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>
60 /* 802.15.4 specific */
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
72 * Transmit return codes: transmit return codes originate from three different
75 * - qdisc return codes
76 * - driver transmit return codes
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
110 typedef enum netdev_tx netdev_tx_t;
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 static inline bool dev_xmit_complete(int rc)
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 if (likely(rc < NET_XMIT_MASK))
131 * Compute the worst case header length according to the protocols
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
139 # define LL_MAX_HEADER 96
142 # define LL_MAX_HEADER 32
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
209 struct rcu_head rcu_head;
212 struct netdev_hw_addr_list {
213 struct list_head list;
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
275 enum netdev_state_t {
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
288 struct netdev_boot_setup {
292 #define NETDEV_BOOT_SETUP_MAX 8
294 int __init netdev_boot_setup(char *str);
297 * Structure for NAPI scheduling similar to tasklet but with weighting
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
306 struct list_head poll_list;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
316 struct net_device *dev;
317 struct sk_buff *gro_list;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
339 typedef enum gro_result gro_result_t;
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
382 enum rx_handler_result {
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
394 static inline bool napi_disable_pending(struct napi_struct *n)
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
400 * napi_schedule_prep - check if napi can be scheduled
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
408 static inline bool napi_schedule_prep(struct napi_struct *n)
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
415 * napi_schedule - schedule NAPI poll
418 * Schedule NAPI poll routine to be called if it is not already
421 static inline void napi_schedule(struct napi_struct *n)
423 if (napi_schedule_prep(n))
428 * napi_schedule_irqoff - schedule NAPI poll
431 * Variant of napi_schedule(), assuming hard irqs are masked.
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
452 * napi_complete - NAPI processing complete
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
458 static inline void napi_complete(struct napi_struct *n)
460 return napi_complete_done(n, 0);
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
470 struct napi_struct *napi_by_id(unsigned int napi_id);
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
476 * generate a new napi_id and store a @napi under it in napi_hash
478 void napi_hash_add(struct napi_struct *napi);
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
487 void napi_hash_del(struct napi_struct *napi);
490 * napi_disable - prevent NAPI from scheduling
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
496 void napi_disable(struct napi_struct *n);
499 * napi_enable - enable NAPI scheduling
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
505 static inline void napi_enable(struct napi_struct *n)
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510 clear_bit(NAPI_STATE_NPSVC, &n->state);
515 * napi_synchronize - wait until NAPI is not running
518 * Wait until NAPI is done being scheduled on this context.
519 * Waits till any outstanding processing completes but
520 * does not disable future activations.
522 static inline void napi_synchronize(const struct napi_struct *n)
524 while (test_bit(NAPI_STATE_SCHED, &n->state))
528 # define napi_synchronize(n) barrier()
531 enum netdev_queue_state_t {
532 __QUEUE_STATE_DRV_XOFF,
533 __QUEUE_STATE_STACK_XOFF,
534 __QUEUE_STATE_FROZEN,
537 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
538 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
539 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
541 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
544 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
549 * netif_tx_* functions below are used to manipulate this flag. The
550 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
551 * queue independently. The netif_xmit_*stopped functions below are called
552 * to check if the queue has been stopped by the driver or stack (either
553 * of the XOFF bits are set in the state). Drivers should not need to call
554 * netif_xmit*stopped functions, they should only be using netif_tx_*.
557 struct netdev_queue {
561 struct net_device *dev;
562 struct Qdisc __rcu *qdisc;
563 struct Qdisc *qdisc_sleeping;
567 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
576 * please use this field instead of dev->trans_start
578 unsigned long trans_start;
581 * Number of TX timeouts for this queue
582 * (/sys/class/net/DEV/Q/trans_timeout)
584 unsigned long trans_timeout;
591 unsigned long tx_maxrate;
592 } ____cacheline_aligned_in_smp;
594 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
596 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
603 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
605 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
612 * This structure holds an RPS map which can be of variable length. The
613 * map is an array of CPUs.
620 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
623 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
624 * tail pointer for that CPU's input queue at the time of last enqueue, and
625 * a hardware filter index.
627 struct rps_dev_flow {
630 unsigned int last_qtail;
632 #define RPS_NO_FILTER 0xffff
635 * The rps_dev_flow_table structure contains a table of flow mappings.
637 struct rps_dev_flow_table {
640 struct rps_dev_flow flows[0];
642 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
643 ((_num) * sizeof(struct rps_dev_flow)))
646 * The rps_sock_flow_table contains mappings of flows to the last CPU
647 * on which they were processed by the application (set in recvmsg).
648 * Each entry is a 32bit value. Upper part is the high order bits
649 * of flow hash, lower part is cpu number.
650 * rps_cpu_mask is used to partition the space, depending on number of
651 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
652 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
653 * meaning we use 32-6=26 bits for the hash.
655 struct rps_sock_flow_table {
658 u32 ents[0] ____cacheline_aligned_in_smp;
660 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
662 #define RPS_NO_CPU 0xffff
664 extern u32 rps_cpu_mask;
665 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 unsigned int index = hash & table->mask;
672 u32 val = hash & ~rps_cpu_mask;
674 /* We only give a hint, preemption can change cpu under us */
675 val |= raw_smp_processor_id();
677 if (table->ents[index] != val)
678 table->ents[index] = val;
682 #ifdef CONFIG_RFS_ACCEL
683 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 #endif /* CONFIG_RPS */
688 /* This structure contains an instance of an RX queue. */
689 struct netdev_rx_queue {
691 struct rps_map __rcu *rps_map;
692 struct rps_dev_flow_table __rcu *rps_flow_table;
695 struct net_device *dev;
696 } ____cacheline_aligned_in_smp;
699 * RX queue sysfs structures and functions.
701 struct rx_queue_attribute {
702 struct attribute attr;
703 ssize_t (*show)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, char *buf);
705 ssize_t (*store)(struct netdev_rx_queue *queue,
706 struct rx_queue_attribute *attr, const char *buf, size_t len);
711 * This structure holds an XPS map which can be of variable length. The
712 * map is an array of queues.
716 unsigned int alloc_len;
720 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
721 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
722 - sizeof(struct xps_map)) / sizeof(u16))
725 * This structure holds all XPS maps for device. Maps are indexed by CPU.
727 struct xps_dev_maps {
729 struct xps_map __rcu *cpu_map[0];
731 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
732 (nr_cpu_ids * sizeof(struct xps_map *)))
733 #endif /* CONFIG_XPS */
735 #define TC_MAX_QUEUE 16
736 #define TC_BITMASK 15
737 /* HW offloaded queuing disciplines txq count and offset maps */
738 struct netdev_tc_txq {
743 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
745 * This structure is to hold information about the device
746 * configured to run FCoE protocol stack.
748 struct netdev_fcoe_hbainfo {
749 char manufacturer[64];
750 char serial_number[64];
751 char hardware_version[64];
752 char driver_version[64];
753 char optionrom_version[64];
754 char firmware_version[64];
756 char model_description[256];
760 #define MAX_PHYS_ITEM_ID_LEN 32
762 /* This structure holds a unique identifier to identify some
763 * physical item (port for example) used by a netdevice.
765 struct netdev_phys_item_id {
766 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
767 unsigned char id_len;
770 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
771 struct netdev_phys_item_id *b)
773 return a->id_len == b->id_len &&
774 memcmp(a->id, b->id, a->id_len) == 0;
777 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
778 struct sk_buff *skb);
781 * This structure defines the management hooks for network devices.
782 * The following hooks can be defined; unless noted otherwise, they are
783 * optional and can be filled with a null pointer.
785 * int (*ndo_init)(struct net_device *dev);
786 * This function is called once when network device is registered.
787 * The network device can use this to any late stage initializaton
788 * or semantic validattion. It can fail with an error code which will
789 * be propogated back to register_netdev
791 * void (*ndo_uninit)(struct net_device *dev);
792 * This function is called when device is unregistered or when registration
793 * fails. It is not called if init fails.
795 * int (*ndo_open)(struct net_device *dev);
796 * This function is called when network device transistions to the up
799 * int (*ndo_stop)(struct net_device *dev);
800 * This function is called when network device transistions to the down
803 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
804 * struct net_device *dev);
805 * Called when a packet needs to be transmitted.
806 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
807 * the queue before that can happen; it's for obsolete devices and weird
808 * corner cases, but the stack really does a non-trivial amount
809 * of useless work if you return NETDEV_TX_BUSY.
810 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
811 * Required can not be NULL.
813 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
814 * void *accel_priv, select_queue_fallback_t fallback);
815 * Called to decide which queue to when device supports multiple
818 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
819 * This function is called to allow device receiver to make
820 * changes to configuration when multicast or promiscious is enabled.
822 * void (*ndo_set_rx_mode)(struct net_device *dev);
823 * This function is called device changes address list filtering.
824 * If driver handles unicast address filtering, it should set
825 * IFF_UNICAST_FLT to its priv_flags.
827 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
828 * This function is called when the Media Access Control address
829 * needs to be changed. If this interface is not defined, the
830 * mac address can not be changed.
832 * int (*ndo_validate_addr)(struct net_device *dev);
833 * Test if Media Access Control address is valid for the device.
835 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
836 * Called when a user request an ioctl which can't be handled by
837 * the generic interface code. If not defined ioctl's return
838 * not supported error code.
840 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
841 * Used to set network devices bus interface parameters. This interface
842 * is retained for legacy reason, new devices should use the bus
843 * interface (PCI) for low level management.
845 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
846 * Called when a user wants to change the Maximum Transfer Unit
847 * of a device. If not defined, any request to change MTU will
848 * will return an error.
850 * void (*ndo_tx_timeout)(struct net_device *dev);
851 * Callback uses when the transmitter has not made any progress
852 * for dev->watchdog ticks.
854 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
855 * struct rtnl_link_stats64 *storage);
856 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
857 * Called when a user wants to get the network device usage
858 * statistics. Drivers must do one of the following:
859 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
860 * rtnl_link_stats64 structure passed by the caller.
861 * 2. Define @ndo_get_stats to update a net_device_stats structure
862 * (which should normally be dev->stats) and return a pointer to
863 * it. The structure may be changed asynchronously only if each
864 * field is written atomically.
865 * 3. Update dev->stats asynchronously and atomically, and define
868 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
869 * If device support VLAN filtering this function is called when a
870 * VLAN id is registered.
872 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
873 * If device support VLAN filtering this function is called when a
874 * VLAN id is unregistered.
876 * void (*ndo_poll_controller)(struct net_device *dev);
878 * SR-IOV management functions.
879 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
880 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
881 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
883 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
884 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
885 * int (*ndo_get_vf_config)(struct net_device *dev,
886 * int vf, struct ifla_vf_info *ivf);
887 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
888 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
889 * struct nlattr *port[]);
891 * Enable or disable the VF ability to query its RSS Redirection Table and
892 * Hash Key. This is needed since on some devices VF share this information
893 * with PF and querying it may adduce a theoretical security risk.
894 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
895 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
896 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
897 * Called to setup 'tc' number of traffic classes in the net device. This
898 * is always called from the stack with the rtnl lock held and netif tx
899 * queues stopped. This allows the netdevice to perform queue management
902 * Fiber Channel over Ethernet (FCoE) offload functions.
903 * int (*ndo_fcoe_enable)(struct net_device *dev);
904 * Called when the FCoE protocol stack wants to start using LLD for FCoE
905 * so the underlying device can perform whatever needed configuration or
906 * initialization to support acceleration of FCoE traffic.
908 * int (*ndo_fcoe_disable)(struct net_device *dev);
909 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
910 * so the underlying device can perform whatever needed clean-ups to
911 * stop supporting acceleration of FCoE traffic.
913 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
914 * struct scatterlist *sgl, unsigned int sgc);
915 * Called when the FCoE Initiator wants to initialize an I/O that
916 * is a possible candidate for Direct Data Placement (DDP). The LLD can
917 * perform necessary setup and returns 1 to indicate the device is set up
918 * successfully to perform DDP on this I/O, otherwise this returns 0.
920 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
921 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
922 * indicated by the FC exchange id 'xid', so the underlying device can
923 * clean up and reuse resources for later DDP requests.
925 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
926 * struct scatterlist *sgl, unsigned int sgc);
927 * Called when the FCoE Target wants to initialize an I/O that
928 * is a possible candidate for Direct Data Placement (DDP). The LLD can
929 * perform necessary setup and returns 1 to indicate the device is set up
930 * successfully to perform DDP on this I/O, otherwise this returns 0.
932 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
933 * struct netdev_fcoe_hbainfo *hbainfo);
934 * Called when the FCoE Protocol stack wants information on the underlying
935 * device. This information is utilized by the FCoE protocol stack to
936 * register attributes with Fiber Channel management service as per the
937 * FC-GS Fabric Device Management Information(FDMI) specification.
939 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
940 * Called when the underlying device wants to override default World Wide
941 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
942 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
943 * protocol stack to use.
946 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
947 * u16 rxq_index, u32 flow_id);
948 * Set hardware filter for RFS. rxq_index is the target queue index;
949 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
950 * Return the filter ID on success, or a negative error code.
952 * Slave management functions (for bridge, bonding, etc).
953 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
954 * Called to make another netdev an underling.
956 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
957 * Called to release previously enslaved netdev.
959 * Feature/offload setting functions.
960 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
961 * netdev_features_t features);
962 * Adjusts the requested feature flags according to device-specific
963 * constraints, and returns the resulting flags. Must not modify
966 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
967 * Called to update device configuration to new features. Passed
968 * feature set might be less than what was returned by ndo_fix_features()).
969 * Must return >0 or -errno if it changed dev->features itself.
971 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
972 * struct net_device *dev,
973 * const unsigned char *addr, u16 vid, u16 flags)
974 * Adds an FDB entry to dev for addr.
975 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
976 * struct net_device *dev,
977 * const unsigned char *addr, u16 vid)
978 * Deletes the FDB entry from dev coresponding to addr.
979 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
980 * struct net_device *dev, struct net_device *filter_dev,
982 * Used to add FDB entries to dump requests. Implementers should add
983 * entries to skb and update idx with the number of entries.
985 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
987 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
988 * struct net_device *dev, u32 filter_mask,
990 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
993 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
994 * Called to change device carrier. Soft-devices (like dummy, team, etc)
995 * which do not represent real hardware may define this to allow their
996 * userspace components to manage their virtual carrier state. Devices
997 * that determine carrier state from physical hardware properties (eg
998 * network cables) or protocol-dependent mechanisms (eg
999 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1001 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1002 * struct netdev_phys_item_id *ppid);
1003 * Called to get ID of physical port of this device. If driver does
1004 * not implement this, it is assumed that the hw is not able to have
1005 * multiple net devices on single physical port.
1007 * void (*ndo_add_vxlan_port)(struct net_device *dev,
1008 * sa_family_t sa_family, __be16 port);
1009 * Called by vxlan to notiy a driver about the UDP port and socket
1010 * address family that vxlan is listnening to. It is called only when
1011 * a new port starts listening. The operation is protected by the
1012 * vxlan_net->sock_lock.
1014 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1015 * sa_family_t sa_family, __be16 port);
1016 * Called by vxlan to notify the driver about a UDP port and socket
1017 * address family that vxlan is not listening to anymore. The operation
1018 * is protected by the vxlan_net->sock_lock.
1020 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1021 * struct net_device *dev)
1022 * Called by upper layer devices to accelerate switching or other
1023 * station functionality into hardware. 'pdev is the lowerdev
1024 * to use for the offload and 'dev' is the net device that will
1025 * back the offload. Returns a pointer to the private structure
1026 * the upper layer will maintain.
1027 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1028 * Called by upper layer device to delete the station created
1029 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1030 * the station and priv is the structure returned by the add
1032 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1033 * struct net_device *dev,
1035 * Callback to use for xmit over the accelerated station. This
1036 * is used in place of ndo_start_xmit on accelerated net
1038 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1039 * struct net_device *dev
1040 * netdev_features_t features);
1041 * Called by core transmit path to determine if device is capable of
1042 * performing offload operations on a given packet. This is to give
1043 * the device an opportunity to implement any restrictions that cannot
1044 * be otherwise expressed by feature flags. The check is called with
1045 * the set of features that the stack has calculated and it returns
1046 * those the driver believes to be appropriate.
1047 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1048 * int queue_index, u32 maxrate);
1049 * Called when a user wants to set a max-rate limitation of specific
1051 * int (*ndo_get_iflink)(const struct net_device *dev);
1052 * Called to get the iflink value of this device.
1053 * void (*ndo_change_proto_down)(struct net_device *dev,
1055 * This function is used to pass protocol port error state information
1056 * to the switch driver. The switch driver can react to the proto_down
1057 * by doing a phys down on the associated switch port.
1058 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1059 * This function is used to get egress tunnel information for given skb.
1060 * This is useful for retrieving outer tunnel header parameters while
1064 struct net_device_ops {
1065 int (*ndo_init)(struct net_device *dev);
1066 void (*ndo_uninit)(struct net_device *dev);
1067 int (*ndo_open)(struct net_device *dev);
1068 int (*ndo_stop)(struct net_device *dev);
1069 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1070 struct net_device *dev);
1071 u16 (*ndo_select_queue)(struct net_device *dev,
1072 struct sk_buff *skb,
1074 select_queue_fallback_t fallback);
1075 void (*ndo_change_rx_flags)(struct net_device *dev,
1077 void (*ndo_set_rx_mode)(struct net_device *dev);
1078 int (*ndo_set_mac_address)(struct net_device *dev,
1080 int (*ndo_validate_addr)(struct net_device *dev);
1081 int (*ndo_do_ioctl)(struct net_device *dev,
1082 struct ifreq *ifr, int cmd);
1083 int (*ndo_set_config)(struct net_device *dev,
1085 int (*ndo_change_mtu)(struct net_device *dev,
1087 int (*ndo_neigh_setup)(struct net_device *dev,
1088 struct neigh_parms *);
1089 void (*ndo_tx_timeout) (struct net_device *dev);
1091 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1092 struct rtnl_link_stats64 *storage);
1093 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1095 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1096 __be16 proto, u16 vid);
1097 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1098 __be16 proto, u16 vid);
1099 #ifdef CONFIG_NET_POLL_CONTROLLER
1100 void (*ndo_poll_controller)(struct net_device *dev);
1101 int (*ndo_netpoll_setup)(struct net_device *dev,
1102 struct netpoll_info *info);
1103 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1105 #ifdef CONFIG_NET_RX_BUSY_POLL
1106 int (*ndo_busy_poll)(struct napi_struct *dev);
1108 int (*ndo_set_vf_mac)(struct net_device *dev,
1109 int queue, u8 *mac);
1110 int (*ndo_set_vf_vlan)(struct net_device *dev,
1111 int queue, u16 vlan, u8 qos);
1112 int (*ndo_set_vf_rate)(struct net_device *dev,
1113 int vf, int min_tx_rate,
1115 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1116 int vf, bool setting);
1117 int (*ndo_set_vf_trust)(struct net_device *dev,
1118 int vf, bool setting);
1119 int (*ndo_get_vf_config)(struct net_device *dev,
1121 struct ifla_vf_info *ivf);
1122 int (*ndo_set_vf_link_state)(struct net_device *dev,
1123 int vf, int link_state);
1124 int (*ndo_get_vf_stats)(struct net_device *dev,
1126 struct ifla_vf_stats
1128 int (*ndo_set_vf_port)(struct net_device *dev,
1130 struct nlattr *port[]);
1131 int (*ndo_get_vf_port)(struct net_device *dev,
1132 int vf, struct sk_buff *skb);
1133 int (*ndo_set_vf_rss_query_en)(
1134 struct net_device *dev,
1135 int vf, bool setting);
1136 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1137 #if IS_ENABLED(CONFIG_FCOE)
1138 int (*ndo_fcoe_enable)(struct net_device *dev);
1139 int (*ndo_fcoe_disable)(struct net_device *dev);
1140 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1142 struct scatterlist *sgl,
1144 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1146 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1148 struct scatterlist *sgl,
1150 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1151 struct netdev_fcoe_hbainfo *hbainfo);
1154 #if IS_ENABLED(CONFIG_LIBFCOE)
1155 #define NETDEV_FCOE_WWNN 0
1156 #define NETDEV_FCOE_WWPN 1
1157 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1158 u64 *wwn, int type);
1161 #ifdef CONFIG_RFS_ACCEL
1162 int (*ndo_rx_flow_steer)(struct net_device *dev,
1163 const struct sk_buff *skb,
1167 int (*ndo_add_slave)(struct net_device *dev,
1168 struct net_device *slave_dev);
1169 int (*ndo_del_slave)(struct net_device *dev,
1170 struct net_device *slave_dev);
1171 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1172 netdev_features_t features);
1173 int (*ndo_set_features)(struct net_device *dev,
1174 netdev_features_t features);
1175 int (*ndo_neigh_construct)(struct neighbour *n);
1176 void (*ndo_neigh_destroy)(struct neighbour *n);
1178 int (*ndo_fdb_add)(struct ndmsg *ndm,
1179 struct nlattr *tb[],
1180 struct net_device *dev,
1181 const unsigned char *addr,
1184 int (*ndo_fdb_del)(struct ndmsg *ndm,
1185 struct nlattr *tb[],
1186 struct net_device *dev,
1187 const unsigned char *addr,
1189 int (*ndo_fdb_dump)(struct sk_buff *skb,
1190 struct netlink_callback *cb,
1191 struct net_device *dev,
1192 struct net_device *filter_dev,
1195 int (*ndo_bridge_setlink)(struct net_device *dev,
1196 struct nlmsghdr *nlh,
1198 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1200 struct net_device *dev,
1203 int (*ndo_bridge_dellink)(struct net_device *dev,
1204 struct nlmsghdr *nlh,
1206 int (*ndo_change_carrier)(struct net_device *dev,
1208 int (*ndo_get_phys_port_id)(struct net_device *dev,
1209 struct netdev_phys_item_id *ppid);
1210 int (*ndo_get_phys_port_name)(struct net_device *dev,
1211 char *name, size_t len);
1212 void (*ndo_add_vxlan_port)(struct net_device *dev,
1213 sa_family_t sa_family,
1215 void (*ndo_del_vxlan_port)(struct net_device *dev,
1216 sa_family_t sa_family,
1219 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1220 struct net_device *dev);
1221 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1224 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1225 struct net_device *dev,
1227 int (*ndo_get_lock_subclass)(struct net_device *dev);
1228 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1229 struct net_device *dev,
1230 netdev_features_t features);
1231 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1234 int (*ndo_get_iflink)(const struct net_device *dev);
1235 int (*ndo_change_proto_down)(struct net_device *dev,
1237 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1238 struct sk_buff *skb);
1242 * enum net_device_priv_flags - &struct net_device priv_flags
1244 * These are the &struct net_device, they are only set internally
1245 * by drivers and used in the kernel. These flags are invisible to
1246 * userspace, this means that the order of these flags can change
1247 * during any kernel release.
1249 * You should have a pretty good reason to be extending these flags.
1251 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1252 * @IFF_EBRIDGE: Ethernet bridging device
1253 * @IFF_BONDING: bonding master or slave
1254 * @IFF_ISATAP: ISATAP interface (RFC4214)
1255 * @IFF_WAN_HDLC: WAN HDLC device
1256 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1258 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1259 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1260 * @IFF_MACVLAN_PORT: device used as macvlan port
1261 * @IFF_BRIDGE_PORT: device used as bridge port
1262 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1263 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1264 * @IFF_UNICAST_FLT: Supports unicast filtering
1265 * @IFF_TEAM_PORT: device used as team port
1266 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1267 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1268 * change when it's running
1269 * @IFF_MACVLAN: Macvlan device
1270 * @IFF_L3MDEV_MASTER: device is an L3 master device
1271 * @IFF_NO_QUEUE: device can run without qdisc attached
1272 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1273 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1275 enum netdev_priv_flags {
1276 IFF_802_1Q_VLAN = 1<<0,
1280 IFF_WAN_HDLC = 1<<4,
1281 IFF_XMIT_DST_RELEASE = 1<<5,
1282 IFF_DONT_BRIDGE = 1<<6,
1283 IFF_DISABLE_NETPOLL = 1<<7,
1284 IFF_MACVLAN_PORT = 1<<8,
1285 IFF_BRIDGE_PORT = 1<<9,
1286 IFF_OVS_DATAPATH = 1<<10,
1287 IFF_TX_SKB_SHARING = 1<<11,
1288 IFF_UNICAST_FLT = 1<<12,
1289 IFF_TEAM_PORT = 1<<13,
1290 IFF_SUPP_NOFCS = 1<<14,
1291 IFF_LIVE_ADDR_CHANGE = 1<<15,
1292 IFF_MACVLAN = 1<<16,
1293 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1294 IFF_IPVLAN_MASTER = 1<<18,
1295 IFF_IPVLAN_SLAVE = 1<<19,
1296 IFF_L3MDEV_MASTER = 1<<20,
1297 IFF_NO_QUEUE = 1<<21,
1298 IFF_OPENVSWITCH = 1<<22,
1299 IFF_L3MDEV_SLAVE = 1<<23,
1302 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1303 #define IFF_EBRIDGE IFF_EBRIDGE
1304 #define IFF_BONDING IFF_BONDING
1305 #define IFF_ISATAP IFF_ISATAP
1306 #define IFF_WAN_HDLC IFF_WAN_HDLC
1307 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1308 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1309 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1310 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1311 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1312 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1313 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1314 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1315 #define IFF_TEAM_PORT IFF_TEAM_PORT
1316 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1317 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1318 #define IFF_MACVLAN IFF_MACVLAN
1319 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1320 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1321 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1322 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1323 #define IFF_NO_QUEUE IFF_NO_QUEUE
1324 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1325 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1328 * struct net_device - The DEVICE structure.
1329 * Actually, this whole structure is a big mistake. It mixes I/O
1330 * data with strictly "high-level" data, and it has to know about
1331 * almost every data structure used in the INET module.
1333 * @name: This is the first field of the "visible" part of this structure
1334 * (i.e. as seen by users in the "Space.c" file). It is the name
1337 * @name_hlist: Device name hash chain, please keep it close to name[]
1338 * @ifalias: SNMP alias
1339 * @mem_end: Shared memory end
1340 * @mem_start: Shared memory start
1341 * @base_addr: Device I/O address
1342 * @irq: Device IRQ number
1344 * @carrier_changes: Stats to monitor carrier on<->off transitions
1346 * @state: Generic network queuing layer state, see netdev_state_t
1347 * @dev_list: The global list of network devices
1348 * @napi_list: List entry, that is used for polling napi devices
1349 * @unreg_list: List entry, that is used, when we are unregistering the
1350 * device, see the function unregister_netdev
1351 * @close_list: List entry, that is used, when we are closing the device
1353 * @adj_list: Directly linked devices, like slaves for bonding
1354 * @all_adj_list: All linked devices, *including* neighbours
1355 * @features: Currently active device features
1356 * @hw_features: User-changeable features
1358 * @wanted_features: User-requested features
1359 * @vlan_features: Mask of features inheritable by VLAN devices
1361 * @hw_enc_features: Mask of features inherited by encapsulating devices
1362 * This field indicates what encapsulation
1363 * offloads the hardware is capable of doing,
1364 * and drivers will need to set them appropriately.
1366 * @mpls_features: Mask of features inheritable by MPLS
1368 * @ifindex: interface index
1369 * @group: The group, that the device belongs to
1371 * @stats: Statistics struct, which was left as a legacy, use
1372 * rtnl_link_stats64 instead
1374 * @rx_dropped: Dropped packets by core network,
1375 * do not use this in drivers
1376 * @tx_dropped: Dropped packets by core network,
1377 * do not use this in drivers
1379 * @wireless_handlers: List of functions to handle Wireless Extensions,
1381 * see <net/iw_handler.h> for details.
1382 * @wireless_data: Instance data managed by the core of wireless extensions
1384 * @netdev_ops: Includes several pointers to callbacks,
1385 * if one wants to override the ndo_*() functions
1386 * @ethtool_ops: Management operations
1387 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1388 * of Layer 2 headers.
1390 * @flags: Interface flags (a la BSD)
1391 * @priv_flags: Like 'flags' but invisible to userspace,
1392 * see if.h for the definitions
1393 * @gflags: Global flags ( kept as legacy )
1394 * @padded: How much padding added by alloc_netdev()
1395 * @operstate: RFC2863 operstate
1396 * @link_mode: Mapping policy to operstate
1397 * @if_port: Selectable AUI, TP, ...
1399 * @mtu: Interface MTU value
1400 * @type: Interface hardware type
1401 * @hard_header_len: Hardware header length, which means that this is the
1402 * minimum size of a packet.
1404 * @needed_headroom: Extra headroom the hardware may need, but not in all
1405 * cases can this be guaranteed
1406 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1407 * cases can this be guaranteed. Some cases also use
1408 * LL_MAX_HEADER instead to allocate the skb
1410 * interface address info:
1412 * @perm_addr: Permanent hw address
1413 * @addr_assign_type: Hw address assignment type
1414 * @addr_len: Hardware address length
1415 * @neigh_priv_len; Used in neigh_alloc(),
1416 * initialized only in atm/clip.c
1417 * @dev_id: Used to differentiate devices that share
1418 * the same link layer address
1419 * @dev_port: Used to differentiate devices that share
1421 * @addr_list_lock: XXX: need comments on this one
1422 * @uc_promisc: Counter, that indicates, that promiscuous mode
1423 * has been enabled due to the need to listen to
1424 * additional unicast addresses in a device that
1425 * does not implement ndo_set_rx_mode()
1426 * @uc: unicast mac addresses
1427 * @mc: multicast mac addresses
1428 * @dev_addrs: list of device hw addresses
1429 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1430 * @promiscuity: Number of times, the NIC is told to work in
1431 * Promiscuous mode, if it becomes 0 the NIC will
1432 * exit from working in Promiscuous mode
1433 * @allmulti: Counter, enables or disables allmulticast mode
1435 * @vlan_info: VLAN info
1436 * @dsa_ptr: dsa specific data
1437 * @tipc_ptr: TIPC specific data
1438 * @atalk_ptr: AppleTalk link
1439 * @ip_ptr: IPv4 specific data
1440 * @dn_ptr: DECnet specific data
1441 * @ip6_ptr: IPv6 specific data
1442 * @ax25_ptr: AX.25 specific data
1443 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1445 * @last_rx: Time of last Rx
1446 * @dev_addr: Hw address (before bcast,
1447 * because most packets are unicast)
1449 * @_rx: Array of RX queues
1450 * @num_rx_queues: Number of RX queues
1451 * allocated at register_netdev() time
1452 * @real_num_rx_queues: Number of RX queues currently active in device
1454 * @rx_handler: handler for received packets
1455 * @rx_handler_data: XXX: need comments on this one
1456 * @ingress_queue: XXX: need comments on this one
1457 * @broadcast: hw bcast address
1459 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1460 * indexed by RX queue number. Assigned by driver.
1461 * This must only be set if the ndo_rx_flow_steer
1462 * operation is defined
1463 * @index_hlist: Device index hash chain
1465 * @_tx: Array of TX queues
1466 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1467 * @real_num_tx_queues: Number of TX queues currently active in device
1468 * @qdisc: Root qdisc from userspace point of view
1469 * @tx_queue_len: Max frames per queue allowed
1470 * @tx_global_lock: XXX: need comments on this one
1472 * @xps_maps: XXX: need comments on this one
1474 * @offload_fwd_mark: Offload device fwding mark
1476 * @trans_start: Time (in jiffies) of last Tx
1477 * @watchdog_timeo: Represents the timeout that is used by
1478 * the watchdog ( see dev_watchdog() )
1479 * @watchdog_timer: List of timers
1481 * @pcpu_refcnt: Number of references to this device
1482 * @todo_list: Delayed register/unregister
1483 * @link_watch_list: XXX: need comments on this one
1485 * @reg_state: Register/unregister state machine
1486 * @dismantle: Device is going to be freed
1487 * @rtnl_link_state: This enum represents the phases of creating
1490 * @destructor: Called from unregister,
1491 * can be used to call free_netdev
1492 * @npinfo: XXX: need comments on this one
1493 * @nd_net: Network namespace this network device is inside
1495 * @ml_priv: Mid-layer private
1496 * @lstats: Loopback statistics
1497 * @tstats: Tunnel statistics
1498 * @dstats: Dummy statistics
1499 * @vstats: Virtual ethernet statistics
1504 * @dev: Class/net/name entry
1505 * @sysfs_groups: Space for optional device, statistics and wireless
1508 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1509 * @rtnl_link_ops: Rtnl_link_ops
1511 * @gso_max_size: Maximum size of generic segmentation offload
1512 * @gso_max_segs: Maximum number of segments that can be passed to the
1514 * @gso_min_segs: Minimum number of segments that can be passed to the
1517 * @dcbnl_ops: Data Center Bridging netlink ops
1518 * @num_tc: Number of traffic classes in the net device
1519 * @tc_to_txq: XXX: need comments on this one
1520 * @prio_tc_map XXX: need comments on this one
1522 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1524 * @priomap: XXX: need comments on this one
1525 * @phydev: Physical device may attach itself
1526 * for hardware timestamping
1528 * @qdisc_tx_busylock: XXX: need comments on this one
1530 * @proto_down: protocol port state information can be sent to the
1531 * switch driver and used to set the phys state of the
1534 * FIXME: cleanup struct net_device such that network protocol info
1539 char name[IFNAMSIZ];
1540 struct hlist_node name_hlist;
1543 * I/O specific fields
1544 * FIXME: Merge these and struct ifmap into one
1546 unsigned long mem_end;
1547 unsigned long mem_start;
1548 unsigned long base_addr;
1551 atomic_t carrier_changes;
1554 * Some hardware also needs these fields (state,dev_list,
1555 * napi_list,unreg_list,close_list) but they are not
1556 * part of the usual set specified in Space.c.
1559 unsigned long state;
1561 struct list_head dev_list;
1562 struct list_head napi_list;
1563 struct list_head unreg_list;
1564 struct list_head close_list;
1565 struct list_head ptype_all;
1566 struct list_head ptype_specific;
1569 struct list_head upper;
1570 struct list_head lower;
1574 struct list_head upper;
1575 struct list_head lower;
1578 netdev_features_t features;
1579 netdev_features_t hw_features;
1580 netdev_features_t wanted_features;
1581 netdev_features_t vlan_features;
1582 netdev_features_t hw_enc_features;
1583 netdev_features_t mpls_features;
1588 struct net_device_stats stats;
1590 atomic_long_t rx_dropped;
1591 atomic_long_t tx_dropped;
1593 #ifdef CONFIG_WIRELESS_EXT
1594 const struct iw_handler_def * wireless_handlers;
1595 struct iw_public_data * wireless_data;
1597 const struct net_device_ops *netdev_ops;
1598 const struct ethtool_ops *ethtool_ops;
1599 #ifdef CONFIG_NET_SWITCHDEV
1600 const struct switchdev_ops *switchdev_ops;
1602 #ifdef CONFIG_NET_L3_MASTER_DEV
1603 const struct l3mdev_ops *l3mdev_ops;
1606 const struct header_ops *header_ops;
1609 unsigned int priv_flags;
1611 unsigned short gflags;
1612 unsigned short padded;
1614 unsigned char operstate;
1615 unsigned char link_mode;
1617 unsigned char if_port;
1621 unsigned short type;
1622 unsigned short hard_header_len;
1624 unsigned short needed_headroom;
1625 unsigned short needed_tailroom;
1627 /* Interface address info. */
1628 unsigned char perm_addr[MAX_ADDR_LEN];
1629 unsigned char addr_assign_type;
1630 unsigned char addr_len;
1631 unsigned short neigh_priv_len;
1632 unsigned short dev_id;
1633 unsigned short dev_port;
1634 spinlock_t addr_list_lock;
1635 unsigned char name_assign_type;
1637 struct netdev_hw_addr_list uc;
1638 struct netdev_hw_addr_list mc;
1639 struct netdev_hw_addr_list dev_addrs;
1642 struct kset *queues_kset;
1644 unsigned int promiscuity;
1645 unsigned int allmulti;
1648 /* Protocol specific pointers */
1650 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1651 struct vlan_info __rcu *vlan_info;
1653 #if IS_ENABLED(CONFIG_NET_DSA)
1654 struct dsa_switch_tree *dsa_ptr;
1656 #if IS_ENABLED(CONFIG_TIPC)
1657 struct tipc_bearer __rcu *tipc_ptr;
1660 struct in_device __rcu *ip_ptr;
1661 struct dn_dev __rcu *dn_ptr;
1662 struct inet6_dev __rcu *ip6_ptr;
1664 struct wireless_dev *ieee80211_ptr;
1665 struct wpan_dev *ieee802154_ptr;
1666 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1667 struct mpls_dev __rcu *mpls_ptr;
1671 * Cache lines mostly used on receive path (including eth_type_trans())
1673 unsigned long last_rx;
1675 /* Interface address info used in eth_type_trans() */
1676 unsigned char *dev_addr;
1680 struct netdev_rx_queue *_rx;
1682 unsigned int num_rx_queues;
1683 unsigned int real_num_rx_queues;
1687 unsigned long gro_flush_timeout;
1688 rx_handler_func_t __rcu *rx_handler;
1689 void __rcu *rx_handler_data;
1691 #ifdef CONFIG_NET_CLS_ACT
1692 struct tcf_proto __rcu *ingress_cl_list;
1694 struct netdev_queue __rcu *ingress_queue;
1695 #ifdef CONFIG_NETFILTER_INGRESS
1696 struct list_head nf_hooks_ingress;
1699 unsigned char broadcast[MAX_ADDR_LEN];
1700 #ifdef CONFIG_RFS_ACCEL
1701 struct cpu_rmap *rx_cpu_rmap;
1703 struct hlist_node index_hlist;
1706 * Cache lines mostly used on transmit path
1708 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1709 unsigned int num_tx_queues;
1710 unsigned int real_num_tx_queues;
1711 struct Qdisc *qdisc;
1712 unsigned long tx_queue_len;
1713 spinlock_t tx_global_lock;
1717 struct xps_dev_maps __rcu *xps_maps;
1720 #ifdef CONFIG_NET_SWITCHDEV
1721 u32 offload_fwd_mark;
1724 /* These may be needed for future network-power-down code. */
1727 * trans_start here is expensive for high speed devices on SMP,
1728 * please use netdev_queue->trans_start instead.
1730 unsigned long trans_start;
1732 struct timer_list watchdog_timer;
1734 int __percpu *pcpu_refcnt;
1735 struct list_head todo_list;
1737 struct list_head link_watch_list;
1739 enum { NETREG_UNINITIALIZED=0,
1740 NETREG_REGISTERED, /* completed register_netdevice */
1741 NETREG_UNREGISTERING, /* called unregister_netdevice */
1742 NETREG_UNREGISTERED, /* completed unregister todo */
1743 NETREG_RELEASED, /* called free_netdev */
1744 NETREG_DUMMY, /* dummy device for NAPI poll */
1750 RTNL_LINK_INITIALIZED,
1751 RTNL_LINK_INITIALIZING,
1752 } rtnl_link_state:16;
1754 void (*destructor)(struct net_device *dev);
1756 #ifdef CONFIG_NETPOLL
1757 struct netpoll_info __rcu *npinfo;
1760 possible_net_t nd_net;
1762 /* mid-layer private */
1765 struct pcpu_lstats __percpu *lstats;
1766 struct pcpu_sw_netstats __percpu *tstats;
1767 struct pcpu_dstats __percpu *dstats;
1768 struct pcpu_vstats __percpu *vstats;
1771 struct garp_port __rcu *garp_port;
1772 struct mrp_port __rcu *mrp_port;
1775 const struct attribute_group *sysfs_groups[4];
1776 const struct attribute_group *sysfs_rx_queue_group;
1778 const struct rtnl_link_ops *rtnl_link_ops;
1780 /* for setting kernel sock attribute on TCP connection setup */
1781 #define GSO_MAX_SIZE 65536
1782 unsigned int gso_max_size;
1783 #define GSO_MAX_SEGS 65535
1787 const struct dcbnl_rtnl_ops *dcbnl_ops;
1790 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1791 u8 prio_tc_map[TC_BITMASK + 1];
1793 #if IS_ENABLED(CONFIG_FCOE)
1794 unsigned int fcoe_ddp_xid;
1796 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1797 struct netprio_map __rcu *priomap;
1799 struct phy_device *phydev;
1800 struct lock_class_key *qdisc_tx_busylock;
1803 #define to_net_dev(d) container_of(d, struct net_device, dev)
1805 #define NETDEV_ALIGN 32
1808 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1810 return dev->prio_tc_map[prio & TC_BITMASK];
1814 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1816 if (tc >= dev->num_tc)
1819 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1824 void netdev_reset_tc(struct net_device *dev)
1827 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1828 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1832 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1834 if (tc >= dev->num_tc)
1837 dev->tc_to_txq[tc].count = count;
1838 dev->tc_to_txq[tc].offset = offset;
1843 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1845 if (num_tc > TC_MAX_QUEUE)
1848 dev->num_tc = num_tc;
1853 int netdev_get_num_tc(struct net_device *dev)
1859 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1862 return &dev->_tx[index];
1865 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1866 const struct sk_buff *skb)
1868 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1871 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1872 void (*f)(struct net_device *,
1873 struct netdev_queue *,
1879 for (i = 0; i < dev->num_tx_queues; i++)
1880 f(dev, &dev->_tx[i], arg);
1883 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1884 struct sk_buff *skb,
1888 * Net namespace inlines
1891 struct net *dev_net(const struct net_device *dev)
1893 return read_pnet(&dev->nd_net);
1897 void dev_net_set(struct net_device *dev, struct net *net)
1899 write_pnet(&dev->nd_net, net);
1902 static inline bool netdev_uses_dsa(struct net_device *dev)
1904 #if IS_ENABLED(CONFIG_NET_DSA)
1905 if (dev->dsa_ptr != NULL)
1906 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1912 * netdev_priv - access network device private data
1913 * @dev: network device
1915 * Get network device private data
1917 static inline void *netdev_priv(const struct net_device *dev)
1919 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1922 /* Set the sysfs physical device reference for the network logical device
1923 * if set prior to registration will cause a symlink during initialization.
1925 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1927 /* Set the sysfs device type for the network logical device to allow
1928 * fine-grained identification of different network device types. For
1929 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1931 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1933 /* Default NAPI poll() weight
1934 * Device drivers are strongly advised to not use bigger value
1936 #define NAPI_POLL_WEIGHT 64
1939 * netif_napi_add - initialize a napi context
1940 * @dev: network device
1941 * @napi: napi context
1942 * @poll: polling function
1943 * @weight: default weight
1945 * netif_napi_add() must be used to initialize a napi context prior to calling
1946 * *any* of the other napi related functions.
1948 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1949 int (*poll)(struct napi_struct *, int), int weight);
1952 * netif_napi_del - remove a napi context
1953 * @napi: napi context
1955 * netif_napi_del() removes a napi context from the network device napi list
1957 void netif_napi_del(struct napi_struct *napi);
1959 struct napi_gro_cb {
1960 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1963 /* Length of frag0. */
1964 unsigned int frag0_len;
1966 /* This indicates where we are processing relative to skb->data. */
1969 /* This is non-zero if the packet cannot be merged with the new skb. */
1972 /* Save the IP ID here and check when we get to the transport layer */
1975 /* Number of segments aggregated. */
1978 /* Start offset for remote checksum offload */
1979 u16 gro_remcsum_start;
1981 /* jiffies when first packet was created/queued */
1984 /* Used in ipv6_gro_receive() and foo-over-udp */
1987 /* This is non-zero if the packet may be of the same flow. */
1990 /* Used in udp_gro_receive */
1993 /* GRO checksum is valid */
1996 /* Number of checksums via CHECKSUM_UNNECESSARY */
2001 #define NAPI_GRO_FREE 1
2002 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2004 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2009 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2012 /* used in skb_gro_receive() slow path */
2013 struct sk_buff *last;
2016 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2018 struct packet_type {
2019 __be16 type; /* This is really htons(ether_type). */
2020 struct net_device *dev; /* NULL is wildcarded here */
2021 int (*func) (struct sk_buff *,
2022 struct net_device *,
2023 struct packet_type *,
2024 struct net_device *);
2025 bool (*id_match)(struct packet_type *ptype,
2027 void *af_packet_priv;
2028 struct list_head list;
2031 struct offload_callbacks {
2032 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2033 netdev_features_t features);
2034 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2035 struct sk_buff *skb);
2036 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2039 struct packet_offload {
2040 __be16 type; /* This is really htons(ether_type). */
2042 struct offload_callbacks callbacks;
2043 struct list_head list;
2048 struct udp_offload_callbacks {
2049 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2050 struct sk_buff *skb,
2051 struct udp_offload *uoff);
2052 int (*gro_complete)(struct sk_buff *skb,
2054 struct udp_offload *uoff);
2057 struct udp_offload {
2060 struct udp_offload_callbacks callbacks;
2063 /* often modified stats are per cpu, other are shared (netdev->stats) */
2064 struct pcpu_sw_netstats {
2069 struct u64_stats_sync syncp;
2072 #define __netdev_alloc_pcpu_stats(type, gfp) \
2074 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2077 for_each_possible_cpu(__cpu) { \
2078 typeof(type) *stat; \
2079 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2080 u64_stats_init(&stat->syncp); \
2086 #define netdev_alloc_pcpu_stats(type) \
2087 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2089 #include <linux/notifier.h>
2091 /* netdevice notifier chain. Please remember to update the rtnetlink
2092 * notification exclusion list in rtnetlink_event() when adding new
2095 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2096 #define NETDEV_DOWN 0x0002
2097 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2098 detected a hardware crash and restarted
2099 - we can use this eg to kick tcp sessions
2101 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2102 #define NETDEV_REGISTER 0x0005
2103 #define NETDEV_UNREGISTER 0x0006
2104 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2105 #define NETDEV_CHANGEADDR 0x0008
2106 #define NETDEV_GOING_DOWN 0x0009
2107 #define NETDEV_CHANGENAME 0x000A
2108 #define NETDEV_FEAT_CHANGE 0x000B
2109 #define NETDEV_BONDING_FAILOVER 0x000C
2110 #define NETDEV_PRE_UP 0x000D
2111 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2112 #define NETDEV_POST_TYPE_CHANGE 0x000F
2113 #define NETDEV_POST_INIT 0x0010
2114 #define NETDEV_UNREGISTER_FINAL 0x0011
2115 #define NETDEV_RELEASE 0x0012
2116 #define NETDEV_NOTIFY_PEERS 0x0013
2117 #define NETDEV_JOIN 0x0014
2118 #define NETDEV_CHANGEUPPER 0x0015
2119 #define NETDEV_RESEND_IGMP 0x0016
2120 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2121 #define NETDEV_CHANGEINFODATA 0x0018
2122 #define NETDEV_BONDING_INFO 0x0019
2123 #define NETDEV_PRECHANGEUPPER 0x001A
2125 int register_netdevice_notifier(struct notifier_block *nb);
2126 int unregister_netdevice_notifier(struct notifier_block *nb);
2128 struct netdev_notifier_info {
2129 struct net_device *dev;
2132 struct netdev_notifier_change_info {
2133 struct netdev_notifier_info info; /* must be first */
2134 unsigned int flags_changed;
2137 struct netdev_notifier_changeupper_info {
2138 struct netdev_notifier_info info; /* must be first */
2139 struct net_device *upper_dev; /* new upper dev */
2140 bool master; /* is upper dev master */
2141 bool linking; /* is the nofication for link or unlink */
2144 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2145 struct net_device *dev)
2150 static inline struct net_device *
2151 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2156 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2159 extern rwlock_t dev_base_lock; /* Device list lock */
2161 #define for_each_netdev(net, d) \
2162 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2163 #define for_each_netdev_reverse(net, d) \
2164 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2165 #define for_each_netdev_rcu(net, d) \
2166 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2167 #define for_each_netdev_safe(net, d, n) \
2168 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2169 #define for_each_netdev_continue(net, d) \
2170 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2171 #define for_each_netdev_continue_rcu(net, d) \
2172 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2173 #define for_each_netdev_in_bond_rcu(bond, slave) \
2174 for_each_netdev_rcu(&init_net, slave) \
2175 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2176 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2178 static inline struct net_device *next_net_device(struct net_device *dev)
2180 struct list_head *lh;
2184 lh = dev->dev_list.next;
2185 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2188 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2190 struct list_head *lh;
2194 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2195 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2198 static inline struct net_device *first_net_device(struct net *net)
2200 return list_empty(&net->dev_base_head) ? NULL :
2201 net_device_entry(net->dev_base_head.next);
2204 static inline struct net_device *first_net_device_rcu(struct net *net)
2206 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2208 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2211 int netdev_boot_setup_check(struct net_device *dev);
2212 unsigned long netdev_boot_base(const char *prefix, int unit);
2213 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2214 const char *hwaddr);
2215 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2216 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2217 void dev_add_pack(struct packet_type *pt);
2218 void dev_remove_pack(struct packet_type *pt);
2219 void __dev_remove_pack(struct packet_type *pt);
2220 void dev_add_offload(struct packet_offload *po);
2221 void dev_remove_offload(struct packet_offload *po);
2223 int dev_get_iflink(const struct net_device *dev);
2224 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2225 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2226 unsigned short mask);
2227 struct net_device *dev_get_by_name(struct net *net, const char *name);
2228 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2229 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2230 int dev_alloc_name(struct net_device *dev, const char *name);
2231 int dev_open(struct net_device *dev);
2232 int dev_close(struct net_device *dev);
2233 int dev_close_many(struct list_head *head, bool unlink);
2234 void dev_disable_lro(struct net_device *dev);
2235 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2236 int dev_queue_xmit(struct sk_buff *skb);
2237 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2238 int register_netdevice(struct net_device *dev);
2239 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2240 void unregister_netdevice_many(struct list_head *head);
2241 static inline void unregister_netdevice(struct net_device *dev)
2243 unregister_netdevice_queue(dev, NULL);
2246 int netdev_refcnt_read(const struct net_device *dev);
2247 void free_netdev(struct net_device *dev);
2248 void netdev_freemem(struct net_device *dev);
2249 void synchronize_net(void);
2250 int init_dummy_netdev(struct net_device *dev);
2252 DECLARE_PER_CPU(int, xmit_recursion);
2253 static inline int dev_recursion_level(void)
2255 return this_cpu_read(xmit_recursion);
2258 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2259 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2260 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2261 int netdev_get_name(struct net *net, char *name, int ifindex);
2262 int dev_restart(struct net_device *dev);
2263 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2265 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2267 return NAPI_GRO_CB(skb)->data_offset;
2270 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2272 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2275 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2277 NAPI_GRO_CB(skb)->data_offset += len;
2280 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2281 unsigned int offset)
2283 return NAPI_GRO_CB(skb)->frag0 + offset;
2286 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2288 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2291 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2292 unsigned int offset)
2294 if (!pskb_may_pull(skb, hlen))
2297 NAPI_GRO_CB(skb)->frag0 = NULL;
2298 NAPI_GRO_CB(skb)->frag0_len = 0;
2299 return skb->data + offset;
2302 static inline void *skb_gro_network_header(struct sk_buff *skb)
2304 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2305 skb_network_offset(skb);
2308 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2309 const void *start, unsigned int len)
2311 if (NAPI_GRO_CB(skb)->csum_valid)
2312 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2313 csum_partial(start, len, 0));
2316 /* GRO checksum functions. These are logical equivalents of the normal
2317 * checksum functions (in skbuff.h) except that they operate on the GRO
2318 * offsets and fields in sk_buff.
2321 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2323 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2325 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2328 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2332 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2333 skb_checksum_start_offset(skb) <
2334 skb_gro_offset(skb)) &&
2335 !skb_at_gro_remcsum_start(skb) &&
2336 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2337 (!zero_okay || check));
2340 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2343 if (NAPI_GRO_CB(skb)->csum_valid &&
2344 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2347 NAPI_GRO_CB(skb)->csum = psum;
2349 return __skb_gro_checksum_complete(skb);
2352 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2354 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2355 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2356 NAPI_GRO_CB(skb)->csum_cnt--;
2358 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2359 * verified a new top level checksum or an encapsulated one
2360 * during GRO. This saves work if we fallback to normal path.
2362 __skb_incr_checksum_unnecessary(skb);
2366 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2369 __sum16 __ret = 0; \
2370 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2371 __ret = __skb_gro_checksum_validate_complete(skb, \
2372 compute_pseudo(skb, proto)); \
2374 __skb_mark_checksum_bad(skb); \
2376 skb_gro_incr_csum_unnecessary(skb); \
2380 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2381 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2383 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2385 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2387 #define skb_gro_checksum_simple_validate(skb) \
2388 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2390 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2392 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2393 !NAPI_GRO_CB(skb)->csum_valid);
2396 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2397 __sum16 check, __wsum pseudo)
2399 NAPI_GRO_CB(skb)->csum = ~pseudo;
2400 NAPI_GRO_CB(skb)->csum_valid = 1;
2403 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2405 if (__skb_gro_checksum_convert_check(skb)) \
2406 __skb_gro_checksum_convert(skb, check, \
2407 compute_pseudo(skb, proto)); \
2410 struct gro_remcsum {
2415 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2421 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2422 unsigned int off, size_t hdrlen,
2423 int start, int offset,
2424 struct gro_remcsum *grc,
2428 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2430 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2433 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2437 ptr = skb_gro_header_fast(skb, off);
2438 if (skb_gro_header_hard(skb, off + plen)) {
2439 ptr = skb_gro_header_slow(skb, off + plen, off);
2444 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2447 /* Adjust skb->csum since we changed the packet */
2448 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2450 grc->offset = off + hdrlen + offset;
2456 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2457 struct gro_remcsum *grc)
2460 size_t plen = grc->offset + sizeof(u16);
2465 ptr = skb_gro_header_fast(skb, grc->offset);
2466 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2467 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2472 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2475 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2476 unsigned short type,
2477 const void *daddr, const void *saddr,
2480 if (!dev->header_ops || !dev->header_ops->create)
2483 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2486 static inline int dev_parse_header(const struct sk_buff *skb,
2487 unsigned char *haddr)
2489 const struct net_device *dev = skb->dev;
2491 if (!dev->header_ops || !dev->header_ops->parse)
2493 return dev->header_ops->parse(skb, haddr);
2496 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2497 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2498 static inline int unregister_gifconf(unsigned int family)
2500 return register_gifconf(family, NULL);
2503 #ifdef CONFIG_NET_FLOW_LIMIT
2504 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2505 struct sd_flow_limit {
2507 unsigned int num_buckets;
2508 unsigned int history_head;
2509 u16 history[FLOW_LIMIT_HISTORY];
2513 extern int netdev_flow_limit_table_len;
2514 #endif /* CONFIG_NET_FLOW_LIMIT */
2517 * Incoming packets are placed on per-cpu queues
2519 struct softnet_data {
2520 struct list_head poll_list;
2521 struct sk_buff_head process_queue;
2524 unsigned int processed;
2525 unsigned int time_squeeze;
2526 unsigned int cpu_collision;
2527 unsigned int received_rps;
2529 struct softnet_data *rps_ipi_list;
2531 #ifdef CONFIG_NET_FLOW_LIMIT
2532 struct sd_flow_limit __rcu *flow_limit;
2534 struct Qdisc *output_queue;
2535 struct Qdisc **output_queue_tailp;
2536 struct sk_buff *completion_queue;
2539 /* Elements below can be accessed between CPUs for RPS */
2540 struct call_single_data csd ____cacheline_aligned_in_smp;
2541 struct softnet_data *rps_ipi_next;
2543 unsigned int input_queue_head;
2544 unsigned int input_queue_tail;
2546 unsigned int dropped;
2547 struct sk_buff_head input_pkt_queue;
2548 struct napi_struct backlog;
2552 static inline void input_queue_head_incr(struct softnet_data *sd)
2555 sd->input_queue_head++;
2559 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2560 unsigned int *qtail)
2563 *qtail = ++sd->input_queue_tail;
2567 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2569 void __netif_schedule(struct Qdisc *q);
2570 void netif_schedule_queue(struct netdev_queue *txq);
2572 static inline void netif_tx_schedule_all(struct net_device *dev)
2576 for (i = 0; i < dev->num_tx_queues; i++)
2577 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2580 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2582 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2586 * netif_start_queue - allow transmit
2587 * @dev: network device
2589 * Allow upper layers to call the device hard_start_xmit routine.
2591 static inline void netif_start_queue(struct net_device *dev)
2593 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2596 static inline void netif_tx_start_all_queues(struct net_device *dev)
2600 for (i = 0; i < dev->num_tx_queues; i++) {
2601 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2602 netif_tx_start_queue(txq);
2606 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2609 * netif_wake_queue - restart transmit
2610 * @dev: network device
2612 * Allow upper layers to call the device hard_start_xmit routine.
2613 * Used for flow control when transmit resources are available.
2615 static inline void netif_wake_queue(struct net_device *dev)
2617 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2620 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2624 for (i = 0; i < dev->num_tx_queues; i++) {
2625 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2626 netif_tx_wake_queue(txq);
2630 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2632 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2636 * netif_stop_queue - stop transmitted packets
2637 * @dev: network device
2639 * Stop upper layers calling the device hard_start_xmit routine.
2640 * Used for flow control when transmit resources are unavailable.
2642 static inline void netif_stop_queue(struct net_device *dev)
2644 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2647 void netif_tx_stop_all_queues(struct net_device *dev);
2649 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2651 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2655 * netif_queue_stopped - test if transmit queue is flowblocked
2656 * @dev: network device
2658 * Test if transmit queue on device is currently unable to send.
2660 static inline bool netif_queue_stopped(const struct net_device *dev)
2662 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2665 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2667 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2671 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2673 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2677 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2679 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2683 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2684 * @dev_queue: pointer to transmit queue
2686 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2687 * to give appropriate hint to the cpu.
2689 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2692 prefetchw(&dev_queue->dql.num_queued);
2697 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2698 * @dev_queue: pointer to transmit queue
2700 * BQL enabled drivers might use this helper in their TX completion path,
2701 * to give appropriate hint to the cpu.
2703 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2706 prefetchw(&dev_queue->dql.limit);
2710 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2714 dql_queued(&dev_queue->dql, bytes);
2716 if (likely(dql_avail(&dev_queue->dql) >= 0))
2719 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2722 * The XOFF flag must be set before checking the dql_avail below,
2723 * because in netdev_tx_completed_queue we update the dql_completed
2724 * before checking the XOFF flag.
2728 /* check again in case another CPU has just made room avail */
2729 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2730 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2735 * netdev_sent_queue - report the number of bytes queued to hardware
2736 * @dev: network device
2737 * @bytes: number of bytes queued to the hardware device queue
2739 * Report the number of bytes queued for sending/completion to the network
2740 * device hardware queue. @bytes should be a good approximation and should
2741 * exactly match netdev_completed_queue() @bytes
2743 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2745 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2748 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2749 unsigned int pkts, unsigned int bytes)
2752 if (unlikely(!bytes))
2755 dql_completed(&dev_queue->dql, bytes);
2758 * Without the memory barrier there is a small possiblity that
2759 * netdev_tx_sent_queue will miss the update and cause the queue to
2760 * be stopped forever
2764 if (dql_avail(&dev_queue->dql) < 0)
2767 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2768 netif_schedule_queue(dev_queue);
2773 * netdev_completed_queue - report bytes and packets completed by device
2774 * @dev: network device
2775 * @pkts: actual number of packets sent over the medium
2776 * @bytes: actual number of bytes sent over the medium
2778 * Report the number of bytes and packets transmitted by the network device
2779 * hardware queue over the physical medium, @bytes must exactly match the
2780 * @bytes amount passed to netdev_sent_queue()
2782 static inline void netdev_completed_queue(struct net_device *dev,
2783 unsigned int pkts, unsigned int bytes)
2785 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2788 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2791 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2797 * netdev_reset_queue - reset the packets and bytes count of a network device
2798 * @dev_queue: network device
2800 * Reset the bytes and packet count of a network device and clear the
2801 * software flow control OFF bit for this network device
2803 static inline void netdev_reset_queue(struct net_device *dev_queue)
2805 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2809 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2810 * @dev: network device
2811 * @queue_index: given tx queue index
2813 * Returns 0 if given tx queue index >= number of device tx queues,
2814 * otherwise returns the originally passed tx queue index.
2816 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2818 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2819 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2820 dev->name, queue_index,
2821 dev->real_num_tx_queues);
2829 * netif_running - test if up
2830 * @dev: network device
2832 * Test if the device has been brought up.
2834 static inline bool netif_running(const struct net_device *dev)
2836 return test_bit(__LINK_STATE_START, &dev->state);
2840 * Routines to manage the subqueues on a device. We only need start
2841 * stop, and a check if it's stopped. All other device management is
2842 * done at the overall netdevice level.
2843 * Also test the device if we're multiqueue.
2847 * netif_start_subqueue - allow sending packets on subqueue
2848 * @dev: network device
2849 * @queue_index: sub queue index
2851 * Start individual transmit queue of a device with multiple transmit queues.
2853 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2855 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2857 netif_tx_start_queue(txq);
2861 * netif_stop_subqueue - stop sending packets on subqueue
2862 * @dev: network device
2863 * @queue_index: sub queue index
2865 * Stop individual transmit queue of a device with multiple transmit queues.
2867 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2869 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2870 netif_tx_stop_queue(txq);
2874 * netif_subqueue_stopped - test status of subqueue
2875 * @dev: network device
2876 * @queue_index: sub queue index
2878 * Check individual transmit queue of a device with multiple transmit queues.
2880 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2883 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2885 return netif_tx_queue_stopped(txq);
2888 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2889 struct sk_buff *skb)
2891 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2894 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2897 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2900 static inline int netif_set_xps_queue(struct net_device *dev,
2901 const struct cpumask *mask,
2908 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
2909 unsigned int num_tx_queues);
2912 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2913 * as a distribution range limit for the returned value.
2915 static inline u16 skb_tx_hash(const struct net_device *dev,
2916 struct sk_buff *skb)
2918 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2922 * netif_is_multiqueue - test if device has multiple transmit queues
2923 * @dev: network device
2925 * Check if device has multiple transmit queues
2927 static inline bool netif_is_multiqueue(const struct net_device *dev)
2929 return dev->num_tx_queues > 1;
2932 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2935 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2937 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2945 static inline unsigned int get_netdev_rx_queue_index(
2946 struct netdev_rx_queue *queue)
2948 struct net_device *dev = queue->dev;
2949 int index = queue - dev->_rx;
2951 BUG_ON(index >= dev->num_rx_queues);
2956 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2957 int netif_get_num_default_rss_queues(void);
2959 enum skb_free_reason {
2960 SKB_REASON_CONSUMED,
2964 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2965 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2968 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2969 * interrupt context or with hardware interrupts being disabled.
2970 * (in_irq() || irqs_disabled())
2972 * We provide four helpers that can be used in following contexts :
2974 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2975 * replacing kfree_skb(skb)
2977 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2978 * Typically used in place of consume_skb(skb) in TX completion path
2980 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2981 * replacing kfree_skb(skb)
2983 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2984 * and consumed a packet. Used in place of consume_skb(skb)
2986 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2988 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2991 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2993 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2996 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2998 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3001 static inline void dev_consume_skb_any(struct sk_buff *skb)
3003 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3006 int netif_rx(struct sk_buff *skb);
3007 int netif_rx_ni(struct sk_buff *skb);
3008 int netif_receive_skb(struct sk_buff *skb);
3009 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3010 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3011 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3012 gro_result_t napi_gro_frags(struct napi_struct *napi);
3013 struct packet_offload *gro_find_receive_by_type(__be16 type);
3014 struct packet_offload *gro_find_complete_by_type(__be16 type);
3016 static inline void napi_free_frags(struct napi_struct *napi)
3018 kfree_skb(napi->skb);
3022 int netdev_rx_handler_register(struct net_device *dev,
3023 rx_handler_func_t *rx_handler,
3024 void *rx_handler_data);
3025 void netdev_rx_handler_unregister(struct net_device *dev);
3027 bool dev_valid_name(const char *name);
3028 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3029 int dev_ethtool(struct net *net, struct ifreq *);
3030 unsigned int dev_get_flags(const struct net_device *);
3031 int __dev_change_flags(struct net_device *, unsigned int flags);
3032 int dev_change_flags(struct net_device *, unsigned int);
3033 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3034 unsigned int gchanges);
3035 int dev_change_name(struct net_device *, const char *);
3036 int dev_set_alias(struct net_device *, const char *, size_t);
3037 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3038 int dev_set_mtu(struct net_device *, int);
3039 void dev_set_group(struct net_device *, int);
3040 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3041 int dev_change_carrier(struct net_device *, bool new_carrier);
3042 int dev_get_phys_port_id(struct net_device *dev,
3043 struct netdev_phys_item_id *ppid);
3044 int dev_get_phys_port_name(struct net_device *dev,
3045 char *name, size_t len);
3046 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3047 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3048 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3049 struct netdev_queue *txq, int *ret);
3050 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3051 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3052 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3054 extern int netdev_budget;
3056 /* Called by rtnetlink.c:rtnl_unlock() */
3057 void netdev_run_todo(void);
3060 * dev_put - release reference to device
3061 * @dev: network device
3063 * Release reference to device to allow it to be freed.
3065 static inline void dev_put(struct net_device *dev)
3067 this_cpu_dec(*dev->pcpu_refcnt);
3071 * dev_hold - get reference to device
3072 * @dev: network device
3074 * Hold reference to device to keep it from being freed.
3076 static inline void dev_hold(struct net_device *dev)
3078 this_cpu_inc(*dev->pcpu_refcnt);
3081 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3082 * and _off may be called from IRQ context, but it is caller
3083 * who is responsible for serialization of these calls.
3085 * The name carrier is inappropriate, these functions should really be
3086 * called netif_lowerlayer_*() because they represent the state of any
3087 * kind of lower layer not just hardware media.
3090 void linkwatch_init_dev(struct net_device *dev);
3091 void linkwatch_fire_event(struct net_device *dev);
3092 void linkwatch_forget_dev(struct net_device *dev);
3095 * netif_carrier_ok - test if carrier present
3096 * @dev: network device
3098 * Check if carrier is present on device
3100 static inline bool netif_carrier_ok(const struct net_device *dev)
3102 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3105 unsigned long dev_trans_start(struct net_device *dev);
3107 void __netdev_watchdog_up(struct net_device *dev);
3109 void netif_carrier_on(struct net_device *dev);
3111 void netif_carrier_off(struct net_device *dev);
3114 * netif_dormant_on - mark device as dormant.
3115 * @dev: network device
3117 * Mark device as dormant (as per RFC2863).
3119 * The dormant state indicates that the relevant interface is not
3120 * actually in a condition to pass packets (i.e., it is not 'up') but is
3121 * in a "pending" state, waiting for some external event. For "on-
3122 * demand" interfaces, this new state identifies the situation where the
3123 * interface is waiting for events to place it in the up state.
3126 static inline void netif_dormant_on(struct net_device *dev)
3128 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3129 linkwatch_fire_event(dev);
3133 * netif_dormant_off - set device as not dormant.
3134 * @dev: network device
3136 * Device is not in dormant state.
3138 static inline void netif_dormant_off(struct net_device *dev)
3140 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3141 linkwatch_fire_event(dev);
3145 * netif_dormant - test if carrier present
3146 * @dev: network device
3148 * Check if carrier is present on device
3150 static inline bool netif_dormant(const struct net_device *dev)
3152 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3157 * netif_oper_up - test if device is operational
3158 * @dev: network device
3160 * Check if carrier is operational
3162 static inline bool netif_oper_up(const struct net_device *dev)
3164 return (dev->operstate == IF_OPER_UP ||
3165 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3169 * netif_device_present - is device available or removed
3170 * @dev: network device
3172 * Check if device has not been removed from system.
3174 static inline bool netif_device_present(struct net_device *dev)
3176 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3179 void netif_device_detach(struct net_device *dev);
3181 void netif_device_attach(struct net_device *dev);
3184 * Network interface message level settings
3188 NETIF_MSG_DRV = 0x0001,
3189 NETIF_MSG_PROBE = 0x0002,
3190 NETIF_MSG_LINK = 0x0004,
3191 NETIF_MSG_TIMER = 0x0008,
3192 NETIF_MSG_IFDOWN = 0x0010,
3193 NETIF_MSG_IFUP = 0x0020,
3194 NETIF_MSG_RX_ERR = 0x0040,
3195 NETIF_MSG_TX_ERR = 0x0080,
3196 NETIF_MSG_TX_QUEUED = 0x0100,
3197 NETIF_MSG_INTR = 0x0200,
3198 NETIF_MSG_TX_DONE = 0x0400,
3199 NETIF_MSG_RX_STATUS = 0x0800,
3200 NETIF_MSG_PKTDATA = 0x1000,
3201 NETIF_MSG_HW = 0x2000,
3202 NETIF_MSG_WOL = 0x4000,
3205 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3206 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3207 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3208 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3209 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3210 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3211 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3212 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3213 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3214 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3215 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3216 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3217 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3218 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3219 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3221 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3224 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3225 return default_msg_enable_bits;
3226 if (debug_value == 0) /* no output */
3228 /* set low N bits */
3229 return (1 << debug_value) - 1;
3232 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3234 spin_lock(&txq->_xmit_lock);
3235 txq->xmit_lock_owner = cpu;
3238 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3240 spin_lock_bh(&txq->_xmit_lock);
3241 txq->xmit_lock_owner = smp_processor_id();
3244 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3246 bool ok = spin_trylock(&txq->_xmit_lock);
3248 txq->xmit_lock_owner = smp_processor_id();
3252 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3254 txq->xmit_lock_owner = -1;
3255 spin_unlock(&txq->_xmit_lock);
3258 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3260 txq->xmit_lock_owner = -1;
3261 spin_unlock_bh(&txq->_xmit_lock);
3264 static inline void txq_trans_update(struct netdev_queue *txq)
3266 if (txq->xmit_lock_owner != -1)
3267 txq->trans_start = jiffies;
3271 * netif_tx_lock - grab network device transmit lock
3272 * @dev: network device
3274 * Get network device transmit lock
3276 static inline void netif_tx_lock(struct net_device *dev)
3281 spin_lock(&dev->tx_global_lock);
3282 cpu = smp_processor_id();
3283 for (i = 0; i < dev->num_tx_queues; i++) {
3284 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3286 /* We are the only thread of execution doing a
3287 * freeze, but we have to grab the _xmit_lock in
3288 * order to synchronize with threads which are in
3289 * the ->hard_start_xmit() handler and already
3290 * checked the frozen bit.
3292 __netif_tx_lock(txq, cpu);
3293 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3294 __netif_tx_unlock(txq);
3298 static inline void netif_tx_lock_bh(struct net_device *dev)
3304 static inline void netif_tx_unlock(struct net_device *dev)
3308 for (i = 0; i < dev->num_tx_queues; i++) {
3309 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3311 /* No need to grab the _xmit_lock here. If the
3312 * queue is not stopped for another reason, we
3315 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3316 netif_schedule_queue(txq);
3318 spin_unlock(&dev->tx_global_lock);
3321 static inline void netif_tx_unlock_bh(struct net_device *dev)
3323 netif_tx_unlock(dev);
3327 #define HARD_TX_LOCK(dev, txq, cpu) { \
3328 if ((dev->features & NETIF_F_LLTX) == 0) { \
3329 __netif_tx_lock(txq, cpu); \
3333 #define HARD_TX_TRYLOCK(dev, txq) \
3334 (((dev->features & NETIF_F_LLTX) == 0) ? \
3335 __netif_tx_trylock(txq) : \
3338 #define HARD_TX_UNLOCK(dev, txq) { \
3339 if ((dev->features & NETIF_F_LLTX) == 0) { \
3340 __netif_tx_unlock(txq); \
3344 static inline void netif_tx_disable(struct net_device *dev)
3350 cpu = smp_processor_id();
3351 for (i = 0; i < dev->num_tx_queues; i++) {
3352 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3354 __netif_tx_lock(txq, cpu);
3355 netif_tx_stop_queue(txq);
3356 __netif_tx_unlock(txq);
3361 static inline void netif_addr_lock(struct net_device *dev)
3363 spin_lock(&dev->addr_list_lock);
3366 static inline void netif_addr_lock_nested(struct net_device *dev)
3368 int subclass = SINGLE_DEPTH_NESTING;
3370 if (dev->netdev_ops->ndo_get_lock_subclass)
3371 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3373 spin_lock_nested(&dev->addr_list_lock, subclass);
3376 static inline void netif_addr_lock_bh(struct net_device *dev)
3378 spin_lock_bh(&dev->addr_list_lock);
3381 static inline void netif_addr_unlock(struct net_device *dev)
3383 spin_unlock(&dev->addr_list_lock);
3386 static inline void netif_addr_unlock_bh(struct net_device *dev)
3388 spin_unlock_bh(&dev->addr_list_lock);
3392 * dev_addrs walker. Should be used only for read access. Call with
3393 * rcu_read_lock held.
3395 #define for_each_dev_addr(dev, ha) \
3396 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3398 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3400 void ether_setup(struct net_device *dev);
3402 /* Support for loadable net-drivers */
3403 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3404 unsigned char name_assign_type,
3405 void (*setup)(struct net_device *),
3406 unsigned int txqs, unsigned int rxqs);
3407 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3408 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3410 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3411 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3414 int register_netdev(struct net_device *dev);
3415 void unregister_netdev(struct net_device *dev);
3417 /* General hardware address lists handling functions */
3418 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3419 struct netdev_hw_addr_list *from_list, int addr_len);
3420 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3421 struct netdev_hw_addr_list *from_list, int addr_len);
3422 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3423 struct net_device *dev,
3424 int (*sync)(struct net_device *, const unsigned char *),
3425 int (*unsync)(struct net_device *,
3426 const unsigned char *));
3427 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3428 struct net_device *dev,
3429 int (*unsync)(struct net_device *,
3430 const unsigned char *));
3431 void __hw_addr_init(struct netdev_hw_addr_list *list);
3433 /* Functions used for device addresses handling */
3434 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3435 unsigned char addr_type);
3436 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3437 unsigned char addr_type);
3438 void dev_addr_flush(struct net_device *dev);
3439 int dev_addr_init(struct net_device *dev);
3441 /* Functions used for unicast addresses handling */
3442 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3443 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3444 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3445 int dev_uc_sync(struct net_device *to, struct net_device *from);
3446 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3447 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3448 void dev_uc_flush(struct net_device *dev);
3449 void dev_uc_init(struct net_device *dev);
3452 * __dev_uc_sync - Synchonize device's unicast list
3453 * @dev: device to sync
3454 * @sync: function to call if address should be added
3455 * @unsync: function to call if address should be removed
3457 * Add newly added addresses to the interface, and release
3458 * addresses that have been deleted.
3460 static inline int __dev_uc_sync(struct net_device *dev,
3461 int (*sync)(struct net_device *,
3462 const unsigned char *),
3463 int (*unsync)(struct net_device *,
3464 const unsigned char *))
3466 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3470 * __dev_uc_unsync - Remove synchronized addresses from device
3471 * @dev: device to sync
3472 * @unsync: function to call if address should be removed
3474 * Remove all addresses that were added to the device by dev_uc_sync().
3476 static inline void __dev_uc_unsync(struct net_device *dev,
3477 int (*unsync)(struct net_device *,
3478 const unsigned char *))
3480 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3483 /* Functions used for multicast addresses handling */
3484 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3485 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3486 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3487 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3488 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3489 int dev_mc_sync(struct net_device *to, struct net_device *from);
3490 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3491 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3492 void dev_mc_flush(struct net_device *dev);
3493 void dev_mc_init(struct net_device *dev);
3496 * __dev_mc_sync - Synchonize device's multicast list
3497 * @dev: device to sync
3498 * @sync: function to call if address should be added
3499 * @unsync: function to call if address should be removed
3501 * Add newly added addresses to the interface, and release
3502 * addresses that have been deleted.
3504 static inline int __dev_mc_sync(struct net_device *dev,
3505 int (*sync)(struct net_device *,
3506 const unsigned char *),
3507 int (*unsync)(struct net_device *,
3508 const unsigned char *))
3510 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3514 * __dev_mc_unsync - Remove synchronized addresses from device
3515 * @dev: device to sync
3516 * @unsync: function to call if address should be removed
3518 * Remove all addresses that were added to the device by dev_mc_sync().
3520 static inline void __dev_mc_unsync(struct net_device *dev,
3521 int (*unsync)(struct net_device *,
3522 const unsigned char *))
3524 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3527 /* Functions used for secondary unicast and multicast support */
3528 void dev_set_rx_mode(struct net_device *dev);
3529 void __dev_set_rx_mode(struct net_device *dev);
3530 int dev_set_promiscuity(struct net_device *dev, int inc);
3531 int dev_set_allmulti(struct net_device *dev, int inc);
3532 void netdev_state_change(struct net_device *dev);
3533 void netdev_notify_peers(struct net_device *dev);
3534 void netdev_features_change(struct net_device *dev);
3535 /* Load a device via the kmod */
3536 void dev_load(struct net *net, const char *name);
3537 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3538 struct rtnl_link_stats64 *storage);
3539 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3540 const struct net_device_stats *netdev_stats);
3542 extern int netdev_max_backlog;
3543 extern int netdev_tstamp_prequeue;
3544 extern int weight_p;
3545 extern int bpf_jit_enable;
3547 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3548 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3549 struct list_head **iter);
3550 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3551 struct list_head **iter);
3553 /* iterate through upper list, must be called under RCU read lock */
3554 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3555 for (iter = &(dev)->adj_list.upper, \
3556 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3558 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3560 /* iterate through upper list, must be called under RCU read lock */
3561 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3562 for (iter = &(dev)->all_adj_list.upper, \
3563 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3565 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3567 void *netdev_lower_get_next_private(struct net_device *dev,
3568 struct list_head **iter);
3569 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3570 struct list_head **iter);
3572 #define netdev_for_each_lower_private(dev, priv, iter) \
3573 for (iter = (dev)->adj_list.lower.next, \
3574 priv = netdev_lower_get_next_private(dev, &(iter)); \
3576 priv = netdev_lower_get_next_private(dev, &(iter)))
3578 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3579 for (iter = &(dev)->adj_list.lower, \
3580 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3582 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3584 void *netdev_lower_get_next(struct net_device *dev,
3585 struct list_head **iter);
3586 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3587 for (iter = &(dev)->adj_list.lower, \
3588 ldev = netdev_lower_get_next(dev, &(iter)); \
3590 ldev = netdev_lower_get_next(dev, &(iter)))
3592 void *netdev_adjacent_get_private(struct list_head *adj_list);
3593 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3594 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3595 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3596 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3597 int netdev_master_upper_dev_link(struct net_device *dev,
3598 struct net_device *upper_dev);
3599 int netdev_master_upper_dev_link_private(struct net_device *dev,
3600 struct net_device *upper_dev,
3602 void netdev_upper_dev_unlink(struct net_device *dev,
3603 struct net_device *upper_dev);
3604 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3605 void *netdev_lower_dev_get_private(struct net_device *dev,
3606 struct net_device *lower_dev);
3608 /* RSS keys are 40 or 52 bytes long */
3609 #define NETDEV_RSS_KEY_LEN 52
3610 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3611 void netdev_rss_key_fill(void *buffer, size_t len);
3613 int dev_get_nest_level(struct net_device *dev,
3614 bool (*type_check)(struct net_device *dev));
3615 int skb_checksum_help(struct sk_buff *skb);
3616 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3617 netdev_features_t features, bool tx_path);
3618 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3619 netdev_features_t features);
3621 struct netdev_bonding_info {
3626 struct netdev_notifier_bonding_info {
3627 struct netdev_notifier_info info; /* must be first */
3628 struct netdev_bonding_info bonding_info;
3631 void netdev_bonding_info_change(struct net_device *dev,
3632 struct netdev_bonding_info *bonding_info);
3635 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3637 return __skb_gso_segment(skb, features, true);
3639 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3641 static inline bool can_checksum_protocol(netdev_features_t features,
3644 return ((features & NETIF_F_GEN_CSUM) ||
3645 ((features & NETIF_F_V4_CSUM) &&
3646 protocol == htons(ETH_P_IP)) ||
3647 ((features & NETIF_F_V6_CSUM) &&
3648 protocol == htons(ETH_P_IPV6)) ||
3649 ((features & NETIF_F_FCOE_CRC) &&
3650 protocol == htons(ETH_P_FCOE)));
3654 void netdev_rx_csum_fault(struct net_device *dev);
3656 static inline void netdev_rx_csum_fault(struct net_device *dev)
3660 /* rx skb timestamps */
3661 void net_enable_timestamp(void);
3662 void net_disable_timestamp(void);
3664 #ifdef CONFIG_PROC_FS
3665 int __init dev_proc_init(void);
3667 #define dev_proc_init() 0
3670 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3671 struct sk_buff *skb, struct net_device *dev,
3674 skb->xmit_more = more ? 1 : 0;
3675 return ops->ndo_start_xmit(skb, dev);
3678 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3679 struct netdev_queue *txq, bool more)
3681 const struct net_device_ops *ops = dev->netdev_ops;
3684 rc = __netdev_start_xmit(ops, skb, dev, more);
3685 if (rc == NETDEV_TX_OK)
3686 txq_trans_update(txq);
3691 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3693 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3696 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3698 return netdev_class_create_file_ns(class_attr, NULL);
3701 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3703 netdev_class_remove_file_ns(class_attr, NULL);
3706 extern struct kobj_ns_type_operations net_ns_type_operations;
3708 const char *netdev_drivername(const struct net_device *dev);
3710 void linkwatch_run_queue(void);
3712 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3713 netdev_features_t f2)
3715 if (f1 & NETIF_F_GEN_CSUM)
3716 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3717 if (f2 & NETIF_F_GEN_CSUM)
3718 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3720 if (f1 & NETIF_F_GEN_CSUM)
3721 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3726 static inline netdev_features_t netdev_get_wanted_features(
3727 struct net_device *dev)
3729 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3731 netdev_features_t netdev_increment_features(netdev_features_t all,
3732 netdev_features_t one, netdev_features_t mask);
3734 /* Allow TSO being used on stacked device :
3735 * Performing the GSO segmentation before last device
3736 * is a performance improvement.
3738 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3739 netdev_features_t mask)
3741 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3744 int __netdev_update_features(struct net_device *dev);
3745 void netdev_update_features(struct net_device *dev);
3746 void netdev_change_features(struct net_device *dev);
3748 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3749 struct net_device *dev);
3751 netdev_features_t passthru_features_check(struct sk_buff *skb,
3752 struct net_device *dev,
3753 netdev_features_t features);
3754 netdev_features_t netif_skb_features(struct sk_buff *skb);
3756 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3758 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3760 /* check flags correspondence */
3761 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3762 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3763 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3764 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3765 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3766 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3767 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3768 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3769 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3770 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3771 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3772 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3773 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3775 return (features & feature) == feature;
3778 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3780 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3781 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3784 static inline bool netif_needs_gso(struct sk_buff *skb,
3785 netdev_features_t features)
3787 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3788 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3789 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3792 static inline void netif_set_gso_max_size(struct net_device *dev,
3795 dev->gso_max_size = size;
3798 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3799 int pulled_hlen, u16 mac_offset,
3802 skb->protocol = protocol;
3803 skb->encapsulation = 1;
3804 skb_push(skb, pulled_hlen);
3805 skb_reset_transport_header(skb);
3806 skb->mac_header = mac_offset;
3807 skb->network_header = skb->mac_header + mac_len;
3808 skb->mac_len = mac_len;
3811 static inline bool netif_is_macvlan(struct net_device *dev)
3813 return dev->priv_flags & IFF_MACVLAN;
3816 static inline bool netif_is_macvlan_port(struct net_device *dev)
3818 return dev->priv_flags & IFF_MACVLAN_PORT;
3821 static inline bool netif_is_ipvlan(struct net_device *dev)
3823 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3826 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3828 return dev->priv_flags & IFF_IPVLAN_MASTER;
3831 static inline bool netif_is_bond_master(struct net_device *dev)
3833 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3836 static inline bool netif_is_bond_slave(struct net_device *dev)
3838 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3841 static inline bool netif_supports_nofcs(struct net_device *dev)
3843 return dev->priv_flags & IFF_SUPP_NOFCS;
3846 static inline bool netif_is_l3_master(const struct net_device *dev)
3848 return dev->priv_flags & IFF_L3MDEV_MASTER;
3851 static inline bool netif_is_l3_slave(const struct net_device *dev)
3853 return dev->priv_flags & IFF_L3MDEV_SLAVE;
3856 static inline bool netif_is_bridge_master(const struct net_device *dev)
3858 return dev->priv_flags & IFF_EBRIDGE;
3861 static inline bool netif_is_bridge_port(const struct net_device *dev)
3863 return dev->priv_flags & IFF_BRIDGE_PORT;
3866 static inline bool netif_is_ovs_master(const struct net_device *dev)
3868 return dev->priv_flags & IFF_OPENVSWITCH;
3871 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3872 static inline void netif_keep_dst(struct net_device *dev)
3874 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3877 extern struct pernet_operations __net_initdata loopback_net_ops;
3879 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3881 /* netdev_printk helpers, similar to dev_printk */
3883 static inline const char *netdev_name(const struct net_device *dev)
3885 if (!dev->name[0] || strchr(dev->name, '%'))
3886 return "(unnamed net_device)";
3890 static inline const char *netdev_reg_state(const struct net_device *dev)
3892 switch (dev->reg_state) {
3893 case NETREG_UNINITIALIZED: return " (uninitialized)";
3894 case NETREG_REGISTERED: return "";
3895 case NETREG_UNREGISTERING: return " (unregistering)";
3896 case NETREG_UNREGISTERED: return " (unregistered)";
3897 case NETREG_RELEASED: return " (released)";
3898 case NETREG_DUMMY: return " (dummy)";
3901 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3902 return " (unknown)";
3906 void netdev_printk(const char *level, const struct net_device *dev,
3907 const char *format, ...);
3909 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3911 void netdev_alert(const struct net_device *dev, const char *format, ...);
3913 void netdev_crit(const struct net_device *dev, const char *format, ...);
3915 void netdev_err(const struct net_device *dev, const char *format, ...);
3917 void netdev_warn(const struct net_device *dev, const char *format, ...);
3919 void netdev_notice(const struct net_device *dev, const char *format, ...);
3921 void netdev_info(const struct net_device *dev, const char *format, ...);
3923 #define MODULE_ALIAS_NETDEV(device) \
3924 MODULE_ALIAS("netdev-" device)
3926 #if defined(CONFIG_DYNAMIC_DEBUG)
3927 #define netdev_dbg(__dev, format, args...) \
3929 dynamic_netdev_dbg(__dev, format, ##args); \
3931 #elif defined(DEBUG)
3932 #define netdev_dbg(__dev, format, args...) \
3933 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3935 #define netdev_dbg(__dev, format, args...) \
3938 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3942 #if defined(VERBOSE_DEBUG)
3943 #define netdev_vdbg netdev_dbg
3946 #define netdev_vdbg(dev, format, args...) \
3949 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3955 * netdev_WARN() acts like dev_printk(), but with the key difference
3956 * of using a WARN/WARN_ON to get the message out, including the
3957 * file/line information and a backtrace.
3959 #define netdev_WARN(dev, format, args...) \
3960 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3961 netdev_reg_state(dev), ##args)
3963 /* netif printk helpers, similar to netdev_printk */
3965 #define netif_printk(priv, type, level, dev, fmt, args...) \
3967 if (netif_msg_##type(priv)) \
3968 netdev_printk(level, (dev), fmt, ##args); \
3971 #define netif_level(level, priv, type, dev, fmt, args...) \
3973 if (netif_msg_##type(priv)) \
3974 netdev_##level(dev, fmt, ##args); \
3977 #define netif_emerg(priv, type, dev, fmt, args...) \
3978 netif_level(emerg, priv, type, dev, fmt, ##args)
3979 #define netif_alert(priv, type, dev, fmt, args...) \
3980 netif_level(alert, priv, type, dev, fmt, ##args)
3981 #define netif_crit(priv, type, dev, fmt, args...) \
3982 netif_level(crit, priv, type, dev, fmt, ##args)
3983 #define netif_err(priv, type, dev, fmt, args...) \
3984 netif_level(err, priv, type, dev, fmt, ##args)
3985 #define netif_warn(priv, type, dev, fmt, args...) \
3986 netif_level(warn, priv, type, dev, fmt, ##args)
3987 #define netif_notice(priv, type, dev, fmt, args...) \
3988 netif_level(notice, priv, type, dev, fmt, ##args)
3989 #define netif_info(priv, type, dev, fmt, args...) \
3990 netif_level(info, priv, type, dev, fmt, ##args)
3992 #if defined(CONFIG_DYNAMIC_DEBUG)
3993 #define netif_dbg(priv, type, netdev, format, args...) \
3995 if (netif_msg_##type(priv)) \
3996 dynamic_netdev_dbg(netdev, format, ##args); \
3998 #elif defined(DEBUG)
3999 #define netif_dbg(priv, type, dev, format, args...) \
4000 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4002 #define netif_dbg(priv, type, dev, format, args...) \
4005 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4010 #if defined(VERBOSE_DEBUG)
4011 #define netif_vdbg netif_dbg
4013 #define netif_vdbg(priv, type, dev, format, args...) \
4016 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4022 * The list of packet types we will receive (as opposed to discard)
4023 * and the routines to invoke.
4025 * Why 16. Because with 16 the only overlap we get on a hash of the
4026 * low nibble of the protocol value is RARP/SNAP/X.25.
4028 * NOTE: That is no longer true with the addition of VLAN tags. Not
4029 * sure which should go first, but I bet it won't make much
4030 * difference if we are running VLANs. The good news is that
4031 * this protocol won't be in the list unless compiled in, so
4032 * the average user (w/out VLANs) will not be adversely affected.
4048 #define PTYPE_HASH_SIZE (16)
4049 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4051 #endif /* _LINUX_NETDEVICE_H */