall_offload
*cls_mall
;
792 struct tc_cls_bpf_offload
*cls_bpf
;
793 struct tc_mqprio_qopt
*mqprio
;
798 /* These structures hold the attributes of xdp state that are being passed
799 * to the netdevice through the xdp op.
801 enum xdp_netdev_command
{
802 /* Set or clear a bpf program used in the earliest stages of packet
803 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
804 * is responsible for calling bpf_prog_put on any old progs that are
805 * stored. In case of error, the callee need not release the new prog
806 * reference, but on success it takes ownership and must bpf_prog_put
807 * when it is no longer used.
810 /* Check if a bpf program is set on the device. The callee should
811 * return true if a program is currently attached and running.
817 enum xdp_netdev_command command
;
820 struct bpf_prog
*prog
;
827 * This structure defines the management hooks for network devices.
828 * The following hooks can be defined; unless noted otherwise, they are
829 * optional and can be filled with a null pointer.
831 * int (*ndo_init)(struct net_device *dev);
832 * This function is called once when a network device is registered.
833 * The network device can use this for any late stage initialization
834 * or semantic validation. It can fail with an error code which will
835 * be propagated back to register_netdev.
837 * void (*ndo_uninit)(struct net_device *dev);
838 * This function is called when device is unregistered or when registration
839 * fails. It is not called if init fails.
841 * int (*ndo_open)(struct net_device *dev);
842 * This function is called when a network device transitions to the up
845 * int (*ndo_stop)(struct net_device *dev);
846 * This function is called when a network device transitions to the down
849 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
850 * struct net_device *dev);
851 * Called when a packet needs to be transmitted.
852 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
853 * the queue before that can happen; it's for obsolete devices and weird
854 * corner cases, but the stack really does a non-trivial amount
855 * of useless work if you return NETDEV_TX_BUSY.
856 * Required; cannot be NULL.
858 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
859 * struct net_device *dev
860 * netdev_features_t features);
861 * Called by core transmit path to determine if device is capable of
862 * performing offload operations on a given packet. This is to give
863 * the device an opportunity to implement any restrictions that cannot
864 * be otherwise expressed by feature flags. The check is called with
865 * the set of features that the stack has calculated and it returns
866 * those the driver believes to be appropriate.
868 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
869 * void *accel_priv, select_queue_fallback_t fallback);
870 * Called to decide which queue to use when device supports multiple
873 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
874 * This function is called to allow device receiver to make
875 * changes to configuration when multicast or promiscuous is enabled.
877 * void (*ndo_set_rx_mode)(struct net_device *dev);
878 * This function is called device changes address list filtering.
879 * If driver handles unicast address filtering, it should set
880 * IFF_UNICAST_FLT in its priv_flags.
882 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
883 * This function is called when the Media Access Control address
884 * needs to be changed. If this interface is not defined, the
885 * MAC address can not be changed.
887 * int (*ndo_validate_addr)(struct net_device *dev);
888 * Test if Media Access Control address is valid for the device.
890 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
891 * Called when a user requests an ioctl which can't be handled by
892 * the generic interface code. If not defined ioctls return
893 * not supported error code.
895 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
896 * Used to set network devices bus interface parameters. This interface
897 * is retained for legacy reasons; new devices should use the bus
898 * interface (PCI) for low level management.
900 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
901 * Called when a user wants to change the Maximum Transfer Unit
902 * of a device. If not defined, any request to change MTU will
903 * will return an error.
905 * void (*ndo_tx_timeout)(struct net_device *dev);
906 * Callback used when the transmitter has not made any progress
907 * for dev->watchdog ticks.
909 * void (*ndo_get_stats64)(struct net_device *dev,
910 * struct rtnl_link_stats64 *storage);
911 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
912 * Called when a user wants to get the network device usage
913 * statistics. Drivers must do one of the following:
914 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
915 * rtnl_link_stats64 structure passed by the caller.
916 * 2. Define @ndo_get_stats to update a net_device_stats structure
917 * (which should normally be dev->stats) and return a pointer to
918 * it. The structure may be changed asynchronously only if each
919 * field is written atomically.
920 * 3. Update dev->stats asynchronously and atomically, and define
923 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
924 * Return true if this device supports offload stats of this attr_id.
926 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
928 * Get statistics for offload operations by attr_id. Write it into the
931 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
932 * If device supports VLAN filtering this function is called when a
933 * VLAN id is registered.
935 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
936 * If device supports VLAN filtering this function is called when a
937 * VLAN id is unregistered.
939 * void (*ndo_poll_controller)(struct net_device *dev);
941 * SR-IOV management functions.
942 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
943 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
944 * u8 qos, __be16 proto);
945 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
947 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
948 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
949 * int (*ndo_get_vf_config)(struct net_device *dev,
950 * int vf, struct ifla_vf_info *ivf);
951 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
952 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
953 * struct nlattr *port[]);
955 * Enable or disable the VF ability to query its RSS Redirection Table and
956 * Hash Key. This is needed since on some devices VF share this information
957 * with PF and querying it may introduce a theoretical security risk.
958 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
959 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
960 * int (*ndo_setup_tc)(struct net_device *dev, u32 handle,
961 * __be16 protocol, struct tc_to_netdev *tc);
962 * Called to setup any 'tc' scheduler, classifier or action on @dev.
963 * This is always called from the stack with the rtnl lock held and netif
964 * tx queues stopped. This allows the netdevice to perform queue
967 * Fiber Channel over Ethernet (FCoE) offload functions.
968 * int (*ndo_fcoe_enable)(struct net_device *dev);
969 * Called when the FCoE protocol stack wants to start using LLD for FCoE
970 * so the underlying device can perform whatever needed configuration or
971 * initialization to support acceleration of FCoE traffic.
973 * int (*ndo_fcoe_disable)(struct net_device *dev);
974 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
975 * so the underlying device can perform whatever needed clean-ups to
976 * stop supporting acceleration of FCoE traffic.
978 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
979 * struct scatterlist *sgl, unsigned int sgc);
980 * Called when the FCoE Initiator wants to initialize an I/O that
981 * is a possible candidate for Direct Data Placement (DDP). The LLD can
982 * perform necessary setup and returns 1 to indicate the device is set up
983 * successfully to perform DDP on this I/O, otherwise this returns 0.
985 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
986 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
987 * indicated by the FC exchange id 'xid', so the underlying device can
988 * clean up and reuse resources for later DDP requests.
990 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
991 * struct scatterlist *sgl, unsigned int sgc);
992 * Called when the FCoE Target wants to initialize an I/O that
993 * is a possible candidate for Direct Data Placement (DDP). The LLD can
994 * perform necessary setup and returns 1 to indicate the device is set up
995 * successfully to perform DDP on this I/O, otherwise this returns 0.
997 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
998 * struct netdev_fcoe_hbainfo *hbainfo);
999 * Called when the FCoE Protocol stack wants information on the underlying
1000 * device. This information is utilized by the FCoE protocol stack to
1001 * register attributes with Fiber Channel management service as per the
1002 * FC-GS Fabric Device Management Information(FDMI) specification.
1004 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1005 * Called when the underlying device wants to override default World Wide
1006 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1007 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1008 * protocol stack to use.
1011 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1012 * u16 rxq_index, u32 flow_id);
1013 * Set hardware filter for RFS. rxq_index is the target queue index;
1014 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1015 * Return the filter ID on success, or a negative error code.
1017 * Slave management functions (for bridge, bonding, etc).
1018 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1019 * Called to make another netdev an underling.
1021 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1022 * Called to release previously enslaved netdev.
1024 * Feature/offload setting functions.
1025 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1026 * netdev_features_t features);
1027 * Adjusts the requested feature flags according to device-specific
1028 * constraints, and returns the resulting flags. Must not modify
1031 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1032 * Called to update device configuration to new features. Passed
1033 * feature set might be less than what was returned by ndo_fix_features()).
1034 * Must return >0 or -errno if it changed dev->features itself.
1036 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1037 * struct net_device *dev,
1038 * const unsigned char *addr, u16 vid, u16 flags)
1039 * Adds an FDB entry to dev for addr.
1040 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1041 * struct net_device *dev,
1042 * const unsigned char *addr, u16 vid)
1043 * Deletes the FDB entry from dev coresponding to addr.
1044 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1045 * struct net_device *dev, struct net_device *filter_dev,
1047 * Used to add FDB entries to dump requests. Implementers should add
1048 * entries to skb and update idx with the number of entries.
1050 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1052 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1053 * struct net_device *dev, u32 filter_mask,
1055 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1058 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1059 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1060 * which do not represent real hardware may define this to allow their
1061 * userspace components to manage their virtual carrier state. Devices
1062 * that determine carrier state from physical hardware properties (eg
1063 * network cables) or protocol-dependent mechanisms (eg
1064 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1066 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1067 * struct netdev_phys_item_id *ppid);
1068 * Called to get ID of physical port of this device. If driver does
1069 * not implement this, it is assumed that the hw is not able to have
1070 * multiple net devices on single physical port.
1072 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1073 * struct udp_tunnel_info *ti);
1074 * Called by UDP tunnel to notify a driver about the UDP port and socket
1075 * address family that a UDP tunnel is listnening to. It is called only
1076 * when a new port starts listening. The operation is protected by the
1079 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1080 * struct udp_tunnel_info *ti);
1081 * Called by UDP tunnel to notify the driver about a UDP port and socket
1082 * address family that the UDP tunnel is not listening to anymore. The
1083 * operation is protected by the RTNL.
1085 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1086 * struct net_device *dev)
1087 * Called by upper layer devices to accelerate switching or other
1088 * station functionality into hardware. 'pdev is the lowerdev
1089 * to use for the offload and 'dev' is the net device that will
1090 * back the offload. Returns a pointer to the private structure
1091 * the upper layer will maintain.
1092 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1093 * Called by upper layer device to delete the station created
1094 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1095 * the station and priv is the structure returned by the add
1097 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1098 * struct net_device *dev,
1100 * Callback to use for xmit over the accelerated station. This
1101 * is used in place of ndo_start_xmit on accelerated net
1103 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1104 * int queue_index, u32 maxrate);
1105 * Called when a user wants to set a max-rate limitation of specific
1107 * int (*ndo_get_iflink)(const struct net_device *dev);
1108 * Called to get the iflink value of this device.
1109 * void (*ndo_change_proto_down)(struct net_device *dev,
1111 * This function is used to pass protocol port error state information
1112 * to the switch driver. The switch driver can react to the proto_down
1113 * by doing a phys down on the associated switch port.
1114 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1115 * This function is used to get egress tunnel information for given skb.
1116 * This is useful for retrieving outer tunnel header parameters while
1118 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1119 * This function is used to specify the headroom that the skb must
1120 * consider when allocation skb during packet reception. Setting
1121 * appropriate rx headroom value allows avoiding skb head copy on
1122 * forward. Setting a negative value resets the rx headroom to the
1124 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1125 * This function is used to set or query state related to XDP on the
1126 * netdevice. See definition of enum xdp_netdev_command for details.
1129 struct net_device_ops
{
1130 int (*ndo_init
)(struct net_device
*dev
);
1131 void (*ndo_uninit
)(struct net_device
*dev
);
1132 int (*ndo_open
)(struct net_device
*dev
);
1133 int (*ndo_stop
)(struct net_device
*dev
);
1134 netdev_tx_t (*ndo_start_xmit
)(struct sk_buff
*skb
,
1135 struct net_device
*dev
);
1136 netdev_features_t (*ndo_features_check
)(struct sk_buff
*skb
,
1137 struct net_device
*dev
,
1138 netdev_features_t features
);
1139 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1140 struct sk_buff
*skb
,
1142 select_queue_fallback_t fallback
);
1143 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1145 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1146 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1148 int (*ndo_validate_addr
)(struct net_device
*dev
);
1149 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1150 struct ifreq
*ifr
, int cmd
);
1151 int (*ndo_set_config
)(struct net_device
*dev
,
1153 int (*ndo_change_mtu
)(struct net_device
*dev
,
1155 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1156 struct neigh_parms
*);
1157 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1159 void (*ndo_get_stats64
)(struct net_device
*dev
,
1160 struct rtnl_link_stats64
*storage
);
1161 bool (*ndo_has_offload_stats
)(const struct net_device
*dev
, int attr_id
);
1162 int (*ndo_get_offload_stats
)(int attr_id
,
1163 const struct net_device
*dev
,
1165 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1167 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1168 __be16 proto
, u16 vid
);
1169 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1170 __be16 proto
, u16 vid
);
1171 #ifdef CONFIG_NET_POLL_CONTROLLER
1172 void (*ndo_poll_controller
)(struct net_device
*dev
);
1173 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1174 struct netpoll_info
*info
);
1175 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1177 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1178 int queue
, u8
*mac
);
1179 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1180 int queue
, u16 vlan
,
1181 u8 qos
, __be16 proto
);
1182 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1183 int vf
, int min_tx_rate
,
1185 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1186 int vf
, bool setting
);
1187 int (*ndo_set_vf_trust
)(struct net_device
*dev
,
1188 int vf
, bool setting
);
1189 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1191 struct ifla_vf_info
*ivf
);
1192 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1193 int vf
, int link_state
);
1194 int (*ndo_get_vf_stats
)(struct net_device
*dev
,
1196 struct ifla_vf_stats
1198 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1200 struct nlattr
*port
[]);
1201 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1202 int vf
, struct sk_buff
*skb
);
1203 int (*ndo_set_vf_guid
)(struct net_device
*dev
,
1206 int (*ndo_set_vf_rss_query_en
)(
1207 struct net_device
*dev
,
1208 int vf
, bool setting
);
1209 int (*ndo_setup_tc
)(struct net_device
*dev
,
1212 struct tc_to_netdev
*tc
);
1213 #if IS_ENABLED(CONFIG_FCOE)
1214 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1215 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1216 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1218 struct scatterlist
*sgl
,
1220 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1222 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1224 struct scatterlist
*sgl
,
1226 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1227 struct netdev_fcoe_hbainfo
*hbainfo
);
1230 #if IS_ENABLED(CONFIG_LIBFCOE)
1231 #define NETDEV_FCOE_WWNN 0
1232 #define NETDEV_FCOE_WWPN 1
1233 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1234 u64
*wwn
, int type
);
1237 #ifdef CONFIG_RFS_ACCEL
1238 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1239 const struct sk_buff
*skb
,
1243 int (*ndo_add_slave
)(struct net_device
*dev
,
1244 struct net_device
*slave_dev
);
1245 int (*ndo_del_slave
)(struct net_device
*dev
,
1246 struct net_device
*slave_dev
);
1247 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1248 netdev_features_t features
);
1249 int (*ndo_set_features
)(struct net_device
*dev
,
1250 netdev_features_t features
);
1251 int (*ndo_neigh_construct
)(struct net_device
*dev
,
1252 struct neighbour
*n
);
1253 void (*ndo_neigh_destroy
)(struct net_device
*dev
,
1254 struct neighbour
*n
);
1256 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1257 struct nlattr
*tb
[],
1258 struct net_device
*dev
,
1259 const unsigned char *addr
,
1262 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1263 struct nlattr
*tb
[],
1264 struct net_device
*dev
,
1265 const unsigned char *addr
,
1267 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1268 struct netlink_callback
*cb
,
1269 struct net_device
*dev
,
1270 struct net_device
*filter_dev
,
1273 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1274 struct nlmsghdr
*nlh
,
1276 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1278 struct net_device
*dev
,
1281 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1282 struct nlmsghdr
*nlh
,
1284 int (*ndo_change_carrier
)(struct net_device
*dev
,
1286 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1287 struct netdev_phys_item_id
*ppid
);
1288 int (*ndo_get_phys_port_name
)(struct net_device
*dev
,
1289 char *name
, size_t len
);
1290 void (*ndo_udp_tunnel_add
)(struct net_device
*dev
,
1291 struct udp_tunnel_info
*ti
);
1292 void (*ndo_udp_tunnel_del
)(struct net_device
*dev
,
1293 struct udp_tunnel_info
*ti
);
1294 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1295 struct net_device
*dev
);
1296 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1299 netdev_tx_t (*ndo_dfwd_start_xmit
) (struct sk_buff
*skb
,
1300 struct net_device
*dev
,
1302 int (*ndo_get_lock_subclass
)(struct net_device
*dev
);
1303 int (*ndo_set_tx_maxrate
)(struct net_device
*dev
,
1306 int (*ndo_get_iflink
)(const struct net_device
*dev
);
1307 int (*ndo_change_proto_down
)(struct net_device
*dev
,
1309 int (*ndo_fill_metadata_dst
)(struct net_device
*dev
,
1310 struct sk_buff
*skb
);
1311 void (*ndo_set_rx_headroom
)(struct net_device
*dev
,
1312 int needed_headroom
);
1313 int (*ndo_xdp
)(struct net_device
*dev
,
1314 struct netdev_xdp
*xdp
);
1318 * enum net_device_priv_flags - &struct net_device priv_flags
1320 * These are the &struct net_device, they are only set internally
1321 * by drivers and used in the kernel. These flags are invisible to
1322 * userspace; this means that the order of these flags can change
1323 * during any kernel release.
1325 * You should have a pretty good reason to be extending these flags.
1327 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1328 * @IFF_EBRIDGE: Ethernet bridging device
1329 * @IFF_BONDING: bonding master or slave
1330 * @IFF_ISATAP: ISATAP interface (RFC4214)
1331 * @IFF_WAN_HDLC: WAN HDLC device
1332 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1334 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1335 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1336 * @IFF_MACVLAN_PORT: device used as macvlan port
1337 * @IFF_BRIDGE_PORT: device used as bridge port
1338 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1339 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1340 * @IFF_UNICAST_FLT: Supports unicast filtering
1341 * @IFF_TEAM_PORT: device used as team port
1342 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1343 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1344 * change when it's running
1345 * @IFF_MACVLAN: Macvlan device
1346 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1347 * underlying stacked devices
1348 * @IFF_IPVLAN_MASTER: IPvlan master device
1349 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1350 * @IFF_L3MDEV_MASTER: device is an L3 master device
1351 * @IFF_NO_QUEUE: device can run without qdisc attached
1352 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1353 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1354 * @IFF_TEAM: device is a team device
1355 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1356 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1357 * entity (i.e. the master device for bridged veth)
1358 * @IFF_MACSEC: device is a MACsec device
1360 enum netdev_priv_flags
{
1361 IFF_802_1Q_VLAN
= 1<<0,
1365 IFF_WAN_HDLC
= 1<<4,
1366 IFF_XMIT_DST_RELEASE
= 1<<5,
1367 IFF_DONT_BRIDGE
= 1<<6,
1368 IFF_DISABLE_NETPOLL
= 1<<7,
1369 IFF_MACVLAN_PORT
= 1<<8,
1370 IFF_BRIDGE_PORT
= 1<<9,
1371 IFF_OVS_DATAPATH
= 1<<10,
1372 IFF_TX_SKB_SHARING
= 1<<11,
1373 IFF_UNICAST_FLT
= 1<<12,
1374 IFF_TEAM_PORT
= 1<<13,
1375 IFF_SUPP_NOFCS
= 1<<14,
1376 IFF_LIVE_ADDR_CHANGE
= 1<<15,
1377 IFF_MACVLAN
= 1<<16,
1378 IFF_XMIT_DST_RELEASE_PERM
= 1<<17,
1379 IFF_IPVLAN_MASTER
= 1<<18,
1380 IFF_IPVLAN_SLAVE
= 1<<19,
1381 IFF_L3MDEV_MASTER
= 1<<20,
1382 IFF_NO_QUEUE
= 1<<21,
1383 IFF_OPENVSWITCH
= 1<<22,
1384 IFF_L3MDEV_SLAVE
= 1<<23,
1386 IFF_RXFH_CONFIGURED
= 1<<25,
1387 IFF_PHONY_HEADROOM
= 1<<26,
1391 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1392 #define IFF_EBRIDGE IFF_EBRIDGE
1393 #define IFF_BONDING IFF_BONDING
1394 #define IFF_ISATAP IFF_ISATAP
1395 #define IFF_WAN_HDLC IFF_WAN_HDLC
1396 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1397 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1398 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1399 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1400 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1401 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1402 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1403 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1404 #define IFF_TEAM_PORT IFF_TEAM_PORT
1405 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1406 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1407 #define IFF_MACVLAN IFF_MACVLAN
1408 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1409 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1410 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1411 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1412 #define IFF_NO_QUEUE IFF_NO_QUEUE
1413 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1414 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1415 #define IFF_TEAM IFF_TEAM
1416 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1417 #define IFF_MACSEC IFF_MACSEC
1420 * struct net_device - The DEVICE structure.
1421 * Actually, this whole structure is a big mistake. It mixes I/O
1422 * data with strictly "high-level" data, and it has to know about
1423 * almost every data structure used in the INET module.
1425 * @name: This is the first field of the "visible" part of this structure
1426 * (i.e. as seen by users in the "Space.c" file). It is the name
1429 * @name_hlist: Device name hash chain, please keep it close to name[]
1430 * @ifalias: SNMP alias
1431 * @mem_end: Shared memory end
1432 * @mem_start: Shared memory start
1433 * @base_addr: Device I/O address
1434 * @irq: Device IRQ number
1436 * @carrier_changes: Stats to monitor carrier on<->off transitions
1438 * @state: Generic network queuing layer state, see netdev_state_t
1439 * @dev_list: The global list of network devices
1440 * @napi_list: List entry used for polling NAPI devices
1441 * @unreg_list: List entry when we are unregistering the
1442 * device; see the function unregister_netdev
1443 * @close_list: List entry used when we are closing the device
1444 * @ptype_all: Device-specific packet handlers for all protocols
1445 * @ptype_specific: Device-specific, protocol-specific packet handlers
1447 * @adj_list: Directly linked devices, like slaves for bonding
1448 * @features: Currently active device features
1449 * @hw_features: User-changeable features
1451 * @wanted_features: User-requested features
1452 * @vlan_features: Mask of features inheritable by VLAN devices
1454 * @hw_enc_features: Mask of features inherited by encapsulating devices
1455 * This field indicates what encapsulation
1456 * offloads the hardware is capable of doing,
1457 * and drivers will need to set them appropriately.
1459 * @mpls_features: Mask of features inheritable by MPLS
1461 * @ifindex: interface index
1462 * @group: The group the device belongs to
1464 * @stats: Statistics struct, which was left as a legacy, use
1465 * rtnl_link_stats64 instead
1467 * @rx_dropped: Dropped packets by core network,
1468 * do not use this in drivers
1469 * @tx_dropped: Dropped packets by core network,
1470 * do not use this in drivers
1471 * @rx_nohandler: nohandler dropped packets by core network on
1472 * inactive devices, do not use this in drivers
1474 * @wireless_handlers: List of functions to handle Wireless Extensions,
1476 * see <net/iw_handler.h> for details.
1477 * @wireless_data: Instance data managed by the core of wireless extensions
1479 * @netdev_ops: Includes several pointers to callbacks,
1480 * if one wants to override the ndo_*() functions
1481 * @ethtool_ops: Management operations
1482 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1483 * discovery handling. Necessary for e.g. 6LoWPAN.
1484 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1485 * of Layer 2 headers.
1487 * @flags: Interface flags (a la BSD)
1488 * @priv_flags: Like 'flags' but invisible to userspace,
1489 * see if.h for the definitions
1490 * @gflags: Global flags ( kept as legacy )
1491 * @padded: How much padding added by alloc_netdev()
1492 * @operstate: RFC2863 operstate
1493 * @link_mode: Mapping policy to operstate
1494 * @if_port: Selectable AUI, TP, ...
1496 * @mtu: Interface MTU value
1497 * @min_mtu: Interface Minimum MTU value
1498 * @max_mtu: Interface Maximum MTU value
1499 * @type: Interface hardware type
1500 * @hard_header_len: Maximum hardware header length.
1501 * @min_header_len: Minimum hardware header length
1503 * @needed_headroom: Extra headroom the hardware may need, but not in all
1504 * cases can this be guaranteed
1505 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1506 * cases can this be guaranteed. Some cases also use
1507 * LL_MAX_HEADER instead to allocate the skb
1509 * interface address info:
1511 * @perm_addr: Permanent hw address
1512 * @addr_assign_type: Hw address assignment type
1513 * @addr_len: Hardware address length
1514 * @neigh_priv_len: Used in neigh_alloc()
1515 * @dev_id: Used to differentiate devices that share
1516 * the same link layer address
1517 * @dev_port: Used to differentiate devices that share
1519 * @addr_list_lock: XXX: need comments on this one
1520 * @uc_promisc: Counter that indicates promiscuous mode
1521 * has been enabled due to the need to listen to
1522 * additional unicast addresses in a device that
1523 * does not implement ndo_set_rx_mode()
1524 * @uc: unicast mac addresses
1525 * @mc: multicast mac addresses
1526 * @dev_addrs: list of device hw addresses
1527 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1528 * @promiscuity: Number of times the NIC is told to work in
1529 * promiscuous mode; if it becomes 0 the NIC will
1530 * exit promiscuous mode
1531 * @allmulti: Counter, enables or disables allmulticast mode
1533 * @vlan_info: VLAN info
1534 * @dsa_ptr: dsa specific data
1535 * @tipc_ptr: TIPC specific data
1536 * @atalk_ptr: AppleTalk link
1537 * @ip_ptr: IPv4 specific data
1538 * @dn_ptr: DECnet specific data
1539 * @ip6_ptr: IPv6 specific data
1540 * @ax25_ptr: AX.25 specific data
1541 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1543 * @dev_addr: Hw address (before bcast,
1544 * because most packets are unicast)
1546 * @_rx: Array of RX queues
1547 * @num_rx_queues: Number of RX queues
1548 * allocated at register_netdev() time
1549 * @real_num_rx_queues: Number of RX queues currently active in device
1551 * @rx_handler: handler for received packets
1552 * @rx_handler_data: XXX: need comments on this one
1553 * @ingress_queue: XXX: need comments on this one
1554 * @broadcast: hw bcast address
1556 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1557 * indexed by RX queue number. Assigned by driver.
1558 * This must only be set if the ndo_rx_flow_steer
1559 * operation is defined
1560 * @index_hlist: Device index hash chain
1562 * @_tx: Array of TX queues
1563 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1564 * @real_num_tx_queues: Number of TX queues currently active in device
1565 * @qdisc: Root qdisc from userspace point of view
1566 * @tx_queue_len: Max frames per queue allowed
1567 * @tx_global_lock: XXX: need comments on this one
1569 * @xps_maps: XXX: need comments on this one
1571 * @watchdog_timeo: Represents the timeout that is used by
1572 * the watchdog (see dev_watchdog())
1573 * @watchdog_timer: List of timers
1575 * @pcpu_refcnt: Number of references to this device
1576 * @todo_list: Delayed register/unregister
1577 * @link_watch_list: XXX: need comments on this one
1579 * @reg_state: Register/unregister state machine
1580 * @dismantle: Device is going to be freed
1581 * @rtnl_link_state: This enum represents the phases of creating
1584 * @destructor: Called from unregister,
1585 * can be used to call free_netdev
1586 * @npinfo: XXX: need comments on this one
1587 * @nd_net: Network namespace this network device is inside
1589 * @ml_priv: Mid-layer private
1590 * @lstats: Loopback statistics
1591 * @tstats: Tunnel statistics
1592 * @dstats: Dummy statistics
1593 * @vstats: Virtual ethernet statistics
1598 * @dev: Class/net/name entry
1599 * @sysfs_groups: Space for optional device, statistics and wireless
1602 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1603 * @rtnl_link_ops: Rtnl_link_ops
1605 * @gso_max_size: Maximum size of generic segmentation offload
1606 * @gso_max_segs: Maximum number of segments that can be passed to the
1609 * @dcbnl_ops: Data Center Bridging netlink ops
1610 * @num_tc: Number of traffic classes in the net device
1611 * @tc_to_txq: XXX: need comments on this one
1612 * @prio_tc_map: XXX: need comments on this one
1614 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1616 * @priomap: XXX: need comments on this one
1617 * @phydev: Physical device may attach itself
1618 * for hardware timestamping
1620 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1621 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1623 * @proto_down: protocol port state information can be sent to the
1624 * switch driver and used to set the phys state of the
1627 * FIXME: cleanup struct net_device such that network protocol info
1632 char name
[IFNAMSIZ
];
1633 struct hlist_node name_hlist
;
1636 * I/O specific fields
1637 * FIXME: Merge these and struct ifmap into one
1639 unsigned long mem_end
;
1640 unsigned long mem_start
;
1641 unsigned long base_addr
;
1644 atomic_t carrier_changes
;
1647 * Some hardware also needs these fields (state,dev_list,
1648 * napi_list,unreg_list,close_list) but they are not
1649 * part of the usual set specified in Space.c.
1652 unsigned long state
;
1654 struct list_head dev_list
;
1655 struct list_head napi_list
;
1656 struct list_head unreg_list
;
1657 struct list_head close_list
;
1658 struct list_head ptype_all
;
1659 struct list_head ptype_specific
;
1662 struct list_head upper
;
1663 struct list_head lower
;
1666 netdev_features_t features
;
1667 netdev_features_t hw_features
;
1668 netdev_features_t wanted_features
;
1669 netdev_features_t vlan_features
;
1670 netdev_features_t hw_enc_features
;
1671 netdev_features_t mpls_features
;
1672 netdev_features_t gso_partial_features
;
1677 struct net_device_stats stats
;
1679 atomic_long_t rx_dropped
;
1680 atomic_long_t tx_dropped
;
1681 atomic_long_t rx_nohandler
;
1683 #ifdef CONFIG_WIRELESS_EXT
1684 const struct iw_handler_def
*wireless_handlers
;
1685 struct iw_public_data
*wireless_data
;
1687 const struct net_device_ops
*netdev_ops
;
1688 const struct ethtool_ops
*ethtool_ops
;
1689 #ifdef CONFIG_NET_SWITCHDEV
1690 const struct switchdev_ops
*switchdev_ops
;
1692 #ifdef CONFIG_NET_L3_MASTER_DEV
1693 const struct l3mdev_ops
*l3mdev_ops
;
1695 #if IS_ENABLED(CONFIG_IPV6)
1696 const struct ndisc_ops
*ndisc_ops
;
1699 const struct header_ops
*header_ops
;
1702 unsigned int priv_flags
;
1704 unsigned short gflags
;
1705 unsigned short padded
;
1707 unsigned char operstate
;
1708 unsigned char link_mode
;
1710 unsigned char if_port
;
1714 unsigned int min_mtu
;
1715 unsigned int max_mtu
;
1716 unsigned short type
;
1717 unsigned short hard_header_len
;
1718 unsigned char min_header_len
;
1720 unsigned short needed_headroom
;
1721 unsigned short needed_tailroom
;
1723 /* Interface address info. */
1724 unsigned char perm_addr
[MAX_ADDR_LEN
];
1725 unsigned char addr_assign_type
;
1726 unsigned char addr_len
;
1727 unsigned short neigh_priv_len
;
1728 unsigned short dev_id
;
1729 unsigned short dev_port
;
1730 spinlock_t addr_list_lock
;
1731 unsigned char name_assign_type
;
1733 struct netdev_hw_addr_list uc
;
1734 struct netdev_hw_addr_list mc
;
1735 struct netdev_hw_addr_list dev_addrs
;
1738 struct kset
*queues_kset
;
1740 unsigned int promiscuity
;
1741 unsigned int allmulti
;
1744 /* Protocol-specific pointers */
1746 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1747 struct vlan_info __rcu
*vlan_info
;
1749 #if IS_ENABLED(CONFIG_NET_DSA)
1750 struct dsa_switch_tree
*dsa_ptr
;
1752 #if IS_ENABLED(CONFIG_TIPC)
1753 struct tipc_bearer __rcu
*tipc_ptr
;
1756 struct in_device __rcu
*ip_ptr
;
1757 struct dn_dev __rcu
*dn_ptr
;
1758 struct inet6_dev __rcu
*ip6_ptr
;
1760 struct wireless_dev
*ieee80211_ptr
;
1761 struct wpan_dev
*ieee802154_ptr
;
1762 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1763 struct mpls_dev __rcu
*mpls_ptr
;
1767 * Cache lines mostly used on receive path (including eth_type_trans())
1769 /* Interface address info used in eth_type_trans() */
1770 unsigned char *dev_addr
;
1773 struct netdev_rx_queue
*_rx
;
1775 unsigned int num_rx_queues
;
1776 unsigned int real_num_rx_queues
;
1779 unsigned long gro_flush_timeout
;
1780 rx_handler_func_t __rcu
*rx_handler
;
1781 void __rcu
*rx_handler_data
;
1783 #ifdef CONFIG_NET_CLS_ACT
1784 struct tcf_proto __rcu
*ingress_cl_list
;
1786 struct netdev_queue __rcu
*ingress_queue
;
1787 #ifdef CONFIG_NETFILTER_INGRESS
1788 struct nf_hook_entry __rcu
*nf_hooks_ingress
;
1791 unsigned char broadcast
[MAX_ADDR_LEN
];
1792 #ifdef CONFIG_RFS_ACCEL
1793 struct cpu_rmap
*rx_cpu_rmap
;
1795 struct hlist_node index_hlist
;
1798 * Cache lines mostly used on transmit path
1800 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1801 unsigned int num_tx_queues
;
1802 unsigned int real_num_tx_queues
;
1803 struct Qdisc
*qdisc
;
1804 #ifdef CONFIG_NET_SCHED
1805 DECLARE_HASHTABLE (qdisc_hash
, 4);
1807 unsigned long tx_queue_len
;
1808 spinlock_t tx_global_lock
;
1812 struct xps_dev_maps __rcu
*xps_maps
;
1814 #ifdef CONFIG_NET_CLS_ACT
1815 struct tcf_proto __rcu
*egress_cl_list
;
1818 /* These may be needed for future network-power-down code. */
1819 struct timer_list watchdog_timer
;
1821 int __percpu
*pcpu_refcnt
;
1822 struct list_head todo_list
;
1824 struct list_head link_watch_list
;
1826 enum { NETREG_UNINITIALIZED
=0,
1827 NETREG_REGISTERED
, /* completed register_netdevice */
1828 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1829 NETREG_UNREGISTERED
, /* completed unregister todo */
1830 NETREG_RELEASED
, /* called free_netdev */
1831 NETREG_DUMMY
, /* dummy device for NAPI poll */
1837 RTNL_LINK_INITIALIZED
,
1838 RTNL_LINK_INITIALIZING
,
1839 } rtnl_link_state
:16;
1841 void (*destructor
)(struct net_device
*dev
);
1843 #ifdef CONFIG_NETPOLL
1844 struct netpoll_info __rcu
*npinfo
;
1847 possible_net_t nd_net
;
1849 /* mid-layer private */
1852 struct pcpu_lstats __percpu
*lstats
;
1853 struct pcpu_sw_netstats __percpu
*tstats
;
1854 struct pcpu_dstats __percpu
*dstats
;
1855 struct pcpu_vstats __percpu
*vstats
;
1858 #if IS_ENABLED(CONFIG_GARP)
1859 struct garp_port __rcu
*garp_port
;
1861 #if IS_ENABLED(CONFIG_MRP)
1862 struct mrp_port __rcu
*mrp_port
;
1866 const struct attribute_group
*sysfs_groups
[4];
1867 const struct attribute_group
*sysfs_rx_queue_group
;
1869 const struct rtnl_link_ops
*rtnl_link_ops
;
1871 /* for setting kernel sock attribute on TCP connection setup */
1872 #define GSO_MAX_SIZE 65536
1873 unsigned int gso_max_size
;
1874 #define GSO_MAX_SEGS 65535
1878 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1881 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1882 u8 prio_tc_map
[TC_BITMASK
+ 1];
1884 #if IS_ENABLED(CONFIG_FCOE)
1885 unsigned int fcoe_ddp_xid
;
1887 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1888 struct netprio_map __rcu
*priomap
;
1890 struct phy_device
*phydev
;
1891 struct lock_class_key
*qdisc_tx_busylock
;
1892 struct lock_class_key
*qdisc_running_key
;
1895 #define to_net_dev(d) container_of(d, struct net_device, dev)
1897 #define NETDEV_ALIGN 32
1900 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1902 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1906 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1908 if (tc
>= dev
->num_tc
)
1911 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1915 int netdev_txq_to_tc(struct net_device
*dev
, unsigned int txq
);
1916 void netdev_reset_tc(struct net_device
*dev
);
1917 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
);
1918 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
);
1921 int netdev_get_num_tc(struct net_device
*dev
)
1927 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1930 return &dev
->_tx
[index
];
1933 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1934 const struct sk_buff
*skb
)
1936 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1939 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1940 void (*f
)(struct net_device
*,
1941 struct netdev_queue
*,
1947 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1948 f(dev
, &dev
->_tx
[i
], arg
);
1951 #define netdev_lockdep_set_classes(dev) \
1953 static struct lock_class_key qdisc_tx_busylock_key; \
1954 static struct lock_class_key qdisc_running_key; \
1955 static struct lock_class_key qdisc_xmit_lock_key; \
1956 static struct lock_class_key dev_addr_list_lock_key; \
1959 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1960 (dev)->qdisc_running_key = &qdisc_running_key; \
1961 lockdep_set_class(&(dev)->addr_list_lock, \
1962 &dev_addr_list_lock_key); \
1963 for (i = 0; i < (dev)->num_tx_queues; i++) \
1964 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1965 &qdisc_xmit_lock_key); \
1968 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1969 struct sk_buff
*skb
,
1972 /* returns the headroom that the master device needs to take in account
1973 * when forwarding to this dev
1975 static inline unsigned netdev_get_fwd_headroom(struct net_device
*dev
)
1977 return dev
->priv_flags
& IFF_PHONY_HEADROOM
? 0 : dev
->needed_headroom
;
1980 static inline void netdev_set_rx_headroom(struct net_device
*dev
, int new_hr
)
1982 if (dev
->netdev_ops
->ndo_set_rx_headroom
)
1983 dev
->netdev_ops
->ndo_set_rx_headroom(dev
, new_hr
);
1986 /* set the device rx headroom to the dev's default */
1987 static inline void netdev_reset_rx_headroom(struct net_device
*dev
)
1989 netdev_set_rx_headroom(dev
, -1);
1993 * Net namespace inlines
1996 struct net
*dev_net(const struct net_device
*dev
)
1998 return read_pnet(&dev
->nd_net
);
2002 void dev_net_set(struct net_device
*dev
, struct net
*net
)
2004 write_pnet(&dev
->nd_net
, net
);
2008 * netdev_priv - access network device private data
2009 * @dev: network device
2011 * Get network device private data
2013 static inline void *netdev_priv(const struct net_device
*dev
)
2015 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
2018 /* Set the sysfs physical device reference for the network logical device
2019 * if set prior to registration will cause a symlink during initialization.
2021 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2023 /* Set the sysfs device type for the network logical device to allow
2024 * fine-grained identification of different network device types. For
2025 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2027 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2029 /* Default NAPI poll() weight
2030 * Device drivers are strongly advised to not use bigger value
2032 #define NAPI_POLL_WEIGHT 64
2035 * netif_napi_add - initialize a NAPI context
2036 * @dev: network device
2037 * @napi: NAPI context
2038 * @poll: polling function
2039 * @weight: default weight
2041 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2042 * *any* of the other NAPI-related functions.
2044 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
2045 int (*poll
)(struct napi_struct
*, int), int weight
);
2048 * netif_tx_napi_add - initialize a NAPI context
2049 * @dev: network device
2050 * @napi: NAPI context
2051 * @poll: polling function
2052 * @weight: default weight
2054 * This variant of netif_napi_add() should be used from drivers using NAPI
2055 * to exclusively poll a TX queue.
2056 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2058 static inline void netif_tx_napi_add(struct net_device
*dev
,
2059 struct napi_struct
*napi
,
2060 int (*poll
)(struct napi_struct
*, int),
2063 set_bit(NAPI_STATE_NO_BUSY_POLL
, &napi
->state
);
2064 netif_napi_add(dev
, napi
, poll
, weight
);
2068 * netif_napi_del - remove a NAPI context
2069 * @napi: NAPI context
2071 * netif_napi_del() removes a NAPI context from the network device NAPI list
2073 void netif_napi_del(struct napi_struct
*napi
);
2075 struct napi_gro_cb
{
2076 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2079 /* Length of frag0. */
2080 unsigned int frag0_len
;
2082 /* This indicates where we are processing relative to skb->data. */
2085 /* This is non-zero if the packet cannot be merged with the new skb. */
2088 /* Save the IP ID here and check when we get to the transport layer */
2091 /* Number of segments aggregated. */
2094 /* Start offset for remote checksum offload */
2095 u16 gro_remcsum_start
;
2097 /* jiffies when first packet was created/queued */
2100 /* Used in ipv6_gro_receive() and foo-over-udp */
2103 /* This is non-zero if the packet may be of the same flow. */
2106 /* Used in tunnel GRO receive */
2109 /* GRO checksum is valid */
2112 /* Number of checksums via CHECKSUM_UNNECESSARY */
2117 #define NAPI_GRO_FREE 1
2118 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2120 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2123 /* Used in GRE, set in fou/gue_gro_receive */
2126 /* Used to determine if flush_id can be ignored */
2129 /* Number of gro_receive callbacks this packet already went through */
2130 u8 recursion_counter
:4;
2134 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2137 /* used in skb_gro_receive() slow path */
2138 struct sk_buff
*last
;
2141 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2143 #define GRO_RECURSION_LIMIT 15
2144 static inline int gro_recursion_inc_test(struct sk_buff
*skb
)
2146 return ++NAPI_GRO_CB(skb
)->recursion_counter
== GRO_RECURSION_LIMIT
;
2149 typedef struct sk_buff
**(*gro_receive_t
)(struct sk_buff
**, struct sk_buff
*);
2150 static inline struct sk_buff
**call_gro_receive(gro_receive_t cb
,
2151 struct sk_buff
**head
,
2152 struct sk_buff
*skb
)
2154 if (unlikely(gro_recursion_inc_test(skb
))) {
2155 NAPI_GRO_CB(skb
)->flush
|= 1;
2159 return cb(head
, skb
);
2162 typedef struct sk_buff
**(*gro_receive_sk_t
)(struct sock
*, struct sk_buff
**,
2164 static inline struct sk_buff
**call_gro_receive_sk(gro_receive_sk_t cb
,
2166 struct sk_buff
**head
,
2167 struct sk_buff
*skb
)
2169 if (unlikely(gro_recursion_inc_test(skb
))) {
2170 NAPI_GRO_CB(skb
)->flush
|= 1;
2174 return cb(sk
, head
, skb
);
2177 struct packet_type
{
2178 __be16 type
; /* This is really htons(ether_type). */
2179 struct net_device
*dev
; /* NULL is wildcarded here */
2180 int (*func
) (struct sk_buff
*,
2181 struct net_device
*,
2182 struct packet_type
*,
2183 struct net_device
*);
2184 bool (*id_match
)(struct packet_type
*ptype
,
2186 void *af_packet_priv
;
2187 struct list_head list
;
2190 struct offload_callbacks
{
2191 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
2192 netdev_features_t features
);
2193 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
2194 struct sk_buff
*skb
);
2195 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
2198 struct packet_offload
{
2199 __be16 type
; /* This is really htons(ether_type). */
2201 struct offload_callbacks callbacks
;
2202 struct list_head list
;
2205 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2206 struct pcpu_sw_netstats
{
2211 struct u64_stats_sync syncp
;
2214 #define __netdev_alloc_pcpu_stats(type, gfp) \
2216 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2219 for_each_possible_cpu(__cpu) { \
2220 typeof(type) *stat; \
2221 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2222 u64_stats_init(&stat->syncp); \
2228 #define netdev_alloc_pcpu_stats(type) \
2229 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2231 enum netdev_lag_tx_type
{
2232 NETDEV_LAG_TX_TYPE_UNKNOWN
,
2233 NETDEV_LAG_TX_TYPE_RANDOM
,
2234 NETDEV_LAG_TX_TYPE_BROADCAST
,
2235 NETDEV_LAG_TX_TYPE_ROUNDROBIN
,
2236 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP
,
2237 NETDEV_LAG_TX_TYPE_HASH
,
2240 struct netdev_lag_upper_info
{
2241 enum netdev_lag_tx_type tx_type
;
2244 struct netdev_lag_lower_state_info
{
2249 #include <linux/notifier.h>
2251 /* netdevice notifier chain. Please remember to update the rtnetlink
2252 * notification exclusion list in rtnetlink_event() when adding new
2255 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2256 #define NETDEV_DOWN 0x0002
2257 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2258 detected a hardware crash and restarted
2259 - we can use this eg to kick tcp sessions
2261 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2262 #define NETDEV_REGISTER 0x0005
2263 #define NETDEV_UNREGISTER 0x0006
2264 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2265 #define NETDEV_CHANGEADDR 0x0008
2266 #define NETDEV_GOING_DOWN 0x0009
2267 #define NETDEV_CHANGENAME 0x000A
2268 #define NETDEV_FEAT_CHANGE 0x000B
2269 #define NETDEV_BONDING_FAILOVER 0x000C
2270 #define NETDEV_PRE_UP 0x000D
2271 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2272 #define NETDEV_POST_TYPE_CHANGE 0x000F
2273 #define NETDEV_POST_INIT 0x0010
2274 #define NETDEV_UNREGISTER_FINAL 0x0011
2275 #define NETDEV_RELEASE 0x0012
2276 #define NETDEV_NOTIFY_PEERS 0x0013
2277 #define NETDEV_JOIN 0x0014
2278 #define NETDEV_CHANGEUPPER 0x0015
2279 #define NETDEV_RESEND_IGMP 0x0016
2280 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2281 #define NETDEV_CHANGEINFODATA 0x0018
2282 #define NETDEV_BONDING_INFO 0x0019
2283 #define NETDEV_PRECHANGEUPPER 0x001A
2284 #define NETDEV_CHANGELOWERSTATE 0x001B
2285 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2286 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2288 int register_netdevice_notifier(struct notifier_block
*nb
);
2289 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2291 struct netdev_notifier_info
{
2292 struct net_device
*dev
;
2295 struct netdev_notifier_change_info
{
2296 struct netdev_notifier_info info
; /* must be first */
2297 unsigned int flags_changed
;
2300 struct netdev_notifier_changeupper_info
{
2301 struct netdev_notifier_info info
; /* must be first */
2302 struct net_device
*upper_dev
; /* new upper dev */
2303 bool master
; /* is upper dev master */
2304 bool linking
; /* is the notification for link or unlink */
2305 void *upper_info
; /* upper dev info */
2308 struct netdev_notifier_changelowerstate_info
{
2309 struct netdev_notifier_info info
; /* must be first */
2310 void *lower_state_info
; /* is lower dev state */
2313 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2314 struct net_device
*dev
)
2319 static inline struct net_device
*
2320 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2325 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2328 extern rwlock_t dev_base_lock
; /* Device list lock */
2330 #define for_each_netdev(net, d) \
2331 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2332 #define for_each_netdev_reverse(net, d) \
2333 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2334 #define for_each_netdev_rcu(net, d) \
2335 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2336 #define for_each_netdev_safe(net, d, n) \
2337 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2338 #define for_each_netdev_continue(net, d) \
2339 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2340 #define for_each_netdev_continue_rcu(net, d) \
2341 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2342 #define for_each_netdev_in_bond_rcu(bond, slave) \
2343 for_each_netdev_rcu(&init_net, slave) \
2344 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2345 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2347 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2349 struct list_head
*lh
;
2353 lh
= dev
->dev_list
.next
;
2354 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2357 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2359 struct list_head
*lh
;
2363 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2364 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2367 static inline struct net_device
*first_net_device(struct net
*net
)
2369 return list_empty(&net
->dev_base_head
) ? NULL
:
2370 net_device_entry(net
->dev_base_head
.next
);
2373 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2375 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2377 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2380 int netdev_boot_setup_check(struct net_device
*dev
);
2381 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2382 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2383 const char *hwaddr
);
2384 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2385 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2386 void dev_add_pack(struct packet_type
*pt
);
2387 void dev_remove_pack(struct packet_type
*pt
);
2388 void __dev_remove_pack(struct packet_type
*pt
);
2389 void dev_add_offload(struct packet_offload
*po
);
2390 void dev_remove_offload(struct packet_offload
*po
);
2392 int dev_get_iflink(const struct net_device
*dev
);
2393 int dev_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
);
2394 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2395 unsigned short mask
);
2396 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2397 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2398 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2399 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2400 int dev_open(struct net_device
*dev
);
2401 int dev_close(struct net_device
*dev
);
2402 int dev_close_many(struct list_head
*head
, bool unlink
);
2403 void dev_disable_lro(struct net_device
*dev
);
2404 int dev_loopback_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*newskb
);
2405 int dev_queue_xmit(struct sk_buff
*skb
);
2406 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2407 int register_netdevice(struct net_device
*dev
);
2408 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2409 void unregister_netdevice_many(struct list_head
*head
);
2410 static inline void unregister_netdevice(struct net_device
*dev
)
2412 unregister_netdevice_queue(dev
, NULL
);
2415 int netdev_refcnt_read(const struct net_device
*dev
);
2416 void free_netdev(struct net_device
*dev
);
2417 void netdev_freemem(struct net_device
*dev
);
2418 void synchronize_net(void);
2419 int init_dummy_netdev(struct net_device
*dev
);
2421 DECLARE_PER_CPU(int, xmit_recursion
);
2422 #define XMIT_RECURSION_LIMIT 10
2424 static inline int dev_recursion_level(void)
2426 return this_cpu_read(xmit_recursion
);
2429 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2430 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2431 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2432 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2433 int dev_restart(struct net_device
*dev
);
2434 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2436 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2438 return NAPI_GRO_CB(skb
)->data_offset
;
2441 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2443 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2446 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2448 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2451 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2452 unsigned int offset
)
2454 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2457 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2459 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2462 static inline void skb_gro_frag0_invalidate(struct sk_buff
*skb
)
2464 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2465 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2468 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2469 unsigned int offset
)
2471 if (!pskb_may_pull(skb
, hlen
))
2474 skb_gro_frag0_invalidate(skb
);
2475 return skb
->data
+ offset
;
2478 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2480 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2481 skb_network_offset(skb
);
2484 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2485 const void *start
, unsigned int len
)
2487 if (NAPI_GRO_CB(skb
)->csum_valid
)
2488 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2489 csum_partial(start
, len
, 0));
2492 /* GRO checksum functions. These are logical equivalents of the normal
2493 * checksum functions (in skbuff.h) except that they operate on the GRO
2494 * offsets and fields in sk_buff.
2497 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2499 static inline bool skb_at_gro_remcsum_start(struct sk_buff
*skb
)
2501 return (NAPI_GRO_CB(skb
)->gro_remcsum_start
== skb_gro_offset(skb
));
2504 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2508 return ((skb
->ip_summed
!= CHECKSUM_PARTIAL
||
2509 skb_checksum_start_offset(skb
) <
2510 skb_gro_offset(skb
)) &&
2511 !skb_at_gro_remcsum_start(skb
) &&
2512 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2513 (!zero_okay
|| check
));
2516 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2519 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2520 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2523 NAPI_GRO_CB(skb
)->csum
= psum
;
2525 return __skb_gro_checksum_complete(skb
);
2528 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2530 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2531 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2532 NAPI_GRO_CB(skb
)->csum_cnt
--;
2534 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2535 * verified a new top level checksum or an encapsulated one
2536 * during GRO. This saves work if we fallback to normal path.
2538 __skb_incr_checksum_unnecessary(skb
);
2542 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2545 __sum16 __ret = 0; \
2546 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2547 __ret = __skb_gro_checksum_validate_complete(skb, \
2548 compute_pseudo(skb, proto)); \
2550 __skb_mark_checksum_bad(skb); \
2552 skb_gro_incr_csum_unnecessary(skb); \
2556 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2557 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2559 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2561 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2563 #define skb_gro_checksum_simple_validate(skb) \
2564 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2566 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2568 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2569 !NAPI_GRO_CB(skb
)->csum_valid
);
2572 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2573 __sum16 check
, __wsum pseudo
)
2575 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2576 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2579 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2581 if (__skb_gro_checksum_convert_check(skb)) \
2582 __skb_gro_checksum_convert(skb, check, \
2583 compute_pseudo(skb, proto)); \
2586 struct gro_remcsum
{
2591 static inline void skb_gro_remcsum_init(struct gro_remcsum
*grc
)
2597 static inline void *skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2598 unsigned int off
, size_t hdrlen
,
2599 int start
, int offset
,
2600 struct gro_remcsum
*grc
,
2604 size_t plen
= hdrlen
+ max_t(size_t, offset
+ sizeof(u16
), start
);
2606 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2609 NAPI_GRO_CB(skb
)->gro_remcsum_start
= off
+ hdrlen
+ start
;
2613 ptr
= skb_gro_header_fast(skb
, off
);
2614 if (skb_gro_header_hard(skb
, off
+ plen
)) {
2615 ptr
= skb_gro_header_slow(skb
, off
+ plen
, off
);
2620 delta
= remcsum_adjust(ptr
+ hdrlen
, NAPI_GRO_CB(skb
)->csum
,
2623 /* Adjust skb->csum since we changed the packet */
2624 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2626 grc
->offset
= off
+ hdrlen
+ offset
;
2632 static inline void skb_gro_remcsum_cleanup(struct sk_buff
*skb
,
2633 struct gro_remcsum
*grc
)
2636 size_t plen
= grc
->offset
+ sizeof(u16
);
2641 ptr
= skb_gro_header_fast(skb
, grc
->offset
);
2642 if (skb_gro_header_hard(skb
, grc
->offset
+ sizeof(u16
))) {
2643 ptr
= skb_gro_header_slow(skb
, plen
, grc
->offset
);
2648 remcsum_unadjust((__sum16
*)ptr
, grc
->delta
);
2651 #ifdef CONFIG_XFRM_OFFLOAD
2652 static inline void skb_gro_flush_final(struct sk_buff
*skb
, struct sk_buff
**pp
, int flush
)
2654 if (PTR_ERR(pp
) != -EINPROGRESS
)
2655 NAPI_GRO_CB(skb
)->flush
|= flush
;
2658 static inline void skb_gro_flush_final(struct sk_buff
*skb
, struct sk_buff
**pp
, int flush
)
2660 NAPI_GRO_CB(skb
)->flush
|= flush
;
2664 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2665 unsigned short type
,
2666 const void *daddr
, const void *saddr
,
2669 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2672 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2675 static inline int dev_parse_header(const struct sk_buff
*skb
,
2676 unsigned char *haddr
)
2678 const struct net_device
*dev
= skb
->dev
;
2680 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2682 return dev
->header_ops
->parse(skb
, haddr
);
2685 /* ll_header must have at least hard_header_len allocated */
2686 static inline bool dev_validate_header(const struct net_device
*dev
,
2687 char *ll_header
, int len
)
2689 if (likely(len
>= dev
->hard_header_len
))
2691 if (len
< dev
->min_header_len
)
2694 if (capable(CAP_SYS_RAWIO
)) {
2695 memset(ll_header
+ len
, 0, dev
->hard_header_len
- len
);
2699 if (dev
->header_ops
&& dev
->header_ops
->validate
)
2700 return dev
->header_ops
->validate(ll_header
, len
);
2705 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2706 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2707 static inline int unregister_gifconf(unsigned int family
)
2709 return register_gifconf(family
, NULL
);
2712 #ifdef CONFIG_NET_FLOW_LIMIT
2713 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2714 struct sd_flow_limit
{
2716 unsigned int num_buckets
;
2717 unsigned int history_head
;
2718 u16 history
[FLOW_LIMIT_HISTORY
];
2722 extern int netdev_flow_limit_table_len
;
2723 #endif /* CONFIG_NET_FLOW_LIMIT */
2726 * Incoming packets are placed on per-CPU queues
2728 struct softnet_data
{
2729 struct list_head poll_list
;
2730 struct sk_buff_head process_queue
;
2733 unsigned int processed
;
2734 unsigned int time_squeeze
;
2735 unsigned int received_rps
;
2737 struct softnet_data
*rps_ipi_list
;
2739 #ifdef CONFIG_NET_FLOW_LIMIT
2740 struct sd_flow_limit __rcu
*flow_limit
;
2742 struct Qdisc
*output_queue
;
2743 struct Qdisc
**output_queue_tailp
;
2744 struct sk_buff
*completion_queue
;
2747 /* input_queue_head should be written by cpu owning this struct,
2748 * and only read by other cpus. Worth using a cache line.
2750 unsigned int input_queue_head ____cacheline_aligned_in_smp
;
2752 /* Elements below can be accessed between CPUs for RPS/RFS */
2753 struct call_single_data csd ____cacheline_aligned_in_smp
;
2754 struct softnet_data
*rps_ipi_next
;
2756 unsigned int input_queue_tail
;
2758 unsigned int dropped
;
2759 struct sk_buff_head input_pkt_queue
;
2760 struct napi_struct backlog
;
2764 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2767 sd
->input_queue_head
++;
2771 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2772 unsigned int *qtail
)
2775 *qtail
= ++sd
->input_queue_tail
;
2779 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2781 void __netif_schedule(struct Qdisc
*q
);
2782 void netif_schedule_queue(struct netdev_queue
*txq
);
2784 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2788 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2789 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2792 static __always_inline
void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2794 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2798 * netif_start_queue - allow transmit
2799 * @dev: network device
2801 * Allow upper layers to call the device hard_start_xmit routine.
2803 static inline void netif_start_queue(struct net_device
*dev
)
2805 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2808 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2812 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2813 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2814 netif_tx_start_queue(txq
);
2818 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2821 * netif_wake_queue - restart transmit
2822 * @dev: network device
2824 * Allow upper layers to call the device hard_start_xmit routine.
2825 * Used for flow control when transmit resources are available.
2827 static inline void netif_wake_queue(struct net_device
*dev
)
2829 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2832 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2836 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2837 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2838 netif_tx_wake_queue(txq
);
2842 static __always_inline
void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2844 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2848 * netif_stop_queue - stop transmitted packets
2849 * @dev: network device
2851 * Stop upper layers calling the device hard_start_xmit routine.
2852 * Used for flow control when transmit resources are unavailable.
2854 static inline void netif_stop_queue(struct net_device
*dev
)
2856 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2859 void netif_tx_stop_all_queues(struct net_device
*dev
);
2861 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2863 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2867 * netif_queue_stopped - test if transmit queue is flowblocked
2868 * @dev: network device
2870 * Test if transmit queue on device is currently unable to send.
2872 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2874 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2877 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2879 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2883 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2885 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2889 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2891 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2895 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2896 * @dev_queue: pointer to transmit queue
2898 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2899 * to give appropriate hint to the CPU.
2901 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2904 prefetchw(&dev_queue
->dql
.num_queued
);
2909 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2910 * @dev_queue: pointer to transmit queue
2912 * BQL enabled drivers might use this helper in their TX completion path,
2913 * to give appropriate hint to the CPU.
2915 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2918 prefetchw(&dev_queue
->dql
.limit
);
2922 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2926 dql_queued(&dev_queue
->dql
, bytes
);
2928 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2931 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2934 * The XOFF flag must be set before checking the dql_avail below,
2935 * because in netdev_tx_completed_queue we update the dql_completed
2936 * before checking the XOFF flag.
2940 /* check again in case another CPU has just made room avail */
2941 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2942 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2947 * netdev_sent_queue - report the number of bytes queued to hardware
2948 * @dev: network device
2949 * @bytes: number of bytes queued to the hardware device queue
2951 * Report the number of bytes queued for sending/completion to the network
2952 * device hardware queue. @bytes should be a good approximation and should
2953 * exactly match netdev_completed_queue() @bytes
2955 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2957 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2960 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2961 unsigned int pkts
, unsigned int bytes
)
2964 if (unlikely(!bytes
))
2967 dql_completed(&dev_queue
->dql
, bytes
);
2970 * Without the memory barrier there is a small possiblity that
2971 * netdev_tx_sent_queue will miss the update and cause the queue to
2972 * be stopped forever
2976 if (dql_avail(&dev_queue
->dql
) < 0)
2979 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2980 netif_schedule_queue(dev_queue
);
2985 * netdev_completed_queue - report bytes and packets completed by device
2986 * @dev: network device
2987 * @pkts: actual number of packets sent over the medium
2988 * @bytes: actual number of bytes sent over the medium
2990 * Report the number of bytes and packets transmitted by the network device
2991 * hardware queue over the physical medium, @bytes must exactly match the
2992 * @bytes amount passed to netdev_sent_queue()
2994 static inline void netdev_completed_queue(struct net_device
*dev
,
2995 unsigned int pkts
, unsigned int bytes
)
2997 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
3000 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
3003 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
3009 * netdev_reset_queue - reset the packets and bytes count of a network device
3010 * @dev_queue: network device
3012 * Reset the bytes and packet count of a network device and clear the
3013 * software flow control OFF bit for this network device
3015 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
3017 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
3021 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3022 * @dev: network device
3023 * @queue_index: given tx queue index
3025 * Returns 0 if given tx queue index >= number of device tx queues,
3026 * otherwise returns the originally passed tx queue index.
3028 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
3030 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
3031 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3032 dev
->name
, queue_index
,
3033 dev
->real_num_tx_queues
);
3041 * netif_running - test if up
3042 * @dev: network device
3044 * Test if the device has been brought up.
3046 static inline bool netif_running(const struct net_device
*dev
)
3048 return test_bit(__LINK_STATE_START
, &dev
->state
);
3052 * Routines to manage the subqueues on a device. We only need start,
3053 * stop, and a check if it's stopped. All other device management is
3054 * done at the overall netdevice level.
3055 * Also test the device if we're multiqueue.
3059 * netif_start_subqueue - allow sending packets on subqueue
3060 * @dev: network device
3061 * @queue_index: sub queue index
3063 * Start individual transmit queue of a device with multiple transmit queues.
3065 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
3067 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3069 netif_tx_start_queue(txq
);
3073 * netif_stop_subqueue - stop sending packets on subqueue
3074 * @dev: network device
3075 * @queue_index: sub queue index
3077 * Stop individual transmit queue of a device with multiple transmit queues.
3079 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
3081 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3082 netif_tx_stop_queue(txq
);
3086 * netif_subqueue_stopped - test status of subqueue
3087 * @dev: network device
3088 * @queue_index: sub queue index
3090 * Check individual transmit queue of a device with multiple transmit queues.
3092 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
3095 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3097 return netif_tx_queue_stopped(txq
);
3100 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
3101 struct sk_buff
*skb
)
3103 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
3107 * netif_wake_subqueue - allow sending packets on subqueue
3108 * @dev: network device
3109 * @queue_index: sub queue index
3111 * Resume individual transmit queue of a device with multiple transmit queues.
3113 static inline void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
)
3115 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3117 netif_tx_wake_queue(txq
);
3121 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
3124 static inline int netif_set_xps_queue(struct net_device
*dev
,
3125 const struct cpumask
*mask
,
3132 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3133 unsigned int num_tx_queues
);
3136 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3137 * as a distribution range limit for the returned value.
3139 static inline u16
skb_tx_hash(const struct net_device
*dev
,
3140 struct sk_buff
*skb
)
3142 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
3146 * netif_is_multiqueue - test if device has multiple transmit queues
3147 * @dev: network device
3149 * Check if device has multiple transmit queues
3151 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
3153 return dev
->num_tx_queues
> 1;
3156 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
3159 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
3161 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
3169 static inline unsigned int get_netdev_rx_queue_index(
3170 struct netdev_rx_queue
*queue
)
3172 struct net_device
*dev
= queue
->dev
;
3173 int index
= queue
- dev
->_rx
;
3175 BUG_ON(index
>= dev
->num_rx_queues
);
3180 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3181 int netif_get_num_default_rss_queues(void);
3183 enum skb_free_reason
{
3184 SKB_REASON_CONSUMED
,
3188 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
3189 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
3192 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3193 * interrupt context or with hardware interrupts being disabled.
3194 * (in_irq() || irqs_disabled())
3196 * We provide four helpers that can be used in following contexts :
3198 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3199 * replacing kfree_skb(skb)
3201 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3202 * Typically used in place of consume_skb(skb) in TX completion path
3204 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3205 * replacing kfree_skb(skb)
3207 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3208 * and consumed a packet. Used in place of consume_skb(skb)
3210 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
3212 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
3215 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
3217 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
3220 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
3222 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
3225 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
3227 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
3230 int netif_rx(struct sk_buff
*skb
);
3231 int netif_rx_ni(struct sk_buff
*skb
);
3232 int netif_receive_skb(struct sk_buff
*skb
);
3233 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
3234 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
3235 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
3236 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
3237 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
3238 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
3240 static inline void napi_free_frags(struct napi_struct
*napi
)
3242 kfree_skb(napi
->skb
);
3246 bool netdev_is_rx_handler_busy(struct net_device
*dev
);
3247 int netdev_rx_handler_register(struct net_device
*dev
,
3248 rx_handler_func_t
*rx_handler
,
3249 void *rx_handler_data
);
3250 void netdev_rx_handler_unregister(struct net_device
*dev
);
3252 bool dev_valid_name(const char *name
);
3253 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
3254 int dev_ethtool(struct net
*net
, struct ifreq
*);
3255 unsigned int dev_get_flags(const struct net_device
*);
3256 int __dev_change_flags(struct net_device
*, unsigned int flags
);
3257 int dev_change_flags(struct net_device
*, unsigned int);
3258 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
3259 unsigned int gchanges
);
3260 int dev_change_name(struct net_device
*, const char *);
3261 int dev_set_alias(struct net_device
*, const char *, size_t);
3262 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
3263 int dev_set_mtu(struct net_device
*, int);
3264 void dev_set_group(struct net_device
*, int);
3265 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
3266 int dev_change_carrier(struct net_device
*, bool new_carrier
);
3267 int dev_get_phys_port_id(struct net_device
*dev
,
3268 struct netdev_phys_item_id
*ppid
);
3269 int dev_get_phys_port_name(struct net_device
*dev
,
3270 char *name
, size_t len
);
3271 int dev_change_proto_down(struct net_device
*dev
, bool proto_down
);
3272 int dev_change_xdp_fd(struct net_device
*dev
, int fd
, u32 flags
);
3273 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
3274 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3275 struct netdev_queue
*txq
, int *ret
);
3276 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3277 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3278 bool is_skb_forwardable(const struct net_device
*dev
,
3279 const struct sk_buff
*skb
);
3281 static __always_inline
int ____dev_forward_skb(struct net_device
*dev
,
3282 struct sk_buff
*skb
)
3284 if (skb_orphan_frags(skb
, GFP_ATOMIC
) ||
3285 unlikely(!is_skb_forwardable(dev
, skb
))) {
3286 atomic_long_inc(&dev
->rx_dropped
);
3291 skb_scrub_packet(skb
, true);
3296 void dev_queue_xmit_nit(struct sk_buff
*skb
, struct net_device
*dev
);
3298 extern int netdev_budget
;
3300 /* Called by rtnetlink.c:rtnl_unlock() */
3301 void netdev_run_todo(void);
3304 * dev_put - release reference to device
3305 * @dev: network device
3307 * Release reference to device to allow it to be freed.
3309 static inline void dev_put(struct net_device
*dev
)
3311 this_cpu_dec(*dev
->pcpu_refcnt
);
3315 * dev_hold - get reference to device
3316 * @dev: network device
3318 * Hold reference to device to keep it from being freed.
3320 static inline void dev_hold(struct net_device
*dev
)
3322 this_cpu_inc(*dev
->pcpu_refcnt
);
3325 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3326 * and _off may be called from IRQ context, but it is caller
3327 * who is responsible for serialization of these calls.
3329 * The name carrier is inappropriate, these functions should really be
3330 * called netif_lowerlayer_*() because they represent the state of any
3331 * kind of lower layer not just hardware media.
3334 void linkwatch_init_dev(struct net_device
*dev
);
3335 void linkwatch_fire_event(struct net_device
*dev
);
3336 void linkwatch_forget_dev(struct net_device
*dev
);
3339 * netif_carrier_ok - test if carrier present
3340 * @dev: network device
3342 * Check if carrier is present on device
3344 static inline bool netif_carrier_ok(const struct net_device
*dev
)
3346 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
3349 unsigned long dev_trans_start(struct net_device
*dev
);
3351 void __netdev_watchdog_up(struct net_device
*dev
);
3353 void netif_carrier_on(struct net_device
*dev
);
3355 void netif_carrier_off(struct net_device
*dev
);
3358 * netif_dormant_on - mark device as dormant.
3359 * @dev: network device
3361 * Mark device as dormant (as per RFC2863).
3363 * The dormant state indicates that the relevant interface is not
3364 * actually in a condition to pass packets (i.e., it is not 'up') but is
3365 * in a "pending" state, waiting for some external event. For "on-
3366 * demand" interfaces, this new state identifies the situation where the
3367 * interface is waiting for events to place it in the up state.
3369 static inline void netif_dormant_on(struct net_device
*dev
)
3371 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3372 linkwatch_fire_event(dev
);
3376 * netif_dormant_off - set device as not dormant.
3377 * @dev: network device
3379 * Device is not in dormant state.
3381 static inline void netif_dormant_off(struct net_device
*dev
)
3383 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3384 linkwatch_fire_event(dev
);
3388 * netif_dormant - test if carrier present
3389 * @dev: network device
3391 * Check if carrier is present on device
3393 static inline bool netif_dormant(const struct net_device
*dev
)
3395 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3400 * netif_oper_up - test if device is operational
3401 * @dev: network device
3403 * Check if carrier is operational
3405 static inline bool netif_oper_up(const struct net_device
*dev
)
3407 return (dev
->operstate
== IF_OPER_UP
||
3408 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3412 * netif_device_present - is device available or removed
3413 * @dev: network device
3415 * Check if device has not been removed from system.
3417 static inline bool netif_device_present(struct net_device
*dev
)
3419 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3422 void netif_device_detach(struct net_device
*dev
);
3424 void netif_device_attach(struct net_device
*dev
);
3427 * Network interface message level settings
3431 NETIF_MSG_DRV
= 0x0001,
3432 NETIF_MSG_PROBE
= 0x0002,
3433 NETIF_MSG_LINK
= 0x0004,
3434 NETIF_MSG_TIMER
= 0x0008,
3435 NETIF_MSG_IFDOWN
= 0x0010,
3436 NETIF_MSG_IFUP
= 0x0020,
3437 NETIF_MSG_RX_ERR
= 0x0040,
3438 NETIF_MSG_TX_ERR
= 0x0080,
3439 NETIF_MSG_TX_QUEUED
= 0x0100,
3440 NETIF_MSG_INTR
= 0x0200,
3441 NETIF_MSG_TX_DONE
= 0x0400,
3442 NETIF_MSG_RX_STATUS
= 0x0800,
3443 NETIF_MSG_PKTDATA
= 0x1000,
3444 NETIF_MSG_HW
= 0x2000,
3445 NETIF_MSG_WOL
= 0x4000,
3448 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3449 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3450 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3451 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3452 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3453 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3454 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3455 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3456 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3457 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3458 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3459 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3460 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3461 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3462 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3464 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3467 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3468 return default_msg_enable_bits
;
3469 if (debug_value
== 0) /* no output */
3471 /* set low N bits */
3472 return (1 << debug_value
) - 1;
3475 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3477 spin_lock(&txq
->_xmit_lock
);
3478 txq
->xmit_lock_owner
= cpu
;
3481 static inline bool __netif_tx_acquire(struct netdev_queue
*txq
)
3483 __acquire(&txq
->_xmit_lock
);
3487 static inline void __netif_tx_release(struct netdev_queue
*txq
)
3489 __release(&txq
->_xmit_lock
);
3492 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3494 spin_lock_bh(&txq
->_xmit_lock
);
3495 txq
->xmit_lock_owner
= smp_processor_id();
3498 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3500 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3502 txq
->xmit_lock_owner
= smp_processor_id();
3506 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3508 txq
->xmit_lock_owner
= -1;
3509 spin_unlock(&txq
->_xmit_lock
);
3512 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3514 txq
->xmit_lock_owner
= -1;
3515 spin_unlock_bh(&txq
->_xmit_lock
);
3518 static inline void txq_trans_update(struct netdev_queue
*txq
)
3520 if (txq
->xmit_lock_owner
!= -1)
3521 txq
->trans_start
= jiffies
;
3524 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3525 static inline void netif_trans_update(struct net_device
*dev
)
3527 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, 0);
3529 if (txq
->trans_start
!= jiffies
)
3530 txq
->trans_start
= jiffies
;
3534 * netif_tx_lock - grab network device transmit lock
3535 * @dev: network device
3537 * Get network device transmit lock
3539 static inline void netif_tx_lock(struct net_device
*dev
)
3544 spin_lock(&dev
->tx_global_lock
);
3545 cpu
= smp_processor_id();
3546 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3547 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3549 /* We are the only thread of execution doing a
3550 * freeze, but we have to grab the _xmit_lock in
3551 * order to synchronize with threads which are in
3552 * the ->hard_start_xmit() handler and already
3553 * checked the frozen bit.
3555 __netif_tx_lock(txq
, cpu
);
3556 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3557 __netif_tx_unlock(txq
);
3561 static inline void netif_tx_lock_bh(struct net_device
*dev
)
3567 static inline void netif_tx_unlock(struct net_device
*dev
)
3571 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3572 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3574 /* No need to grab the _xmit_lock here. If the
3575 * queue is not stopped for another reason, we
3578 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3579 netif_schedule_queue(txq
);
3581 spin_unlock(&dev
->tx_global_lock
);
3584 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
3586 netif_tx_unlock(dev
);
3590 #define HARD_TX_LOCK(dev, txq, cpu) { \
3591 if ((dev->features & NETIF_F_LLTX) == 0) { \
3592 __netif_tx_lock(txq, cpu); \
3594 __netif_tx_acquire(txq); \
3598 #define HARD_TX_TRYLOCK(dev, txq) \
3599 (((dev->features & NETIF_F_LLTX) == 0) ? \
3600 __netif_tx_trylock(txq) : \
3601 __netif_tx_acquire(txq))
3603 #define HARD_TX_UNLOCK(dev, txq) { \
3604 if ((dev->features & NETIF_F_LLTX) == 0) { \
3605 __netif_tx_unlock(txq); \
3607 __netif_tx_release(txq); \
3611 static inline void netif_tx_disable(struct net_device
*dev
)
3617 cpu
= smp_processor_id();
3618 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3619 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3621 __netif_tx_lock(txq
, cpu
);
3622 netif_tx_stop_queue(txq
);
3623 __netif_tx_unlock(txq
);
3628 static inline void netif_addr_lock(struct net_device
*dev
)
3630 spin_lock(&dev
->addr_list_lock
);
3633 static inline void netif_addr_lock_nested(struct net_device
*dev
)
3635 int subclass
= SINGLE_DEPTH_NESTING
;
3637 if (dev
->netdev_ops
->ndo_get_lock_subclass
)
3638 subclass
= dev
->netdev_ops
->ndo_get_lock_subclass(dev
);
3640 spin_lock_nested(&dev
->addr_list_lock
, subclass
);
3643 static inline void netif_addr_lock_bh(struct net_device
*dev
)
3645 spin_lock_bh(&dev
->addr_list_lock
);
3648 static inline void netif_addr_unlock(struct net_device
*dev
)
3650 spin_unlock(&dev
->addr_list_lock
);
3653 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
3655 spin_unlock_bh(&dev
->addr_list_lock
);
3659 * dev_addrs walker. Should be used only for read access. Call with
3660 * rcu_read_lock held.
3662 #define for_each_dev_addr(dev, ha) \
3663 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3665 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3667 void ether_setup(struct net_device
*dev
);
3669 /* Support for loadable net-drivers */
3670 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
3671 unsigned char name_assign_type
,
3672 void (*setup
)(struct net_device
*),
3673 unsigned int txqs
, unsigned int rxqs
);
3674 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3675 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3677 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3678 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3681 int register_netdev(struct net_device
*dev
);
3682 void unregister_netdev(struct net_device
*dev
);
3684 /* General hardware address lists handling functions */
3685 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
3686 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3687 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
3688 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3689 int __hw_addr_sync_dev(struct netdev_hw_addr_list
*list
,
3690 struct net_device
*dev
,
3691 int (*sync
)(struct net_device
*, const unsigned char *),
3692 int (*unsync
)(struct net_device
*,
3693 const unsigned char *));
3694 void __hw_addr_unsync_dev(struct netdev_hw_addr_list
*list
,
3695 struct net_device
*dev
,
3696 int (*unsync
)(struct net_device
*,
3697 const unsigned char *));
3698 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
3700 /* Functions used for device addresses handling */
3701 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
3702 unsigned char addr_type
);
3703 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
3704 unsigned char addr_type
);
3705 void dev_addr_flush(struct net_device
*dev
);
3706 int dev_addr_init(struct net_device
*dev
);
3708 /* Functions used for unicast addresses handling */
3709 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
3710 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3711 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
3712 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
3713 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3714 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
3715 void dev_uc_flush(struct net_device
*dev
);
3716 void dev_uc_init(struct net_device
*dev
);
3719 * __dev_uc_sync - Synchonize device's unicast list
3720 * @dev: device to sync
3721 * @sync: function to call if address should be added
3722 * @unsync: function to call if address should be removed
3724 * Add newly added addresses to the interface, and release
3725 * addresses that have been deleted.
3727 static inline int __dev_uc_sync(struct net_device
*dev
,
3728 int (*sync
)(struct net_device
*,
3729 const unsigned char *),
3730 int (*unsync
)(struct net_device
*,
3731 const unsigned char *))
3733 return __hw_addr_sync_dev(&dev
->uc
, dev
, sync
, unsync
);
3737 * __dev_uc_unsync - Remove synchronized addresses from device
3738 * @dev: device to sync
3739 * @unsync: function to call if address should be removed
3741 * Remove all addresses that were added to the device by dev_uc_sync().
3743 static inline void __dev_uc_unsync(struct net_device
*dev
,
3744 int (*unsync
)(struct net_device
*,
3745 const unsigned char *))
3747 __hw_addr_unsync_dev(&dev
->uc
, dev
, unsync
);
3750 /* Functions used for multicast addresses handling */
3751 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
3752 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
3753 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3754 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
3755 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
3756 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
3757 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3758 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
3759 void dev_mc_flush(struct net_device
*dev
);
3760 void dev_mc_init(struct net_device
*dev
);
3763 * __dev_mc_sync - Synchonize device's multicast list
3764 * @dev: device to sync
3765 * @sync: function to call if address should be added
3766 * @unsync: function to call if address should be removed
3768 * Add newly added addresses to the interface, and release
3769 * addresses that have been deleted.
3771 static inline int __dev_mc_sync(struct net_device
*dev
,
3772 int (*sync
)(struct net_device
*,
3773 const unsigned char *),
3774 int (*unsync
)(struct net_device
*,
3775 const unsigned char *))
3777 return __hw_addr_sync_dev(&dev
->mc
, dev
, sync
, unsync
);
3781 * __dev_mc_unsync - Remove synchronized addresses from device
3782 * @dev: device to sync
3783 * @unsync: function to call if address should be removed
3785 * Remove all addresses that were added to the device by dev_mc_sync().
3787 static inline void __dev_mc_unsync(struct net_device
*dev
,
3788 int (*unsync
)(struct net_device
*,
3789 const unsigned char *))
3791 __hw_addr_unsync_dev(&dev
->mc
, dev
, unsync
);
3794 /* Functions used for secondary unicast and multicast support */
3795 void dev_set_rx_mode(struct net_device
*dev
);
3796 void __dev_set_rx_mode(struct net_device
*dev
);
3797 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
3798 int dev_set_allmulti(struct net_device
*dev
, int inc
);
3799 void netdev_state_change(struct net_device
*dev
);
3800 void netdev_notify_peers(struct net_device
*dev
);
3801 void netdev_features_change(struct net_device
*dev
);
3802 /* Load a device via the kmod */
3803 void dev_load(struct net
*net
, const char *name
);
3804 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
3805 struct rtnl_link_stats64
*storage
);
3806 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
3807 const struct net_device_stats
*netdev_stats
);
3809 extern int netdev_max_backlog
;
3810 extern int netdev_tstamp_prequeue
;
3811 extern int weight_p
;
3812 extern int dev_weight_rx_bias
;
3813 extern int dev_weight_tx_bias
;
3814 extern int dev_rx_weight
;
3815 extern int dev_tx_weight
;
3817 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3818 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3819 struct list_head
**iter
);
3820 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3821 struct list_head
**iter
);
3823 /* iterate through upper list, must be called under RCU read lock */
3824 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3825 for (iter = &(dev)->adj_list.upper, \
3826 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3828 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3830 int netdev_walk_all_upper_dev_rcu(struct net_device
*dev
,
3831 int (*fn
)(struct net_device
*upper_dev
,
3835 bool netdev_has_upper_dev_all_rcu(struct net_device
*dev
,
3836 struct net_device
*upper_dev
);
3838 void *netdev_lower_get_next_private(struct net_device
*dev
,
3839 struct list_head
**iter
);
3840 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3841 struct list_head
**iter
);
3843 #define netdev_for_each_lower_private(dev, priv, iter) \
3844 for (iter = (dev)->adj_list.lower.next, \
3845 priv = netdev_lower_get_next_private(dev, &(iter)); \
3847 priv = netdev_lower_get_next_private(dev, &(iter)))
3849 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3850 for (iter = &(dev)->adj_list.lower, \
3851 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3853 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3855 void *netdev_lower_get_next(struct net_device
*dev
,
3856 struct list_head
**iter
);
3858 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3859 for (iter = (dev)->adj_list.lower.next, \
3860 ldev = netdev_lower_get_next(dev, &(iter)); \
3862 ldev = netdev_lower_get_next(dev, &(iter)))
3864 struct net_device
*netdev_all_lower_get_next(struct net_device
*dev
,
3865 struct list_head
**iter
);
3866 struct net_device
*netdev_all_lower_get_next_rcu(struct net_device
*dev
,
3867 struct list_head
**iter
);
3869 int netdev_walk_all_lower_dev(struct net_device
*dev
,
3870 int (*fn
)(struct net_device
*lower_dev
,
3873 int netdev_walk_all_lower_dev_rcu(struct net_device
*dev
,
3874 int (*fn
)(struct net_device
*lower_dev
,
3878 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3879 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3880 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3881 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3882 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
3883 int netdev_master_upper_dev_link(struct net_device
*dev
,
3884 struct net_device
*upper_dev
,
3885 void *upper_priv
, void *upper_info
);
3886 void netdev_upper_dev_unlink(struct net_device
*dev
,
3887 struct net_device
*upper_dev
);
3888 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3889 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3890 struct net_device
*lower_dev
);
3891 void netdev_lower_state_changed(struct net_device
*lower_dev
,
3892 void *lower_state_info
);
3894 /* RSS keys are 40 or 52 bytes long */
3895 #define NETDEV_RSS_KEY_LEN 52
3896 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
] __read_mostly
;
3897 void netdev_rss_key_fill(void *buffer
, size_t len
);
3899 int dev_get_nest_level(struct net_device
*dev
);
3900 int skb_checksum_help(struct sk_buff
*skb
);
3901 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3902 netdev_features_t features
, bool tx_path
);
3903 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3904 netdev_features_t features
);
3906 struct netdev_bonding_info
{
3911 struct netdev_notifier_bonding_info
{
3912 struct netdev_notifier_info info
; /* must be first */
3913 struct netdev_bonding_info bonding_info
;
3916 void netdev_bonding_info_change(struct net_device
*dev
,
3917 struct netdev_bonding_info
*bonding_info
);
3920 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3922 return __skb_gso_segment(skb
, features
, true);
3924 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3926 static inline bool can_checksum_protocol(netdev_features_t features
,
3929 if (protocol
== htons(ETH_P_FCOE
))
3930 return !!(features
& NETIF_F_FCOE_CRC
);
3932 /* Assume this is an IP checksum (not SCTP CRC) */
3934 if (features
& NETIF_F_HW_CSUM
) {
3935 /* Can checksum everything */
3940 case htons(ETH_P_IP
):
3941 return !!(features
& NETIF_F_IP_CSUM
);
3942 case htons(ETH_P_IPV6
):
3943 return !!(features
& NETIF_F_IPV6_CSUM
);
3950 void netdev_rx_csum_fault(struct net_device
*dev
);
3952 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3956 /* rx skb timestamps */
3957 void net_enable_timestamp(void);
3958 void net_disable_timestamp(void);
3960 #ifdef CONFIG_PROC_FS
3961 int __init
dev_proc_init(void);
3963 #define dev_proc_init() 0
3966 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
3967 struct sk_buff
*skb
, struct net_device
*dev
,
3970 skb
->xmit_more
= more
? 1 : 0;
3971 return ops
->ndo_start_xmit(skb
, dev
);
3974 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3975 struct netdev_queue
*txq
, bool more
)
3977 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3980 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
3981 if (rc
== NETDEV_TX_OK
)
3982 txq_trans_update(txq
);
3987 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3989 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3992 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3994 return netdev_class_create_file_ns(class_attr
, NULL
);
3997 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3999 netdev_class_remove_file_ns(class_attr
, NULL
);
4002 extern struct kobj_ns_type_operations net_ns_type_operations
;
4004 const char *netdev_drivername(const struct net_device
*dev
);
4006 void linkwatch_run_queue(void);
4008 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
4009 netdev_features_t f2
)
4011 if ((f1
^ f2
) & NETIF_F_HW_CSUM
) {
4012 if (f1
& NETIF_F_HW_CSUM
)
4013 f1
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
4015 f2
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
4021 static inline netdev_features_t
netdev_get_wanted_features(
4022 struct net_device
*dev
)
4024 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
4026 netdev_features_t
netdev_increment_features(netdev_features_t all
,
4027 netdev_features_t one
, netdev_features_t mask
);
4029 /* Allow TSO being used on stacked device :
4030 * Performing the GSO segmentation before last device
4031 * is a performance improvement.
4033 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
4034 netdev_features_t mask
)
4036 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
4039 int __netdev_update_features(struct net_device
*dev
);
4040 void netdev_update_features(struct net_device
*dev
);
4041 void netdev_change_features(struct net_device
*dev
);
4043 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
4044 struct net_device
*dev
);
4046 netdev_features_t
passthru_features_check(struct sk_buff
*skb
,
4047 struct net_device
*dev
,
4048 netdev_features_t features
);
4049 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
4051 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
4053 netdev_features_t feature
= (netdev_features_t
)gso_type
<< NETIF_F_GSO_SHIFT
;
4055 /* check flags correspondence */
4056 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
4057 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
4058 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
4059 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
4060 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID
!= (NETIF_F_TSO_MANGLEID
>> NETIF_F_GSO_SHIFT
));
4061 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
4062 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
4063 BUILD_BUG_ON(SKB_GSO_GRE
!= (NETIF_F_GSO_GRE
>> NETIF_F_GSO_SHIFT
));
4064 BUILD_BUG_ON(SKB_GSO_GRE_CSUM
!= (NETIF_F_GSO_GRE_CSUM
>> NETIF_F_GSO_SHIFT
));
4065 BUILD_BUG_ON(SKB_GSO_IPXIP4
!= (NETIF_F_GSO_IPXIP4
>> NETIF_F_GSO_SHIFT
));
4066 BUILD_BUG_ON(SKB_GSO_IPXIP6
!= (NETIF_F_GSO_IPXIP6
>> NETIF_F_GSO_SHIFT
));
4067 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL
!= (NETIF_F_GSO_UDP_TUNNEL
>> NETIF_F_GSO_SHIFT
));
4068 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM
!= (NETIF_F_GSO_UDP_TUNNEL_CSUM
>> NETIF_F_GSO_SHIFT
));
4069 BUILD_BUG_ON(SKB_GSO_PARTIAL
!= (NETIF_F_GSO_PARTIAL
>> NETIF_F_GSO_SHIFT
));
4070 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
4071 BUILD_BUG_ON(SKB_GSO_SCTP
!= (NETIF_F_GSO_SCTP
>> NETIF_F_GSO_SHIFT
));
4073 return (features
& feature
) == feature
;
4076 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
4078 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
4079 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
4082 static inline bool netif_needs_gso(struct sk_buff
*skb
,
4083 netdev_features_t features
)
4085 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
4086 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
4087 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
4090 static inline void netif_set_gso_max_size(struct net_device
*dev
,
4093 dev
->gso_max_size
= size
;
4096 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
4097 int pulled_hlen
, u16 mac_offset
,
4100 skb
->protocol
= protocol
;
4101 skb
->encapsulation
= 1;
4102 skb_push(skb
, pulled_hlen
);
4103 skb_reset_transport_header(skb
);
4104 skb
->mac_header
= mac_offset
;
4105 skb
->network_header
= skb
->mac_header
+ mac_len
;
4106 skb
->mac_len
= mac_len
;
4109 static inline bool netif_is_macsec(const struct net_device
*dev
)
4111 return dev
->priv_flags
& IFF_MACSEC
;
4114 static inline bool netif_is_macvlan(const struct net_device
*dev
)
4116 return dev
->priv_flags
& IFF_MACVLAN
;
4119 static inline bool netif_is_macvlan_port(const struct net_device
*dev
)
4121 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
4124 static inline bool netif_is_ipvlan(const struct net_device
*dev
)
4126 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
4129 static inline bool netif_is_ipvlan_port(const struct net_device
*dev
)
4131 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
4134 static inline bool netif_is_bond_master(const struct net_device
*dev
)
4136 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
4139 static inline bool netif_is_bond_slave(const struct net_device
*dev
)
4141 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
4144 static inline bool netif_supports_nofcs(struct net_device
*dev
)
4146 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
4149 static inline bool netif_is_l3_master(const struct net_device
*dev
)
4151 return dev
->priv_flags
& IFF_L3MDEV_MASTER
;
4154 static inline bool netif_is_l3_slave(const struct net_device
*dev
)
4156 return dev
->priv_flags
& IFF_L3MDEV_SLAVE
;
4159 static inline bool netif_is_bridge_master(const struct net_device
*dev
)
4161 return dev
->priv_flags
& IFF_EBRIDGE
;
4164 static inline bool netif_is_bridge_port(const struct net_device
*dev
)
4166 return dev
->priv_flags
& IFF_BRIDGE_PORT
;
4169 static inline bool netif_is_ovs_master(const struct net_device
*dev
)
4171 return dev
->priv_flags
& IFF_OPENVSWITCH
;
4174 static inline bool netif_is_ovs_port(const struct net_device
*dev
)
4176 return dev
->priv_flags
& IFF_OVS_DATAPATH
;
4179 static inline bool netif_is_team_master(const struct net_device
*dev
)
4181 return dev
->priv_flags
& IFF_TEAM
;
4184 static inline bool netif_is_team_port(const struct net_device
*dev
)
4186 return dev
->priv_flags
& IFF_TEAM_PORT
;
4189 static inline bool netif_is_lag_master(const struct net_device
*dev
)
4191 return netif_is_bond_master(dev
) || netif_is_team_master(dev
);
4194 static inline bool netif_is_lag_port(const struct net_device
*dev
)
4196 return netif_is_bond_slave(dev
) || netif_is_team_port(dev
);
4199 static inline bool netif_is_rxfh_configured(const struct net_device
*dev
)
4201 return dev
->priv_flags
& IFF_RXFH_CONFIGURED
;
4204 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4205 static inline void netif_keep_dst(struct net_device
*dev
)
4207 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
4210 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4211 static inline bool netif_reduces_vlan_mtu(struct net_device
*dev
)
4213 /* TODO: reserve and use an additional IFF bit, if we get more users */
4214 return dev
->priv_flags
& IFF_MACSEC
;
4217 extern struct pernet_operations __net_initdata loopback_net_ops
;
4219 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4221 /* netdev_printk helpers, similar to dev_printk */
4223 static inline const char *netdev_name(const struct net_device
*dev
)
4225 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
4226 return "(unnamed net_device)";
4230 static inline const char *netdev_reg_state(const struct net_device
*dev
)
4232 switch (dev
->reg_state
) {
4233 case NETREG_UNINITIALIZED
: return " (uninitialized)";
4234 case NETREG_REGISTERED
: return "";
4235 case NETREG_UNREGISTERING
: return " (unregistering)";
4236 case NETREG_UNREGISTERED
: return " (unregistered)";
4237 case NETREG_RELEASED
: return " (released)";
4238 case NETREG_DUMMY
: return " (dummy)";
4241 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
4242 return " (unknown)";
4246 void netdev_printk(const char *level
, const struct net_device
*dev
,
4247 const char *format
, ...);
4249 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
4251 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
4253 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
4255 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
4257 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
4259 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
4261 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
4263 #define MODULE_ALIAS_NETDEV(device) \
4264 MODULE_ALIAS("netdev-" device)
4266 #if defined(CONFIG_DYNAMIC_DEBUG)
4267 #define netdev_dbg(__dev, format, args...) \
4269 dynamic_netdev_dbg(__dev, format, ##args); \
4271 #elif defined(DEBUG)
4272 #define netdev_dbg(__dev, format, args...) \
4273 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4275 #define netdev_dbg(__dev, format, args...) \
4278 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4282 #if defined(VERBOSE_DEBUG)
4283 #define netdev_vdbg netdev_dbg
4286 #define netdev_vdbg(dev, format, args...) \
4289 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4295 * netdev_WARN() acts like dev_printk(), but with the key difference
4296 * of using a WARN/WARN_ON to get the message out, including the
4297 * file/line information and a backtrace.
4299 #define netdev_WARN(dev, format, args...) \
4300 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4301 netdev_reg_state(dev), ##args)
4303 /* netif printk helpers, similar to netdev_printk */
4305 #define netif_printk(priv, type, level, dev, fmt, args...) \
4307 if (netif_msg_##type(priv)) \
4308 netdev_printk(level, (dev), fmt, ##args); \
4311 #define netif_level(level, priv, type, dev, fmt, args...) \
4313 if (netif_msg_##type(priv)) \
4314 netdev_##level(dev, fmt, ##args); \
4317 #define netif_emerg(priv, type, dev, fmt, args...) \
4318 netif_level(emerg, priv, type, dev, fmt, ##args)
4319 #define netif_alert(priv, type, dev, fmt, args...) \
4320 netif_level(alert, priv, type, dev, fmt, ##args)
4321 #define netif_crit(priv, type, dev, fmt, args...) \
4322 netif_level(crit, priv, type, dev, fmt, ##args)
4323 #define netif_err(priv, type, dev, fmt, args...) \
4324 netif_level(err, priv, type, dev, fmt, ##args)
4325 #define netif_warn(priv, type, dev, fmt, args...) \
4326 netif_level(warn, priv, type, dev, fmt, ##args)
4327 #define netif_notice(priv, type, dev, fmt, args...) \
4328 netif_level(notice, priv, type, dev, fmt, ##args)
4329 #define netif_info(priv, type, dev, fmt, args...) \
4330 netif_level(info, priv, type, dev, fmt, ##args)
4332 #if defined(CONFIG_DYNAMIC_DEBUG)
4333 #define netif_dbg(priv, type, netdev, format, args...) \
4335 if (netif_msg_##type(priv)) \
4336 dynamic_netdev_dbg(netdev, format, ##args); \
4338 #elif defined(DEBUG)
4339 #define netif_dbg(priv, type, dev, format, args...) \
4340 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4342 #define netif_dbg(priv, type, dev, format, args...) \
4345 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4350 /* if @cond then downgrade to debug, else print at @level */
4351 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4354 netif_dbg(priv, type, netdev, fmt, ##args); \
4356 netif_ ## level(priv, type, netdev, fmt, ##args); \
4359 #if defined(VERBOSE_DEBUG)
4360 #define netif_vdbg netif_dbg
4362 #define netif_vdbg(priv, type, dev, format, args...) \
4365 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4371 * The list of packet types we will receive (as opposed to discard)
4372 * and the routines to invoke.
4374 * Why 16. Because with 16 the only overlap we get on a hash of the
4375 * low nibble of the protocol value is RARP/SNAP/X.25.
4377 * NOTE: That is no longer true with the addition of VLAN tags. Not
4378 * sure which should go first, but I bet it won't make much
4379 * difference if we are running VLANs. The good news is that
4380 * this protocol won't be in the list unless compiled in, so
4381 * the average user (w/out VLANs) will not be adversely affected.
4397 #define PTYPE_HASH_SIZE (16)
4398 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4400 #endif /* _LINUX_NETDEVICE_H */