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(), 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*/
109 NETDEV_TX_LOCKED
= 0x20, /* driver tx lock was already taken */
111 typedef enum netdev_tx netdev_tx_t
;
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117 static inline bool dev_xmit_complete(int rc
)
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 if (likely(rc
< NET_XMIT_MASK
))
132 * Compute the worst case header length according to the protocols
136 #if defined(CONFIG_HYPERV_NET)
137 # define LL_MAX_HEADER 128
138 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
139 # if defined(CONFIG_MAC80211_MESH)
140 # define LL_MAX_HEADER 128
142 # define LL_MAX_HEADER 96
145 # define LL_MAX_HEADER 32
148 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
150 #define MAX_HEADER LL_MAX_HEADER
152 #define MAX_HEADER (LL_MAX_HEADER + 48)
156 * Old network device statistics. Fields are native words
157 * (unsigned long) so they can be read and written atomically.
160 struct net_device_stats
{
161 unsigned long rx_packets
;
162 unsigned long tx_packets
;
163 unsigned long rx_bytes
;
164 unsigned long tx_bytes
;
165 unsigned long rx_errors
;
166 unsigned long tx_errors
;
167 unsigned long rx_dropped
;
168 unsigned long tx_dropped
;
169 unsigned long multicast
;
170 unsigned long collisions
;
171 unsigned long rx_length_errors
;
172 unsigned long rx_over_errors
;
173 unsigned long rx_crc_errors
;
174 unsigned long rx_frame_errors
;
175 unsigned long rx_fifo_errors
;
176 unsigned long rx_missed_errors
;
177 unsigned long tx_aborted_errors
;
178 unsigned long tx_carrier_errors
;
179 unsigned long tx_fifo_errors
;
180 unsigned long tx_heartbeat_errors
;
181 unsigned long tx_window_errors
;
182 unsigned long rx_compressed
;
183 unsigned long tx_compressed
;
187 #include <linux/cache.h>
188 #include <linux/skbuff.h>
191 #include <linux/static_key.h>
192 extern struct static_key rps_needed
;
199 struct netdev_hw_addr
{
200 struct list_head list
;
201 unsigned char addr
[MAX_ADDR_LEN
];
203 #define NETDEV_HW_ADDR_T_LAN 1
204 #define NETDEV_HW_ADDR_T_SAN 2
205 #define NETDEV_HW_ADDR_T_SLAVE 3
206 #define NETDEV_HW_ADDR_T_UNICAST 4
207 #define NETDEV_HW_ADDR_T_MULTICAST 5
212 struct rcu_head rcu_head
;
215 struct netdev_hw_addr_list
{
216 struct list_head list
;
220 #define netdev_hw_addr_list_count(l) ((l)->count)
221 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222 #define netdev_hw_addr_list_for_each(ha, l) \
223 list_for_each_entry(ha, &(l)->list, list)
225 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
227 #define netdev_for_each_uc_addr(ha, dev) \
228 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
230 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
232 #define netdev_for_each_mc_addr(ha, dev) \
233 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
240 /* cached hardware header; allow for machine alignment needs. */
241 #define HH_DATA_MOD 16
242 #define HH_DATA_OFF(__len) \
243 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
244 #define HH_DATA_ALIGN(__len) \
245 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 unsigned long hh_data
[HH_DATA_ALIGN(LL_MAX_HEADER
) / sizeof(long)];
249 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
251 * dev->hard_header_len ? (dev->hard_header_len +
252 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254 * We could use other alignment values, but we must maintain the
255 * relationship HH alignment <= LL alignment.
257 #define LL_RESERVED_SPACE(dev) \
258 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
259 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
263 int (*create
) (struct sk_buff
*skb
, struct net_device
*dev
,
264 unsigned short type
, const void *daddr
,
265 const void *saddr
, unsigned int len
);
266 int (*parse
)(const struct sk_buff
*skb
, unsigned char *haddr
);
267 int (*cache
)(const struct neighbour
*neigh
, struct hh_cache
*hh
, __be16 type
);
268 void (*cache_update
)(struct hh_cache
*hh
,
269 const struct net_device
*dev
,
270 const unsigned char *haddr
);
271 bool (*validate
)(const char *ll_header
, unsigned int len
);
274 /* These flag bits are private to the generic network queueing
275 * layer, they may not be explicitly referenced by any other
279 enum netdev_state_t
{
281 __LINK_STATE_PRESENT
,
282 __LINK_STATE_NOCARRIER
,
283 __LINK_STATE_LINKWATCH_PENDING
,
284 __LINK_STATE_DORMANT
,
289 * This structure holds at boot time configured netdevice settings. They
290 * are then used in the device probing.
292 struct netdev_boot_setup
{
296 #define NETDEV_BOOT_SETUP_MAX 8
298 int __init
netdev_boot_setup(char *str
);
301 * Structure for NAPI scheduling similar to tasklet but with weighting
304 /* The poll_list must only be managed by the entity which
305 * changes the state of the NAPI_STATE_SCHED bit. This means
306 * whoever atomically sets that bit can add this napi_struct
307 * to the per-cpu poll_list, and whoever clears that bit
308 * can remove from the list right before clearing the bit.
310 struct list_head poll_list
;
314 unsigned int gro_count
;
315 int (*poll
)(struct napi_struct
*, int);
316 #ifdef CONFIG_NETPOLL
317 spinlock_t poll_lock
;
320 struct net_device
*dev
;
321 struct sk_buff
*gro_list
;
323 struct hrtimer timer
;
324 struct list_head dev_list
;
325 struct hlist_node napi_hash_node
;
326 unsigned int napi_id
;
330 NAPI_STATE_SCHED
, /* Poll is scheduled */
331 NAPI_STATE_DISABLE
, /* Disable pending */
332 NAPI_STATE_NPSVC
, /* Netpoll - don't dequeue from poll_list */
333 NAPI_STATE_HASHED
, /* In NAPI hash (busy polling possible) */
334 NAPI_STATE_NO_BUSY_POLL
,/* Do not add in napi_hash, no busy polling */
344 typedef enum gro_result gro_result_t
;
347 * enum rx_handler_result - Possible return values for rx_handlers.
348 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
350 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
351 * case skb->dev was changed by rx_handler.
352 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
353 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
355 * rx_handlers are functions called from inside __netif_receive_skb(), to do
356 * special processing of the skb, prior to delivery to protocol handlers.
358 * Currently, a net_device can only have a single rx_handler registered. Trying
359 * to register a second rx_handler will return -EBUSY.
361 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
362 * To unregister a rx_handler on a net_device, use
363 * netdev_rx_handler_unregister().
365 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
368 * If the rx_handler consumed to skb in some way, it should return
369 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
370 * the skb to be delivered in some other ways.
372 * If the rx_handler changed skb->dev, to divert the skb to another
373 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
374 * new device will be called if it exists.
376 * If the rx_handler consider the skb should be ignored, it should return
377 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
378 * are registered on exact device (ptype->dev == skb->dev).
380 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
381 * delivered, it should return RX_HANDLER_PASS.
383 * A device without a registered rx_handler will behave as if rx_handler
384 * returned RX_HANDLER_PASS.
387 enum rx_handler_result
{
393 typedef enum rx_handler_result rx_handler_result_t
;
394 typedef rx_handler_result_t
rx_handler_func_t(struct sk_buff
**pskb
);
396 void __napi_schedule(struct napi_struct
*n
);
397 void __napi_schedule_irqoff(struct napi_struct
*n
);
399 static inline bool napi_disable_pending(struct napi_struct
*n
)
401 return test_bit(NAPI_STATE_DISABLE
, &n
->state
);
405 * napi_schedule_prep - check if napi can be scheduled
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
413 static inline bool napi_schedule_prep(struct napi_struct
*n
)
415 return !napi_disable_pending(n
) &&
416 !test_and_set_bit(NAPI_STATE_SCHED
, &n
->state
);
420 * napi_schedule - schedule NAPI poll
423 * Schedule NAPI poll routine to be called if it is not already
426 static inline void napi_schedule(struct napi_struct
*n
)
428 if (napi_schedule_prep(n
))
433 * napi_schedule_irqoff - schedule NAPI poll
436 * Variant of napi_schedule(), assuming hard irqs are masked.
438 static inline void napi_schedule_irqoff(struct napi_struct
*n
)
440 if (napi_schedule_prep(n
))
441 __napi_schedule_irqoff(n
);
444 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
445 static inline bool napi_reschedule(struct napi_struct
*napi
)
447 if (napi_schedule_prep(napi
)) {
448 __napi_schedule(napi
);
454 void __napi_complete(struct napi_struct
*n
);
455 void napi_complete_done(struct napi_struct
*n
, int work_done
);
457 * napi_complete - NAPI processing complete
460 * Mark NAPI processing as complete.
461 * Consider using napi_complete_done() instead.
463 static inline void napi_complete(struct napi_struct
*n
)
465 return napi_complete_done(n
, 0);
469 * napi_hash_add - add a NAPI to global hashtable
470 * @napi: napi context
472 * generate a new napi_id and store a @napi under it in napi_hash
473 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
474 * Note: This is normally automatically done from netif_napi_add(),
475 * so might disappear in a future linux version.
477 void napi_hash_add(struct napi_struct
*napi
);
480 * napi_hash_del - remove a NAPI from global table
481 * @napi: napi context
483 * Warning: caller must observe rcu grace period
484 * before freeing memory containing @napi, if
485 * this function returns true.
486 * Note: core networking stack automatically calls it
487 * from netif_napi_del()
488 * Drivers might want to call this helper to combine all
489 * the needed rcu grace periods into a single one.
491 bool napi_hash_del(struct napi_struct
*napi
);
494 * napi_disable - prevent NAPI from scheduling
497 * Stop NAPI from being scheduled on this context.
498 * Waits till any outstanding processing completes.
500 void napi_disable(struct napi_struct
*n
);
503 * napi_enable - enable NAPI scheduling
506 * Resume NAPI from being scheduled on this context.
507 * Must be paired with napi_disable.
509 static inline void napi_enable(struct napi_struct
*n
)
511 BUG_ON(!test_bit(NAPI_STATE_SCHED
, &n
->state
));
512 smp_mb__before_atomic();
513 clear_bit(NAPI_STATE_SCHED
, &n
->state
);
514 clear_bit(NAPI_STATE_NPSVC
, &n
->state
);
518 * napi_synchronize - wait until NAPI is not running
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
525 static inline void napi_synchronize(const struct napi_struct
*n
)
527 if (IS_ENABLED(CONFIG_SMP
))
528 while (test_bit(NAPI_STATE_SCHED
, &n
->state
))
534 enum netdev_queue_state_t
{
535 __QUEUE_STATE_DRV_XOFF
,
536 __QUEUE_STATE_STACK_XOFF
,
537 __QUEUE_STATE_FROZEN
,
540 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
541 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
544 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
547 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
552 * netif_tx_* functions below are used to manipulate this flag. The
553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554 * queue independently. The netif_xmit_*stopped functions below are called
555 * to check if the queue has been stopped by the driver or stack (either
556 * of the XOFF bits are set in the state). Drivers should not need to call
557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
560 struct netdev_queue
{
564 struct net_device
*dev
;
565 struct Qdisc __rcu
*qdisc
;
566 struct Qdisc
*qdisc_sleeping
;
570 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp
;
579 * please use this field instead of dev->trans_start
581 unsigned long trans_start
;
584 * Number of TX timeouts for this queue
585 * (/sys/class/net/DEV/Q/trans_timeout)
587 unsigned long trans_timeout
;
594 unsigned long tx_maxrate
;
595 } ____cacheline_aligned_in_smp
;
597 static inline int netdev_queue_numa_node_read(const struct netdev_queue
*q
)
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
606 static inline void netdev_queue_numa_node_write(struct netdev_queue
*q
, int node
)
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
623 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
630 struct rps_dev_flow
{
633 unsigned int last_qtail
;
635 #define RPS_NO_FILTER 0xffff
638 * The rps_dev_flow_table structure contains a table of flow mappings.
640 struct rps_dev_flow_table
{
643 struct rps_dev_flow flows
[0];
645 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
646 ((_num) * sizeof(struct rps_dev_flow)))
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
651 * Each entry is a 32bit value. Upper part is the high order bits
652 * of flow hash, lower part is cpu number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
658 struct rps_sock_flow_table
{
661 u32 ents
[0] ____cacheline_aligned_in_smp
;
663 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665 #define RPS_NO_CPU 0xffff
667 extern u32 rps_cpu_mask
;
668 extern struct rps_sock_flow_table __rcu
*rps_sock_flow_table
;
670 static inline void rps_record_sock_flow(struct rps_sock_flow_table
*table
,
674 unsigned int index
= hash
& table
->mask
;
675 u32 val
= hash
& ~rps_cpu_mask
;
677 /* We only give a hint, preemption can change cpu under us */
678 val
|= raw_smp_processor_id();
680 if (table
->ents
[index
] != val
)
681 table
->ents
[index
] = val
;
685 #ifdef CONFIG_RFS_ACCEL
686 bool rps_may_expire_flow(struct net_device
*dev
, u16 rxq_index
, u32 flow_id
,
689 #endif /* CONFIG_RPS */
691 /* This structure contains an instance of an RX queue. */
692 struct netdev_rx_queue
{
694 struct rps_map __rcu
*rps_map
;
695 struct rps_dev_flow_table __rcu
*rps_flow_table
;
698 struct net_device
*dev
;
699 } ____cacheline_aligned_in_smp
;
702 * RX queue sysfs structures and functions.
704 struct rx_queue_attribute
{
705 struct attribute attr
;
706 ssize_t (*show
)(struct netdev_rx_queue
*queue
,
707 struct rx_queue_attribute
*attr
, char *buf
);
708 ssize_t (*store
)(struct netdev_rx_queue
*queue
,
709 struct rx_queue_attribute
*attr
, const char *buf
, size_t len
);
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
719 unsigned int alloc_len
;
723 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
724 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
730 struct xps_dev_maps
{
732 struct xps_map __rcu
*cpu_map
[0];
734 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * sizeof(struct xps_map *)))
736 #endif /* CONFIG_XPS */
738 #define TC_MAX_QUEUE 16
739 #define TC_BITMASK 15
740 /* HW offloaded queuing disciplines txq count and offset maps */
741 struct netdev_tc_txq
{
746 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
751 struct netdev_fcoe_hbainfo
{
752 char manufacturer
[64];
753 char serial_number
[64];
754 char hardware_version
[64];
755 char driver_version
[64];
756 char optionrom_version
[64];
757 char firmware_version
[64];
759 char model_description
[256];
763 #define MAX_PHYS_ITEM_ID_LEN 32
765 /* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
768 struct netdev_phys_item_id
{
769 unsigned char id
[MAX_PHYS_ITEM_ID_LEN
];
770 unsigned char id_len
;
773 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id
*a
,
774 struct netdev_phys_item_id
*b
)
776 return a
->id_len
== b
->id_len
&&
777 memcmp(a
->id
, b
->id
, a
->id_len
) == 0;
780 typedef u16 (*select_queue_fallback_t
)(struct net_device
*dev
,
781 struct sk_buff
*skb
);
783 /* These structures hold the attributes of qdisc and classifiers
784 * that are being passed to the netdevice through the setup_tc op.
792 struct tc_cls_u32_offload
;
794 struct tc_to_netdev
{
798 struct tc_cls_u32_offload
*cls_u32
;
799 struct tc_cls_flower_offload
*cls_flower
;
805 * This structure defines the management hooks for network devices.
806 * The following hooks can be defined; unless noted otherwise, they are
807 * optional and can be filled with a null pointer.
809 * int (*ndo_init)(struct net_device *dev);
810 * This function is called once when network device is registered.
811 * The network device can use this to any late stage initializaton
812 * or semantic validattion. It can fail with an error code which will
813 * be propogated back to register_netdev
815 * void (*ndo_uninit)(struct net_device *dev);
816 * This function is called when device is unregistered or when registration
817 * fails. It is not called if init fails.
819 * int (*ndo_open)(struct net_device *dev);
820 * This function is called when network device transistions to the up
823 * int (*ndo_stop)(struct net_device *dev);
824 * This function is called when network device transistions to the down
827 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
828 * struct net_device *dev);
829 * Called when a packet needs to be transmitted.
830 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
831 * the queue before that can happen; it's for obsolete devices and weird
832 * corner cases, but the stack really does a non-trivial amount
833 * of useless work if you return NETDEV_TX_BUSY.
834 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
835 * Required can not be NULL.
837 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
838 * netdev_features_t features);
839 * Adjusts the requested feature flags according to device-specific
840 * constraints, and returns the resulting flags. Must not modify
843 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
844 * void *accel_priv, select_queue_fallback_t fallback);
845 * Called to decide which queue to when device supports multiple
848 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
849 * This function is called to allow device receiver to make
850 * changes to configuration when multicast or promiscious is enabled.
852 * void (*ndo_set_rx_mode)(struct net_device *dev);
853 * This function is called device changes address list filtering.
854 * If driver handles unicast address filtering, it should set
855 * IFF_UNICAST_FLT to its priv_flags.
857 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
858 * This function is called when the Media Access Control address
859 * needs to be changed. If this interface is not defined, the
860 * mac address can not be changed.
862 * int (*ndo_validate_addr)(struct net_device *dev);
863 * Test if Media Access Control address is valid for the device.
865 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
866 * Called when a user request an ioctl which can't be handled by
867 * the generic interface code. If not defined ioctl's return
868 * not supported error code.
870 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
871 * Used to set network devices bus interface parameters. This interface
872 * is retained for legacy reason, new devices should use the bus
873 * interface (PCI) for low level management.
875 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
876 * Called when a user wants to change the Maximum Transfer Unit
877 * of a device. If not defined, any request to change MTU will
878 * will return an error.
880 * void (*ndo_tx_timeout)(struct net_device *dev);
881 * Callback uses when the transmitter has not made any progress
882 * for dev->watchdog ticks.
884 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
885 * struct rtnl_link_stats64 *storage);
886 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
887 * Called when a user wants to get the network device usage
888 * statistics. Drivers must do one of the following:
889 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
890 * rtnl_link_stats64 structure passed by the caller.
891 * 2. Define @ndo_get_stats to update a net_device_stats structure
892 * (which should normally be dev->stats) and return a pointer to
893 * it. The structure may be changed asynchronously only if each
894 * field is written atomically.
895 * 3. Update dev->stats asynchronously and atomically, and define
898 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
899 * If device support VLAN filtering this function is called when a
900 * VLAN id is registered.
902 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
903 * If device support VLAN filtering this function is called when a
904 * VLAN id is unregistered.
906 * void (*ndo_poll_controller)(struct net_device *dev);
908 * SR-IOV management functions.
909 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
910 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
911 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
913 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
914 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
915 * int (*ndo_get_vf_config)(struct net_device *dev,
916 * int vf, struct ifla_vf_info *ivf);
917 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
918 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
919 * struct nlattr *port[]);
921 * Enable or disable the VF ability to query its RSS Redirection Table and
922 * Hash Key. This is needed since on some devices VF share this information
923 * with PF and querying it may adduce a theoretical security risk.
924 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
925 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
926 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
927 * Called to setup 'tc' number of traffic classes in the net device. This
928 * is always called from the stack with the rtnl lock held and netif tx
929 * queues stopped. This allows the netdevice to perform queue management
932 * Fiber Channel over Ethernet (FCoE) offload functions.
933 * int (*ndo_fcoe_enable)(struct net_device *dev);
934 * Called when the FCoE protocol stack wants to start using LLD for FCoE
935 * so the underlying device can perform whatever needed configuration or
936 * initialization to support acceleration of FCoE traffic.
938 * int (*ndo_fcoe_disable)(struct net_device *dev);
939 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
940 * so the underlying device can perform whatever needed clean-ups to
941 * stop supporting acceleration of FCoE traffic.
943 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
944 * struct scatterlist *sgl, unsigned int sgc);
945 * Called when the FCoE Initiator wants to initialize an I/O that
946 * is a possible candidate for Direct Data Placement (DDP). The LLD can
947 * perform necessary setup and returns 1 to indicate the device is set up
948 * successfully to perform DDP on this I/O, otherwise this returns 0.
950 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
951 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
952 * indicated by the FC exchange id 'xid', so the underlying device can
953 * clean up and reuse resources for later DDP requests.
955 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
956 * struct scatterlist *sgl, unsigned int sgc);
957 * Called when the FCoE Target wants to initialize an I/O that
958 * is a possible candidate for Direct Data Placement (DDP). The LLD can
959 * perform necessary setup and returns 1 to indicate the device is set up
960 * successfully to perform DDP on this I/O, otherwise this returns 0.
962 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
963 * struct netdev_fcoe_hbainfo *hbainfo);
964 * Called when the FCoE Protocol stack wants information on the underlying
965 * device. This information is utilized by the FCoE protocol stack to
966 * register attributes with Fiber Channel management service as per the
967 * FC-GS Fabric Device Management Information(FDMI) specification.
969 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
970 * Called when the underlying device wants to override default World Wide
971 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
972 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
973 * protocol stack to use.
976 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
977 * u16 rxq_index, u32 flow_id);
978 * Set hardware filter for RFS. rxq_index is the target queue index;
979 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
980 * Return the filter ID on success, or a negative error code.
982 * Slave management functions (for bridge, bonding, etc).
983 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
984 * Called to make another netdev an underling.
986 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
987 * Called to release previously enslaved netdev.
989 * Feature/offload setting functions.
990 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
991 * Called to update device configuration to new features. Passed
992 * feature set might be less than what was returned by ndo_fix_features()).
993 * Must return >0 or -errno if it changed dev->features itself.
995 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
996 * struct net_device *dev,
997 * const unsigned char *addr, u16 vid, u16 flags)
998 * Adds an FDB entry to dev for addr.
999 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1000 * struct net_device *dev,
1001 * const unsigned char *addr, u16 vid)
1002 * Deletes the FDB entry from dev coresponding to addr.
1003 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1004 * struct net_device *dev, struct net_device *filter_dev,
1006 * Used to add FDB entries to dump requests. Implementers should add
1007 * entries to skb and update idx with the number of entries.
1009 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1011 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1012 * struct net_device *dev, u32 filter_mask,
1014 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1017 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1018 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1019 * which do not represent real hardware may define this to allow their
1020 * userspace components to manage their virtual carrier state. Devices
1021 * that determine carrier state from physical hardware properties (eg
1022 * network cables) or protocol-dependent mechanisms (eg
1023 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1025 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1026 * struct netdev_phys_item_id *ppid);
1027 * Called to get ID of physical port of this device. If driver does
1028 * not implement this, it is assumed that the hw is not able to have
1029 * multiple net devices on single physical port.
1031 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1032 * sa_family_t sa_family, __be16 port);
1033 * Called by vxlan to notiy a driver about the UDP port and socket
1034 * address family that vxlan is listnening to. It is called only when
1035 * a new port starts listening. The operation is protected by the
1036 * vxlan_net->sock_lock.
1038 * void (*ndo_add_geneve_port)(struct net_device *dev,
1039 * sa_family_t sa_family, __be16 port);
1040 * Called by geneve to notify a driver about the UDP port and socket
1041 * address family that geneve is listnening to. It is called only when
1042 * a new port starts listening. The operation is protected by the
1043 * geneve_net->sock_lock.
1045 * void (*ndo_del_geneve_port)(struct net_device *dev,
1046 * sa_family_t sa_family, __be16 port);
1047 * Called by geneve to notify the driver about a UDP port and socket
1048 * address family that geneve is not listening to anymore. The operation
1049 * is protected by the geneve_net->sock_lock.
1051 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1052 * sa_family_t sa_family, __be16 port);
1053 * Called by vxlan to notify the driver about a UDP port and socket
1054 * address family that vxlan is not listening to anymore. The operation
1055 * is protected by the vxlan_net->sock_lock.
1057 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1058 * struct net_device *dev)
1059 * Called by upper layer devices to accelerate switching or other
1060 * station functionality into hardware. 'pdev is the lowerdev
1061 * to use for the offload and 'dev' is the net device that will
1062 * back the offload. Returns a pointer to the private structure
1063 * the upper layer will maintain.
1064 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1065 * Called by upper layer device to delete the station created
1066 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1067 * the station and priv is the structure returned by the add
1069 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1070 * struct net_device *dev,
1072 * Callback to use for xmit over the accelerated station. This
1073 * is used in place of ndo_start_xmit on accelerated net
1075 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1076 * struct net_device *dev
1077 * netdev_features_t features);
1078 * Called by core transmit path to determine if device is capable of
1079 * performing offload operations on a given packet. This is to give
1080 * the device an opportunity to implement any restrictions that cannot
1081 * be otherwise expressed by feature flags. The check is called with
1082 * the set of features that the stack has calculated and it returns
1083 * those the driver believes to be appropriate.
1084 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1085 * int queue_index, u32 maxrate);
1086 * Called when a user wants to set a max-rate limitation of specific
1088 * int (*ndo_get_iflink)(const struct net_device *dev);
1089 * Called to get the iflink value of this device.
1090 * void (*ndo_change_proto_down)(struct net_device *dev,
1092 * This function is used to pass protocol port error state information
1093 * to the switch driver. The switch driver can react to the proto_down
1094 * by doing a phys down on the associated switch port.
1095 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1096 * This function is used to get egress tunnel information for given skb.
1097 * This is useful for retrieving outer tunnel header parameters while
1099 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1100 * This function is used to specify the headroom that the skb must
1101 * consider when allocation skb during packet reception. Setting
1102 * appropriate rx headroom value allows avoiding skb head copy on
1103 * forward. Setting a negative value reset the rx headroom to the
1107 struct net_device_ops
{
1108 int (*ndo_init
)(struct net_device
*dev
);
1109 void (*ndo_uninit
)(struct net_device
*dev
);
1110 int (*ndo_open
)(struct net_device
*dev
);
1111 int (*ndo_stop
)(struct net_device
*dev
);
1112 netdev_tx_t (*ndo_start_xmit
)(struct sk_buff
*skb
,
1113 struct net_device
*dev
);
1114 netdev_features_t (*ndo_features_check
)(struct sk_buff
*skb
,
1115 struct net_device
*dev
,
1116 netdev_features_t features
);
1117 u16 (*ndo_select_queue
)(struct net_device
*dev
,
1118 struct sk_buff
*skb
,
1120 select_queue_fallback_t fallback
);
1121 void (*ndo_change_rx_flags
)(struct net_device
*dev
,
1123 void (*ndo_set_rx_mode
)(struct net_device
*dev
);
1124 int (*ndo_set_mac_address
)(struct net_device
*dev
,
1126 int (*ndo_validate_addr
)(struct net_device
*dev
);
1127 int (*ndo_do_ioctl
)(struct net_device
*dev
,
1128 struct ifreq
*ifr
, int cmd
);
1129 int (*ndo_set_config
)(struct net_device
*dev
,
1131 int (*ndo_change_mtu
)(struct net_device
*dev
,
1133 int (*ndo_neigh_setup
)(struct net_device
*dev
,
1134 struct neigh_parms
*);
1135 void (*ndo_tx_timeout
) (struct net_device
*dev
);
1137 struct rtnl_link_stats64
* (*ndo_get_stats64
)(struct net_device
*dev
,
1138 struct rtnl_link_stats64
*storage
);
1139 struct net_device_stats
* (*ndo_get_stats
)(struct net_device
*dev
);
1141 int (*ndo_vlan_rx_add_vid
)(struct net_device
*dev
,
1142 __be16 proto
, u16 vid
);
1143 int (*ndo_vlan_rx_kill_vid
)(struct net_device
*dev
,
1144 __be16 proto
, u16 vid
);
1145 #ifdef CONFIG_NET_POLL_CONTROLLER
1146 void (*ndo_poll_controller
)(struct net_device
*dev
);
1147 int (*ndo_netpoll_setup
)(struct net_device
*dev
,
1148 struct netpoll_info
*info
);
1149 void (*ndo_netpoll_cleanup
)(struct net_device
*dev
);
1151 #ifdef CONFIG_NET_RX_BUSY_POLL
1152 int (*ndo_busy_poll
)(struct napi_struct
*dev
);
1154 int (*ndo_set_vf_mac
)(struct net_device
*dev
,
1155 int queue
, u8
*mac
);
1156 int (*ndo_set_vf_vlan
)(struct net_device
*dev
,
1157 int queue
, u16 vlan
, u8 qos
);
1158 int (*ndo_set_vf_rate
)(struct net_device
*dev
,
1159 int vf
, int min_tx_rate
,
1161 int (*ndo_set_vf_spoofchk
)(struct net_device
*dev
,
1162 int vf
, bool setting
);
1163 int (*ndo_set_vf_trust
)(struct net_device
*dev
,
1164 int vf
, bool setting
);
1165 int (*ndo_get_vf_config
)(struct net_device
*dev
,
1167 struct ifla_vf_info
*ivf
);
1168 int (*ndo_set_vf_link_state
)(struct net_device
*dev
,
1169 int vf
, int link_state
);
1170 int (*ndo_get_vf_stats
)(struct net_device
*dev
,
1172 struct ifla_vf_stats
1174 int (*ndo_set_vf_port
)(struct net_device
*dev
,
1176 struct nlattr
*port
[]);
1177 int (*ndo_get_vf_port
)(struct net_device
*dev
,
1178 int vf
, struct sk_buff
*skb
);
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_L3MDEV_MASTER: device is an L3 master device
1324 * @IFF_NO_QUEUE: device can run without qdisc attached
1325 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1326 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1327 * @IFF_TEAM: device is a team device
1328 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1329 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1330 * entity (i.e. the master device for bridged veth)
1331 * @IFF_MACSEC: device is a MACsec device
1333 enum netdev_priv_flags
{
1334 IFF_802_1Q_VLAN
= 1<<0,
1338 IFF_WAN_HDLC
= 1<<4,
1339 IFF_XMIT_DST_RELEASE
= 1<<5,
1340 IFF_DONT_BRIDGE
= 1<<6,
1341 IFF_DISABLE_NETPOLL
= 1<<7,
1342 IFF_MACVLAN_PORT
= 1<<8,
1343 IFF_BRIDGE_PORT
= 1<<9,
1344 IFF_OVS_DATAPATH
= 1<<10,
1345 IFF_TX_SKB_SHARING
= 1<<11,
1346 IFF_UNICAST_FLT
= 1<<12,
1347 IFF_TEAM_PORT
= 1<<13,
1348 IFF_SUPP_NOFCS
= 1<<14,
1349 IFF_LIVE_ADDR_CHANGE
= 1<<15,
1350 IFF_MACVLAN
= 1<<16,
1351 IFF_XMIT_DST_RELEASE_PERM
= 1<<17,
1352 IFF_IPVLAN_MASTER
= 1<<18,
1353 IFF_IPVLAN_SLAVE
= 1<<19,
1354 IFF_L3MDEV_MASTER
= 1<<20,
1355 IFF_NO_QUEUE
= 1<<21,
1356 IFF_OPENVSWITCH
= 1<<22,
1357 IFF_L3MDEV_SLAVE
= 1<<23,
1359 IFF_RXFH_CONFIGURED
= 1<<25,
1360 IFF_PHONY_HEADROOM
= 1<<26,
1364 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1365 #define IFF_EBRIDGE IFF_EBRIDGE
1366 #define IFF_BONDING IFF_BONDING
1367 #define IFF_ISATAP IFF_ISATAP
1368 #define IFF_WAN_HDLC IFF_WAN_HDLC
1369 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1370 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1371 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1372 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1373 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1374 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1375 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1376 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1377 #define IFF_TEAM_PORT IFF_TEAM_PORT
1378 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1379 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1380 #define IFF_MACVLAN IFF_MACVLAN
1381 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1382 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1383 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1384 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1385 #define IFF_NO_QUEUE IFF_NO_QUEUE
1386 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1387 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1388 #define IFF_TEAM IFF_TEAM
1389 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1390 #define IFF_MACSEC IFF_MACSEC
1393 * struct net_device - The DEVICE structure.
1394 * Actually, this whole structure is a big mistake. It mixes I/O
1395 * data with strictly "high-level" data, and it has to know about
1396 * almost every data structure used in the INET module.
1398 * @name: This is the first field of the "visible" part of this structure
1399 * (i.e. as seen by users in the "Space.c" file). It is the name
1402 * @name_hlist: Device name hash chain, please keep it close to name[]
1403 * @ifalias: SNMP alias
1404 * @mem_end: Shared memory end
1405 * @mem_start: Shared memory start
1406 * @base_addr: Device I/O address
1407 * @irq: Device IRQ number
1409 * @carrier_changes: Stats to monitor carrier on<->off transitions
1411 * @state: Generic network queuing layer state, see netdev_state_t
1412 * @dev_list: The global list of network devices
1413 * @napi_list: List entry, that is used for polling napi devices
1414 * @unreg_list: List entry, that is used, when we are unregistering the
1415 * device, see the function unregister_netdev
1416 * @close_list: List entry, that is used, when we are closing the device
1418 * @adj_list: Directly linked devices, like slaves for bonding
1419 * @all_adj_list: All linked devices, *including* neighbours
1420 * @features: Currently active device features
1421 * @hw_features: User-changeable features
1423 * @wanted_features: User-requested features
1424 * @vlan_features: Mask of features inheritable by VLAN devices
1426 * @hw_enc_features: Mask of features inherited by encapsulating devices
1427 * This field indicates what encapsulation
1428 * offloads the hardware is capable of doing,
1429 * and drivers will need to set them appropriately.
1431 * @mpls_features: Mask of features inheritable by MPLS
1433 * @ifindex: interface index
1434 * @group: The group, that the device belongs to
1436 * @stats: Statistics struct, which was left as a legacy, use
1437 * rtnl_link_stats64 instead
1439 * @rx_dropped: Dropped packets by core network,
1440 * do not use this in drivers
1441 * @tx_dropped: Dropped packets by core network,
1442 * do not use this in drivers
1443 * @rx_nohandler: nohandler dropped packets by core network on
1444 * inactive devices, do not use this in drivers
1446 * @wireless_handlers: List of functions to handle Wireless Extensions,
1448 * see <net/iw_handler.h> for details.
1449 * @wireless_data: Instance data managed by the core of wireless extensions
1451 * @netdev_ops: Includes several pointers to callbacks,
1452 * if one wants to override the ndo_*() functions
1453 * @ethtool_ops: Management operations
1454 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1455 * of Layer 2 headers.
1457 * @flags: Interface flags (a la BSD)
1458 * @priv_flags: Like 'flags' but invisible to userspace,
1459 * see if.h for the definitions
1460 * @gflags: Global flags ( kept as legacy )
1461 * @padded: How much padding added by alloc_netdev()
1462 * @operstate: RFC2863 operstate
1463 * @link_mode: Mapping policy to operstate
1464 * @if_port: Selectable AUI, TP, ...
1466 * @mtu: Interface MTU value
1467 * @type: Interface hardware type
1468 * @hard_header_len: Maximum hardware header length.
1470 * @needed_headroom: Extra headroom the hardware may need, but not in all
1471 * cases can this be guaranteed
1472 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1473 * cases can this be guaranteed. Some cases also use
1474 * LL_MAX_HEADER instead to allocate the skb
1476 * interface address info:
1478 * @perm_addr: Permanent hw address
1479 * @addr_assign_type: Hw address assignment type
1480 * @addr_len: Hardware address length
1481 * @neigh_priv_len; Used in neigh_alloc(),
1482 * initialized only in atm/clip.c
1483 * @dev_id: Used to differentiate devices that share
1484 * the same link layer address
1485 * @dev_port: Used to differentiate devices that share
1487 * @addr_list_lock: XXX: need comments on this one
1488 * @uc_promisc: Counter, that indicates, that promiscuous mode
1489 * has been enabled due to the need to listen to
1490 * additional unicast addresses in a device that
1491 * does not implement ndo_set_rx_mode()
1492 * @uc: unicast mac addresses
1493 * @mc: multicast mac addresses
1494 * @dev_addrs: list of device hw addresses
1495 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1496 * @promiscuity: Number of times, the NIC is told to work in
1497 * Promiscuous mode, if it becomes 0 the NIC will
1498 * exit from working in Promiscuous mode
1499 * @allmulti: Counter, enables or disables allmulticast mode
1501 * @vlan_info: VLAN info
1502 * @dsa_ptr: dsa specific data
1503 * @tipc_ptr: TIPC specific data
1504 * @atalk_ptr: AppleTalk link
1505 * @ip_ptr: IPv4 specific data
1506 * @dn_ptr: DECnet specific data
1507 * @ip6_ptr: IPv6 specific data
1508 * @ax25_ptr: AX.25 specific data
1509 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1511 * @last_rx: Time of last Rx
1512 * @dev_addr: Hw address (before bcast,
1513 * because most packets are unicast)
1515 * @_rx: Array of RX queues
1516 * @num_rx_queues: Number of RX queues
1517 * allocated at register_netdev() time
1518 * @real_num_rx_queues: Number of RX queues currently active in device
1520 * @rx_handler: handler for received packets
1521 * @rx_handler_data: XXX: need comments on this one
1522 * @ingress_queue: XXX: need comments on this one
1523 * @broadcast: hw bcast address
1525 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1526 * indexed by RX queue number. Assigned by driver.
1527 * This must only be set if the ndo_rx_flow_steer
1528 * operation is defined
1529 * @index_hlist: Device index hash chain
1531 * @_tx: Array of TX queues
1532 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1533 * @real_num_tx_queues: Number of TX queues currently active in device
1534 * @qdisc: Root qdisc from userspace point of view
1535 * @tx_queue_len: Max frames per queue allowed
1536 * @tx_global_lock: XXX: need comments on this one
1538 * @xps_maps: XXX: need comments on this one
1540 * @offload_fwd_mark: Offload device fwding mark
1542 * @trans_start: Time (in jiffies) of last Tx
1543 * @watchdog_timeo: Represents the timeout that is used by
1544 * the watchdog ( see dev_watchdog() )
1545 * @watchdog_timer: List of timers
1547 * @pcpu_refcnt: Number of references to this device
1548 * @todo_list: Delayed register/unregister
1549 * @link_watch_list: XXX: need comments on this one
1551 * @reg_state: Register/unregister state machine
1552 * @dismantle: Device is going to be freed
1553 * @rtnl_link_state: This enum represents the phases of creating
1556 * @destructor: Called from unregister,
1557 * can be used to call free_netdev
1558 * @npinfo: XXX: need comments on this one
1559 * @nd_net: Network namespace this network device is inside
1561 * @ml_priv: Mid-layer private
1562 * @lstats: Loopback statistics
1563 * @tstats: Tunnel statistics
1564 * @dstats: Dummy statistics
1565 * @vstats: Virtual ethernet statistics
1570 * @dev: Class/net/name entry
1571 * @sysfs_groups: Space for optional device, statistics and wireless
1574 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1575 * @rtnl_link_ops: Rtnl_link_ops
1577 * @gso_max_size: Maximum size of generic segmentation offload
1578 * @gso_max_segs: Maximum number of segments that can be passed to the
1580 * @gso_min_segs: Minimum number of segments that can be passed to the
1583 * @dcbnl_ops: Data Center Bridging netlink ops
1584 * @num_tc: Number of traffic classes in the net device
1585 * @tc_to_txq: XXX: need comments on this one
1586 * @prio_tc_map XXX: need comments on this one
1588 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1590 * @priomap: XXX: need comments on this one
1591 * @phydev: Physical device may attach itself
1592 * for hardware timestamping
1594 * @qdisc_tx_busylock: XXX: need comments on this one
1596 * @proto_down: protocol port state information can be sent to the
1597 * switch driver and used to set the phys state of the
1600 * FIXME: cleanup struct net_device such that network protocol info
1605 char name
[IFNAMSIZ
];
1606 struct hlist_node name_hlist
;
1609 * I/O specific fields
1610 * FIXME: Merge these and struct ifmap into one
1612 unsigned long mem_end
;
1613 unsigned long mem_start
;
1614 unsigned long base_addr
;
1617 atomic_t carrier_changes
;
1620 * Some hardware also needs these fields (state,dev_list,
1621 * napi_list,unreg_list,close_list) but they are not
1622 * part of the usual set specified in Space.c.
1625 unsigned long state
;
1627 struct list_head dev_list
;
1628 struct list_head napi_list
;
1629 struct list_head unreg_list
;
1630 struct list_head close_list
;
1631 struct list_head ptype_all
;
1632 struct list_head ptype_specific
;
1635 struct list_head upper
;
1636 struct list_head lower
;
1640 struct list_head upper
;
1641 struct list_head lower
;
1644 netdev_features_t features
;
1645 netdev_features_t hw_features
;
1646 netdev_features_t wanted_features
;
1647 netdev_features_t vlan_features
;
1648 netdev_features_t hw_enc_features
;
1649 netdev_features_t mpls_features
;
1654 struct net_device_stats stats
;
1656 atomic_long_t rx_dropped
;
1657 atomic_long_t tx_dropped
;
1658 atomic_long_t rx_nohandler
;
1660 #ifdef CONFIG_WIRELESS_EXT
1661 const struct iw_handler_def
* wireless_handlers
;
1662 struct iw_public_data
* wireless_data
;
1664 const struct net_device_ops
*netdev_ops
;
1665 const struct ethtool_ops
*ethtool_ops
;
1666 #ifdef CONFIG_NET_SWITCHDEV
1667 const struct switchdev_ops
*switchdev_ops
;
1669 #ifdef CONFIG_NET_L3_MASTER_DEV
1670 const struct l3mdev_ops
*l3mdev_ops
;
1673 const struct header_ops
*header_ops
;
1676 unsigned int priv_flags
;
1678 unsigned short gflags
;
1679 unsigned short padded
;
1681 unsigned char operstate
;
1682 unsigned char link_mode
;
1684 unsigned char if_port
;
1688 unsigned short type
;
1689 unsigned short hard_header_len
;
1691 unsigned short needed_headroom
;
1692 unsigned short needed_tailroom
;
1694 /* Interface address info. */
1695 unsigned char perm_addr
[MAX_ADDR_LEN
];
1696 unsigned char addr_assign_type
;
1697 unsigned char addr_len
;
1698 unsigned short neigh_priv_len
;
1699 unsigned short dev_id
;
1700 unsigned short dev_port
;
1701 spinlock_t addr_list_lock
;
1702 unsigned char name_assign_type
;
1704 struct netdev_hw_addr_list uc
;
1705 struct netdev_hw_addr_list mc
;
1706 struct netdev_hw_addr_list dev_addrs
;
1709 struct kset
*queues_kset
;
1711 unsigned int promiscuity
;
1712 unsigned int allmulti
;
1715 /* Protocol specific pointers */
1717 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1718 struct vlan_info __rcu
*vlan_info
;
1720 #if IS_ENABLED(CONFIG_NET_DSA)
1721 struct dsa_switch_tree
*dsa_ptr
;
1723 #if IS_ENABLED(CONFIG_TIPC)
1724 struct tipc_bearer __rcu
*tipc_ptr
;
1727 struct in_device __rcu
*ip_ptr
;
1728 struct dn_dev __rcu
*dn_ptr
;
1729 struct inet6_dev __rcu
*ip6_ptr
;
1731 struct wireless_dev
*ieee80211_ptr
;
1732 struct wpan_dev
*ieee802154_ptr
;
1733 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1734 struct mpls_dev __rcu
*mpls_ptr
;
1738 * Cache lines mostly used on receive path (including eth_type_trans())
1740 unsigned long last_rx
;
1742 /* Interface address info used in eth_type_trans() */
1743 unsigned char *dev_addr
;
1747 struct netdev_rx_queue
*_rx
;
1749 unsigned int num_rx_queues
;
1750 unsigned int real_num_rx_queues
;
1754 unsigned long gro_flush_timeout
;
1755 rx_handler_func_t __rcu
*rx_handler
;
1756 void __rcu
*rx_handler_data
;
1758 #ifdef CONFIG_NET_CLS_ACT
1759 struct tcf_proto __rcu
*ingress_cl_list
;
1761 struct netdev_queue __rcu
*ingress_queue
;
1762 #ifdef CONFIG_NETFILTER_INGRESS
1763 struct list_head nf_hooks_ingress
;
1766 unsigned char broadcast
[MAX_ADDR_LEN
];
1767 #ifdef CONFIG_RFS_ACCEL
1768 struct cpu_rmap
*rx_cpu_rmap
;
1770 struct hlist_node index_hlist
;
1773 * Cache lines mostly used on transmit path
1775 struct netdev_queue
*_tx ____cacheline_aligned_in_smp
;
1776 unsigned int num_tx_queues
;
1777 unsigned int real_num_tx_queues
;
1778 struct Qdisc
*qdisc
;
1779 unsigned long tx_queue_len
;
1780 spinlock_t tx_global_lock
;
1784 struct xps_dev_maps __rcu
*xps_maps
;
1786 #ifdef CONFIG_NET_CLS_ACT
1787 struct tcf_proto __rcu
*egress_cl_list
;
1789 #ifdef CONFIG_NET_SWITCHDEV
1790 u32 offload_fwd_mark
;
1793 /* These may be needed for future network-power-down code. */
1796 * trans_start here is expensive for high speed devices on SMP,
1797 * please use netdev_queue->trans_start instead.
1799 unsigned long trans_start
;
1801 struct timer_list watchdog_timer
;
1803 int __percpu
*pcpu_refcnt
;
1804 struct list_head todo_list
;
1806 struct list_head link_watch_list
;
1808 enum { NETREG_UNINITIALIZED
=0,
1809 NETREG_REGISTERED
, /* completed register_netdevice */
1810 NETREG_UNREGISTERING
, /* called unregister_netdevice */
1811 NETREG_UNREGISTERED
, /* completed unregister todo */
1812 NETREG_RELEASED
, /* called free_netdev */
1813 NETREG_DUMMY
, /* dummy device for NAPI poll */
1819 RTNL_LINK_INITIALIZED
,
1820 RTNL_LINK_INITIALIZING
,
1821 } rtnl_link_state
:16;
1823 void (*destructor
)(struct net_device
*dev
);
1825 #ifdef CONFIG_NETPOLL
1826 struct netpoll_info __rcu
*npinfo
;
1829 possible_net_t nd_net
;
1831 /* mid-layer private */
1834 struct pcpu_lstats __percpu
*lstats
;
1835 struct pcpu_sw_netstats __percpu
*tstats
;
1836 struct pcpu_dstats __percpu
*dstats
;
1837 struct pcpu_vstats __percpu
*vstats
;
1840 struct garp_port __rcu
*garp_port
;
1841 struct mrp_port __rcu
*mrp_port
;
1844 const struct attribute_group
*sysfs_groups
[4];
1845 const struct attribute_group
*sysfs_rx_queue_group
;
1847 const struct rtnl_link_ops
*rtnl_link_ops
;
1849 /* for setting kernel sock attribute on TCP connection setup */
1850 #define GSO_MAX_SIZE 65536
1851 unsigned int gso_max_size
;
1852 #define GSO_MAX_SEGS 65535
1856 const struct dcbnl_rtnl_ops
*dcbnl_ops
;
1859 struct netdev_tc_txq tc_to_txq
[TC_MAX_QUEUE
];
1860 u8 prio_tc_map
[TC_BITMASK
+ 1];
1862 #if IS_ENABLED(CONFIG_FCOE)
1863 unsigned int fcoe_ddp_xid
;
1865 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1866 struct netprio_map __rcu
*priomap
;
1868 struct phy_device
*phydev
;
1869 struct lock_class_key
*qdisc_tx_busylock
;
1872 #define to_net_dev(d) container_of(d, struct net_device, dev)
1874 #define NETDEV_ALIGN 32
1877 int netdev_get_prio_tc_map(const struct net_device
*dev
, u32 prio
)
1879 return dev
->prio_tc_map
[prio
& TC_BITMASK
];
1883 int netdev_set_prio_tc_map(struct net_device
*dev
, u8 prio
, u8 tc
)
1885 if (tc
>= dev
->num_tc
)
1888 dev
->prio_tc_map
[prio
& TC_BITMASK
] = tc
& TC_BITMASK
;
1893 void netdev_reset_tc(struct net_device
*dev
)
1896 memset(dev
->tc_to_txq
, 0, sizeof(dev
->tc_to_txq
));
1897 memset(dev
->prio_tc_map
, 0, sizeof(dev
->prio_tc_map
));
1901 int netdev_set_tc_queue(struct net_device
*dev
, u8 tc
, u16 count
, u16 offset
)
1903 if (tc
>= dev
->num_tc
)
1906 dev
->tc_to_txq
[tc
].count
= count
;
1907 dev
->tc_to_txq
[tc
].offset
= offset
;
1912 int netdev_set_num_tc(struct net_device
*dev
, u8 num_tc
)
1914 if (num_tc
> TC_MAX_QUEUE
)
1917 dev
->num_tc
= num_tc
;
1922 int netdev_get_num_tc(struct net_device
*dev
)
1928 struct netdev_queue
*netdev_get_tx_queue(const struct net_device
*dev
,
1931 return &dev
->_tx
[index
];
1934 static inline struct netdev_queue
*skb_get_tx_queue(const struct net_device
*dev
,
1935 const struct sk_buff
*skb
)
1937 return netdev_get_tx_queue(dev
, skb_get_queue_mapping(skb
));
1940 static inline void netdev_for_each_tx_queue(struct net_device
*dev
,
1941 void (*f
)(struct net_device
*,
1942 struct netdev_queue
*,
1948 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
1949 f(dev
, &dev
->_tx
[i
], arg
);
1952 struct netdev_queue
*netdev_pick_tx(struct net_device
*dev
,
1953 struct sk_buff
*skb
,
1956 /* returns the headroom that the master device needs to take in account
1957 * when forwarding to this dev
1959 static inline unsigned netdev_get_fwd_headroom(struct net_device
*dev
)
1961 return dev
->priv_flags
& IFF_PHONY_HEADROOM
? 0 : dev
->needed_headroom
;
1964 static inline void netdev_set_rx_headroom(struct net_device
*dev
, int new_hr
)
1966 if (dev
->netdev_ops
->ndo_set_rx_headroom
)
1967 dev
->netdev_ops
->ndo_set_rx_headroom(dev
, new_hr
);
1970 /* set the device rx headroom to the dev's default */
1971 static inline void netdev_reset_rx_headroom(struct net_device
*dev
)
1973 netdev_set_rx_headroom(dev
, -1);
1977 * Net namespace inlines
1980 struct net
*dev_net(const struct net_device
*dev
)
1982 return read_pnet(&dev
->nd_net
);
1986 void dev_net_set(struct net_device
*dev
, struct net
*net
)
1988 write_pnet(&dev
->nd_net
, net
);
1991 static inline bool netdev_uses_dsa(struct net_device
*dev
)
1993 #if IS_ENABLED(CONFIG_NET_DSA)
1994 if (dev
->dsa_ptr
!= NULL
)
1995 return dsa_uses_tagged_protocol(dev
->dsa_ptr
);
2001 * netdev_priv - access network device private data
2002 * @dev: network device
2004 * Get network device private data
2006 static inline void *netdev_priv(const struct net_device
*dev
)
2008 return (char *)dev
+ ALIGN(sizeof(struct net_device
), NETDEV_ALIGN
);
2011 /* Set the sysfs physical device reference for the network logical device
2012 * if set prior to registration will cause a symlink during initialization.
2014 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2016 /* Set the sysfs device type for the network logical device to allow
2017 * fine-grained identification of different network device types. For
2018 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
2020 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2022 /* Default NAPI poll() weight
2023 * Device drivers are strongly advised to not use bigger value
2025 #define NAPI_POLL_WEIGHT 64
2028 * netif_napi_add - initialize a napi context
2029 * @dev: network device
2030 * @napi: napi context
2031 * @poll: polling function
2032 * @weight: default weight
2034 * netif_napi_add() must be used to initialize a napi context prior to calling
2035 * *any* of the other napi related functions.
2037 void netif_napi_add(struct net_device
*dev
, struct napi_struct
*napi
,
2038 int (*poll
)(struct napi_struct
*, int), int weight
);
2041 * netif_tx_napi_add - initialize a napi context
2042 * @dev: network device
2043 * @napi: napi context
2044 * @poll: polling function
2045 * @weight: default weight
2047 * This variant of netif_napi_add() should be used from drivers using NAPI
2048 * to exclusively poll a TX queue.
2049 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2051 static inline void netif_tx_napi_add(struct net_device
*dev
,
2052 struct napi_struct
*napi
,
2053 int (*poll
)(struct napi_struct
*, int),
2056 set_bit(NAPI_STATE_NO_BUSY_POLL
, &napi
->state
);
2057 netif_napi_add(dev
, napi
, poll
, weight
);
2061 * netif_napi_del - remove a napi context
2062 * @napi: napi context
2064 * netif_napi_del() removes a napi context from the network device napi list
2066 void netif_napi_del(struct napi_struct
*napi
);
2068 struct napi_gro_cb
{
2069 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2072 /* Length of frag0. */
2073 unsigned int frag0_len
;
2075 /* This indicates where we are processing relative to skb->data. */
2078 /* This is non-zero if the packet cannot be merged with the new skb. */
2081 /* Save the IP ID here and check when we get to the transport layer */
2084 /* Number of segments aggregated. */
2087 /* Start offset for remote checksum offload */
2088 u16 gro_remcsum_start
;
2090 /* jiffies when first packet was created/queued */
2093 /* Used in ipv6_gro_receive() and foo-over-udp */
2096 /* This is non-zero if the packet may be of the same flow. */
2099 /* Used in udp_gro_receive */
2102 /* GRO checksum is valid */
2105 /* Number of checksums via CHECKSUM_UNNECESSARY */
2110 #define NAPI_GRO_FREE 1
2111 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2113 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2118 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2121 /* used in skb_gro_receive() slow path */
2122 struct sk_buff
*last
;
2125 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2127 struct packet_type
{
2128 __be16 type
; /* This is really htons(ether_type). */
2129 struct net_device
*dev
; /* NULL is wildcarded here */
2130 int (*func
) (struct sk_buff
*,
2131 struct net_device
*,
2132 struct packet_type
*,
2133 struct net_device
*);
2134 bool (*id_match
)(struct packet_type
*ptype
,
2136 void *af_packet_priv
;
2137 struct list_head list
;
2140 struct offload_callbacks
{
2141 struct sk_buff
*(*gso_segment
)(struct sk_buff
*skb
,
2142 netdev_features_t features
);
2143 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
2144 struct sk_buff
*skb
);
2145 int (*gro_complete
)(struct sk_buff
*skb
, int nhoff
);
2148 struct packet_offload
{
2149 __be16 type
; /* This is really htons(ether_type). */
2151 struct offload_callbacks callbacks
;
2152 struct list_head list
;
2157 struct udp_offload_callbacks
{
2158 struct sk_buff
**(*gro_receive
)(struct sk_buff
**head
,
2159 struct sk_buff
*skb
,
2160 struct udp_offload
*uoff
);
2161 int (*gro_complete
)(struct sk_buff
*skb
,
2163 struct udp_offload
*uoff
);
2166 struct udp_offload
{
2169 struct udp_offload_callbacks callbacks
;
2172 /* often modified stats are per cpu, other are shared (netdev->stats) */
2173 struct pcpu_sw_netstats
{
2178 struct u64_stats_sync syncp
;
2181 #define __netdev_alloc_pcpu_stats(type, gfp) \
2183 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2186 for_each_possible_cpu(__cpu) { \
2187 typeof(type) *stat; \
2188 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2189 u64_stats_init(&stat->syncp); \
2195 #define netdev_alloc_pcpu_stats(type) \
2196 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2198 enum netdev_lag_tx_type
{
2199 NETDEV_LAG_TX_TYPE_UNKNOWN
,
2200 NETDEV_LAG_TX_TYPE_RANDOM
,
2201 NETDEV_LAG_TX_TYPE_BROADCAST
,
2202 NETDEV_LAG_TX_TYPE_ROUNDROBIN
,
2203 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP
,
2204 NETDEV_LAG_TX_TYPE_HASH
,
2207 struct netdev_lag_upper_info
{
2208 enum netdev_lag_tx_type tx_type
;
2211 struct netdev_lag_lower_state_info
{
2216 #include <linux/notifier.h>
2218 /* netdevice notifier chain. Please remember to update the rtnetlink
2219 * notification exclusion list in rtnetlink_event() when adding new
2222 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2223 #define NETDEV_DOWN 0x0002
2224 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2225 detected a hardware crash and restarted
2226 - we can use this eg to kick tcp sessions
2228 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2229 #define NETDEV_REGISTER 0x0005
2230 #define NETDEV_UNREGISTER 0x0006
2231 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2232 #define NETDEV_CHANGEADDR 0x0008
2233 #define NETDEV_GOING_DOWN 0x0009
2234 #define NETDEV_CHANGENAME 0x000A
2235 #define NETDEV_FEAT_CHANGE 0x000B
2236 #define NETDEV_BONDING_FAILOVER 0x000C
2237 #define NETDEV_PRE_UP 0x000D
2238 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2239 #define NETDEV_POST_TYPE_CHANGE 0x000F
2240 #define NETDEV_POST_INIT 0x0010
2241 #define NETDEV_UNREGISTER_FINAL 0x0011
2242 #define NETDEV_RELEASE 0x0012
2243 #define NETDEV_NOTIFY_PEERS 0x0013
2244 #define NETDEV_JOIN 0x0014
2245 #define NETDEV_CHANGEUPPER 0x0015
2246 #define NETDEV_RESEND_IGMP 0x0016
2247 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2248 #define NETDEV_CHANGEINFODATA 0x0018
2249 #define NETDEV_BONDING_INFO 0x0019
2250 #define NETDEV_PRECHANGEUPPER 0x001A
2251 #define NETDEV_CHANGELOWERSTATE 0x001B
2253 int register_netdevice_notifier(struct notifier_block
*nb
);
2254 int unregister_netdevice_notifier(struct notifier_block
*nb
);
2256 struct netdev_notifier_info
{
2257 struct net_device
*dev
;
2260 struct netdev_notifier_change_info
{
2261 struct netdev_notifier_info info
; /* must be first */
2262 unsigned int flags_changed
;
2265 struct netdev_notifier_changeupper_info
{
2266 struct netdev_notifier_info info
; /* must be first */
2267 struct net_device
*upper_dev
; /* new upper dev */
2268 bool master
; /* is upper dev master */
2269 bool linking
; /* is the nofication for link or unlink */
2270 void *upper_info
; /* upper dev info */
2273 struct netdev_notifier_changelowerstate_info
{
2274 struct netdev_notifier_info info
; /* must be first */
2275 void *lower_state_info
; /* is lower dev state */
2278 static inline void netdev_notifier_info_init(struct netdev_notifier_info
*info
,
2279 struct net_device
*dev
)
2284 static inline struct net_device
*
2285 netdev_notifier_info_to_dev(const struct netdev_notifier_info
*info
)
2290 int call_netdevice_notifiers(unsigned long val
, struct net_device
*dev
);
2293 extern rwlock_t dev_base_lock
; /* Device list lock */
2295 #define for_each_netdev(net, d) \
2296 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2297 #define for_each_netdev_reverse(net, d) \
2298 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2299 #define for_each_netdev_rcu(net, d) \
2300 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2301 #define for_each_netdev_safe(net, d, n) \
2302 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2303 #define for_each_netdev_continue(net, d) \
2304 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2305 #define for_each_netdev_continue_rcu(net, d) \
2306 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2307 #define for_each_netdev_in_bond_rcu(bond, slave) \
2308 for_each_netdev_rcu(&init_net, slave) \
2309 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2310 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2312 static inline struct net_device
*next_net_device(struct net_device
*dev
)
2314 struct list_head
*lh
;
2318 lh
= dev
->dev_list
.next
;
2319 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2322 static inline struct net_device
*next_net_device_rcu(struct net_device
*dev
)
2324 struct list_head
*lh
;
2328 lh
= rcu_dereference(list_next_rcu(&dev
->dev_list
));
2329 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2332 static inline struct net_device
*first_net_device(struct net
*net
)
2334 return list_empty(&net
->dev_base_head
) ? NULL
:
2335 net_device_entry(net
->dev_base_head
.next
);
2338 static inline struct net_device
*first_net_device_rcu(struct net
*net
)
2340 struct list_head
*lh
= rcu_dereference(list_next_rcu(&net
->dev_base_head
));
2342 return lh
== &net
->dev_base_head
? NULL
: net_device_entry(lh
);
2345 int netdev_boot_setup_check(struct net_device
*dev
);
2346 unsigned long netdev_boot_base(const char *prefix
, int unit
);
2347 struct net_device
*dev_getbyhwaddr_rcu(struct net
*net
, unsigned short type
,
2348 const char *hwaddr
);
2349 struct net_device
*dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2350 struct net_device
*__dev_getfirstbyhwtype(struct net
*net
, unsigned short type
);
2351 void dev_add_pack(struct packet_type
*pt
);
2352 void dev_remove_pack(struct packet_type
*pt
);
2353 void __dev_remove_pack(struct packet_type
*pt
);
2354 void dev_add_offload(struct packet_offload
*po
);
2355 void dev_remove_offload(struct packet_offload
*po
);
2357 int dev_get_iflink(const struct net_device
*dev
);
2358 int dev_fill_metadata_dst(struct net_device
*dev
, struct sk_buff
*skb
);
2359 struct net_device
*__dev_get_by_flags(struct net
*net
, unsigned short flags
,
2360 unsigned short mask
);
2361 struct net_device
*dev_get_by_name(struct net
*net
, const char *name
);
2362 struct net_device
*dev_get_by_name_rcu(struct net
*net
, const char *name
);
2363 struct net_device
*__dev_get_by_name(struct net
*net
, const char *name
);
2364 int dev_alloc_name(struct net_device
*dev
, const char *name
);
2365 int dev_open(struct net_device
*dev
);
2366 int dev_close(struct net_device
*dev
);
2367 int dev_close_many(struct list_head
*head
, bool unlink
);
2368 void dev_disable_lro(struct net_device
*dev
);
2369 int dev_loopback_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*newskb
);
2370 int dev_queue_xmit(struct sk_buff
*skb
);
2371 int dev_queue_xmit_accel(struct sk_buff
*skb
, void *accel_priv
);
2372 int register_netdevice(struct net_device
*dev
);
2373 void unregister_netdevice_queue(struct net_device
*dev
, struct list_head
*head
);
2374 void unregister_netdevice_many(struct list_head
*head
);
2375 static inline void unregister_netdevice(struct net_device
*dev
)
2377 unregister_netdevice_queue(dev
, NULL
);
2380 int netdev_refcnt_read(const struct net_device
*dev
);
2381 void free_netdev(struct net_device
*dev
);
2382 void netdev_freemem(struct net_device
*dev
);
2383 void synchronize_net(void);
2384 int init_dummy_netdev(struct net_device
*dev
);
2386 DECLARE_PER_CPU(int, xmit_recursion
);
2387 static inline int dev_recursion_level(void)
2389 return this_cpu_read(xmit_recursion
);
2392 struct net_device
*dev_get_by_index(struct net
*net
, int ifindex
);
2393 struct net_device
*__dev_get_by_index(struct net
*net
, int ifindex
);
2394 struct net_device
*dev_get_by_index_rcu(struct net
*net
, int ifindex
);
2395 int netdev_get_name(struct net
*net
, char *name
, int ifindex
);
2396 int dev_restart(struct net_device
*dev
);
2397 int skb_gro_receive(struct sk_buff
**head
, struct sk_buff
*skb
);
2399 static inline unsigned int skb_gro_offset(const struct sk_buff
*skb
)
2401 return NAPI_GRO_CB(skb
)->data_offset
;
2404 static inline unsigned int skb_gro_len(const struct sk_buff
*skb
)
2406 return skb
->len
- NAPI_GRO_CB(skb
)->data_offset
;
2409 static inline void skb_gro_pull(struct sk_buff
*skb
, unsigned int len
)
2411 NAPI_GRO_CB(skb
)->data_offset
+= len
;
2414 static inline void *skb_gro_header_fast(struct sk_buff
*skb
,
2415 unsigned int offset
)
2417 return NAPI_GRO_CB(skb
)->frag0
+ offset
;
2420 static inline int skb_gro_header_hard(struct sk_buff
*skb
, unsigned int hlen
)
2422 return NAPI_GRO_CB(skb
)->frag0_len
< hlen
;
2425 static inline void *skb_gro_header_slow(struct sk_buff
*skb
, unsigned int hlen
,
2426 unsigned int offset
)
2428 if (!pskb_may_pull(skb
, hlen
))
2431 NAPI_GRO_CB(skb
)->frag0
= NULL
;
2432 NAPI_GRO_CB(skb
)->frag0_len
= 0;
2433 return skb
->data
+ offset
;
2436 static inline void *skb_gro_network_header(struct sk_buff
*skb
)
2438 return (NAPI_GRO_CB(skb
)->frag0
?: skb
->data
) +
2439 skb_network_offset(skb
);
2442 static inline void skb_gro_postpull_rcsum(struct sk_buff
*skb
,
2443 const void *start
, unsigned int len
)
2445 if (NAPI_GRO_CB(skb
)->csum_valid
)
2446 NAPI_GRO_CB(skb
)->csum
= csum_sub(NAPI_GRO_CB(skb
)->csum
,
2447 csum_partial(start
, len
, 0));
2450 /* GRO checksum functions. These are logical equivalents of the normal
2451 * checksum functions (in skbuff.h) except that they operate on the GRO
2452 * offsets and fields in sk_buff.
2455 __sum16
__skb_gro_checksum_complete(struct sk_buff
*skb
);
2457 static inline bool skb_at_gro_remcsum_start(struct sk_buff
*skb
)
2459 return (NAPI_GRO_CB(skb
)->gro_remcsum_start
== skb_gro_offset(skb
));
2462 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff
*skb
,
2466 return ((skb
->ip_summed
!= CHECKSUM_PARTIAL
||
2467 skb_checksum_start_offset(skb
) <
2468 skb_gro_offset(skb
)) &&
2469 !skb_at_gro_remcsum_start(skb
) &&
2470 NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2471 (!zero_okay
|| check
));
2474 static inline __sum16
__skb_gro_checksum_validate_complete(struct sk_buff
*skb
,
2477 if (NAPI_GRO_CB(skb
)->csum_valid
&&
2478 !csum_fold(csum_add(psum
, NAPI_GRO_CB(skb
)->csum
)))
2481 NAPI_GRO_CB(skb
)->csum
= psum
;
2483 return __skb_gro_checksum_complete(skb
);
2486 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff
*skb
)
2488 if (NAPI_GRO_CB(skb
)->csum_cnt
> 0) {
2489 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2490 NAPI_GRO_CB(skb
)->csum_cnt
--;
2492 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2493 * verified a new top level checksum or an encapsulated one
2494 * during GRO. This saves work if we fallback to normal path.
2496 __skb_incr_checksum_unnecessary(skb
);
2500 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2503 __sum16 __ret = 0; \
2504 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2505 __ret = __skb_gro_checksum_validate_complete(skb, \
2506 compute_pseudo(skb, proto)); \
2508 __skb_mark_checksum_bad(skb); \
2510 skb_gro_incr_csum_unnecessary(skb); \
2514 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2515 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2517 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2519 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2521 #define skb_gro_checksum_simple_validate(skb) \
2522 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2524 static inline bool __skb_gro_checksum_convert_check(struct sk_buff
*skb
)
2526 return (NAPI_GRO_CB(skb
)->csum_cnt
== 0 &&
2527 !NAPI_GRO_CB(skb
)->csum_valid
);
2530 static inline void __skb_gro_checksum_convert(struct sk_buff
*skb
,
2531 __sum16 check
, __wsum pseudo
)
2533 NAPI_GRO_CB(skb
)->csum
= ~pseudo
;
2534 NAPI_GRO_CB(skb
)->csum_valid
= 1;
2537 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2539 if (__skb_gro_checksum_convert_check(skb)) \
2540 __skb_gro_checksum_convert(skb, check, \
2541 compute_pseudo(skb, proto)); \
2544 struct gro_remcsum
{
2549 static inline void skb_gro_remcsum_init(struct gro_remcsum
*grc
)
2555 static inline void *skb_gro_remcsum_process(struct sk_buff
*skb
, void *ptr
,
2556 unsigned int off
, size_t hdrlen
,
2557 int start
, int offset
,
2558 struct gro_remcsum
*grc
,
2562 size_t plen
= hdrlen
+ max_t(size_t, offset
+ sizeof(u16
), start
);
2564 BUG_ON(!NAPI_GRO_CB(skb
)->csum_valid
);
2567 NAPI_GRO_CB(skb
)->gro_remcsum_start
= off
+ hdrlen
+ start
;
2571 ptr
= skb_gro_header_fast(skb
, off
);
2572 if (skb_gro_header_hard(skb
, off
+ plen
)) {
2573 ptr
= skb_gro_header_slow(skb
, off
+ plen
, off
);
2578 delta
= remcsum_adjust(ptr
+ hdrlen
, NAPI_GRO_CB(skb
)->csum
,
2581 /* Adjust skb->csum since we changed the packet */
2582 NAPI_GRO_CB(skb
)->csum
= csum_add(NAPI_GRO_CB(skb
)->csum
, delta
);
2584 grc
->offset
= off
+ hdrlen
+ offset
;
2590 static inline void skb_gro_remcsum_cleanup(struct sk_buff
*skb
,
2591 struct gro_remcsum
*grc
)
2594 size_t plen
= grc
->offset
+ sizeof(u16
);
2599 ptr
= skb_gro_header_fast(skb
, grc
->offset
);
2600 if (skb_gro_header_hard(skb
, grc
->offset
+ sizeof(u16
))) {
2601 ptr
= skb_gro_header_slow(skb
, plen
, grc
->offset
);
2606 remcsum_unadjust((__sum16
*)ptr
, grc
->delta
);
2609 struct skb_csum_offl_spec
{
2623 bool __skb_csum_offload_chk(struct sk_buff
*skb
,
2624 const struct skb_csum_offl_spec
*spec
,
2625 bool *csum_encapped
,
2628 static inline bool skb_csum_offload_chk(struct sk_buff
*skb
,
2629 const struct skb_csum_offl_spec
*spec
,
2630 bool *csum_encapped
,
2633 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
)
2636 return __skb_csum_offload_chk(skb
, spec
, csum_encapped
, csum_help
);
2639 static inline bool skb_csum_offload_chk_help(struct sk_buff
*skb
,
2640 const struct skb_csum_offl_spec
*spec
)
2644 return skb_csum_offload_chk(skb
, spec
, &csum_encapped
, true);
2647 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff
*skb
)
2649 static const struct skb_csum_offl_spec csum_offl_spec
= {
2651 .ip_options_okay
= 1,
2658 return skb_csum_offload_chk_help(skb
, &csum_offl_spec
);
2661 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff
*skb
)
2663 static const struct skb_csum_offl_spec csum_offl_spec
= {
2665 .ip_options_okay
= 1,
2671 return skb_csum_offload_chk_help(skb
, &csum_offl_spec
);
2674 static inline int dev_hard_header(struct sk_buff
*skb
, struct net_device
*dev
,
2675 unsigned short type
,
2676 const void *daddr
, const void *saddr
,
2679 if (!dev
->header_ops
|| !dev
->header_ops
->create
)
2682 return dev
->header_ops
->create(skb
, dev
, type
, daddr
, saddr
, len
);
2685 static inline int dev_parse_header(const struct sk_buff
*skb
,
2686 unsigned char *haddr
)
2688 const struct net_device
*dev
= skb
->dev
;
2690 if (!dev
->header_ops
|| !dev
->header_ops
->parse
)
2692 return dev
->header_ops
->parse(skb
, haddr
);
2695 /* ll_header must have at least hard_header_len allocated */
2696 static inline bool dev_validate_header(const struct net_device
*dev
,
2697 char *ll_header
, int len
)
2699 if (likely(len
>= dev
->hard_header_len
))
2702 if (capable(CAP_SYS_RAWIO
)) {
2703 memset(ll_header
+ len
, 0, dev
->hard_header_len
- len
);
2707 if (dev
->header_ops
&& dev
->header_ops
->validate
)
2708 return dev
->header_ops
->validate(ll_header
, len
);
2713 typedef int gifconf_func_t(struct net_device
* dev
, char __user
* bufptr
, int len
);
2714 int register_gifconf(unsigned int family
, gifconf_func_t
*gifconf
);
2715 static inline int unregister_gifconf(unsigned int family
)
2717 return register_gifconf(family
, NULL
);
2720 #ifdef CONFIG_NET_FLOW_LIMIT
2721 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2722 struct sd_flow_limit
{
2724 unsigned int num_buckets
;
2725 unsigned int history_head
;
2726 u16 history
[FLOW_LIMIT_HISTORY
];
2730 extern int netdev_flow_limit_table_len
;
2731 #endif /* CONFIG_NET_FLOW_LIMIT */
2734 * Incoming packets are placed on per-cpu queues
2736 struct softnet_data
{
2737 struct list_head poll_list
;
2738 struct sk_buff_head process_queue
;
2741 unsigned int processed
;
2742 unsigned int time_squeeze
;
2743 unsigned int cpu_collision
;
2744 unsigned int received_rps
;
2746 struct softnet_data
*rps_ipi_list
;
2748 #ifdef CONFIG_NET_FLOW_LIMIT
2749 struct sd_flow_limit __rcu
*flow_limit
;
2751 struct Qdisc
*output_queue
;
2752 struct Qdisc
**output_queue_tailp
;
2753 struct sk_buff
*completion_queue
;
2756 /* Elements below can be accessed between CPUs for RPS */
2757 struct call_single_data csd ____cacheline_aligned_in_smp
;
2758 struct softnet_data
*rps_ipi_next
;
2760 unsigned int input_queue_head
;
2761 unsigned int input_queue_tail
;
2763 unsigned int dropped
;
2764 struct sk_buff_head input_pkt_queue
;
2765 struct napi_struct backlog
;
2769 static inline void input_queue_head_incr(struct softnet_data
*sd
)
2772 sd
->input_queue_head
++;
2776 static inline void input_queue_tail_incr_save(struct softnet_data
*sd
,
2777 unsigned int *qtail
)
2780 *qtail
= ++sd
->input_queue_tail
;
2784 DECLARE_PER_CPU_ALIGNED(struct softnet_data
, softnet_data
);
2786 void __netif_schedule(struct Qdisc
*q
);
2787 void netif_schedule_queue(struct netdev_queue
*txq
);
2789 static inline void netif_tx_schedule_all(struct net_device
*dev
)
2793 for (i
= 0; i
< dev
->num_tx_queues
; i
++)
2794 netif_schedule_queue(netdev_get_tx_queue(dev
, i
));
2797 static inline void netif_tx_start_queue(struct netdev_queue
*dev_queue
)
2799 clear_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2803 * netif_start_queue - allow transmit
2804 * @dev: network device
2806 * Allow upper layers to call the device hard_start_xmit routine.
2808 static inline void netif_start_queue(struct net_device
*dev
)
2810 netif_tx_start_queue(netdev_get_tx_queue(dev
, 0));
2813 static inline void netif_tx_start_all_queues(struct net_device
*dev
)
2817 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2818 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2819 netif_tx_start_queue(txq
);
2823 void netif_tx_wake_queue(struct netdev_queue
*dev_queue
);
2826 * netif_wake_queue - restart transmit
2827 * @dev: network device
2829 * Allow upper layers to call the device hard_start_xmit routine.
2830 * Used for flow control when transmit resources are available.
2832 static inline void netif_wake_queue(struct net_device
*dev
)
2834 netif_tx_wake_queue(netdev_get_tx_queue(dev
, 0));
2837 static inline void netif_tx_wake_all_queues(struct net_device
*dev
)
2841 for (i
= 0; i
< dev
->num_tx_queues
; i
++) {
2842 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, i
);
2843 netif_tx_wake_queue(txq
);
2847 static inline void netif_tx_stop_queue(struct netdev_queue
*dev_queue
)
2849 set_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2853 * netif_stop_queue - stop transmitted packets
2854 * @dev: network device
2856 * Stop upper layers calling the device hard_start_xmit routine.
2857 * Used for flow control when transmit resources are unavailable.
2859 static inline void netif_stop_queue(struct net_device
*dev
)
2861 netif_tx_stop_queue(netdev_get_tx_queue(dev
, 0));
2864 void netif_tx_stop_all_queues(struct net_device
*dev
);
2866 static inline bool netif_tx_queue_stopped(const struct netdev_queue
*dev_queue
)
2868 return test_bit(__QUEUE_STATE_DRV_XOFF
, &dev_queue
->state
);
2872 * netif_queue_stopped - test if transmit queue is flowblocked
2873 * @dev: network device
2875 * Test if transmit queue on device is currently unable to send.
2877 static inline bool netif_queue_stopped(const struct net_device
*dev
)
2879 return netif_tx_queue_stopped(netdev_get_tx_queue(dev
, 0));
2882 static inline bool netif_xmit_stopped(const struct netdev_queue
*dev_queue
)
2884 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF
;
2888 netif_xmit_frozen_or_stopped(const struct netdev_queue
*dev_queue
)
2890 return dev_queue
->state
& QUEUE_STATE_ANY_XOFF_OR_FROZEN
;
2894 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue
*dev_queue
)
2896 return dev_queue
->state
& QUEUE_STATE_DRV_XOFF_OR_FROZEN
;
2900 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2901 * @dev_queue: pointer to transmit queue
2903 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2904 * to give appropriate hint to the cpu.
2906 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue
*dev_queue
)
2909 prefetchw(&dev_queue
->dql
.num_queued
);
2914 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2915 * @dev_queue: pointer to transmit queue
2917 * BQL enabled drivers might use this helper in their TX completion path,
2918 * to give appropriate hint to the cpu.
2920 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue
*dev_queue
)
2923 prefetchw(&dev_queue
->dql
.limit
);
2927 static inline void netdev_tx_sent_queue(struct netdev_queue
*dev_queue
,
2931 dql_queued(&dev_queue
->dql
, bytes
);
2933 if (likely(dql_avail(&dev_queue
->dql
) >= 0))
2936 set_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2939 * The XOFF flag must be set before checking the dql_avail below,
2940 * because in netdev_tx_completed_queue we update the dql_completed
2941 * before checking the XOFF flag.
2945 /* check again in case another CPU has just made room avail */
2946 if (unlikely(dql_avail(&dev_queue
->dql
) >= 0))
2947 clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
);
2952 * netdev_sent_queue - report the number of bytes queued to hardware
2953 * @dev: network device
2954 * @bytes: number of bytes queued to the hardware device queue
2956 * Report the number of bytes queued for sending/completion to the network
2957 * device hardware queue. @bytes should be a good approximation and should
2958 * exactly match netdev_completed_queue() @bytes
2960 static inline void netdev_sent_queue(struct net_device
*dev
, unsigned int bytes
)
2962 netdev_tx_sent_queue(netdev_get_tx_queue(dev
, 0), bytes
);
2965 static inline void netdev_tx_completed_queue(struct netdev_queue
*dev_queue
,
2966 unsigned int pkts
, unsigned int bytes
)
2969 if (unlikely(!bytes
))
2972 dql_completed(&dev_queue
->dql
, bytes
);
2975 * Without the memory barrier there is a small possiblity that
2976 * netdev_tx_sent_queue will miss the update and cause the queue to
2977 * be stopped forever
2981 if (dql_avail(&dev_queue
->dql
) < 0)
2984 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF
, &dev_queue
->state
))
2985 netif_schedule_queue(dev_queue
);
2990 * netdev_completed_queue - report bytes and packets completed by device
2991 * @dev: network device
2992 * @pkts: actual number of packets sent over the medium
2993 * @bytes: actual number of bytes sent over the medium
2995 * Report the number of bytes and packets transmitted by the network device
2996 * hardware queue over the physical medium, @bytes must exactly match the
2997 * @bytes amount passed to netdev_sent_queue()
2999 static inline void netdev_completed_queue(struct net_device
*dev
,
3000 unsigned int pkts
, unsigned int bytes
)
3002 netdev_tx_completed_queue(netdev_get_tx_queue(dev
, 0), pkts
, bytes
);
3005 static inline void netdev_tx_reset_queue(struct netdev_queue
*q
)
3008 clear_bit(__QUEUE_STATE_STACK_XOFF
, &q
->state
);
3014 * netdev_reset_queue - reset the packets and bytes count of a network device
3015 * @dev_queue: network device
3017 * Reset the bytes and packet count of a network device and clear the
3018 * software flow control OFF bit for this network device
3020 static inline void netdev_reset_queue(struct net_device
*dev_queue
)
3022 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue
, 0));
3026 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3027 * @dev: network device
3028 * @queue_index: given tx queue index
3030 * Returns 0 if given tx queue index >= number of device tx queues,
3031 * otherwise returns the originally passed tx queue index.
3033 static inline u16
netdev_cap_txqueue(struct net_device
*dev
, u16 queue_index
)
3035 if (unlikely(queue_index
>= dev
->real_num_tx_queues
)) {
3036 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3037 dev
->name
, queue_index
,
3038 dev
->real_num_tx_queues
);
3046 * netif_running - test if up
3047 * @dev: network device
3049 * Test if the device has been brought up.
3051 static inline bool netif_running(const struct net_device
*dev
)
3053 return test_bit(__LINK_STATE_START
, &dev
->state
);
3057 * Routines to manage the subqueues on a device. We only need start
3058 * stop, and a check if it's stopped. All other device management is
3059 * done at the overall netdevice level.
3060 * Also test the device if we're multiqueue.
3064 * netif_start_subqueue - allow sending packets on subqueue
3065 * @dev: network device
3066 * @queue_index: sub queue index
3068 * Start individual transmit queue of a device with multiple transmit queues.
3070 static inline void netif_start_subqueue(struct net_device
*dev
, u16 queue_index
)
3072 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3074 netif_tx_start_queue(txq
);
3078 * netif_stop_subqueue - stop sending packets on subqueue
3079 * @dev: network device
3080 * @queue_index: sub queue index
3082 * Stop individual transmit queue of a device with multiple transmit queues.
3084 static inline void netif_stop_subqueue(struct net_device
*dev
, u16 queue_index
)
3086 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3087 netif_tx_stop_queue(txq
);
3091 * netif_subqueue_stopped - test status of subqueue
3092 * @dev: network device
3093 * @queue_index: sub queue index
3095 * Check individual transmit queue of a device with multiple transmit queues.
3097 static inline bool __netif_subqueue_stopped(const struct net_device
*dev
,
3100 struct netdev_queue
*txq
= netdev_get_tx_queue(dev
, queue_index
);
3102 return netif_tx_queue_stopped(txq
);
3105 static inline bool netif_subqueue_stopped(const struct net_device
*dev
,
3106 struct sk_buff
*skb
)
3108 return __netif_subqueue_stopped(dev
, skb_get_queue_mapping(skb
));
3111 void netif_wake_subqueue(struct net_device
*dev
, u16 queue_index
);
3114 int netif_set_xps_queue(struct net_device
*dev
, const struct cpumask
*mask
,
3117 static inline int netif_set_xps_queue(struct net_device
*dev
,
3118 const struct cpumask
*mask
,
3125 u16
__skb_tx_hash(const struct net_device
*dev
, struct sk_buff
*skb
,
3126 unsigned int num_tx_queues
);
3129 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3130 * as a distribution range limit for the returned value.
3132 static inline u16
skb_tx_hash(const struct net_device
*dev
,
3133 struct sk_buff
*skb
)
3135 return __skb_tx_hash(dev
, skb
, dev
->real_num_tx_queues
);
3139 * netif_is_multiqueue - test if device has multiple transmit queues
3140 * @dev: network device
3142 * Check if device has multiple transmit queues
3144 static inline bool netif_is_multiqueue(const struct net_device
*dev
)
3146 return dev
->num_tx_queues
> 1;
3149 int netif_set_real_num_tx_queues(struct net_device
*dev
, unsigned int txq
);
3152 int netif_set_real_num_rx_queues(struct net_device
*dev
, unsigned int rxq
);
3154 static inline int netif_set_real_num_rx_queues(struct net_device
*dev
,
3162 static inline unsigned int get_netdev_rx_queue_index(
3163 struct netdev_rx_queue
*queue
)
3165 struct net_device
*dev
= queue
->dev
;
3166 int index
= queue
- dev
->_rx
;
3168 BUG_ON(index
>= dev
->num_rx_queues
);
3173 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3174 int netif_get_num_default_rss_queues(void);
3176 enum skb_free_reason
{
3177 SKB_REASON_CONSUMED
,
3181 void __dev_kfree_skb_irq(struct sk_buff
*skb
, enum skb_free_reason reason
);
3182 void __dev_kfree_skb_any(struct sk_buff
*skb
, enum skb_free_reason reason
);
3185 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3186 * interrupt context or with hardware interrupts being disabled.
3187 * (in_irq() || irqs_disabled())
3189 * We provide four helpers that can be used in following contexts :
3191 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3192 * replacing kfree_skb(skb)
3194 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3195 * Typically used in place of consume_skb(skb) in TX completion path
3197 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3198 * replacing kfree_skb(skb)
3200 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3201 * and consumed a packet. Used in place of consume_skb(skb)
3203 static inline void dev_kfree_skb_irq(struct sk_buff
*skb
)
3205 __dev_kfree_skb_irq(skb
, SKB_REASON_DROPPED
);
3208 static inline void dev_consume_skb_irq(struct sk_buff
*skb
)
3210 __dev_kfree_skb_irq(skb
, SKB_REASON_CONSUMED
);
3213 static inline void dev_kfree_skb_any(struct sk_buff
*skb
)
3215 __dev_kfree_skb_any(skb
, SKB_REASON_DROPPED
);
3218 static inline void dev_consume_skb_any(struct sk_buff
*skb
)
3220 __dev_kfree_skb_any(skb
, SKB_REASON_CONSUMED
);
3223 int netif_rx(struct sk_buff
*skb
);
3224 int netif_rx_ni(struct sk_buff
*skb
);
3225 int netif_receive_skb(struct sk_buff
*skb
);
3226 gro_result_t
napi_gro_receive(struct napi_struct
*napi
, struct sk_buff
*skb
);
3227 void napi_gro_flush(struct napi_struct
*napi
, bool flush_old
);
3228 struct sk_buff
*napi_get_frags(struct napi_struct
*napi
);
3229 gro_result_t
napi_gro_frags(struct napi_struct
*napi
);
3230 struct packet_offload
*gro_find_receive_by_type(__be16 type
);
3231 struct packet_offload
*gro_find_complete_by_type(__be16 type
);
3233 static inline void napi_free_frags(struct napi_struct
*napi
)
3235 kfree_skb(napi
->skb
);
3239 int netdev_rx_handler_register(struct net_device
*dev
,
3240 rx_handler_func_t
*rx_handler
,
3241 void *rx_handler_data
);
3242 void netdev_rx_handler_unregister(struct net_device
*dev
);
3244 bool dev_valid_name(const char *name
);
3245 int dev_ioctl(struct net
*net
, unsigned int cmd
, void __user
*);
3246 int dev_ethtool(struct net
*net
, struct ifreq
*);
3247 unsigned int dev_get_flags(const struct net_device
*);
3248 int __dev_change_flags(struct net_device
*, unsigned int flags
);
3249 int dev_change_flags(struct net_device
*, unsigned int);
3250 void __dev_notify_flags(struct net_device
*, unsigned int old_flags
,
3251 unsigned int gchanges
);
3252 int dev_change_name(struct net_device
*, const char *);
3253 int dev_set_alias(struct net_device
*, const char *, size_t);
3254 int dev_change_net_namespace(struct net_device
*, struct net
*, const char *);
3255 int dev_set_mtu(struct net_device
*, int);
3256 void dev_set_group(struct net_device
*, int);
3257 int dev_set_mac_address(struct net_device
*, struct sockaddr
*);
3258 int dev_change_carrier(struct net_device
*, bool new_carrier
);
3259 int dev_get_phys_port_id(struct net_device
*dev
,
3260 struct netdev_phys_item_id
*ppid
);
3261 int dev_get_phys_port_name(struct net_device
*dev
,
3262 char *name
, size_t len
);
3263 int dev_change_proto_down(struct net_device
*dev
, bool proto_down
);
3264 struct sk_buff
*validate_xmit_skb_list(struct sk_buff
*skb
, struct net_device
*dev
);
3265 struct sk_buff
*dev_hard_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3266 struct netdev_queue
*txq
, int *ret
);
3267 int __dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3268 int dev_forward_skb(struct net_device
*dev
, struct sk_buff
*skb
);
3269 bool is_skb_forwardable(struct net_device
*dev
, struct sk_buff
*skb
);
3271 extern int netdev_budget
;
3273 /* Called by rtnetlink.c:rtnl_unlock() */
3274 void netdev_run_todo(void);
3277 * dev_put - release reference to device
3278 * @dev: network device
3280 * Release reference to device to allow it to be freed.
3282 static inline void dev_put(struct net_device
*dev
)
3284 this_cpu_dec(*dev
->pcpu_refcnt
);
3288 * dev_hold - get reference to device
3289 * @dev: network device
3291 * Hold reference to device to keep it from being freed.
3293 static inline void dev_hold(struct net_device
*dev
)
3295 this_cpu_inc(*dev
->pcpu_refcnt
);
3298 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3299 * and _off may be called from IRQ context, but it is caller
3300 * who is responsible for serialization of these calls.
3302 * The name carrier is inappropriate, these functions should really be
3303 * called netif_lowerlayer_*() because they represent the state of any
3304 * kind of lower layer not just hardware media.
3307 void linkwatch_init_dev(struct net_device
*dev
);
3308 void linkwatch_fire_event(struct net_device
*dev
);
3309 void linkwatch_forget_dev(struct net_device
*dev
);
3312 * netif_carrier_ok - test if carrier present
3313 * @dev: network device
3315 * Check if carrier is present on device
3317 static inline bool netif_carrier_ok(const struct net_device
*dev
)
3319 return !test_bit(__LINK_STATE_NOCARRIER
, &dev
->state
);
3322 unsigned long dev_trans_start(struct net_device
*dev
);
3324 void __netdev_watchdog_up(struct net_device
*dev
);
3326 void netif_carrier_on(struct net_device
*dev
);
3328 void netif_carrier_off(struct net_device
*dev
);
3331 * netif_dormant_on - mark device as dormant.
3332 * @dev: network device
3334 * Mark device as dormant (as per RFC2863).
3336 * The dormant state indicates that the relevant interface is not
3337 * actually in a condition to pass packets (i.e., it is not 'up') but is
3338 * in a "pending" state, waiting for some external event. For "on-
3339 * demand" interfaces, this new state identifies the situation where the
3340 * interface is waiting for events to place it in the up state.
3343 static inline void netif_dormant_on(struct net_device
*dev
)
3345 if (!test_and_set_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3346 linkwatch_fire_event(dev
);
3350 * netif_dormant_off - set device as not dormant.
3351 * @dev: network device
3353 * Device is not in dormant state.
3355 static inline void netif_dormant_off(struct net_device
*dev
)
3357 if (test_and_clear_bit(__LINK_STATE_DORMANT
, &dev
->state
))
3358 linkwatch_fire_event(dev
);
3362 * netif_dormant - test if carrier present
3363 * @dev: network device
3365 * Check if carrier is present on device
3367 static inline bool netif_dormant(const struct net_device
*dev
)
3369 return test_bit(__LINK_STATE_DORMANT
, &dev
->state
);
3374 * netif_oper_up - test if device is operational
3375 * @dev: network device
3377 * Check if carrier is operational
3379 static inline bool netif_oper_up(const struct net_device
*dev
)
3381 return (dev
->operstate
== IF_OPER_UP
||
3382 dev
->operstate
== IF_OPER_UNKNOWN
/* backward compat */);
3386 * netif_device_present - is device available or removed
3387 * @dev: network device
3389 * Check if device has not been removed from system.
3391 static inline bool netif_device_present(struct net_device
*dev
)
3393 return test_bit(__LINK_STATE_PRESENT
, &dev
->state
);
3396 void netif_device_detach(struct net_device
*dev
);
3398 void netif_device_attach(struct net_device
*dev
);
3401 * Network interface message level settings
3405 NETIF_MSG_DRV
= 0x0001,
3406 NETIF_MSG_PROBE
= 0x0002,
3407 NETIF_MSG_LINK
= 0x0004,
3408 NETIF_MSG_TIMER
= 0x0008,
3409 NETIF_MSG_IFDOWN
= 0x0010,
3410 NETIF_MSG_IFUP
= 0x0020,
3411 NETIF_MSG_RX_ERR
= 0x0040,
3412 NETIF_MSG_TX_ERR
= 0x0080,
3413 NETIF_MSG_TX_QUEUED
= 0x0100,
3414 NETIF_MSG_INTR
= 0x0200,
3415 NETIF_MSG_TX_DONE
= 0x0400,
3416 NETIF_MSG_RX_STATUS
= 0x0800,
3417 NETIF_MSG_PKTDATA
= 0x1000,
3418 NETIF_MSG_HW
= 0x2000,
3419 NETIF_MSG_WOL
= 0x4000,
3422 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3423 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3424 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3425 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3426 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3427 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3428 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3429 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3430 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3431 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3432 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3433 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3434 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3435 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3436 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3438 static inline u32
netif_msg_init(int debug_value
, int default_msg_enable_bits
)
3441 if (debug_value
< 0 || debug_value
>= (sizeof(u32
) * 8))
3442 return default_msg_enable_bits
;
3443 if (debug_value
== 0) /* no output */
3445 /* set low N bits */
3446 return (1 << debug_value
) - 1;
3449 static inline void __netif_tx_lock(struct netdev_queue
*txq
, int cpu
)
3451 spin_lock(&txq
->_xmit_lock
);
3452 txq
->xmit_lock_owner
= cpu
;
3455 static inline void __netif_tx_lock_bh(struct netdev_queue
*txq
)
3457 spin_lock_bh(&txq
->_xmit_lock
);
3458 txq
->xmit_lock_owner
= smp_processor_id();
3461 static inline bool __netif_tx_trylock(struct netdev_queue
*txq
)
3463 bool ok
= spin_trylock(&txq
->_xmit_lock
);
3465 txq
->xmit_lock_owner
= smp_processor_id();
3469 static inline void __netif_tx_unlock(struct netdev_queue
*txq
)
3471 txq
->xmit_lock_owner
= -1;
3472 spin_unlock(&txq
->_xmit_lock
);
3475 static inline void __netif_tx_unlock_bh(struct netdev_queue
*txq
)
3477 txq
->xmit_lock_owner
= -1;
3478 spin_unlock_bh(&txq
->_xmit_lock
);
3481 static inline void txq_trans_update(struct netdev_queue
*txq
)
3483 if (txq
->xmit_lock_owner
!= -1)
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
;
3762 extern int bpf_jit_enable
;
3764 bool netdev_has_upper_dev(struct net_device
*dev
, struct net_device
*upper_dev
);
3765 struct net_device
*netdev_upper_get_next_dev_rcu(struct net_device
*dev
,
3766 struct list_head
**iter
);
3767 struct net_device
*netdev_all_upper_get_next_dev_rcu(struct net_device
*dev
,
3768 struct list_head
**iter
);
3770 /* iterate through upper list, must be called under RCU read lock */
3771 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3772 for (iter = &(dev)->adj_list.upper, \
3773 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3775 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3777 /* iterate through upper list, must be called under RCU read lock */
3778 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3779 for (iter = &(dev)->all_adj_list.upper, \
3780 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3782 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3784 void *netdev_lower_get_next_private(struct net_device
*dev
,
3785 struct list_head
**iter
);
3786 void *netdev_lower_get_next_private_rcu(struct net_device
*dev
,
3787 struct list_head
**iter
);
3789 #define netdev_for_each_lower_private(dev, priv, iter) \
3790 for (iter = (dev)->adj_list.lower.next, \
3791 priv = netdev_lower_get_next_private(dev, &(iter)); \
3793 priv = netdev_lower_get_next_private(dev, &(iter)))
3795 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3796 for (iter = &(dev)->adj_list.lower, \
3797 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3799 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3801 void *netdev_lower_get_next(struct net_device
*dev
,
3802 struct list_head
**iter
);
3803 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3804 for (iter = (dev)->adj_list.lower.next, \
3805 ldev = netdev_lower_get_next(dev, &(iter)); \
3807 ldev = netdev_lower_get_next(dev, &(iter)))
3809 void *netdev_adjacent_get_private(struct list_head
*adj_list
);
3810 void *netdev_lower_get_first_private_rcu(struct net_device
*dev
);
3811 struct net_device
*netdev_master_upper_dev_get(struct net_device
*dev
);
3812 struct net_device
*netdev_master_upper_dev_get_rcu(struct net_device
*dev
);
3813 int netdev_upper_dev_link(struct net_device
*dev
, struct net_device
*upper_dev
);
3814 int netdev_master_upper_dev_link(struct net_device
*dev
,
3815 struct net_device
*upper_dev
,
3816 void *upper_priv
, void *upper_info
);
3817 void netdev_upper_dev_unlink(struct net_device
*dev
,
3818 struct net_device
*upper_dev
);
3819 void netdev_adjacent_rename_links(struct net_device
*dev
, char *oldname
);
3820 void *netdev_lower_dev_get_private(struct net_device
*dev
,
3821 struct net_device
*lower_dev
);
3822 void netdev_lower_state_changed(struct net_device
*lower_dev
,
3823 void *lower_state_info
);
3825 /* RSS keys are 40 or 52 bytes long */
3826 #define NETDEV_RSS_KEY_LEN 52
3827 extern u8 netdev_rss_key
[NETDEV_RSS_KEY_LEN
] __read_mostly
;
3828 void netdev_rss_key_fill(void *buffer
, size_t len
);
3830 int dev_get_nest_level(struct net_device
*dev
,
3831 bool (*type_check
)(const struct net_device
*dev
));
3832 int skb_checksum_help(struct sk_buff
*skb
);
3833 struct sk_buff
*__skb_gso_segment(struct sk_buff
*skb
,
3834 netdev_features_t features
, bool tx_path
);
3835 struct sk_buff
*skb_mac_gso_segment(struct sk_buff
*skb
,
3836 netdev_features_t features
);
3838 struct netdev_bonding_info
{
3843 struct netdev_notifier_bonding_info
{
3844 struct netdev_notifier_info info
; /* must be first */
3845 struct netdev_bonding_info bonding_info
;
3848 void netdev_bonding_info_change(struct net_device
*dev
,
3849 struct netdev_bonding_info
*bonding_info
);
3852 struct sk_buff
*skb_gso_segment(struct sk_buff
*skb
, netdev_features_t features
)
3854 return __skb_gso_segment(skb
, features
, true);
3856 __be16
skb_network_protocol(struct sk_buff
*skb
, int *depth
);
3858 static inline bool can_checksum_protocol(netdev_features_t features
,
3861 if (protocol
== htons(ETH_P_FCOE
))
3862 return !!(features
& NETIF_F_FCOE_CRC
);
3864 /* Assume this is an IP checksum (not SCTP CRC) */
3866 if (features
& NETIF_F_HW_CSUM
) {
3867 /* Can checksum everything */
3872 case htons(ETH_P_IP
):
3873 return !!(features
& NETIF_F_IP_CSUM
);
3874 case htons(ETH_P_IPV6
):
3875 return !!(features
& NETIF_F_IPV6_CSUM
);
3881 /* Map an ethertype into IP protocol if possible */
3882 static inline int eproto_to_ipproto(int eproto
)
3885 case htons(ETH_P_IP
):
3887 case htons(ETH_P_IPV6
):
3888 return IPPROTO_IPV6
;
3895 void netdev_rx_csum_fault(struct net_device
*dev
);
3897 static inline void netdev_rx_csum_fault(struct net_device
*dev
)
3901 /* rx skb timestamps */
3902 void net_enable_timestamp(void);
3903 void net_disable_timestamp(void);
3905 #ifdef CONFIG_PROC_FS
3906 int __init
dev_proc_init(void);
3908 #define dev_proc_init() 0
3911 static inline netdev_tx_t
__netdev_start_xmit(const struct net_device_ops
*ops
,
3912 struct sk_buff
*skb
, struct net_device
*dev
,
3915 skb
->xmit_more
= more
? 1 : 0;
3916 return ops
->ndo_start_xmit(skb
, dev
);
3919 static inline netdev_tx_t
netdev_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
,
3920 struct netdev_queue
*txq
, bool more
)
3922 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3925 rc
= __netdev_start_xmit(ops
, skb
, dev
, more
);
3926 if (rc
== NETDEV_TX_OK
)
3927 txq_trans_update(txq
);
3932 int netdev_class_create_file_ns(struct class_attribute
*class_attr
,
3934 void netdev_class_remove_file_ns(struct class_attribute
*class_attr
,
3937 static inline int netdev_class_create_file(struct class_attribute
*class_attr
)
3939 return netdev_class_create_file_ns(class_attr
, NULL
);
3942 static inline void netdev_class_remove_file(struct class_attribute
*class_attr
)
3944 netdev_class_remove_file_ns(class_attr
, NULL
);
3947 extern struct kobj_ns_type_operations net_ns_type_operations
;
3949 const char *netdev_drivername(const struct net_device
*dev
);
3951 void linkwatch_run_queue(void);
3953 static inline netdev_features_t
netdev_intersect_features(netdev_features_t f1
,
3954 netdev_features_t f2
)
3956 if ((f1
^ f2
) & NETIF_F_HW_CSUM
) {
3957 if (f1
& NETIF_F_HW_CSUM
)
3958 f1
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
3960 f2
|= (NETIF_F_IP_CSUM
|NETIF_F_IPV6_CSUM
);
3966 static inline netdev_features_t
netdev_get_wanted_features(
3967 struct net_device
*dev
)
3969 return (dev
->features
& ~dev
->hw_features
) | dev
->wanted_features
;
3971 netdev_features_t
netdev_increment_features(netdev_features_t all
,
3972 netdev_features_t one
, netdev_features_t mask
);
3974 /* Allow TSO being used on stacked device :
3975 * Performing the GSO segmentation before last device
3976 * is a performance improvement.
3978 static inline netdev_features_t
netdev_add_tso_features(netdev_features_t features
,
3979 netdev_features_t mask
)
3981 return netdev_increment_features(features
, NETIF_F_ALL_TSO
, mask
);
3984 int __netdev_update_features(struct net_device
*dev
);
3985 void netdev_update_features(struct net_device
*dev
);
3986 void netdev_change_features(struct net_device
*dev
);
3988 void netif_stacked_transfer_operstate(const struct net_device
*rootdev
,
3989 struct net_device
*dev
);
3991 netdev_features_t
passthru_features_check(struct sk_buff
*skb
,
3992 struct net_device
*dev
,
3993 netdev_features_t features
);
3994 netdev_features_t
netif_skb_features(struct sk_buff
*skb
);
3996 static inline bool net_gso_ok(netdev_features_t features
, int gso_type
)
3998 netdev_features_t feature
= gso_type
<< NETIF_F_GSO_SHIFT
;
4000 /* check flags correspondence */
4001 BUILD_BUG_ON(SKB_GSO_TCPV4
!= (NETIF_F_TSO
>> NETIF_F_GSO_SHIFT
));
4002 BUILD_BUG_ON(SKB_GSO_UDP
!= (NETIF_F_UFO
>> NETIF_F_GSO_SHIFT
));
4003 BUILD_BUG_ON(SKB_GSO_DODGY
!= (NETIF_F_GSO_ROBUST
>> NETIF_F_GSO_SHIFT
));
4004 BUILD_BUG_ON(SKB_GSO_TCP_ECN
!= (NETIF_F_TSO_ECN
>> 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_IPIP
!= (NETIF_F_GSO_IPIP
>> NETIF_F_GSO_SHIFT
));
4010 BUILD_BUG_ON(SKB_GSO_SIT
!= (NETIF_F_GSO_SIT
>> 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_TUNNEL_REMCSUM
!= (NETIF_F_GSO_TUNNEL_REMCSUM
>> NETIF_F_GSO_SHIFT
));
4015 return (features
& feature
) == feature
;
4018 static inline bool skb_gso_ok(struct sk_buff
*skb
, netdev_features_t features
)
4020 return net_gso_ok(features
, skb_shinfo(skb
)->gso_type
) &&
4021 (!skb_has_frag_list(skb
) || (features
& NETIF_F_FRAGLIST
));
4024 static inline bool netif_needs_gso(struct sk_buff
*skb
,
4025 netdev_features_t features
)
4027 return skb_is_gso(skb
) && (!skb_gso_ok(skb
, features
) ||
4028 unlikely((skb
->ip_summed
!= CHECKSUM_PARTIAL
) &&
4029 (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)));
4032 static inline void netif_set_gso_max_size(struct net_device
*dev
,
4035 dev
->gso_max_size
= size
;
4038 static inline void skb_gso_error_unwind(struct sk_buff
*skb
, __be16 protocol
,
4039 int pulled_hlen
, u16 mac_offset
,
4042 skb
->protocol
= protocol
;
4043 skb
->encapsulation
= 1;
4044 skb_push(skb
, pulled_hlen
);
4045 skb_reset_transport_header(skb
);
4046 skb
->mac_header
= mac_offset
;
4047 skb
->network_header
= skb
->mac_header
+ mac_len
;
4048 skb
->mac_len
= mac_len
;
4051 static inline bool netif_is_macsec(const struct net_device
*dev
)
4053 return dev
->priv_flags
& IFF_MACSEC
;
4056 static inline bool netif_is_macvlan(const struct net_device
*dev
)
4058 return dev
->priv_flags
& IFF_MACVLAN
;
4061 static inline bool netif_is_macvlan_port(const struct net_device
*dev
)
4063 return dev
->priv_flags
& IFF_MACVLAN_PORT
;
4066 static inline bool netif_is_ipvlan(const struct net_device
*dev
)
4068 return dev
->priv_flags
& IFF_IPVLAN_SLAVE
;
4071 static inline bool netif_is_ipvlan_port(const struct net_device
*dev
)
4073 return dev
->priv_flags
& IFF_IPVLAN_MASTER
;
4076 static inline bool netif_is_bond_master(const struct net_device
*dev
)
4078 return dev
->flags
& IFF_MASTER
&& dev
->priv_flags
& IFF_BONDING
;
4081 static inline bool netif_is_bond_slave(const struct net_device
*dev
)
4083 return dev
->flags
& IFF_SLAVE
&& dev
->priv_flags
& IFF_BONDING
;
4086 static inline bool netif_supports_nofcs(struct net_device
*dev
)
4088 return dev
->priv_flags
& IFF_SUPP_NOFCS
;
4091 static inline bool netif_is_l3_master(const struct net_device
*dev
)
4093 return dev
->priv_flags
& IFF_L3MDEV_MASTER
;
4096 static inline bool netif_is_l3_slave(const struct net_device
*dev
)
4098 return dev
->priv_flags
& IFF_L3MDEV_SLAVE
;
4101 static inline bool netif_is_bridge_master(const struct net_device
*dev
)
4103 return dev
->priv_flags
& IFF_EBRIDGE
;
4106 static inline bool netif_is_bridge_port(const struct net_device
*dev
)
4108 return dev
->priv_flags
& IFF_BRIDGE_PORT
;
4111 static inline bool netif_is_ovs_master(const struct net_device
*dev
)
4113 return dev
->priv_flags
& IFF_OPENVSWITCH
;
4116 static inline bool netif_is_team_master(const struct net_device
*dev
)
4118 return dev
->priv_flags
& IFF_TEAM
;
4121 static inline bool netif_is_team_port(const struct net_device
*dev
)
4123 return dev
->priv_flags
& IFF_TEAM_PORT
;
4126 static inline bool netif_is_lag_master(const struct net_device
*dev
)
4128 return netif_is_bond_master(dev
) || netif_is_team_master(dev
);
4131 static inline bool netif_is_lag_port(const struct net_device
*dev
)
4133 return netif_is_bond_slave(dev
) || netif_is_team_port(dev
);
4136 static inline bool netif_is_rxfh_configured(const struct net_device
*dev
)
4138 return dev
->priv_flags
& IFF_RXFH_CONFIGURED
;
4141 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4142 static inline void netif_keep_dst(struct net_device
*dev
)
4144 dev
->priv_flags
&= ~(IFF_XMIT_DST_RELEASE
| IFF_XMIT_DST_RELEASE_PERM
);
4147 extern struct pernet_operations __net_initdata loopback_net_ops
;
4149 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4151 /* netdev_printk helpers, similar to dev_printk */
4153 static inline const char *netdev_name(const struct net_device
*dev
)
4155 if (!dev
->name
[0] || strchr(dev
->name
, '%'))
4156 return "(unnamed net_device)";
4160 static inline const char *netdev_reg_state(const struct net_device
*dev
)
4162 switch (dev
->reg_state
) {
4163 case NETREG_UNINITIALIZED
: return " (uninitialized)";
4164 case NETREG_REGISTERED
: return "";
4165 case NETREG_UNREGISTERING
: return " (unregistering)";
4166 case NETREG_UNREGISTERED
: return " (unregistered)";
4167 case NETREG_RELEASED
: return " (released)";
4168 case NETREG_DUMMY
: return " (dummy)";
4171 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev
->name
, dev
->reg_state
);
4172 return " (unknown)";
4176 void netdev_printk(const char *level
, const struct net_device
*dev
,
4177 const char *format
, ...);
4179 void netdev_emerg(const struct net_device
*dev
, const char *format
, ...);
4181 void netdev_alert(const struct net_device
*dev
, const char *format
, ...);
4183 void netdev_crit(const struct net_device
*dev
, const char *format
, ...);
4185 void netdev_err(const struct net_device
*dev
, const char *format
, ...);
4187 void netdev_warn(const struct net_device
*dev
, const char *format
, ...);
4189 void netdev_notice(const struct net_device
*dev
, const char *format
, ...);
4191 void netdev_info(const struct net_device
*dev
, const char *format
, ...);
4193 #define MODULE_ALIAS_NETDEV(device) \
4194 MODULE_ALIAS("netdev-" device)
4196 #if defined(CONFIG_DYNAMIC_DEBUG)
4197 #define netdev_dbg(__dev, format, args...) \
4199 dynamic_netdev_dbg(__dev, format, ##args); \
4201 #elif defined(DEBUG)
4202 #define netdev_dbg(__dev, format, args...) \
4203 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4205 #define netdev_dbg(__dev, format, args...) \
4208 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4212 #if defined(VERBOSE_DEBUG)
4213 #define netdev_vdbg netdev_dbg
4216 #define netdev_vdbg(dev, format, args...) \
4219 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4225 * netdev_WARN() acts like dev_printk(), but with the key difference
4226 * of using a WARN/WARN_ON to get the message out, including the
4227 * file/line information and a backtrace.
4229 #define netdev_WARN(dev, format, args...) \
4230 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4231 netdev_reg_state(dev), ##args)
4233 /* netif printk helpers, similar to netdev_printk */
4235 #define netif_printk(priv, type, level, dev, fmt, args...) \
4237 if (netif_msg_##type(priv)) \
4238 netdev_printk(level, (dev), fmt, ##args); \
4241 #define netif_level(level, priv, type, dev, fmt, args...) \
4243 if (netif_msg_##type(priv)) \
4244 netdev_##level(dev, fmt, ##args); \
4247 #define netif_emerg(priv, type, dev, fmt, args...) \
4248 netif_level(emerg, priv, type, dev, fmt, ##args)
4249 #define netif_alert(priv, type, dev, fmt, args...) \
4250 netif_level(alert, priv, type, dev, fmt, ##args)
4251 #define netif_crit(priv, type, dev, fmt, args...) \
4252 netif_level(crit, priv, type, dev, fmt, ##args)
4253 #define netif_err(priv, type, dev, fmt, args...) \
4254 netif_level(err, priv, type, dev, fmt, ##args)
4255 #define netif_warn(priv, type, dev, fmt, args...) \
4256 netif_level(warn, priv, type, dev, fmt, ##args)
4257 #define netif_notice(priv, type, dev, fmt, args...) \
4258 netif_level(notice, priv, type, dev, fmt, ##args)
4259 #define netif_info(priv, type, dev, fmt, args...) \
4260 netif_level(info, priv, type, dev, fmt, ##args)
4262 #if defined(CONFIG_DYNAMIC_DEBUG)
4263 #define netif_dbg(priv, type, netdev, format, args...) \
4265 if (netif_msg_##type(priv)) \
4266 dynamic_netdev_dbg(netdev, format, ##args); \
4268 #elif defined(DEBUG)
4269 #define netif_dbg(priv, type, dev, format, args...) \
4270 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4272 #define netif_dbg(priv, type, dev, format, args...) \
4275 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4280 #if defined(VERBOSE_DEBUG)
4281 #define netif_vdbg netif_dbg
4283 #define netif_vdbg(priv, type, dev, format, args...) \
4286 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4292 * The list of packet types we will receive (as opposed to discard)
4293 * and the routines to invoke.
4295 * Why 16. Because with 16 the only overlap we get on a hash of the
4296 * low nibble of the protocol value is RARP/SNAP/X.25.
4298 * NOTE: That is no longer true with the addition of VLAN tags. Not
4299 * sure which should go first, but I bet it won't make much
4300 * difference if we are running VLANs. The good news is that
4301 * this protocol won't be in the list unless compiled in, so
4302 * the average user (w/out VLANs) will not be adversely affected.
4318 #define PTYPE_HASH_SIZE (16)
4319 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4321 #endif /* _LINUX_NETDEVICE_H */