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>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
61 /* 802.15.4 specific */
65 void netdev_set_default_ethtool_ops(struct net_device
*dev
,
66 const struct ethtool_ops
*ops
);
68 /* Backlog congestion levels */
69 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70 #define NET_RX_DROP 1 /* packet dropped */
73 * Transmit return codes: transmit return codes originate from three different
76 * - qdisc return codes
77 * - driver transmit return codes
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
84 * higher layers. Virtual network devices transmit synchronously; in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
86 * others are propagated to higher layers.
89 /* qdisc ->enqueue() return codes. */
90 #define NET_XMIT_SUCCESS 0x00
91 #define NET_XMIT_DROP 0x01 /* skb dropped */
92 #define NET_XMIT_CN 0x02 /* congestion notification */
93 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
96 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
99 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
100 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
102 /* Driver transmit return codes */
103 #define NETDEV_TX_MASK 0xf0
106 __NETDEV_TX_MIN
= INT_MIN
, /* make sure enum is signed */
107 NETDEV_TX_OK
= 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY
= 0x10, /* driver tx path was busy*/
110 typedef enum netdev_tx netdev_tx_t
;
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 static inline bool dev_xmit_complete(int rc
)
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 if (likely(rc
< NET_XMIT_MASK
))
131 * Compute the worst-case header length according to the protocols
135 #if defined(CONFIG_HYPERV_NET)
136 # define LL_MAX_HEADER 128
137 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
138 # if defined(CONFIG_MAC80211_MESH)
139 # define LL_MAX_HEADER 128
141 # define LL_MAX_HEADER 96
144 # define LL_MAX_HEADER 32
147 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
148 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
149 #define MAX_HEADER LL_MAX_HEADER
151 #define MAX_HEADER (LL_MAX_HEADER + 48)
155 * Old network device statistics. Fields are native words
156 * (unsigned long) so they can be read and written atomically.
159 struct net_device_stats
{
160 unsigned long rx_packets
;
161 unsigned long tx_packets
;
162 unsigned long rx_bytes
;
163 unsigned long tx_bytes
;
164 unsigned long rx_errors
;
165 unsigned long tx_errors
;
166 unsigned long rx_dropped
;
167 unsigned long tx_dropped
;
168 unsigned long multicast
;
169 unsigned long collisions
;
170 unsigned long rx_length_errors
;
171 unsigned long rx_over_errors
;
172 unsigned long rx_crc_errors
;
173 unsigned long rx_frame_errors
;
174 unsigned long rx_fifo_errors
;
175 unsigned long rx_missed_errors
;
176 unsigned long tx_aborted_errors
;
177 unsigned long tx_carrier_errors
;
178 unsigned long tx_fifo_errors
;
179 unsigned long tx_heartbeat_errors
;
180 unsigned long tx_window_errors
;
181 unsigned long rx_compressed
;
182 unsigned long tx_compressed
;
186 #include <linux/cache.h>
187 #include <linux/skbuff.h>
190 #include <linux/static_key.h>
191 extern struct static_key rps_needed
;
198 struct netdev_hw_addr
{
199 struct list_head list
;
200 unsigned char addr
[MAX_ADDR_LEN
];
202 #define NETDEV_HW_ADDR_T_LAN 1
203 #define NETDEV_HW_ADDR_T_SAN 2
204 #define NETDEV_HW_ADDR_T_SLAVE 3
205 #define NETDEV_HW_ADDR_T_UNICAST 4
206 #define NETDEV_HW_ADDR_T_MULTICAST 5
211 struct rcu_head rcu_head
;
214 struct netdev_hw_addr_list
{
215 struct list_head list
;
219 #define netdev_hw_addr_list_count(l) ((l)->count)
220 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
221 #define netdev_hw_addr_list_for_each(ha, l) \
222 list_for_each_entry(ha, &(l)->list, list)
224 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
225 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
226 #define netdev_for_each_uc_addr(ha, dev) \
227 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
229 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
230 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
231 #define netdev_for_each_mc_addr(ha, dev) \
232 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
239 /* cached hardware header; allow for machine alignment needs. */
240 #define HH_DATA_MOD 16
241 #define HH_DATA_OFF(__len) \
242 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
243 #define HH_DATA_ALIGN(__len) \
244 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
245 unsigned long hh_data
[HH_DATA_ALIGN(LL_MAX_HEADER
) / sizeof(long)];
248 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
250 * dev->hard_header_len ? (dev->hard_header_len +
251 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253 * We could use other alignment values, but we must maintain the
254 * relationship HH alignment <= LL alignment.
256 #define LL_RESERVED_SPACE(dev) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
259 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 int (*create
) (struct sk_buff
*skb
, struct net_device
*dev
,
263 unsigned short type
, const void *daddr
,
264 const void *saddr
, unsigned int len
);
265 int (*parse
)(const struct sk_buff
*skb
, unsigned char *haddr
);
266 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
267 void (*cache_update
)(struct hh_cache
*hh
,
268 const struct net_device
*dev
,
269 const unsigned char *haddr
);
270 bool (*validate
)(const char *ll_header
, unsigned int len
);
273 /* These flag bits are private to the generic network queueing
274 * layer; they may not be explicitly referenced by any other
278 enum netdev_state_t
{
280 __LINK_STATE_PRESENT
,
281 __LINK_STATE_NOCARRIER
,
282 __LINK_STATE_LINKWATCH_PENDING
,
283 __LINK_STATE_DORMANT
,
288 * This structure holds boot-time configured netdevice settings. They
289 * are then used in the device probing.
291 struct netdev_boot_setup
{
295 #define NETDEV_BOOT_SETUP_MAX 8
297 int __init
netdev_boot_setup(char *str
);
300 * Structure for NAPI scheduling similar to tasklet but with weighting
303 /* The poll_list must only be managed by the entity which
304 * changes the state of the NAPI_STATE_SCHED bit. This means
305 * whoever atomically sets that bit can add this napi_struct
306 * to the per-CPU poll_list, and whoever clears that bit
307 * can remove from the list right before clearing the bit.
309 struct list_head poll_list
;
313 unsigned int gro_count
;
314 int (*poll
)(struct napi_struct
*, int);
315 #ifdef CONFIG_NETPOLL
316 spinlock_t poll_lock
;
319 struct net_device
*dev
;
320 struct sk_buff
*gro_list
;
322 struct hrtimer timer
;
323 struct list_head dev_list
;
324 struct hlist_node napi_hash_node
;
325 unsigned int napi_id
;
329 NAPI_STATE_SCHED
, /* Poll is scheduled */
330 NAPI_STATE_DISABLE
, /* Disable pending */
331 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
332 NAPI_STATE_HASHED
, /* In NAPI hash (busy polling possible) */
333 NAPI_STATE_NO_BUSY_POLL
,/* Do not add in napi_hash, no busy polling */
343 typedef enum gro_result gro_result_t
;
346 * enum rx_handler_result - Possible return values for rx_handlers.
347 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
349 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
350 * case skb->dev was changed by rx_handler.
351 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
352 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
354 * rx_handlers are functions called from inside __netif_receive_skb(), to do
355 * special processing of the skb, prior to delivery to protocol handlers.
357 * Currently, a net_device can only have a single rx_handler registered. Trying
358 * to register a second rx_handler will return -EBUSY.
360 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
361 * To unregister a rx_handler on a net_device, use
362 * netdev_rx_handler_unregister().
364 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
367 * If the rx_handler consumed the skb in some way, it should return
368 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
369 * the skb to be delivered in some other way.
371 * If the rx_handler changed skb->dev, to divert the skb to another
372 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
373 * new device will be called if it exists.
375 * If the rx_handler decides the skb should be ignored, it should return
376 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
377 * are registered on exact device (ptype->dev == skb->dev).
379 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
380 * delivered, it should return RX_HANDLER_PASS.
382 * A device without a registered rx_handler will behave as if rx_handler
383 * returned RX_HANDLER_PASS.
386 enum rx_handler_result
{
392 typedef enum rx_handler_result rx_handler_result_t
;
393 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
395 void __napi_schedule(struct napi_struct
*n
);
396 void __napi_schedule_irqoff(struct napi_struct
*n
);
398 static inline bool napi_disable_pending(struct napi_struct
*n
)
400 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
404 * napi_schedule_prep - check if NAPI can be scheduled
407 * Test if NAPI routine is already running, and if not mark
408 * it as running. This is used as a condition variable to
409 * insure only one NAPI poll instance runs. We also make
410 * sure there is no pending NAPI disable.
412 static inline bool napi_schedule_prep(struct napi_struct
*n
)
414 return !napi_disable_pending(n
) &&
415 !test_and_set_bit(NAPI_STATE_SCHED
, &n
->state
);
419 * napi_schedule - schedule NAPI poll
422 * Schedule NAPI poll routine to be called if it is not already
425 static inline void napi_schedule(struct napi_struct
*n
)
427 if (napi_schedule_prep(n
))
432 * napi_schedule_irqoff - schedule NAPI poll
435 * Variant of napi_schedule(), assuming hard irqs are masked.
437 static inline void napi_schedule_irqoff(struct napi_struct
*n
)
439 if (napi_schedule_prep(n
))
440 __napi_schedule_irqoff(n
);
443 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
444 static inline bool napi_reschedule(struct napi_struct
*napi
)
446 if (napi_schedule_prep(napi
)) {
447 __napi_schedule(napi
);
453 void __napi_complete(struct napi_struct
*n
);
454 void napi_complete_done(struct napi_struct
*n
, int work_done
);
456 * napi_complete - NAPI processing complete
459 * Mark NAPI processing as complete.
460 * Consider using napi_complete_done() instead.
462 static inline void napi_complete(struct napi_struct
*n
)
464 return napi_complete_done(n
, 0);
468 * napi_hash_add - add a NAPI to global hashtable
469 * @napi: NAPI context
471 * Generate a new napi_id and store a @napi under it in napi_hash.
472 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
473 * Note: This is normally automatically done from netif_napi_add(),
474 * so might disappear in a future Linux version.
476 void napi_hash_add(struct napi_struct
*napi
);
479 * napi_hash_del - remove a NAPI from global table
480 * @napi: NAPI context
482 * Warning: caller must observe RCU grace period
483 * before freeing memory containing @napi, if
484 * this function returns true.
485 * Note: core networking stack automatically calls it
486 * from netif_napi_del().
487 * Drivers might want to call this helper to combine all
488 * the needed RCU grace periods into a single one.
490 bool napi_hash_del(struct napi_struct
*napi
);
493 * napi_disable - prevent NAPI from scheduling
496 * Stop NAPI from being scheduled on this context.
497 * Waits till any outstanding processing completes.
499 void napi_disable(struct napi_struct
*n
);
502 * napi_enable - enable NAPI scheduling
505 * Resume NAPI from being scheduled on this context.
506 * Must be paired with napi_disable.
508 static inline void napi_enable(struct napi_struct
*n
)
510 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
511 smp_mb__before_atomic();
512 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
513 clear_bit(NAPI_STATE_NPSVC
, &n
->state
);
517 * napi_synchronize - wait until NAPI is not running
520 * Wait until NAPI is done being scheduled on this context.
521 * Waits till any outstanding processing completes but
522 * does not disable future activations.
524 static inline void napi_synchronize(const struct napi_struct
*n
)
526 if (IS_ENABLED(CONFIG_SMP
))
527 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
533 enum netdev_queue_state_t
{
534 __QUEUE_STATE_DRV_XOFF
,
535 __QUEUE_STATE_STACK_XOFF
,
536 __QUEUE_STATE_FROZEN
,
539 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
540 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
551 * netif_tx_* functions below are used to manipulate this flag. The
552 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
553 * queue independently. The netif_xmit_*stopped functions below are called
554 * to check if the queue has been stopped by the driver or stack (either
555 * of the XOFF bits are set in the state). Drivers should not need to call
556 * netif_xmit*stopped functions, they should only be using netif_tx_*.
559 struct netdev_queue
{
563 struct net_device
*dev
;
564 struct Qdisc __rcu
*qdisc
;
565 struct Qdisc
*qdisc_sleeping
;
569 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
572 unsigned long tx_maxrate
;
574 * Number of TX timeouts for this queue
575 * (/sys/class/net/DEV/Q/trans_timeout)
577 unsigned long trans_timeout
;
581 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
584 * Time (in jiffies) of last Tx
586 unsigned long trans_start
;
593 } ____cacheline_aligned_in_smp
;
595 static inline int netdev_queue_numa_node_read(const struct netdev_queue
*q
)
597 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
604 static inline void netdev_queue_numa_node_write(struct netdev_queue
*q
, int node
)
606 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
613 * This structure holds an RPS map which can be of variable length. The
614 * map is an array of CPUs.
621 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
624 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
625 * tail pointer for that CPU's input queue at the time of last enqueue, and
626 * a hardware filter index.
628 struct rps_dev_flow
{
631 unsigned int last_qtail
;
633 #define RPS_NO_FILTER 0xffff
636 * The rps_dev_flow_table structure contains a table of flow mappings.
638 struct rps_dev_flow_table
{
641 struct rps_dev_flow flows
[0];
643 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
644 ((_num) * sizeof(struct rps_dev_flow)))
647 * The rps_sock_flow_table contains mappings of flows to the last CPU
648 * on which they were processed by the application (set in recvmsg).
649 * Each entry is a 32bit value. Upper part is the high-order bits
650 * of flow hash, lower part is CPU number.
651 * rps_cpu_mask is used to partition the space, depending on number of
652 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
653 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
654 * meaning we use 32-6=26 bits for the hash.
656 struct rps_sock_flow_table
{
659 u32 ents
[0] ____cacheline_aligned_in_smp
;
661 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
663 #define RPS_NO_CPU 0xffff
665 extern u32 rps_cpu_mask
;
666 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
668 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
672 unsigned int index
= hash
& table
->mask
;
673 u32 val
= hash
& ~rps_cpu_mask
;
675 /* We only give a hint, preemption can change CPU under us */
676 val
|= raw_smp_processor_id();
678 if (table
->ents
[index
] != val
)
679 table
->ents
[index
] = val
;
683 #ifdef CONFIG_RFS_ACCEL
684 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
687 #endif /* CONFIG_RPS */
689 /* This structure contains an instance of an RX queue. */
690 struct netdev_rx_queue
{
692 struct rps_map __rcu
*rps_map
;
693 struct rps_dev_flow_table __rcu
*rps_flow_table
;
696 struct net_device
*dev
;
697 } ____cacheline_aligned_in_smp
;
700 * RX queue sysfs structures and functions.
702 struct rx_queue_attribute
{
703 struct attribute attr
;
704 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
705 struct rx_queue_attribute
*attr
, char *buf
);
706 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
707 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
712 * This structure holds an XPS map which can be of variable length. The
713 * map is an array of queues.
717 unsigned int alloc_len
;
721 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
722 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
723 - sizeof(struct xps_map)) / sizeof(u16))
726 * This structure holds all XPS maps for device. Maps are indexed by CPU.
728 struct xps_dev_maps
{
730 struct xps_map __rcu
*cpu_map
[0];
732 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
733 (nr_cpu_ids * sizeof(struct xps_map *)))
734 #endif /* CONFIG_XPS */
736 #define TC_MAX_QUEUE 16
737 #define TC_BITMASK 15
738 /* HW offloaded queuing disciplines txq count and offset maps */
739 struct netdev_tc_txq
{
744 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746 * This structure is to hold information about the device
747 * configured to run FCoE protocol stack.
749 struct netdev_fcoe_hbainfo
{
750 char manufacturer
[64];
751 char serial_number
[64];
752 char hardware_version
[64];
753 char driver_version
[64];
754 char optionrom_version
[64];
755 char firmware_version
[64];
757 char model_description
[256];
761 #define MAX_PHYS_ITEM_ID_LEN 32
763 /* This structure holds a unique identifier to identify some
764 * physical item (port for example) used by a netdevice.
766 struct netdev_phys_item_id
{
767 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
768 unsigned char id_len
;
771 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id
*a
,
772 struct netdev_phys_item_id
*b
)
774 return a
->id_len
== b
->id_len
&&
775 memcmp(a
->id
, b
->id
, a
->id_len
) == 0;
778 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
779 struct sk_buff
*skb
);
781 /* These structures hold the attributes of qdisc and classifiers
782 * that are being passed to the netdevice through the setup_tc op.
790 struct tc_cls_u32_offload
;
792 struct tc_to_netdev
{
796 struct tc_cls_u32_offload
*cls_u32
;
797 struct tc_cls_flower_offload
*cls_flower
;
803 * This structure defines the management hooks for network devices.
804 * The following hooks can be defined; unless noted otherwise, they are
805 * optional and can be filled with a null pointer.
807 * int (*ndo_init)(struct net_device *dev);
808 * This function is called once when a network device is registered.
809 * The network device can use this for any late stage initialization
810 * or semantic validation. It can fail with an error code which will
811 * be propagated back to register_netdev.
813 * void (*ndo_uninit)(struct net_device *dev);
814 * This function is called when device is unregistered or when registration
815 * fails. It is not called if init fails.
817 * int (*ndo_open)(struct net_device *dev);
818 * This function is called when a network device transitions to the up
821 * int (*ndo_stop)(struct net_device *dev);
822 * This function is called when a network device transitions to the down
825 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
826 * struct net_device *dev);
827 * Called when a packet needs to be transmitted.
828 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
829 * the queue before that can happen; it's for obsolete devices and weird
830 * corner cases, but the stack really does a non-trivial amount
831 * of useless work if you return NETDEV_TX_BUSY.
832 * Required; cannot be NULL.
834 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
835 * netdev_features_t features);
836 * Adjusts the requested feature flags according to device-specific
837 * constraints, and returns the resulting flags. Must not modify
840 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
841 * void *accel_priv, select_queue_fallback_t fallback);
842 * Called to decide which queue to use when device supports multiple
845 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
846 * This function is called to allow device receiver to make
847 * changes to configuration when multicast or promiscuous is enabled.
849 * void (*ndo_set_rx_mode)(struct net_device *dev);
850 * This function is called device changes address list filtering.
851 * If driver handles unicast address filtering, it should set
852 * IFF_UNICAST_FLT in its priv_flags.
854 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
855 * This function is called when the Media Access Control address
856 * needs to be changed. If this interface is not defined, the
857 * MAC address can not be changed.
859 * int (*ndo_validate_addr)(struct net_device *dev);
860 * Test if Media Access Control address is valid for the device.
862 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
863 * Called when a user requests an ioctl which can't be handled by
864 * the generic interface code. If not defined ioctls return
865 * not supported error code.
867 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
868 * Used to set network devices bus interface parameters. This interface
869 * is retained for legacy reasons; new devices should use the bus
870 * interface (PCI) for low level management.
872 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
873 * Called when a user wants to change the Maximum Transfer Unit
874 * of a device. If not defined, any request to change MTU will
875 * will return an error.
877 * void (*ndo_tx_timeout)(struct net_device *dev);
878 * Callback used when the transmitter has not made any progress
879 * for dev->watchdog ticks.
881 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
882 * struct rtnl_link_stats64 *storage);
883 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
884 * Called when a user wants to get the network device usage
885 * statistics. Drivers must do one of the following:
886 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
887 * rtnl_link_stats64 structure passed by the caller.
888 * 2. Define @ndo_get_stats to update a net_device_stats structure
889 * (which should normally be dev->stats) and return a pointer to
890 * it. The structure may be changed asynchronously only if each
891 * field is written atomically.
892 * 3. Update dev->stats asynchronously and atomically, and define
895 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
896 * If device supports VLAN filtering this function is called when a
897 * VLAN id is registered.
899 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
900 * If device supports VLAN filtering this function is called when a
901 * VLAN id is unregistered.
903 * void (*ndo_poll_controller)(struct net_device *dev);
905 * SR-IOV management functions.
906 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
907 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
908 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
910 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
911 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
912 * int (*ndo_get_vf_config)(struct net_device *dev,
913 * int vf, struct ifla_vf_info *ivf);
914 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
915 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
916 * struct nlattr *port[]);
918 * Enable or disable the VF ability to query its RSS Redirection Table and
919 * Hash Key. This is needed since on some devices VF share this information
920 * with PF and querying it may introduce a theoretical security risk.
921 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
922 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
923 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
924 * Called to setup 'tc' number of traffic classes in the net device. This
925 * is always called from the stack with the rtnl lock held and netif tx
926 * queues stopped. This allows the netdevice to perform queue management
929 * Fiber Channel over Ethernet (FCoE) offload functions.
930 * int (*ndo_fcoe_enable)(struct net_device *dev);
931 * Called when the FCoE protocol stack wants to start using LLD for FCoE
932 * so the underlying device can perform whatever needed configuration or
933 * initialization to support acceleration of FCoE traffic.
935 * int (*ndo_fcoe_disable)(struct net_device *dev);
936 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
937 * so the underlying device can perform whatever needed clean-ups to
938 * stop supporting acceleration of FCoE traffic.
940 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
941 * struct scatterlist *sgl, unsigned int sgc);
942 * Called when the FCoE Initiator wants to initialize an I/O that
943 * is a possible candidate for Direct Data Placement (DDP). The LLD can
944 * perform necessary setup and returns 1 to indicate the device is set up
945 * successfully to perform DDP on this I/O, otherwise this returns 0.
947 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
948 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
949 * indicated by the FC exchange id 'xid', so the underlying device can
950 * clean up and reuse resources for later DDP requests.
952 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
953 * struct scatterlist *sgl, unsigned int sgc);
954 * Called when the FCoE Target wants to initialize an I/O that
955 * is a possible candidate for Direct Data Placement (DDP). The LLD can
956 * perform necessary setup and returns 1 to indicate the device is set up
957 * successfully to perform DDP on this I/O, otherwise this returns 0.
959 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
960 * struct netdev_fcoe_hbainfo *hbainfo);
961 * Called when the FCoE Protocol stack wants information on the underlying
962 * device. This information is utilized by the FCoE protocol stack to
963 * register attributes with Fiber Channel management service as per the
964 * FC-GS Fabric Device Management Information(FDMI) specification.
966 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
967 * Called when the underlying device wants to override default World Wide
968 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
969 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
970 * protocol stack to use.
973 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
974 * u16 rxq_index, u32 flow_id);
975 * Set hardware filter for RFS. rxq_index is the target queue index;
976 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
977 * Return the filter ID on success, or a negative error code.
979 * Slave management functions (for bridge, bonding, etc).
980 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
981 * Called to make another netdev an underling.
983 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
984 * Called to release previously enslaved netdev.
986 * Feature/offload setting functions.
987 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
988 * Called to update device configuration to new features. Passed
989 * feature set might be less than what was returned by ndo_fix_features()).
990 * Must return >0 or -errno if it changed dev->features itself.
992 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
993 * struct net_device *dev,
994 * const unsigned char *addr, u16 vid, u16 flags)
995 * Adds an FDB entry to dev for addr.
996 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
997 * struct net_device *dev,
998 * const unsigned char *addr, u16 vid)
999 * Deletes the FDB entry from dev coresponding to addr.
1000 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1001 * struct net_device *dev, struct net_device *filter_dev,
1003 * Used to add FDB entries to dump requests. Implementers should add
1004 * entries to skb and update idx with the number of entries.
1006 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1008 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1009 * struct net_device *dev, u32 filter_mask,
1011 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1014 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1015 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1016 * which do not represent real hardware may define this to allow their
1017 * userspace components to manage their virtual carrier state. Devices
1018 * that determine carrier state from physical hardware properties (eg
1019 * network cables) or protocol-dependent mechanisms (eg
1020 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1022 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1023 * struct netdev_phys_item_id *ppid);
1024 * Called to get ID of physical port of this device. If driver does
1025 * not implement this, it is assumed that the hw is not able to have
1026 * multiple net devices on single physical port.
1028 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1029 * sa_family_t sa_family, __be16 port);
1030 * Called by vxlan to notify a driver about the UDP port and socket
1031 * address family that vxlan is listening to. It is called only when
1032 * a new port starts listening. The operation is protected by the
1033 * vxlan_net->sock_lock.
1035 * void (*ndo_add_geneve_port)(struct net_device *dev,
1036 * sa_family_t sa_family, __be16 port);
1037 * Called by geneve to notify a driver about the UDP port and socket
1038 * address family that geneve is listnening to. It is called only when
1039 * a new port starts listening. The operation is protected by the
1040 * geneve_net->sock_lock.
1042 * void (*ndo_del_geneve_port)(struct net_device *dev,
1043 * sa_family_t sa_family, __be16 port);
1044 * Called by geneve to notify the driver about a UDP port and socket
1045 * address family that geneve is not listening to anymore. The operation
1046 * is protected by the geneve_net->sock_lock.
1048 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1049 * sa_family_t sa_family, __be16 port);
1050 * Called by vxlan to notify the driver about a UDP port and socket
1051 * address family that vxlan is not listening to anymore. The operation
1052 * is protected by the vxlan_net->sock_lock.
1054 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1055 * struct net_device *dev)
1056 * Called by upper layer devices to accelerate switching or other
1057 * station functionality into hardware. 'pdev is the lowerdev
1058 * to use for the offload and 'dev' is the net device that will
1059 * back the offload. Returns a pointer to the private structure
1060 * the upper layer will maintain.
1061 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1062 * Called by upper layer device to delete the station created
1063 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1064 * the station and priv is the structure returned by the add
1066 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1067 * struct net_device *dev,
1069 * Callback to use for xmit over the accelerated station. This
1070 * is used in place of ndo_start_xmit on accelerated net
1072 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1073 * struct net_device *dev
1074 * netdev_features_t features);
1075 * Called by core transmit path to determine if device is capable of
1076 * performing offload operations on a given packet. This is to give
1077 * the device an opportunity to implement any restrictions that cannot
1078 * be otherwise expressed by feature flags. The check is called with
1079 * the set of features that the stack has calculated and it returns
1080 * those the driver believes to be appropriate.
1081 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1082 * int queue_index, u32 maxrate);
1083 * Called when a user wants to set a max-rate limitation of specific
1085 * int (*ndo_get_iflink)(const struct net_device *dev);
1086 * Called to get the iflink value of this device.
1087 * void (*ndo_change_proto_down)(struct net_device *dev,
1089 * This function is used to pass protocol port error state information
1090 * to the switch driver. The switch driver can react to the proto_down
1091 * by doing a phys down on the associated switch port.
1092 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1093 * This function is used to get egress tunnel information for given skb.
1094 * This is useful for retrieving outer tunnel header parameters while
1096 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1097 * This function is used to specify the headroom that the skb must
1098 * consider when allocation skb during packet reception. Setting
1099 * appropriate rx headroom value allows avoiding skb head copy on
1100 * forward. Setting a negative value resets the rx headroom to the
1104 struct net_device_ops
{
1105 int (*ndo_init
)(struct net_device
*dev
);
1106 void (*ndo_uninit
)(struct net_device
*dev
);
1107 int (*ndo_open
)(struct net_device
*dev
);
1108 int (*ndo_stop
)(struct net_device
*dev
);
1109 netdev_tx_t (*ndo_start_xmit
)(struct sk_buff
*skb
,
1110 struct net_device
*dev
);
1111 netdev_features_t (*ndo_features_check
)(struct sk_buff
*skb
,
1112 struct net_device
*dev
,
1113 netdev_features_t features
);
1114 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1115 struct sk_buff
*skb
,
1117 select_queue_fallback_t fallback
);
1118 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1120 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1121 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1123 int (*ndo_validate_addr
)(struct net_device
*dev
);
1124 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1125 struct ifreq
*ifr
, int cmd
);
1126 int (*ndo_set_config
)(struct net_device
*dev
,
1128 int (*ndo_change_mtu
)(struct net_device
*dev
,
1130 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1131 struct neigh_parms
*);
1132 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1134 struct rtnl_link_stats64
* (*ndo_get_stats64
)(struct net_device
*dev
,
1135 struct rtnl_link_stats64
*storage
);
1136 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1138 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1139 __be16 proto
, u16 vid
);
1140 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1141 __be16 proto
, u16 vid
);
1142 #ifdef CONFIG_NET_POLL_CONTROLLER
1143 void (*ndo_poll_controller
)(struct net_device
*dev
);
1144 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1145 struct netpoll_info
*info
);
1146 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1148 #ifdef CONFIG_NET_RX_BUSY_POLL
1149 int (*ndo_busy_poll
)(struct napi_struct
*dev
);
1151 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1152 int queue
, u8
*mac
);
1153 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1154 int queue
, u16 vlan
, u8 qos
);
1155 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1156 int vf
, int min_tx_rate
,
1158 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1159 int vf
, bool setting
);
1160 int (*ndo_set_vf_trust
)(struct net_device
*dev
,
1161 int vf
, bool setting
);
1162 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1164 struct ifla_vf_info
*ivf
);
1165 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1166 int vf
, int link_state
);
1167 int (*ndo_get_vf_stats
)(struct net_device
*dev
,
1169 struct ifla_vf_stats
1171 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1173 struct nlattr
*port
[]);
1174 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1175 int vf
, struct sk_buff
*skb
);
1176 int (*ndo_set_vf_guid
)(struct net_device
*dev
,
1179 int (*ndo_set_vf_rss_query_en
)(
1180 struct net_device
*dev
,
1181 int vf
, bool setting
);
1182 int (*ndo_setup_tc
)(struct net_device
*dev
,
1185 struct tc_to_netdev
*tc
);
1186 #if IS_ENABLED(CONFIG_FCOE)
1187 int (*ndo_fcoe_enable
)(struct net_device
*dev
);
1188 int (*ndo_fcoe_disable
)(struct net_device
*dev
);
1189 int (*ndo_fcoe_ddp_setup
)(struct net_device
*dev
,
1191 struct scatterlist
*sgl
,
1193 int (*ndo_fcoe_ddp_done
)(struct net_device
*dev
,
1195 int (*ndo_fcoe_ddp_target
)(struct net_device
*dev
,
1197 struct scatterlist
*sgl
,
1199 int (*ndo_fcoe_get_hbainfo
)(struct net_device
*dev
,
1200 struct netdev_fcoe_hbainfo
*hbainfo
);
1203 #if IS_ENABLED(CONFIG_LIBFCOE)
1204 #define NETDEV_FCOE_WWNN 0
1205 #define NETDEV_FCOE_WWPN 1
1206 int (*ndo_fcoe_get_wwn
)(struct net_device
*dev
,
1207 u64
*wwn
, int type
);
1210 #ifdef CONFIG_RFS_ACCEL
1211 int (*ndo_rx_flow_steer
)(struct net_device
*dev
,
1212 const struct sk_buff
*skb
,
1216 int (*ndo_add_slave
)(struct net_device
*dev
,
1217 struct net_device
*slave_dev
);
1218 int (*ndo_del_slave
)(struct net_device
*dev
,
1219 struct net_device
*slave_dev
);
1220 netdev_features_t (*ndo_fix_features
)(struct net_device
*dev
,
1221 netdev_features_t features
);
1222 int (*ndo_set_features
)(struct net_device
*dev
,
1223 netdev_features_t features
);
1224 int (*ndo_neigh_construct
)(struct neighbour
*n
);
1225 void (*ndo_neigh_destroy
)(struct neighbour
*n
);
1227 int (*ndo_fdb_add
)(struct ndmsg
*ndm
,
1228 struct nlattr
*tb
[],
1229 struct net_device
*dev
,
1230 const unsigned char *addr
,
1233 int (*ndo_fdb_del
)(struct ndmsg
*ndm
,
1234 struct nlattr
*tb
[],
1235 struct net_device
*dev
,
1236 const unsigned char *addr
,
1238 int (*ndo_fdb_dump
)(struct sk_buff
*skb
,
1239 struct netlink_callback
*cb
,
1240 struct net_device
*dev
,
1241 struct net_device
*filter_dev
,
1244 int (*ndo_bridge_setlink
)(struct net_device
*dev
,
1245 struct nlmsghdr
*nlh
,
1247 int (*ndo_bridge_getlink
)(struct sk_buff
*skb
,
1249 struct net_device
*dev
,
1252 int (*ndo_bridge_dellink
)(struct net_device
*dev
,
1253 struct nlmsghdr
*nlh
,
1255 int (*ndo_change_carrier
)(struct net_device
*dev
,
1257 int (*ndo_get_phys_port_id
)(struct net_device
*dev
,
1258 struct netdev_phys_item_id
*ppid
);
1259 int (*ndo_get_phys_port_name
)(struct net_device
*dev
,
1260 char *name
, size_t len
);
1261 void (*ndo_add_vxlan_port
)(struct net_device
*dev
,
1262 sa_family_t sa_family
,
1264 void (*ndo_del_vxlan_port
)(struct net_device
*dev
,
1265 sa_family_t sa_family
,
1267 void (*ndo_add_geneve_port
)(struct net_device
*dev
,
1268 sa_family_t sa_family
,
1270 void (*ndo_del_geneve_port
)(struct net_device
*dev
,
1271 sa_family_t sa_family
,
1273 void* (*ndo_dfwd_add_station
)(struct net_device
*pdev
,
1274 struct net_device
*dev
);
1275 void (*ndo_dfwd_del_station
)(struct net_device
*pdev
,
1278 netdev_tx_t (*ndo_dfwd_start_xmit
) (struct sk_buff
*skb
,
1279 struct net_device
*dev
,
1281 int (*ndo_get_lock_subclass
)(struct net_device
*dev
);
1282 int (*ndo_set_tx_maxrate
)(struct net_device
*dev
,
1285 int (*ndo_get_iflink
)(const struct net_device
*dev
);
1286 int (*ndo_change_proto_down
)(struct net_device
*dev
,
1288 int (*ndo_fill_metadata_dst
)(struct net_device
*dev
,
1289 struct sk_buff
*skb
);
1290 void (*ndo_set_rx_headroom
)(struct net_device
*dev
,
1291 int needed_headroom
);
1295 * enum net_device_priv_flags - &struct net_device priv_flags
1297 * These are the &struct net_device, they are only set internally
1298 * by drivers and used in the kernel. These flags are invisible to
1299 * userspace; this means that the order of these flags can change
1300 * during any kernel release.
1302 * You should have a pretty good reason to be extending these flags.
1304 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1305 * @IFF_EBRIDGE: Ethernet bridging device
1306 * @IFF_BONDING: bonding master or slave
1307 * @IFF_ISATAP: ISATAP interface (RFC4214)
1308 * @IFF_WAN_HDLC: WAN HDLC device
1309 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1311 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1312 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1313 * @IFF_MACVLAN_PORT: device used as macvlan port
1314 * @IFF_BRIDGE_PORT: device used as bridge port
1315 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1316 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1317 * @IFF_UNICAST_FLT: Supports unicast filtering
1318 * @IFF_TEAM_PORT: device used as team port
1319 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1320 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1321 * change when it's running
1322 * @IFF_MACVLAN: Macvlan device
1323 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1324 * underlying stacked devices
1325 * @IFF_IPVLAN_MASTER: IPvlan master device
1326 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1327 * @IFF_L3MDEV_MASTER: device is an L3 master device
1328 * @IFF_NO_QUEUE: device can run without qdisc attached
1329 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1330 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1331 * @IFF_TEAM: device is a team device
1332 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1333 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1334 * entity (i.e. the master device for bridged veth)
1335 * @IFF_MACSEC: device is a MACsec device
1337 enum netdev_priv_flags
{
1338 IFF_802_1Q_VLAN
= 1<<0,
1342 IFF_WAN_HDLC
= 1<<4,
1343 IFF_XMIT_DST_RELEASE
= 1<<5,
1344 IFF_DONT_BRIDGE
= 1<<6,
1345 IFF_DISABLE_NETPOLL
= 1<<7,
1346 IFF_MACVLAN_PORT
= 1<<8,
1347 IFF_BRIDGE_PORT
= 1<<9,
1348 IFF_OVS_DATAPATH
= 1<<10,
1349 IFF_TX_SKB_SHARING
= 1<<11,
1350 IFF_UNICAST_FLT
= 1<<12,
1351 IFF_TEAM_PORT
= 1<<13,
1352 IFF_SUPP_NOFCS
= 1<<14,
1353 IFF_LIVE_ADDR_CHANGE
= 1<<15,
1354 IFF_MACVLAN
= 1<<16,
1355 IFF_XMIT_DST_RELEASE_PERM
= 1<<17,
1356 IFF_IPVLAN_MASTER
= 1<<18,
1357 IFF_IPVLAN_SLAVE
= 1<<19,
1358 IFF_L3MDEV_MASTER
= 1<<20,
1359 IFF_NO_QUEUE
= 1<<21,
1360 IFF_OPENVSWITCH
= 1<<22,
1361 IFF_L3MDEV_SLAVE
= 1<<23,
1363 IFF_RXFH_CONFIGURED
= 1<<25,
1364 IFF_PHONY_HEADROOM
= 1<<26,
1368 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1369 #define IFF_EBRIDGE IFF_EBRIDGE
1370 #define IFF_BONDING IFF_BONDING
1371 #define IFF_ISATAP IFF_ISATAP
1372 #define IFF_WAN_HDLC IFF_WAN_HDLC
1373 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1374 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1375 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1376 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1377 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1378 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1379 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1380 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1381 #define IFF_TEAM_PORT IFF_TEAM_PORT
1382 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1383 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1384 #define IFF_MACVLAN IFF_MACVLAN
1385 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1386 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1387 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1388 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1389 #define IFF_NO_QUEUE IFF_NO_QUEUE
1390 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1391 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1392 #define IFF_TEAM IFF_TEAM
1393 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1394 #define IFF_MACSEC IFF_MACSEC
1397 * struct net_device - The DEVICE structure.
1398 * Actually, this whole structure is a big mistake. It mixes I/O
1399 * data with strictly "high-level" data, and it has to know about
1400 * almost every data structure used in the INET module.
1402 * @name: This is the first field of the "visible" part of this structure
1403 * (i.e. as seen by users in the "Space.c" file). It is the name
1406 * @name_hlist: Device name hash chain, please keep it close to name[]
1407 * @ifalias: SNMP alias
1408 * @mem_end: Shared memory end
1409 * @mem_start: Shared memory start
1410 * @base_addr: Device I/O address
1411 * @irq: Device IRQ number
1413 * @carrier_changes: Stats to monitor carrier on<->off transitions
1415 * @state: Generic network queuing layer state, see netdev_state_t
1416 * @dev_list: The global list of network devices
1417 * @napi_list: List entry used for polling NAPI devices
1418 * @unreg_list: List entry when we are unregistering the
1419 * device; see the function unregister_netdev
1420 * @close_list: List entry used when we are closing the device
1421 * @ptype_all: Device-specific packet handlers for all protocols
1422 * @ptype_specific: Device-specific, protocol-specific packet handlers
1424 * @adj_list: Directly linked devices, like slaves for bonding
1425 * @all_adj_list: All linked devices, *including* neighbours
1426 * @features: Currently active device features
1427 * @hw_features: User-changeable features
1429 * @wanted_features: User-requested features
1430 * @vlan_features: Mask of features inheritable by VLAN devices
1432 * @hw_enc_features: Mask of features inherited by encapsulating devices
1433 * This field indicates what encapsulation
1434 * offloads the hardware is capable of doing,
1435 * and drivers will need to set them appropriately.
1437 * @mpls_features: Mask of features inheritable by MPLS
1439 * @ifindex: interface index
1440 * @group: The group the device belongs to
1442 * @stats: Statistics struct, which was left as a legacy, use
1443 * rtnl_link_stats64 instead
1445 * @rx_dropped: Dropped packets by core network,
1446 * do not use this in drivers
1447 * @tx_dropped: Dropped packets by core network,
1448 * do not use this in drivers
1449 * @rx_nohandler: nohandler dropped packets by core network on
1450 * inactive devices, do not use this in drivers
1452 * @wireless_handlers: List of functions to handle Wireless Extensions,
1454 * see <net/iw_handler.h> for details.
1455 * @wireless_data: Instance data managed by the core of wireless extensions
1457 * @netdev_ops: Includes several pointers to callbacks,
1458 * if one wants to override the ndo_*() functions
1459 * @ethtool_ops: Management operations
1460 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1461 * of Layer 2 headers.
1463 * @flags: Interface flags (a la BSD)
1464 * @priv_flags: Like 'flags' but invisible to userspace,
1465 * see if.h for the definitions
1466 * @gflags: Global flags ( kept as legacy )
1467 * @padded: How much padding added by alloc_netdev()
1468 * @operstate: RFC2863 operstate
1469 * @link_mode: Mapping policy to operstate
1470 * @if_port: Selectable AUI, TP, ...
1472 * @mtu: Interface MTU value
1473 * @type: Interface hardware type
1474 * @hard_header_len: Maximum hardware header length.
1476 * @needed_headroom: Extra headroom the hardware may need, but not in all
1477 * cases can this be guaranteed
1478 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1479 * cases can this be guaranteed. Some cases also use
1480 * LL_MAX_HEADER instead to allocate the skb
1482 * interface address info:
1484 * @perm_addr: Permanent hw address
1485 * @addr_assign_type: Hw address assignment type
1486 * @addr_len: Hardware address length
1487 * @neigh_priv_len; Used in neigh_alloc(),
1488 * initialized only in atm/clip.c
1489 * @dev_id: Used to differentiate devices that share
1490 * the same link layer address
1491 * @dev_port: Used to differentiate devices that share
1493 * @addr_list_lock: XXX: need comments on this one
1494 * @uc_promisc: Counter that indicates promiscuous mode
1495 * has been enabled due to the need to listen to
1496 * additional unicast addresses in a device that
1497 * does not implement ndo_set_rx_mode()
1498 * @uc: unicast mac addresses
1499 * @mc: multicast mac addresses
1500 * @dev_addrs: list of device hw addresses
1501 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1502 * @promiscuity: Number of times the NIC is told to work in
1503 * promiscuous mode; if it becomes 0 the NIC will
1504 * exit promiscuous mode
1505 * @allmulti: Counter, enables or disables allmulticast mode
1507 * @vlan_info: VLAN info
1508 * @dsa_ptr: dsa specific data
1509 * @tipc_ptr: TIPC specific data
1510 * @atalk_ptr: AppleTalk link
1511 * @ip_ptr: IPv4 specific data
1512 * @dn_ptr: DECnet specific data
1513 * @ip6_ptr: IPv6 specific data
1514 * @ax25_ptr: AX.25 specific data
1515 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1517 * @last_rx: Time of last Rx
1518 * @dev_addr: Hw address (before bcast,
1519 * because most packets are unicast)
1521 * @_rx: Array of RX queues
1522 * @num_rx_queues: Number of RX queues
1523 * allocated at register_netdev() time
1524 * @real_num_rx_queues: Number of RX queues currently active in device
1526 * @rx_handler: handler for received packets
1527 * @rx_handler_data: XXX: need comments on this one
1528 * @ingress_queue: XXX: need comments on this one
1529 * @broadcast: hw bcast address
1531 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1532 * indexed by RX queue number. Assigned by driver.
1533 * This must only be set if the ndo_rx_flow_steer
1534 * operation is defined
1535 * @index_hlist: Device index hash chain
1537 * @_tx: Array of TX queues
1538 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1539 * @real_num_tx_queues: Number of TX queues currently active in device
1540 * @qdisc: Root qdisc from userspace point of view
1541 * @tx_queue_len: Max frames per queue allowed
1542 * @tx_global_lock: XXX: need comments on this one
1544 * @xps_maps: XXX: need comments on this one
1546 * @offload_fwd_mark: Offload device fwding mark
1548 * @watchdog_timeo: Represents the timeout that is used by
1549 * the watchdog (see dev_watchdog())
1550 * @watchdog_timer: List of timers
1552 * @pcpu_refcnt: Number of references to this device
1553 * @todo_list: Delayed register/unregister
1554 * @link_watch_list: XXX: need comments on this one
1556 * @reg_state: Register/unregister state machine
1557 * @dismantle: Device is going to be freed
1558 * @rtnl_link_state: This enum represents the phases of creating
1561 * @destructor: Called from unregister,
1562 * can be used to call free_netdev
1563 * @npinfo: XXX: need comments on this one
1564 * @nd_net: Network namespace this network device is inside
1566 * @ml_priv: Mid-layer private
1567 * @lstats: Loopback statistics
1568 * @tstats: Tunnel statistics
1569 * @dstats: Dummy statistics
1570 * @vstats: Virtual ethernet statistics
1575 * @dev: Class/net/name entry
1576 * @sysfs_groups: Space for optional device, statistics and wireless
1579 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1580 * @rtnl_link_ops: Rtnl_link_ops
1582 * @gso_max_size: Maximum size of generic segmentation offload
1583 * @gso_max_segs: Maximum number of segments that can be passed to the
1586 * @dcbnl_ops: Data Center Bridging netlink ops
1587 * @num_tc: Number of traffic classes in the net device
1588 * @tc_to_txq: XXX: need comments on this one
1589 * @prio_tc_map XXX: need comments on this one
1591 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1593 * @priomap: XXX: need comments on this one
1594 * @phydev: Physical device may attach itself
1595 * for hardware timestamping
1597 * @qdisc_tx_busylock: XXX: need comments on this one
1599 * @proto_down: protocol port state information can be sent to the
1600 * switch driver and used to set the phys state of the
1603 * FIXME: cleanup struct net_device such that network protocol info
1608 char name
[IFNAMSIZ
];
1609 struct hlist_node name_hlist
;
1612 * I/O specific fields
1613 * FIXME: Merge these and struct ifmap into one
1615 unsigned long mem_end
;
1616 unsigned long mem_start
;
1617 unsigned long base_addr
;
1620 atomic_t carrier_changes
;
1623 * Some hardware also needs these fields (state,dev_list,
1624 * napi_list,unreg_list,close_list) but they are not
1625 * part of the usual set specified in Space.c.
1628 unsigned long state
;
1630 struct list_head dev_list
;
1631 struct list_head napi_list
;
1632 struct list_head unreg_list
;
1633 struct list_head close_list
;
1634 struct list_head ptype_all
;
1635 struct list_head ptype_specific
;
1638 struct list_head upper
;
1639 struct list_head lower
;
1643 struct list_head upper
;
1644 struct list_head lower
;
1647 netdev_features_t features
;
1648 netdev_features_t hw_features
;
1649 netdev_features_t wanted_features
;
1650 netdev_features_t vlan_features
;
1651 netdev_features_t hw_enc_features
;
1652 netdev_features_t mpls_features
;
1653 netdev_features_t gso_partial_features
;
1658 struct net_device_stats stats
;
1660 atomic_long_t rx_dropped
;
1661 atomic_long_t tx_dropped
;
1662 atomic_long_t rx_nohandler
;
1664 #ifdef CONFIG_WIRELESS_EXT
1665 const struct iw_handler_def
*wireless_handlers
;
1666 struct iw_public_data
*wireless_data
;
1668 const struct net_device_ops
*netdev_ops
;
1669 const struct ethtool_ops
*ethtool_ops
;
1670 #ifdef CONFIG_NET_SWITCHDEV
1671 const struct switchdev_ops
*switchdev_ops
;
1673 #ifdef CONFIG_NET_L3_MASTER_DEV
1674 const struct l3mdev_ops
*l3mdev_ops
;
1677 const struct header_ops
*header_ops
;
1680 unsigned int priv_flags
;
1682 unsigned short gflags
;
1683 unsigned short padded
;
1685 unsigned char operstate
;
1686 unsigned char link_mode
;
1688 unsigned char if_port
;
1692 unsigned short type
;
1693 unsigned short hard_header_len
;
1695 unsigned short needed_headroom
;
1696 unsigned short needed_tailroom
;
1698 /* Interface address info. */
1699 unsigned char perm_addr
[MAX_ADDR_LEN
];
1700 unsigned char addr_assign_type
;
1701 unsigned char addr_len
;
1702 unsigned short neigh_priv_len
;
1703 unsigned short dev_id
;
1704 unsigned short dev_port
;
1705 spinlock_t addr_list_lock
;
1706 unsigned char name_assign_type
;
1708 struct netdev_hw_addr_list uc
;
1709 struct netdev_hw_addr_list mc
;
1710 struct netdev_hw_addr_list dev_addrs
;
1713 struct kset
*queues_kset
;
1715 unsigned int promiscuity
;
1716 unsigned int allmulti
;
1719 /* Protocol-specific pointers */
1721 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1722 struct vlan_info __rcu
*vlan_info
;
1724 #if IS_ENABLED(CONFIG_NET_DSA)
1725 struct dsa_switch_tree
*dsa_ptr
;
1727 #if IS_ENABLED(CONFIG_TIPC)
1728 struct tipc_bearer __rcu
*tipc_ptr
;
1731 struct in_device __rcu
*ip_ptr
;
1732 struct dn_dev __rcu
*dn_ptr
;
1733 struct inet6_dev __rcu
*ip6_ptr
;
1735 struct wireless_dev
*ieee80211_ptr
;
1736 struct wpan_dev
*ieee802154_ptr
;
1737 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1738 struct mpls_dev __rcu
*mpls_ptr
;
1742 * Cache lines mostly used on receive path (including eth_type_trans())
1744 unsigned long last_rx
;
1746 /* Interface address info used in eth_type_trans() */
1747 unsigned char *dev_addr
;
1750 struct netdev_rx_queue
*_rx
;
1752 unsigned int num_rx_queues
;
1753 unsigned int real_num_rx_queues
;
1756 unsigned long gro_flush_timeout
;
1757 rx_handler_func_t __rcu
*rx_handler
;
1758 void __rcu
*rx_handler_data
;
1760 #ifdef CONFIG_NET_CLS_ACT
1761 struct tcf_proto __rcu
*ingress_cl_list
;
1763 struct netdev_queue __rcu
*ingress_queue
;
1764 #ifdef CONFIG_NETFILTER_INGRESS
1765 struct list_head nf_hooks_ingress
;
1768 unsigned char broadcast
[MAX_ADDR_LEN
];
1769 #ifdef CONFIG_RFS_ACCEL
1770 struct cpu_rmap
*rx_cpu_rmap
;
1772 struct hlist_node index_hlist
;
1775 * Cache lines mostly used on transmit path
1777 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1778 unsigned int num_tx_queues
;
1779 unsigned int real_num_tx_queues
;
1780 struct Qdisc
*qdisc
;
1781 unsigned long tx_queue_len
;
1782 spinlock_t tx_global_lock
;
1786 struct xps_dev_maps __rcu
*xps_maps
;
1788 #ifdef CONFIG_NET_CLS_ACT
1789 struct tcf_proto __rcu
*egress_cl_list
;
1791 #ifdef CONFIG_NET_SWITCHDEV
1792 u32 offload_fwd_mark
;
1795 /* These may be needed for future network-power-down code. */
1796 struct timer_list watchdog_timer
;
1798 int __percpu
*pcpu_refcnt
;
1799 struct list_head todo_list
;
1801 struct list_head link_watch_list
;
1803 enum { NETREG_UNINITIALIZED
=0,
1804 NETREG_REGISTERED
, /* completed register_netdevice */
1805 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1806 NETREG_UNREGISTERED
, /* completed unregister todo */
1807 NETREG_RELEASED
, /* called free_netdev */
1808 NETREG_DUMMY
, /* dummy device for NAPI poll */
1814 RTNL_LINK_INITIALIZED
,
1815 RTNL_LINK_INITIALIZING
,
1816 } rtnl_link_state
:16;
1818 void (*destructor
)(struct net_device
*dev
);
1820 #ifdef CONFIG_NETPOLL
1821 struct netpoll_info __rcu
*npinfo
;
1824 possible_net_t nd_net
;
1826 /* mid-layer private */
1829 struct pcpu_lstats __percpu
*lstats
;
1830 struct pcpu_sw_netstats __percpu
*tstats
;
1831 struct pcpu_dstats __percpu
*dstats
;
1832 struct pcpu_vstats __percpu
*vstats
;
1835 struct garp_port __rcu
*garp_port
;
1836 struct mrp_port __rcu
*mrp_port
;
1839 const struct attribute_group
*sysfs_groups
[4];
1840 const struct attribute_group
*sysfs_rx_queue_group
;
1842 const struct rtnl_link_ops
*rtnl_link_ops
;
1844 /* for setting kernel sock attribute on TCP connection setup */
1845 #define GSO_MAX_SIZE 65536
1846 unsigned int gso_max_size
;
1847 #define GSO_MAX_SEGS 65535
1851 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1854 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1855 u8 prio_tc_map
[TC_BITMASK
+ 1];
1857 #if IS_ENABLED(CONFIG_FCOE)
1858 unsigned int fcoe_ddp_xid
;
1860 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1861 struct netprio_map __rcu
*priomap
;
1863 struct phy_device
*phydev
;
1864 struct lock_class_key
*qdisc_tx_busylock
;
1867 #define to_net_dev(d) container_of(d, struct net_device, dev)
1869 #define NETDEV_ALIGN 32
1872 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1874 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1878 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1880 if (tc
>= dev
->num_tc
)
1883 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1888 void netdev_reset_tc(struct net_device
*dev
)
1891 memset(dev
->tc_to_txq
, 0, sizeof(dev
->tc_to_txq
));
1892 memset(dev
->prio_tc_map
, 0, sizeof(dev
->prio_tc_map
));
1896 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
)
1898 if (tc
>= dev
->num_tc
)
1901 dev
->tc_to_txq
[tc
].count
= count
;
1902 dev
->tc_to_txq
[tc
].offset
= offset
;
1907 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
)
1909 if (num_tc
> TC_MAX_QUEUE
)
1912 dev
->num_tc
= num_tc
;
1917 int netdev_get_num_tc(struct net_device
*dev
)
1923 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1926 return &dev
->_tx
[index
];
1929 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1930 const struct sk_buff
*skb
)
1932 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1935 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1936 void (*f
)(struct net_device
*,
1937 struct netdev_queue
*,
1943 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1944 f(dev
, &dev
->_tx
[i
], arg
);
1947 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1948 struct sk_buff
*skb
,
1951 /* returns the headroom that the master device needs to take in account
1952 * when forwarding to this dev
1954 static inline unsigned netdev_get_fwd_headroom(struct net_device
*dev
)
1956 return dev
->priv_flags
& IFF_PHONY_HEADROOM
? 0 : dev
->needed_headroom
;
1959 static inline void netdev_set_rx_headroom(struct net_device
*dev
, int new_hr
)
1961 if (dev
->netdev_ops
->ndo_set_rx_headroom
)
1962 dev
->netdev_ops
->ndo_set_rx_headroom(dev
, new_hr
);
1965 /* set the device rx headroom to the dev's default */
1966 static inline void netdev_reset_rx_headroom(struct net_device
*dev
)
1968 netdev_set_rx_headroom(dev
, -1);
1972 * Net namespace inlines
1975 struct net
*dev_net(const struct net_device
*dev
)
1977 return read_pnet(&dev
->nd_net
);
1981 void dev_net_set(struct net_device
*dev
, struct net
*net
)
1983 write_pnet(&dev
->nd_net
, net
);
1986 static inline bool netdev_uses_dsa(struct net_device
*dev
)
1988 #if IS_ENABLED(CONFIG_NET_DSA)
1989 if (dev
->dsa_ptr
!= NULL
)
1990 return dsa_uses_tagged_protocol(dev
->dsa_ptr
);
1996 * netdev_priv - access network device private data
1997 * @dev: network device
1999 * Get network device private data
2001 static inline void *netdev_priv(const struct net_device
*dev
)
2003 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
2006 /* Set the sysfs physical device reference for the network logical device
2007 * if set prior to registration will cause a symlink during initialization.
2009 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2011 /* Set the sysfs device type for the network logical device to allow
2012 * fine-grained identification of different network device types. For
2013 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2015 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2017 /* Default NAPI poll() weight
2018 * Device drivers are strongly advised to not use bigger value
2020 #define NAPI_POLL_WEIGHT 64
2023 * netif_napi_add - initialize a NAPI context
2024 * @dev: network device
2025 * @napi: NAPI context
2026 * @poll: polling function
2027 * @weight: default weight
2029 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2030 * *any* of the other NAPI-related functions.
2032 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
2033 int (*poll
)(struct napi_struct
*, int), int weight
);
2036 * netif_tx_napi_add - initialize a NAPI context
2037 * @dev: network device
2038 * @napi: NAPI context
2039 * @poll: polling function
2040 * @weight: default weight
2042 * This variant of netif_napi_add() should be used from drivers using NAPI
2043 * to exclusively poll a TX queue.
2044 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2046 static inline void netif_tx_napi_add(struct net_device
*dev
,
2047 struct napi_struct
*napi
,
2048 int (*poll
)(struct napi_struct
*, int),
2051 set_bit(NAPI_STATE_NO_BUSY_POLL
, &napi
->state
);
2052 netif_napi_add(dev
, napi
, poll
, weight
);
2056 * netif_napi_del - remove a NAPI context
2057 * @napi: NAPI context
2059 * netif_napi_del() removes a NAPI context from the network device NAPI list
2061 void netif_napi_del(struct napi_struct
*napi
);
2063 struct napi_gro_cb
{
2064 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2067 /* Length of frag0. */
2068 unsigned int frag0_len
;
2070 /* This indicates where we are processing relative to skb->data. */
2073 /* This is non-zero if the packet cannot be merged with the new skb. */
2076 /* Save the IP ID here and check when we get to the transport layer */
2079 /* Number of segments aggregated. */
2082 /* Start offset for remote checksum offload */
2083 u16 gro_remcsum_start
;
2085 /* jiffies when first packet was created/queued */
2088 /* Used in ipv6_gro_receive() and foo-over-udp */
2091 /* This is non-zero if the packet may be of the same flow. */
2094 /* Used in tunnel GRO receive */
2097 /* GRO checksum is valid */
2100 /* Number of checksums via CHECKSUM_UNNECESSARY */
2105 #define NAPI_GRO_FREE 1
2106 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2108 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2111 /* Used in GRE, set in fou/gue_gro_receive */
2114 /* Used to determine if flush_id can be ignored */
2119 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2122 /* used in skb_gro_receive() slow path */
2123 struct sk_buff
*last
;
2126 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2128 struct packet_type
{
2129 __be16 type
; /* This is really htons(ether_type). */
2130 struct net_device
*dev
; /* NULL is wildcarded here */
2131 int (*func
) (struct sk_buff
*,
2132 struct net_device
*,
2133 struct packet_type
*,
2134 struct net_device
*);
2135 bool (*id_match
)(struct packet_type
*ptype
,
2137 void *af_packet_priv
;
2138 struct list_head list
;
2141 struct offload_callbacks
{
2142 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
2143 netdev_features_t features
);
2144 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
2145 struct sk_buff
*skb
);
2146 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
2149 struct packet_offload
{
2150 __be16 type
; /* This is really htons(ether_type). */
2152 struct offload_callbacks callbacks
;
2153 struct list_head list
;
2156 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2157 struct pcpu_sw_netstats
{
2162 struct u64_stats_sync syncp
;
2165 #define __netdev_alloc_pcpu_stats(type, gfp) \
2167 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2170 for_each_possible_cpu(__cpu) { \
2171 typeof(type) *stat; \
2172 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2173 u64_stats_init(&stat->syncp); \
2179 #define netdev_alloc_pcpu_stats(type) \
2180 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2182 enum netdev_lag_tx_type
{
2183 NETDEV_LAG_TX_TYPE_UNKNOWN
,
2184 NETDEV_LAG_TX_TYPE_RANDOM
,
2185 NETDEV_LAG_TX_TYPE_BROADCAST
,
2186 NETDEV_LAG_TX_TYPE_ROUNDROBIN
,
2187 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP
,
2188 NETDEV_LAG_TX_TYPE_HASH
,
2191 struct netdev_lag_upper_info
{
2192 enum netdev_lag_tx_type tx_type
;
2195 struct netdev_lag_lower_state_info
{
2200 #include <linux/notifier.h>
2202 /* netdevice notifier chain. Please remember to update the rtnetlink
2203 * notification exclusion list in rtnetlink_event() when adding new
2206 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2207 #define NETDEV_DOWN 0x0002
2208 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2209 detected a hardware crash and restarted
2210 - we can use this eg to kick tcp sessions
2212 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2213 #define NETDEV_REGISTER 0x0005
2214 #define NETDEV_UNREGISTER 0x0006
2215 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2216 #define NETDEV_CHANGEADDR 0x0008
2217 #define NETDEV_GOING_DOWN 0x0009
2218 #define NETDEV_CHANGENAME 0x000A
2219 #define NETDEV_FEAT_CHANGE 0x000B
2220 #define NETDEV_BONDING_FAILOVER 0x000C
2221 #define NETDEV_PRE_UP 0x000D
2222 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2223 #define NETDEV_POST_TYPE_CHANGE 0x000F
2224 #define NETDEV_POST_INIT 0x0010
2225 #define NETDEV_UNREGISTER_FINAL 0x0011
2226 #define NETDEV_RELEASE 0x0012
2227 #define NETDEV_NOTIFY_PEERS 0x0013
2228 #define NETDEV_JOIN 0x0014
2229 #define NETDEV_CHANGEUPPER 0x0015
2230 #define NETDEV_RESEND_IGMP 0x0016
2231 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2232 #define NETDEV_CHANGEINFODATA 0x0018
2233 #define NETDEV_BONDING_INFO 0x0019
2234 #define NETDEV_PRECHANGEUPPER 0x001A
2235 #define NETDEV_CHANGELOWERSTATE 0x001B
2236 #define NETDEV_OFFLOAD_PUSH_VXLAN 0x001C
2237 #define NETDEV_OFFLOAD_PUSH_GENEVE 0x001D
2239 int register_netdevice_notifier(struct notifier_block
*nb
);
2240 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2242 struct netdev_notifier_info
{
2243 struct net_device
*dev
;
2246 struct netdev_notifier_change_info
{
2247 struct netdev_notifier_info info
; /* must be first */
2248 unsigned int flags_changed
;
2251 struct netdev_notifier_changeupper_info
{
2252 struct netdev_notifier_info info
; /* must be first */
2253 struct net_device
*upper_dev
; /* new upper dev */
2254 bool master
; /* is upper dev master */
2255 bool linking
; /* is the notification for link or unlink */
2256 void *upper_info
; /* upper dev info */
2259 struct netdev_notifier_changelowerstate_info
{
2260 struct netdev_notifier_info info
; /* must be first */
2261 void *lower_state_info
; /* is lower dev state */
2264 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2265 struct net_device
*dev
)
2270 static inline struct net_device
*
2271 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2276 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2279 extern rwlock_t dev_base_lock
; /* Device list lock */
2281 #define for_each_netdev(net, d) \
2282 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2283 #define for_each_netdev_reverse(net, d) \
2284 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2285 #define for_each_netdev_rcu(net, d) \
2286 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2287 #define for_each_netdev_safe(net, d, n) \
2288 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2289 #define for_each_netdev_continue(net, d) \
2290 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2291 #define for_each_netdev_continue_rcu(net, d) \
2292 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2293 #define for_each_netdev_in_bond_rcu(bond, slave) \
2294 for_each_netdev_rcu(&init_net, slave) \
2295 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2296 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2298 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2300 struct list_head
*lh
;
2304 lh
= dev
->dev_list
.next
;
2305 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2308 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2310 struct list_head
*lh
;
2314 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2315 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2318 static inline struct net_device
*first_net_device(struct net
*net
)
2320 return list_empty(&net
->dev_base_head
) ? NULL
:
2321 net_device_entry(net
->dev_base_head
.next
);
2324 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2326 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2328 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2331 int netdev_boot_setup_check(struct net_device
*dev
);
2332 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2333 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2334 const char *hwaddr
);
2335 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2336 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2337 void dev_add_pack(struct packet_type
*pt
);
2338 void dev_remove_pack(struct packet_type
*pt
);
2339 void __dev_remove_pack(struct packet_type
*pt
);
2340 void dev_add_offload(struct packet_offload
*po
);
2341 void dev_remove_offload(struct packet_offload
*po
);
2343 int dev_get_iflink(const struct net_device
*dev
);
2344 int dev_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
);
2345 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2346 unsigned short mask
);
2347 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2348 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2349 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2350 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2351 int dev_open(struct net_device
*dev
);
2352 int dev_close(struct net_device
*dev
);
2353 int dev_close_many(struct list_head
*head
, bool unlink
);
2354 void dev_disable_lro(struct net_device
*dev
);
2355 int dev_loopback_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*newskb
);
2356 int dev_queue_xmit(struct sk_buff
*skb
);
2357 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2358 int register_netdevice(struct net_device
*dev
);
2359 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2360 void unregister_netdevice_many(struct list_head
*head
);
2361 static inline void unregister_netdevice(struct net_device
*dev
)
2363 unregister_netdevice_queue(dev
, NULL
);
2366 int netdev_refcnt_read(const struct net_device
*dev
);
2367 void free_netdev(struct net_device
*dev
);
2368 void netdev_freemem(struct net_device
*dev
);
2369 void synchronize_net(void);
2370 int init_dummy_netdev(struct net_device
*dev
);
2372 DECLARE_PER_CPU(int, xmit_recursion
);
2373 static inline int dev_recursion_level(void)
2375 return this_cpu_read(xmit_recursion
);
2378 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2379 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2380 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2381 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2382 int dev_restart(struct net_device
*dev
);
2383 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2385 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2387 return NAPI_GRO_CB(skb
)->data_offset
;
2390 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2392 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2395 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2397 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2400 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2401 unsigned int offset
)
2403 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2406 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2408 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2411 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2412 unsigned int offset
)
2414 if (!pskb_may_pull(skb
, hlen
))
2417 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2418 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2419 return skb
->data
+ offset
;
2422 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2424 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2425 skb_network_offset(skb
);
2428 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2429 const void *start
, unsigned int len
)
2431 if (NAPI_GRO_CB(skb
)->csum_valid
)
2432 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2433 csum_partial(start
, len
, 0));
2436 /* GRO checksum functions. These are logical equivalents of the normal
2437 * checksum functions (in skbuff.h) except that they operate on the GRO
2438 * offsets and fields in sk_buff.
2441 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2443 static inline bool skb_at_gro_remcsum_start(struct sk_buff
*skb
)
2445 return (NAPI_GRO_CB(skb
)->gro_remcsum_start
== skb_gro_offset(skb
));
2448 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2452 return ((skb
->ip_summed
!= CHECKSUM_PARTIAL
||
2453 skb_checksum_start_offset(skb
) <
2454 skb_gro_offset(skb
)) &&
2455 !skb_at_gro_remcsum_start(skb
) &&
2456 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2457 (!zero_okay
|| check
));
2460 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2463 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2464 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2467 NAPI_GRO_CB(skb
)->csum
= psum
;
2469 return __skb_gro_checksum_complete(skb
);
2472 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2474 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2475 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2476 NAPI_GRO_CB(skb
)->csum_cnt
--;
2478 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2479 * verified a new top level checksum or an encapsulated one
2480 * during GRO. This saves work if we fallback to normal path.
2482 __skb_incr_checksum_unnecessary(skb
);
2486 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2489 __sum16 __ret = 0; \
2490 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2491 __ret = __skb_gro_checksum_validate_complete(skb, \
2492 compute_pseudo(skb, proto)); \
2494 __skb_mark_checksum_bad(skb); \
2496 skb_gro_incr_csum_unnecessary(skb); \
2500 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2501 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2503 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2505 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2507 #define skb_gro_checksum_simple_validate(skb) \
2508 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2510 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2512 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2513 !NAPI_GRO_CB(skb
)->csum_valid
);
2516 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2517 __sum16 check
, __wsum pseudo
)
2519 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2520 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2523 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2525 if (__skb_gro_checksum_convert_check(skb)) \
2526 __skb_gro_checksum_convert(skb, check, \
2527 compute_pseudo(skb, proto)); \
2530 struct gro_remcsum
{
2535 static inline void skb_gro_remcsum_init(struct gro_remcsum
*grc
)
2541 static inline void *skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2542 unsigned int off
, size_t hdrlen
,
2543 int start
, int offset
,
2544 struct gro_remcsum
*grc
,
2548 size_t plen
= hdrlen
+ max_t(size_t, offset
+ sizeof(u16
), start
);
2550 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2553 NAPI_GRO_CB(skb
)->gro_remcsum_start
= off
+ hdrlen
+ start
;
2557 ptr
= skb_gro_header_fast(skb
, off
);
2558 if (skb_gro_header_hard(skb
, off
+ plen
)) {
2559 ptr
= skb_gro_header_slow(skb
, off
+ plen
, off
);
2564 delta
= remcsum_adjust(ptr
+ hdrlen
, NAPI_GRO_CB(skb
)->csum
,
2567 /* Adjust skb->csum since we changed the packet */
2568 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2570 grc
->offset
= off
+ hdrlen
+ offset
;
2576 static inline void skb_gro_remcsum_cleanup(struct sk_buff
*skb
,
2577 struct gro_remcsum
*grc
)
2580 size_t plen
= grc
->offset
+ sizeof(u16
);
2585 ptr
= skb_gro_header_fast(skb
, grc
->offset
);
2586 if (skb_gro_header_hard(skb
, grc
->offset
+ sizeof(u16
))) {
2587 ptr
= skb_gro_header_slow(skb
, plen
, grc
->offset
);
2592 remcsum_unadjust((__sum16
*)ptr
, grc
->delta
);
2595 struct skb_csum_offl_spec
{
2609 bool __skb_csum_offload_chk(struct sk_buff
*skb
,
2610 const struct skb_csum_offl_spec
*spec
,
2611 bool *csum_encapped
,
2614 static inline bool skb_csum_offload_chk(struct sk_buff
*skb
,
2615 const struct skb_csum_offl_spec
*spec
,
2616 bool *csum_encapped
,
2619 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
2622 return __skb_csum_offload_chk(skb
, spec
, csum_encapped
, csum_help
);
2625 static inline bool skb_csum_offload_chk_help(struct sk_buff
*skb
,
2626 const struct skb_csum_offl_spec
*spec
)
2630 return skb_csum_offload_chk(skb
, spec
, &csum_encapped
, true);
2633 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff
*skb
)
2635 static const struct skb_csum_offl_spec csum_offl_spec
= {
2637 .ip_options_okay
= 1,
2644 return skb_csum_offload_chk_help(skb
, &csum_offl_spec
);
2647 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff
*skb
)
2649 static const struct skb_csum_offl_spec csum_offl_spec
= {
2651 .ip_options_okay
= 1,
2657 return skb_csum_offload_chk_help(skb
, &csum_offl_spec
);
2660 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2661 unsigned short type
,
2662 const void *daddr
, const void *saddr
,
2665 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2668 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2671 static inline int dev_parse_header(const struct sk_buff
*skb
,
2672 unsigned char *haddr
)
2674 const struct net_device
*dev
= skb
->dev
;
2676 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2678 return dev
->header_ops
->parse(skb
, haddr
);
2681 /* ll_header must have at least hard_header_len allocated */
2682 static inline bool dev_validate_header(const struct net_device
*dev
,
2683 char *ll_header
, int len
)
2685 if (likely(len
>= dev
->hard_header_len
))
2688 if (capable(CAP_SYS_RAWIO
)) {
2689 memset(ll_header
+ len
, 0, dev
->hard_header_len
- len
);
2693 if (dev
->header_ops
&& dev
->header_ops
->validate
)
2694 return dev
->header_ops
->validate(ll_header
, len
);
2699 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2700 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2701 static inline int unregister_gifconf(unsigned int family
)
2703 return register_gifconf(family
, NULL
);
2706 #ifdef CONFIG_NET_FLOW_LIMIT
2707 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2708 struct sd_flow_limit
{
2710 unsigned int num_buckets
;
2711 unsigned int history_head
;
2712 u16 history
[FLOW_LIMIT_HISTORY
];
2716 extern int netdev_flow_limit_table_len
;
2717 #endif /* CONFIG_NET_FLOW_LIMIT */
2720 * Incoming packets are placed on per-CPU queues
2722 struct softnet_data
{
2723 struct list_head poll_list
;
2724 struct sk_buff_head process_queue
;
2727 unsigned int processed
;
2728 unsigned int time_squeeze
;
2729 unsigned int received_rps
;
2731 struct softnet_data
*rps_ipi_list
;
2733 #ifdef CONFIG_NET_FLOW_LIMIT
2734 struct sd_flow_limit __rcu
*flow_limit
;
2736 struct Qdisc
*output_queue
;
2737 struct Qdisc
**output_queue_tailp
;
2738 struct sk_buff
*completion_queue
;
2741 /* input_queue_head should be written by cpu owning this struct,
2742 * and only read by other cpus. Worth using a cache line.
2744 unsigned int input_queue_head ____cacheline_aligned_in_smp
;
2746 /* Elements below can be accessed between CPUs for RPS/RFS */
2747 struct call_single_data csd ____cacheline_aligned_in_smp
;
2748 struct softnet_data
*rps_ipi_next
;
2750 unsigned int input_queue_tail
;
2752 unsigned int dropped
;
2753 struct sk_buff_head input_pkt_queue
;
2754 struct napi_struct backlog
;
2758 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2761 sd
->input_queue_head
++;
2765 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2766 unsigned int *qtail
)
2769 *qtail
= ++sd
->input_queue_tail
;
2773 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2775 void __netif_schedule(struct Qdisc
*q
);
2776 void netif_schedule_queue(struct netdev_queue
*txq
);
2778 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2782 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2783 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2786 static __always_inline
void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2788 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2792 * netif_start_queue - allow transmit
2793 * @dev: network device
2795 * Allow upper layers to call the device hard_start_xmit routine.
2797 static inline void netif_start_queue(struct net_device
*dev
)
2799 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2802 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2806 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2807 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2808 netif_tx_start_queue(txq
);
2812 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2815 * netif_wake_queue - restart transmit
2816 * @dev: network device
2818 * Allow upper layers to call the device hard_start_xmit routine.
2819 * Used for flow control when transmit resources are available.
2821 static inline void netif_wake_queue(struct net_device
*dev
)
2823 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2826 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2830 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2831 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2832 netif_tx_wake_queue(txq
);
2836 static __always_inline
void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2838 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2842 * netif_stop_queue - stop transmitted packets
2843 * @dev: network device
2845 * Stop upper layers calling the device hard_start_xmit routine.
2846 * Used for flow control when transmit resources are unavailable.
2848 static inline void netif_stop_queue(struct net_device
*dev
)
2850 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2853 void netif_tx_stop_all_queues(struct net_device
*dev
);
2855 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2857 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2861 * netif_queue_stopped - test if transmit queue is flowblocked
2862 * @dev: network device
2864 * Test if transmit queue on device is currently unable to send.
2866 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2868 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2871 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2873 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2877 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2879 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2883 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2885 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2889 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2890 * @dev_queue: pointer to transmit queue
2892 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2893 * to give appropriate hint to the CPU.
2895 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2898 prefetchw(&dev_queue
->dql
.num_queued
);
2903 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2904 * @dev_queue: pointer to transmit queue
2906 * BQL enabled drivers might use this helper in their TX completion path,
2907 * to give appropriate hint to the CPU.
2909 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2912 prefetchw(&dev_queue
->dql
.limit
);
2916 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2920 dql_queued(&dev_queue
->dql
, bytes
);
2922 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2925 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2928 * The XOFF flag must be set before checking the dql_avail below,
2929 * because in netdev_tx_completed_queue we update the dql_completed
2930 * before checking the XOFF flag.
2934 /* check again in case another CPU has just made room avail */
2935 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2936 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2941 * netdev_sent_queue - report the number of bytes queued to hardware
2942 * @dev: network device
2943 * @bytes: number of bytes queued to the hardware device queue
2945 * Report the number of bytes queued for sending/completion to the network
2946 * device hardware queue. @bytes should be a good approximation and should
2947 * exactly match netdev_completed_queue() @bytes
2949 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2951 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2954 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2955 unsigned int pkts
, unsigned int bytes
)
2958 if (unlikely(!bytes
))
2961 dql_completed(&dev_queue
->dql
, bytes
);
2964 * Without the memory barrier there is a small possiblity that
2965 * netdev_tx_sent_queue will miss the update and cause the queue to
2966 * be stopped forever
2970 if (dql_avail(&dev_queue
->dql
) < 0)
2973 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2974 netif_schedule_queue(dev_queue
);
2979 * netdev_completed_queue - report bytes and packets completed by device
2980 * @dev: network device
2981 * @pkts: actual number of packets sent over the medium
2982 * @bytes: actual number of bytes sent over the medium
2984 * Report the number of bytes and packets transmitted by the network device
2985 * hardware queue over the physical medium, @bytes must exactly match the
2986 * @bytes amount passed to netdev_sent_queue()
2988 static inline void netdev_completed_queue(struct net_device
*dev
,
2989 unsigned int pkts
, unsigned int bytes
)
2991 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
2994 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
2997 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
3003 * netdev_reset_queue - reset the packets and bytes count of a network device
3004 * @dev_queue: network device
3006 * Reset the bytes and packet count of a network device and clear the
3007 * software flow control OFF bit for this network device
3009 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
3011 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
3015 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3016 * @dev: network device
3017 * @queue_index: given tx queue index
3019 * Returns 0 if given tx queue index >= number of device tx queues,
3020 * otherwise returns the originally passed tx queue index.
3022 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
3024 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
3025 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3026 dev
->name
, queue_index
,
3027 dev
->real_num_tx_queues
);
3035 * netif_running - test if up
3036 * @dev: network device
3038 * Test if the device has been brought up.
3040 static inline bool netif_running(const struct net_device
*dev
)
3042 return test_bit(__LINK_STATE_START
, &dev
->state
);
3046 * Routines to manage the subqueues on a device. We only need start,
3047 * stop, and a check if it's stopped. All other device management is
3048 * done at the overall netdevice level.
3049 * Also test the device if we're multiqueue.
3053 * netif_start_subqueue - allow sending packets on subqueue
3054 * @dev: network device
3055 * @queue_index: sub queue index
3057 * Start individual transmit queue of a device with multiple transmit queues.
3059 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
3061 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3063 netif_tx_start_queue(txq
);
3067 * netif_stop_subqueue - stop sending packets on subqueue
3068 * @dev: network device
3069 * @queue_index: sub queue index
3071 * Stop individual transmit queue of a device with multiple transmit queues.
3073 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
3075 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3076 netif_tx_stop_queue(txq
);
3080 * netif_subqueue_stopped - test status of subqueue
3081 * @dev: network device
3082 * @queue_index: sub queue index
3084 * Check individual transmit queue of a device with multiple transmit queues.
3086 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
3089 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3091 return netif_tx_queue_stopped(txq
);
3094 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
3095 struct sk_buff
*skb
)
3097 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
3100 void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
);
3103 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
3106 static inline int netif_set_xps_queue(struct net_device
*dev
,
3107 const struct cpumask
*mask
,
3114 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3115 unsigned int num_tx_queues
);
3118 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3119 * as a distribution range limit for the returned value.
3121 static inline u16
skb_tx_hash(const struct net_device
*dev
,
3122 struct sk_buff
*skb
)
3124 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
3128 * netif_is_multiqueue - test if device has multiple transmit queues
3129 * @dev: network device
3131 * Check if device has multiple transmit queues
3133 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
3135 return dev
->num_tx_queues
> 1;
3138 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
3141 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
3143 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
3151 static inline unsigned int get_netdev_rx_queue_index(
3152 struct netdev_rx_queue
*queue
)
3154 struct net_device
*dev
= queue
->dev
;
3155 int index
= queue
- dev
->_rx
;
3157 BUG_ON(index
>= dev
->num_rx_queues
);
3162 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3163 int netif_get_num_default_rss_queues(void);
3165 enum skb_free_reason
{
3166 SKB_REASON_CONSUMED
,
3170 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
3171 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
3174 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3175 * interrupt context or with hardware interrupts being disabled.
3176 * (in_irq() || irqs_disabled())
3178 * We provide four helpers that can be used in following contexts :
3180 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3181 * replacing kfree_skb(skb)
3183 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3184 * Typically used in place of consume_skb(skb) in TX completion path
3186 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3187 * replacing kfree_skb(skb)
3189 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3190 * and consumed a packet. Used in place of consume_skb(skb)
3192 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
3194 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
3197 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
3199 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
3202 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
3204 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
3207 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
3209 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
3212 int netif_rx(struct sk_buff
*skb
);
3213 int netif_rx_ni(struct sk_buff
*skb
);
3214 int netif_receive_skb(struct sk_buff
*skb
);
3215 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
3216 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
3217 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
3218 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
3219 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
3220 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
3222 static inline void napi_free_frags(struct napi_struct
*napi
)
3224 kfree_skb(napi
->skb
);
3228 int netdev_rx_handler_register(struct net_device
*dev
,
3229 rx_handler_func_t
*rx_handler
,
3230 void *rx_handler_data
);
3231 void netdev_rx_handler_unregister(struct net_device
*dev
);
3233 bool dev_valid_name(const char *name
);
3234 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
3235 int dev_ethtool(struct net
*net
, struct ifreq
*);
3236 unsigned int dev_get_flags(const struct net_device
*);
3237 int __dev_change_flags(struct net_device
*, unsigned int flags
);
3238 int dev_change_flags(struct net_device
*, unsigned int);
3239 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
3240 unsigned int gchanges
);
3241 int dev_change_name(struct net_device
*, const char *);
3242 int dev_set_alias(struct net_device
*, const char *, size_t);
3243 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
3244 int dev_set_mtu(struct net_device
*, int);
3245 void dev_set_group(struct net_device
*, int);
3246 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
3247 int dev_change_carrier(struct net_device
*, bool new_carrier
);
3248 int dev_get_phys_port_id(struct net_device
*dev
,
3249 struct netdev_phys_item_id
*ppid
);
3250 int dev_get_phys_port_name(struct net_device
*dev
,
3251 char *name
, size_t len
);
3252 int dev_change_proto_down(struct net_device
*dev
, bool proto_down
);
3253 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
3254 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3255 struct netdev_queue
*txq
, int *ret
);
3256 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3257 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3258 bool is_skb_forwardable(const struct net_device
*dev
,
3259 const struct sk_buff
*skb
);
3261 void dev_queue_xmit_nit(struct sk_buff
*skb
, struct net_device
*dev
);
3263 extern int netdev_budget
;
3265 /* Called by rtnetlink.c:rtnl_unlock() */
3266 void netdev_run_todo(void);
3269 * dev_put - release reference to device
3270 * @dev: network device
3272 * Release reference to device to allow it to be freed.
3274 static inline void dev_put(struct net_device
*dev
)
3276 this_cpu_dec(*dev
->pcpu_refcnt
);
3280 * dev_hold - get reference to device
3281 * @dev: network device
3283 * Hold reference to device to keep it from being freed.
3285 static inline void dev_hold(struct net_device
*dev
)
3287 this_cpu_inc(*dev
->pcpu_refcnt
);
3290 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3291 * and _off may be called from IRQ context, but it is caller
3292 * who is responsible for serialization of these calls.
3294 * The name carrier is inappropriate, these functions should really be
3295 * called netif_lowerlayer_*() because they represent the state of any
3296 * kind of lower layer not just hardware media.
3299 void linkwatch_init_dev(struct net_device
*dev
);
3300 void linkwatch_fire_event(struct net_device
*dev
);
3301 void linkwatch_forget_dev(struct net_device
*dev
);
3304 * netif_carrier_ok - test if carrier present
3305 * @dev: network device
3307 * Check if carrier is present on device
3309 static inline bool netif_carrier_ok(const struct net_device
*dev
)
3311 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
3314 unsigned long dev_trans_start(struct net_device
*dev
);
3316 void __netdev_watchdog_up(struct net_device
*dev
);
3318 void netif_carrier_on(struct net_device
*dev
);
3320 void netif_carrier_off(struct net_device
*dev
);
3323 * netif_dormant_on - mark device as dormant.
3324 * @dev: network device
3326 * Mark device as dormant (as per RFC2863).
3328 * The dormant state indicates that the relevant interface is not
3329 * actually in a condition to pass packets (i.e., it is not 'up') but is
3330 * in a "pending" state, waiting for some external event. For "on-
3331 * demand" interfaces, this new state identifies the situation where the
3332 * interface is waiting for events to place it in the up state.
3334 static inline void netif_dormant_on(struct net_device
*dev
)
3336 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3337 linkwatch_fire_event(dev
);
3341 * netif_dormant_off - set device as not dormant.
3342 * @dev: network device
3344 * Device is not in dormant state.
3346 static inline void netif_dormant_off(struct net_device
*dev
)
3348 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3349 linkwatch_fire_event(dev
);
3353 * netif_dormant - test if carrier present
3354 * @dev: network device
3356 * Check if carrier is present on device
3358 static inline bool netif_dormant(const struct net_device
*dev
)
3360 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3365 * netif_oper_up - test if device is operational
3366 * @dev: network device
3368 * Check if carrier is operational
3370 static inline bool netif_oper_up(const struct net_device
*dev
)
3372 return (dev
->operstate
== IF_OPER_UP
||
3373 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3377 * netif_device_present - is device available or removed
3378 * @dev: network device
3380 * Check if device has not been removed from system.
3382 static inline bool netif_device_present(struct net_device
*dev
)
3384 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3387 void netif_device_detach(struct net_device
*dev
);
3389 void netif_device_attach(struct net_device
*dev
);
3392 * Network interface message level settings
3396 NETIF_MSG_DRV
= 0x0001,
3397 NETIF_MSG_PROBE
= 0x0002,
3398 NETIF_MSG_LINK
= 0x0004,
3399 NETIF_MSG_TIMER
= 0x0008,
3400 NETIF_MSG_IFDOWN
= 0x0010,
3401 NETIF_MSG_IFUP
= 0x0020,
3402 NETIF_MSG_RX_ERR
= 0x0040,
3403 NETIF_MSG_TX_ERR
= 0x0080,
3404 NETIF_MSG_TX_QUEUED
= 0x0100,
3405 NETIF_MSG_INTR
= 0x0200,
3406 NETIF_MSG_TX_DONE
= 0x0400,
3407 NETIF_MSG_RX_STATUS
= 0x0800,
3408 NETIF_MSG_PKTDATA
= 0x1000,
3409 NETIF_MSG_HW
= 0x2000,
3410 NETIF_MSG_WOL
= 0x4000,
3413 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3414 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3415 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3416 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3417 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3418 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3419 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3420 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3421 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3422 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3423 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3424 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3425 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3426 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3427 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3429 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3432 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3433 return default_msg_enable_bits
;
3434 if (debug_value
== 0) /* no output */
3436 /* set low N bits */
3437 return (1 << debug_value
) - 1;
3440 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3442 spin_lock(&txq
->_xmit_lock
);
3443 txq
->xmit_lock_owner
= cpu
;
3446 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3448 spin_lock_bh(&txq
->_xmit_lock
);
3449 txq
->xmit_lock_owner
= smp_processor_id();
3452 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3454 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3456 txq
->xmit_lock_owner
= smp_processor_id();
3460 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3462 txq
->xmit_lock_owner
= -1;
3463 spin_unlock(&txq
->_xmit_lock
);
3466 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3468 txq
->xmit_lock_owner
= -1;
3469 spin_unlock_bh(&txq
->_xmit_lock
);
3472 static inline void txq_trans_update(struct netdev_queue
*txq
)
3474 if (txq
->xmit_lock_owner
!= -1)
3475 txq
->trans_start
= jiffies
;
3478 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3479 static inline void netif_trans_update(struct net_device
*dev
)
3481 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, 0);
3483 if (txq
->trans_start
!= jiffies
)
3484 txq
->trans_start
= jiffies
;
3488 * netif_tx_lock - grab network device transmit lock
3489 * @dev: network device
3491 * Get network device transmit lock
3493 static inline void netif_tx_lock(struct net_device
*dev
)
3498 spin_lock(&dev
->tx_global_lock
);
3499 cpu
= smp_processor_id();
3500 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3501 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3503 /* We are the only thread of execution doing a
3504 * freeze, but we have to grab the _xmit_lock in
3505 * order to synchronize with threads which are in
3506 * the ->hard_start_xmit() handler and already
3507 * checked the frozen bit.
3509 __netif_tx_lock(txq
, cpu
);
3510 set_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3511 __netif_tx_unlock(txq
);
3515 static inline void netif_tx_lock_bh(struct net_device
*dev
)
3521 static inline void netif_tx_unlock(struct net_device
*dev
)
3525 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3526 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3528 /* No need to grab the _xmit_lock here. If the
3529 * queue is not stopped for another reason, we
3532 clear_bit(__QUEUE_STATE_FROZEN
, &txq
->state
);
3533 netif_schedule_queue(txq
);
3535 spin_unlock(&dev
->tx_global_lock
);
3538 static inline void netif_tx_unlock_bh(struct net_device
*dev
)
3540 netif_tx_unlock(dev
);
3544 #define HARD_TX_LOCK(dev, txq, cpu) { \
3545 if ((dev->features & NETIF_F_LLTX) == 0) { \
3546 __netif_tx_lock(txq, cpu); \
3550 #define HARD_TX_TRYLOCK(dev, txq) \
3551 (((dev->features & NETIF_F_LLTX) == 0) ? \
3552 __netif_tx_trylock(txq) : \
3555 #define HARD_TX_UNLOCK(dev, txq) { \
3556 if ((dev->features & NETIF_F_LLTX) == 0) { \
3557 __netif_tx_unlock(txq); \
3561 static inline void netif_tx_disable(struct net_device
*dev
)
3567 cpu
= smp_processor_id();
3568 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
3569 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
3571 __netif_tx_lock(txq
, cpu
);
3572 netif_tx_stop_queue(txq
);
3573 __netif_tx_unlock(txq
);
3578 static inline void netif_addr_lock(struct net_device
*dev
)
3580 spin_lock(&dev
->addr_list_lock
);
3583 static inline void netif_addr_lock_nested(struct net_device
*dev
)
3585 int subclass
= SINGLE_DEPTH_NESTING
;
3587 if (dev
->netdev_ops
->ndo_get_lock_subclass
)
3588 subclass
= dev
->netdev_ops
->ndo_get_lock_subclass(dev
);
3590 spin_lock_nested(&dev
->addr_list_lock
, subclass
);
3593 static inline void netif_addr_lock_bh(struct net_device
*dev
)
3595 spin_lock_bh(&dev
->addr_list_lock
);
3598 static inline void netif_addr_unlock(struct net_device
*dev
)
3600 spin_unlock(&dev
->addr_list_lock
);
3603 static inline void netif_addr_unlock_bh(struct net_device
*dev
)
3605 spin_unlock_bh(&dev
->addr_list_lock
);
3609 * dev_addrs walker. Should be used only for read access. Call with
3610 * rcu_read_lock held.
3612 #define for_each_dev_addr(dev, ha) \
3613 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3615 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3617 void ether_setup(struct net_device
*dev
);
3619 /* Support for loadable net-drivers */
3620 struct net_device
*alloc_netdev_mqs(int sizeof_priv
, const char *name
,
3621 unsigned char name_assign_type
,
3622 void (*setup
)(struct net_device
*),
3623 unsigned int txqs
, unsigned int rxqs
);
3624 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3625 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3627 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3628 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3631 int register_netdev(struct net_device
*dev
);
3632 void unregister_netdev(struct net_device
*dev
);
3634 /* General hardware address lists handling functions */
3635 int __hw_addr_sync(struct netdev_hw_addr_list
*to_list
,
3636 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3637 void __hw_addr_unsync(struct netdev_hw_addr_list
*to_list
,
3638 struct netdev_hw_addr_list
*from_list
, int addr_len
);
3639 int __hw_addr_sync_dev(struct netdev_hw_addr_list
*list
,
3640 struct net_device
*dev
,
3641 int (*sync
)(struct net_device
*, const unsigned char *),
3642 int (*unsync
)(struct net_device
*,
3643 const unsigned char *));
3644 void __hw_addr_unsync_dev(struct netdev_hw_addr_list
*list
,
3645 struct net_device
*dev
,
3646 int (*unsync
)(struct net_device
*,
3647 const unsigned char *));
3648 void __hw_addr_init(struct netdev_hw_addr_list
*list
);
3650 /* Functions used for device addresses handling */
3651 int dev_addr_add(struct net_device
*dev
, const unsigned char *addr
,
3652 unsigned char addr_type
);
3653 int dev_addr_del(struct net_device
*dev
, const unsigned char *addr
,
3654 unsigned char addr_type
);
3655 void dev_addr_flush(struct net_device
*dev
);
3656 int dev_addr_init(struct net_device
*dev
);
3658 /* Functions used for unicast addresses handling */
3659 int dev_uc_add(struct net_device
*dev
, const unsigned char *addr
);
3660 int dev_uc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3661 int dev_uc_del(struct net_device
*dev
, const unsigned char *addr
);
3662 int dev_uc_sync(struct net_device
*to
, struct net_device
*from
);
3663 int dev_uc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3664 void dev_uc_unsync(struct net_device
*to
, struct net_device
*from
);
3665 void dev_uc_flush(struct net_device
*dev
);
3666 void dev_uc_init(struct net_device
*dev
);
3669 * __dev_uc_sync - Synchonize device's unicast list
3670 * @dev: device to sync
3671 * @sync: function to call if address should be added
3672 * @unsync: function to call if address should be removed
3674 * Add newly added addresses to the interface, and release
3675 * addresses that have been deleted.
3677 static inline int __dev_uc_sync(struct net_device
*dev
,
3678 int (*sync
)(struct net_device
*,
3679 const unsigned char *),
3680 int (*unsync
)(struct net_device
*,
3681 const unsigned char *))
3683 return __hw_addr_sync_dev(&dev
->uc
, dev
, sync
, unsync
);
3687 * __dev_uc_unsync - Remove synchronized addresses from device
3688 * @dev: device to sync
3689 * @unsync: function to call if address should be removed
3691 * Remove all addresses that were added to the device by dev_uc_sync().
3693 static inline void __dev_uc_unsync(struct net_device
*dev
,
3694 int (*unsync
)(struct net_device
*,
3695 const unsigned char *))
3697 __hw_addr_unsync_dev(&dev
->uc
, dev
, unsync
);
3700 /* Functions used for multicast addresses handling */
3701 int dev_mc_add(struct net_device
*dev
, const unsigned char *addr
);
3702 int dev_mc_add_global(struct net_device
*dev
, const unsigned char *addr
);
3703 int dev_mc_add_excl(struct net_device
*dev
, const unsigned char *addr
);
3704 int dev_mc_del(struct net_device
*dev
, const unsigned char *addr
);
3705 int dev_mc_del_global(struct net_device
*dev
, const unsigned char *addr
);
3706 int dev_mc_sync(struct net_device
*to
, struct net_device
*from
);
3707 int dev_mc_sync_multiple(struct net_device
*to
, struct net_device
*from
);
3708 void dev_mc_unsync(struct net_device
*to
, struct net_device
*from
);
3709 void dev_mc_flush(struct net_device
*dev
);
3710 void dev_mc_init(struct net_device
*dev
);
3713 * __dev_mc_sync - Synchonize device's multicast list
3714 * @dev: device to sync
3715 * @sync: function to call if address should be added
3716 * @unsync: function to call if address should be removed
3718 * Add newly added addresses to the interface, and release
3719 * addresses that have been deleted.
3721 static inline int __dev_mc_sync(struct net_device
*dev
,
3722 int (*sync
)(struct net_device
*,
3723 const unsigned char *),
3724 int (*unsync
)(struct net_device
*,
3725 const unsigned char *))
3727 return __hw_addr_sync_dev(&dev
->mc
, dev
, sync
, unsync
);
3731 * __dev_mc_unsync - Remove synchronized addresses from device
3732 * @dev: device to sync
3733 * @unsync: function to call if address should be removed
3735 * Remove all addresses that were added to the device by dev_mc_sync().
3737 static inline void __dev_mc_unsync(struct net_device
*dev
,
3738 int (*unsync
)(struct net_device
*,
3739 const unsigned char *))
3741 __hw_addr_unsync_dev(&dev
->mc
, dev
, unsync
);
3744 /* Functions used for secondary unicast and multicast support */
3745 void dev_set_rx_mode(struct net_device
*dev
);
3746 void __dev_set_rx_mode(struct net_device
*dev
);
3747 int dev_set_promiscuity(struct net_device
*dev
, int inc
);
3748 int dev_set_allmulti(struct net_device
*dev
, int inc
);
3749 void netdev_state_change(struct net_device
*dev
);
3750 void netdev_notify_peers(struct net_device
*dev
);
3751 void netdev_features_change(struct net_device
*dev
);
3752 /* Load a device via the kmod */
3753 void dev_load(struct net
*net
, const char *name
);
3754 struct rtnl_link_stats64
*dev_get_stats(struct net_device
*dev
,
3755 struct rtnl_link_stats64
*storage
);
3756 void netdev_stats_to_stats64(struct rtnl_link_stats64
*stats64
,
3757 const struct net_device_stats
*netdev_stats
);
3759 extern int netdev_max_backlog
;
3760 extern int netdev_tstamp_prequeue
;
3761 extern int weight_p
;
3763 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3764 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3765 struct list_head
**iter
);
3766 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3767 struct list_head
**iter
);
3769 /* iterate through upper list, must be called under RCU read lock */
3770 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3771 for (iter = &(dev)->adj_list.upper, \
3772 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3774 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3776 /* iterate through upper list, must be called under RCU read lock */
3777 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3778 for (iter = &(dev)->all_adj_list.upper, \
3779 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3781 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3783 void *netdev_lower_get_next_private(struct net_device
*dev
,
3784 struct list_head
**iter
);
3785 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3786 struct list_head
**iter
);
3788 #define netdev_for_each_lower_private(dev, priv, iter) \
3789 for (iter = (dev)->adj_list.lower.next, \
3790 priv = netdev_lower_get_next_private(dev, &(iter)); \
3792 priv = netdev_lower_get_next_private(dev, &(iter)))
3794 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3795 for (iter = &(dev)->adj_list.lower, \
3796 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3798 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3800 void *netdev_lower_get_next(struct net_device
*dev
,
3801 struct list_head
**iter
);
3802 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3803 for (iter = (dev)->adj_list.lower.next, \
3804 ldev = netdev_lower_get_next(dev, &(iter)); \
3806 ldev = netdev_lower_get_next(dev, &(iter)))
3808 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3809 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3810 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3811 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3812 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
3813 int netdev_master_upper_dev_link(struct net_device
*dev
,
3814 struct net_device
*upper_dev
,
3815 void *upper_priv
, void *upper_info
);
3816 void netdev_upper_dev_unlink(struct net_device
*dev
,
3817 struct net_device
*upper_dev
);
3818 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3819 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3820 struct net_device
*lower_dev
);
3821 void netdev_lower_state_changed(struct net_device
*lower_dev
,
3822 void *lower_state_info
);
3824 /* RSS keys are 40 or 52 bytes long */
3825 #define NETDEV_RSS_KEY_LEN 52
3826 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
] __read_mostly
;
3827 void netdev_rss_key_fill(void *buffer
, size_t len
);
3829 int dev_get_nest_level(struct net_device
*dev
,
3830 bool (*type_check
)(const struct net_device
*dev
));
3831 int skb_checksum_help(struct sk_buff
*skb
);
3832 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3833 netdev_features_t features
, bool tx_path
);
3834 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3835 netdev_features_t features
);
3837 struct netdev_bonding_info
{
3842 struct netdev_notifier_bonding_info
{
3843 struct netdev_notifier_info info
; /* must be first */
3844 struct netdev_bonding_info bonding_info
;
3847 void netdev_bonding_info_change(struct net_device
*dev
,
3848 struct netdev_bonding_info
*bonding_info
);
3851 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3853 return __skb_gso_segment(skb
, features
, true);
3855 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3857 static inline bool can_checksum_protocol(netdev_features_t features
,
3860 if (protocol
== htons(ETH_P_FCOE
))
3861 return !!(features
& NETIF_F_FCOE_CRC
);
3863 /* Assume this is an IP checksum (not SCTP CRC) */
3865 if (features
& NETIF_F_HW_CSUM
) {
3866 /* Can checksum everything */
3871 case htons(ETH_P_IP
):
3872 return !!(features
& NETIF_F_IP_CSUM
);
3873 case htons(ETH_P_IPV6
):
3874 return !!(features
& NETIF_F_IPV6_CSUM
);
3880 /* Map an ethertype into IP protocol if possible */
3881 static inline int eproto_to_ipproto(int eproto
)
3884 case htons(ETH_P_IP
):
3886 case htons(ETH_P_IPV6
):
3887 return IPPROTO_IPV6
;
3894 void netdev_rx_csum_fault(struct net_device
*dev
);
3896 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3900 /* rx skb timestamps */
3901 void net_enable_timestamp(void);
3902 void net_disable_timestamp(void);
3904 #ifdef CONFIG_PROC_FS
3905 int __init
dev_proc_init(void);
3907 #define dev_proc_init() 0
3910 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
3911 struct sk_buff
*skb
, struct net_device
*dev
,
3914 skb
->xmit_more
= more
? 1 : 0;
3915 return ops
->ndo_start_xmit(skb
, dev
);
3918 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3919 struct netdev_queue
*txq
, bool more
)
3921 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3924 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
3925 if (rc
== NETDEV_TX_OK
)
3926 txq_trans_update(txq
);
3931 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3933 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3936 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3938 return netdev_class_create_file_ns(class_attr
, NULL
);
3941 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3943 netdev_class_remove_file_ns(class_attr
, NULL
);
3946 extern struct kobj_ns_type_operations net_ns_type_operations
;
3948 const char *netdev_drivername(const struct net_device
*dev
);
3950 void linkwatch_run_queue(void);
3952 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
3953 netdev_features_t f2
)
3955 if ((f1
^ f2
) & NETIF_F_HW_CSUM
) {
3956 if (f1
& NETIF_F_HW_CSUM
)
3957 f1
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
3959 f2
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
3965 static inline netdev_features_t
netdev_get_wanted_features(
3966 struct net_device
*dev
)
3968 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
3970 netdev_features_t
netdev_increment_features(netdev_features_t all
,
3971 netdev_features_t one
, netdev_features_t mask
);
3973 /* Allow TSO being used on stacked device :
3974 * Performing the GSO segmentation before last device
3975 * is a performance improvement.
3977 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
3978 netdev_features_t mask
)
3980 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
3983 int __netdev_update_features(struct net_device
*dev
);
3984 void netdev_update_features(struct net_device
*dev
);
3985 void netdev_change_features(struct net_device
*dev
);
3987 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
3988 struct net_device
*dev
);
3990 netdev_features_t
passthru_features_check(struct sk_buff
*skb
,
3991 struct net_device
*dev
,
3992 netdev_features_t features
);
3993 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
3995 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
3997 netdev_features_t feature
= (netdev_features_t
)gso_type
<< NETIF_F_GSO_SHIFT
;
3999 /* check flags correspondence */
4000 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
4001 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
4002 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
4003 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> NETIF_F_GSO_SHIFT
));
4004 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID
!= (NETIF_F_TSO_MANGLEID
>> NETIF_F_GSO_SHIFT
));
4005 BUILD_BUG_ON(SKB_GSO_TCPV6
!= (NETIF_F_TSO6
>> NETIF_F_GSO_SHIFT
));
4006 BUILD_BUG_ON(SKB_GSO_FCOE
!= (NETIF_F_FSO
>> NETIF_F_GSO_SHIFT
));
4007 BUILD_BUG_ON(SKB_GSO_GRE
!= (NETIF_F_GSO_GRE
>> NETIF_F_GSO_SHIFT
));
4008 BUILD_BUG_ON(SKB_GSO_GRE_CSUM
!= (NETIF_F_GSO_GRE_CSUM
>> NETIF_F_GSO_SHIFT
));
4009 BUILD_BUG_ON(SKB_GSO_IPXIP4
!= (NETIF_F_GSO_IPXIP4
>> NETIF_F_GSO_SHIFT
));
4010 BUILD_BUG_ON(SKB_GSO_IPXIP6
!= (NETIF_F_GSO_IPXIP6
>> NETIF_F_GSO_SHIFT
));
4011 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL
!= (NETIF_F_GSO_UDP_TUNNEL
>> NETIF_F_GSO_SHIFT
));
4012 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM
!= (NETIF_F_GSO_UDP_TUNNEL_CSUM
>> NETIF_F_GSO_SHIFT
));
4013 BUILD_BUG_ON(SKB_GSO_PARTIAL
!= (NETIF_F_GSO_PARTIAL
>> NETIF_F_GSO_SHIFT
));
4014 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
4016 return (features
& feature
) == feature
;
4019 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
4021 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
4022 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
4025 static inline bool netif_needs_gso(struct sk_buff
*skb
,
4026 netdev_features_t features
)
4028 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
4029 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
4030 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
4033 static inline void netif_set_gso_max_size(struct net_device
*dev
,
4036 dev
->gso_max_size
= size
;
4039 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
4040 int pulled_hlen
, u16 mac_offset
,
4043 skb
->protocol
= protocol
;
4044 skb
->encapsulation
= 1;
4045 skb_push(skb
, pulled_hlen
);
4046 skb_reset_transport_header(skb
);
4047 skb
->mac_header
= mac_offset
;
4048 skb
->network_header
= skb
->mac_header
+ mac_len
;
4049 skb
->mac_len
= mac_len
;
4052 static inline bool netif_is_macsec(const struct net_device
*dev
)
4054 return dev
->priv_flags
& IFF_MACSEC
;
4057 static inline bool netif_is_macvlan(const struct net_device
*dev
)
4059 return dev
->priv_flags
& IFF_MACVLAN
;
4062 static inline bool netif_is_macvlan_port(const struct net_device
*dev
)
4064 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
4067 static inline bool netif_is_ipvlan(const struct net_device
*dev
)
4069 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
4072 static inline bool netif_is_ipvlan_port(const struct net_device
*dev
)
4074 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
4077 static inline bool netif_is_bond_master(const struct net_device
*dev
)
4079 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
4082 static inline bool netif_is_bond_slave(const struct net_device
*dev
)
4084 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
4087 static inline bool netif_supports_nofcs(struct net_device
*dev
)
4089 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
4092 static inline bool netif_is_l3_master(const struct net_device
*dev
)
4094 return dev
->priv_flags
& IFF_L3MDEV_MASTER
;
4097 static inline bool netif_is_l3_slave(const struct net_device
*dev
)
4099 return dev
->priv_flags
& IFF_L3MDEV_SLAVE
;
4102 static inline bool netif_is_bridge_master(const struct net_device
*dev
)
4104 return dev
->priv_flags
& IFF_EBRIDGE
;
4107 static inline bool netif_is_bridge_port(const struct net_device
*dev
)
4109 return dev
->priv_flags
& IFF_BRIDGE_PORT
;
4112 static inline bool netif_is_ovs_master(const struct net_device
*dev
)
4114 return dev
->priv_flags
& IFF_OPENVSWITCH
;
4117 static inline bool netif_is_team_master(const struct net_device
*dev
)
4119 return dev
->priv_flags
& IFF_TEAM
;
4122 static inline bool netif_is_team_port(const struct net_device
*dev
)
4124 return dev
->priv_flags
& IFF_TEAM_PORT
;
4127 static inline bool netif_is_lag_master(const struct net_device
*dev
)
4129 return netif_is_bond_master(dev
) || netif_is_team_master(dev
);
4132 static inline bool netif_is_lag_port(const struct net_device
*dev
)
4134 return netif_is_bond_slave(dev
) || netif_is_team_port(dev
);
4137 static inline bool netif_is_rxfh_configured(const struct net_device
*dev
)
4139 return dev
->priv_flags
& IFF_RXFH_CONFIGURED
;
4142 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4143 static inline void netif_keep_dst(struct net_device
*dev
)
4145 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
4148 extern struct pernet_operations __net_initdata loopback_net_ops
;
4150 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4152 /* netdev_printk helpers, similar to dev_printk */
4154 static inline const char *netdev_name(const struct net_device
*dev
)
4156 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
4157 return "(unnamed net_device)";
4161 static inline const char *netdev_reg_state(const struct net_device
*dev
)
4163 switch (dev
->reg_state
) {
4164 case NETREG_UNINITIALIZED
: return " (uninitialized)";
4165 case NETREG_REGISTERED
: return "";
4166 case NETREG_UNREGISTERING
: return " (unregistering)";
4167 case NETREG_UNREGISTERED
: return " (unregistered)";
4168 case NETREG_RELEASED
: return " (released)";
4169 case NETREG_DUMMY
: return " (dummy)";
4172 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
4173 return " (unknown)";
4177 void netdev_printk(const char *level
, const struct net_device
*dev
,
4178 const char *format
, ...);
4180 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
4182 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
4184 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
4186 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
4188 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
4190 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
4192 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
4194 #define MODULE_ALIAS_NETDEV(device) \
4195 MODULE_ALIAS("netdev-" device)
4197 #if defined(CONFIG_DYNAMIC_DEBUG)
4198 #define netdev_dbg(__dev, format, args...) \
4200 dynamic_netdev_dbg(__dev, format, ##args); \
4202 #elif defined(DEBUG)
4203 #define netdev_dbg(__dev, format, args...) \
4204 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4206 #define netdev_dbg(__dev, format, args...) \
4209 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4213 #if defined(VERBOSE_DEBUG)
4214 #define netdev_vdbg netdev_dbg
4217 #define netdev_vdbg(dev, format, args...) \
4220 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4226 * netdev_WARN() acts like dev_printk(), but with the key difference
4227 * of using a WARN/WARN_ON to get the message out, including the
4228 * file/line information and a backtrace.
4230 #define netdev_WARN(dev, format, args...) \
4231 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4232 netdev_reg_state(dev), ##args)
4234 /* netif printk helpers, similar to netdev_printk */
4236 #define netif_printk(priv, type, level, dev, fmt, args...) \
4238 if (netif_msg_##type(priv)) \
4239 netdev_printk(level, (dev), fmt, ##args); \
4242 #define netif_level(level, priv, type, dev, fmt, args...) \
4244 if (netif_msg_##type(priv)) \
4245 netdev_##level(dev, fmt, ##args); \
4248 #define netif_emerg(priv, type, dev, fmt, args...) \
4249 netif_level(emerg, priv, type, dev, fmt, ##args)
4250 #define netif_alert(priv, type, dev, fmt, args...) \
4251 netif_level(alert, priv, type, dev, fmt, ##args)
4252 #define netif_crit(priv, type, dev, fmt, args...) \
4253 netif_level(crit, priv, type, dev, fmt, ##args)
4254 #define netif_err(priv, type, dev, fmt, args...) \
4255 netif_level(err, priv, type, dev, fmt, ##args)
4256 #define netif_warn(priv, type, dev, fmt, args...) \
4257 netif_level(warn, priv, type, dev, fmt, ##args)
4258 #define netif_notice(priv, type, dev, fmt, args...) \
4259 netif_level(notice, priv, type, dev, fmt, ##args)
4260 #define netif_info(priv, type, dev, fmt, args...) \
4261 netif_level(info, priv, type, dev, fmt, ##args)
4263 #if defined(CONFIG_DYNAMIC_DEBUG)
4264 #define netif_dbg(priv, type, netdev, format, args...) \
4266 if (netif_msg_##type(priv)) \
4267 dynamic_netdev_dbg(netdev, format, ##args); \
4269 #elif defined(DEBUG)
4270 #define netif_dbg(priv, type, dev, format, args...) \
4271 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4273 #define netif_dbg(priv, type, dev, format, args...) \
4276 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4281 #if defined(VERBOSE_DEBUG)
4282 #define netif_vdbg netif_dbg
4284 #define netif_vdbg(priv, type, dev, format, args...) \
4287 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4293 * The list of packet types we will receive (as opposed to discard)
4294 * and the routines to invoke.
4296 * Why 16. Because with 16 the only overlap we get on a hash of the
4297 * low nibble of the protocol value is RARP/SNAP/X.25.
4299 * NOTE: That is no longer true with the addition of VLAN tags. Not
4300 * sure which should go first, but I bet it won't make much
4301 * difference if we are running VLANs. The good news is that
4302 * this protocol won't be in the list unless compiled in, so
4303 * the average user (w/out VLANs) will not be adversely affected.
4319 #define PTYPE_HASH_SIZE (16)
4320 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4322 #endif /* _LINUX_NETDEVICE_H */