2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the Interfaces handler.
8 * Version: @(#)dev.h 1.0.10 08/12/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
23 * Moved to /usr/include/linux for NET3
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
46 #include <net/dcbnl.h>
48 #include <net/netprio_cgroup.h>
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
62 /* 802.15.4 specific */
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 const struct ethtool_ops *ops);
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
74 #define NET_RX_DROP 1 /* packet dropped */
77 * Transmit return codes: transmit return codes originate from three different
80 * - qdisc return codes
81 * - driver transmit return codes
84 * Drivers are allowed to return any one of those in their hard_start_xmit()
85 * function. Real network devices commonly used with qdiscs should only return
86 * the driver transmit return codes though - when qdiscs are used, the actual
87 * transmission happens asynchronously, so the value is not propagated to
88 * higher layers. Virtual network devices transmit synchronously; in this case
89 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90 * others are propagated to higher layers.
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS 0x00
95 #define NET_XMIT_DROP 0x01 /* skb dropped */
96 #define NET_XMIT_CN 0x02 /* congestion notification */
97 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100 * indicates that the device will soon be dropping packets, or already drops
101 * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK 0xf0
109 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
110 NETDEV_TX_OK = 0x00, /* driver took care of packet */
111 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
113 typedef enum netdev_tx netdev_tx_t;
116 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
119 static inline bool dev_xmit_complete(int rc)
122 * Positive cases with an skb consumed by a driver:
123 * - successful transmission (rc == NETDEV_TX_OK)
124 * - error while transmitting (rc < 0)
125 * - error while queueing to a different device (rc & NET_XMIT_MASK)
127 if (likely(rc < NET_XMIT_MASK))
134 * Compute the worst-case header length according to the protocols
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 # define LL_MAX_HEADER 128
144 # define LL_MAX_HEADER 96
147 # define LL_MAX_HEADER 32
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 * Old network device statistics. Fields are native words
159 * (unsigned long) so they can be read and written atomically.
162 struct net_device_stats {
163 unsigned long rx_packets;
164 unsigned long tx_packets;
165 unsigned long rx_bytes;
166 unsigned long tx_bytes;
167 unsigned long rx_errors;
168 unsigned long tx_errors;
169 unsigned long rx_dropped;
170 unsigned long tx_dropped;
171 unsigned long multicast;
172 unsigned long collisions;
173 unsigned long rx_length_errors;
174 unsigned long rx_over_errors;
175 unsigned long rx_crc_errors;
176 unsigned long rx_frame_errors;
177 unsigned long rx_fifo_errors;
178 unsigned long rx_missed_errors;
179 unsigned long tx_aborted_errors;
180 unsigned long tx_carrier_errors;
181 unsigned long tx_fifo_errors;
182 unsigned long tx_heartbeat_errors;
183 unsigned long tx_window_errors;
184 unsigned long rx_compressed;
185 unsigned long tx_compressed;
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
201 struct netdev_hw_addr {
202 struct list_head list;
203 unsigned char addr[MAX_ADDR_LEN];
205 #define NETDEV_HW_ADDR_T_LAN 1
206 #define NETDEV_HW_ADDR_T_SAN 2
207 #define NETDEV_HW_ADDR_T_SLAVE 3
208 #define NETDEV_HW_ADDR_T_UNICAST 4
209 #define NETDEV_HW_ADDR_T_MULTICAST 5
214 struct rcu_head rcu_head;
217 struct netdev_hw_addr_list {
218 struct list_head list;
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 list_for_each_entry(ha, &(l)->list, list)
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
242 /* cached hardware header; allow for machine alignment needs. */
243 #define HH_DATA_MOD 16
244 #define HH_DATA_OFF(__len) \
245 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
246 #define HH_DATA_ALIGN(__len) \
247 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
248 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
251 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253 * dev->hard_header_len ? (dev->hard_header_len +
254 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256 * We could use other alignment values, but we must maintain the
257 * relationship HH alignment <= LL alignment.
259 #define LL_RESERVED_SPACE(dev) \
260 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
262 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
265 int (*create) (struct sk_buff *skb, struct net_device *dev,
266 unsigned short type, const void *daddr,
267 const void *saddr, unsigned int len);
268 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
269 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
270 void (*cache_update)(struct hh_cache *hh,
271 const struct net_device *dev,
272 const unsigned char *haddr);
273 bool (*validate)(const char *ll_header, unsigned int len);
276 /* These flag bits are private to the generic network queueing
277 * layer; they may not be explicitly referenced by any other
281 enum netdev_state_t {
283 __LINK_STATE_PRESENT,
284 __LINK_STATE_NOCARRIER,
285 __LINK_STATE_LINKWATCH_PENDING,
286 __LINK_STATE_DORMANT,
291 * This structure holds boot-time configured netdevice settings. They
292 * are then used in the device probing.
294 struct netdev_boot_setup {
298 #define NETDEV_BOOT_SETUP_MAX 8
300 int __init netdev_boot_setup(char *str);
303 * Structure for NAPI scheduling similar to tasklet but with weighting
306 /* The poll_list must only be managed by the entity which
307 * changes the state of the NAPI_STATE_SCHED bit. This means
308 * whoever atomically sets that bit can add this napi_struct
309 * to the per-CPU poll_list, and whoever clears that bit
310 * can remove from the list right before clearing the bit.
312 struct list_head poll_list;
316 unsigned int gro_count;
317 int (*poll)(struct napi_struct *, int);
318 #ifdef CONFIG_NETPOLL
319 spinlock_t poll_lock;
322 struct net_device *dev;
323 struct sk_buff *gro_list;
325 struct hrtimer timer;
326 struct list_head dev_list;
327 struct hlist_node napi_hash_node;
328 unsigned int napi_id;
332 NAPI_STATE_SCHED, /* Poll is scheduled */
333 NAPI_STATE_DISABLE, /* Disable pending */
334 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
335 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
336 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
346 typedef enum gro_result gro_result_t;
349 * enum rx_handler_result - Possible return values for rx_handlers.
350 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
352 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
353 * case skb->dev was changed by rx_handler.
354 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
355 * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
357 * rx_handlers are functions called from inside __netif_receive_skb(), to do
358 * special processing of the skb, prior to delivery to protocol handlers.
360 * Currently, a net_device can only have a single rx_handler registered. Trying
361 * to register a second rx_handler will return -EBUSY.
363 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
364 * To unregister a rx_handler on a net_device, use
365 * netdev_rx_handler_unregister().
367 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
370 * If the rx_handler consumed the skb in some way, it should return
371 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
372 * the skb to be delivered in some other way.
374 * If the rx_handler changed skb->dev, to divert the skb to another
375 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
376 * new device will be called if it exists.
378 * If the rx_handler decides the skb should be ignored, it should return
379 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
380 * are registered on exact device (ptype->dev == skb->dev).
382 * If the rx_handler didn't change skb->dev, but wants the skb to be normally
383 * delivered, it should return RX_HANDLER_PASS.
385 * A device without a registered rx_handler will behave as if rx_handler
386 * returned RX_HANDLER_PASS.
389 enum rx_handler_result {
395 typedef enum rx_handler_result rx_handler_result_t;
396 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
398 void __napi_schedule(struct napi_struct *n);
399 void __napi_schedule_irqoff(struct napi_struct *n);
401 static inline bool napi_disable_pending(struct napi_struct *n)
403 return test_bit(NAPI_STATE_DISABLE, &n->state);
407 * napi_schedule_prep - check if NAPI can be scheduled
410 * Test if NAPI routine is already running, and if not mark
411 * it as running. This is used as a condition variable to
412 * insure only one NAPI poll instance runs. We also make
413 * sure there is no pending NAPI disable.
415 static inline bool napi_schedule_prep(struct napi_struct *n)
417 return !napi_disable_pending(n) &&
418 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
422 * napi_schedule - schedule NAPI poll
425 * Schedule NAPI poll routine to be called if it is not already
428 static inline void napi_schedule(struct napi_struct *n)
430 if (napi_schedule_prep(n))
435 * napi_schedule_irqoff - schedule NAPI poll
438 * Variant of napi_schedule(), assuming hard irqs are masked.
440 static inline void napi_schedule_irqoff(struct napi_struct *n)
442 if (napi_schedule_prep(n))
443 __napi_schedule_irqoff(n);
446 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
447 static inline bool napi_reschedule(struct napi_struct *napi)
449 if (napi_schedule_prep(napi)) {
450 __napi_schedule(napi);
456 void __napi_complete(struct napi_struct *n);
457 void napi_complete_done(struct napi_struct *n, int work_done);
459 * napi_complete - NAPI processing complete
462 * Mark NAPI processing as complete.
463 * Consider using napi_complete_done() instead.
465 static inline void napi_complete(struct napi_struct *n)
467 return napi_complete_done(n, 0);
471 * napi_hash_add - add a NAPI to global hashtable
472 * @napi: NAPI context
474 * Generate a new napi_id and store a @napi under it in napi_hash.
475 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
476 * Note: This is normally automatically done from netif_napi_add(),
477 * so might disappear in a future Linux version.
479 void napi_hash_add(struct napi_struct *napi);
482 * napi_hash_del - remove a NAPI from global table
483 * @napi: NAPI context
485 * Warning: caller must observe RCU grace period
486 * before freeing memory containing @napi, if
487 * this function returns true.
488 * Note: core networking stack automatically calls it
489 * from netif_napi_del().
490 * Drivers might want to call this helper to combine all
491 * the needed RCU grace periods into a single one.
493 bool napi_hash_del(struct napi_struct *napi);
496 * napi_disable - prevent NAPI from scheduling
499 * Stop NAPI from being scheduled on this context.
500 * Waits till any outstanding processing completes.
502 void napi_disable(struct napi_struct *n);
505 * napi_enable - enable NAPI scheduling
508 * Resume NAPI from being scheduled on this context.
509 * Must be paired with napi_disable.
511 static inline void napi_enable(struct napi_struct *n)
513 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 smp_mb__before_atomic();
515 clear_bit(NAPI_STATE_SCHED, &n->state);
516 clear_bit(NAPI_STATE_NPSVC, &n->state);
520 * napi_synchronize - wait until NAPI is not running
523 * Wait until NAPI is done being scheduled on this context.
524 * Waits till any outstanding processing completes but
525 * does not disable future activations.
527 static inline void napi_synchronize(const struct napi_struct *n)
529 if (IS_ENABLED(CONFIG_SMP))
530 while (test_bit(NAPI_STATE_SCHED, &n->state))
536 enum netdev_queue_state_t {
537 __QUEUE_STATE_DRV_XOFF,
538 __QUEUE_STATE_STACK_XOFF,
539 __QUEUE_STATE_FROZEN,
542 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
546 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
553 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
554 * netif_tx_* functions below are used to manipulate this flag. The
555 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556 * queue independently. The netif_xmit_*stopped functions below are called
557 * to check if the queue has been stopped by the driver or stack (either
558 * of the XOFF bits are set in the state). Drivers should not need to call
559 * netif_xmit*stopped functions, they should only be using netif_tx_*.
562 struct netdev_queue {
566 struct net_device *dev;
567 struct Qdisc __rcu *qdisc;
568 struct Qdisc *qdisc_sleeping;
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
575 unsigned long tx_maxrate;
577 * Number of TX timeouts for this queue
578 * (/sys/class/net/DEV/Q/trans_timeout)
580 unsigned long trans_timeout;
584 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
587 * Time (in jiffies) of last Tx
589 unsigned long trans_start;
596 } ____cacheline_aligned_in_smp;
598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
616 * This structure holds an RPS map which can be of variable length. The
617 * map is an array of CPUs.
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
627 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628 * tail pointer for that CPU's input queue at the time of last enqueue, and
629 * a hardware filter index.
631 struct rps_dev_flow {
634 unsigned int last_qtail;
636 #define RPS_NO_FILTER 0xffff
639 * The rps_dev_flow_table structure contains a table of flow mappings.
641 struct rps_dev_flow_table {
644 struct rps_dev_flow flows[0];
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647 ((_num) * sizeof(struct rps_dev_flow)))
650 * The rps_sock_flow_table contains mappings of flows to the last CPU
651 * on which they were processed by the application (set in recvmsg).
652 * Each entry is a 32bit value. Upper part is the high-order bits
653 * of flow hash, lower part is CPU number.
654 * rps_cpu_mask is used to partition the space, depending on number of
655 * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656 * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
657 * meaning we use 32-6=26 bits for the hash.
659 struct rps_sock_flow_table {
662 u32 ents[0] ____cacheline_aligned_in_smp;
664 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
666 #define RPS_NO_CPU 0xffff
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
675 unsigned int index = hash & table->mask;
676 u32 val = hash & ~rps_cpu_mask;
678 /* We only give a hint, preemption can change CPU under us */
679 val |= raw_smp_processor_id();
681 if (table->ents[index] != val)
682 table->ents[index] = val;
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
690 #endif /* CONFIG_RPS */
692 /* This structure contains an instance of an RX queue. */
693 struct netdev_rx_queue {
695 struct rps_map __rcu *rps_map;
696 struct rps_dev_flow_table __rcu *rps_flow_table;
699 struct net_device *dev;
700 } ____cacheline_aligned_in_smp;
703 * RX queue sysfs structures and functions.
705 struct rx_queue_attribute {
706 struct attribute attr;
707 ssize_t (*show)(struct netdev_rx_queue *queue,
708 struct rx_queue_attribute *attr, char *buf);
709 ssize_t (*store)(struct netdev_rx_queue *queue,
710 struct rx_queue_attribute *attr, const char *buf, size_t len);
715 * This structure holds an XPS map which can be of variable length. The
716 * map is an array of queues.
720 unsigned int alloc_len;
724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726 - sizeof(struct xps_map)) / sizeof(u16))
729 * This structure holds all XPS maps for device. Maps are indexed by CPU.
731 struct xps_dev_maps {
733 struct xps_map __rcu *cpu_map[0];
735 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
736 (nr_cpu_ids * sizeof(struct xps_map *)))
737 #endif /* CONFIG_XPS */
739 #define TC_MAX_QUEUE 16
740 #define TC_BITMASK 15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
749 * This structure is to hold information about the device
750 * configured to run FCoE protocol stack.
752 struct netdev_fcoe_hbainfo {
753 char manufacturer[64];
754 char serial_number[64];
755 char hardware_version[64];
756 char driver_version[64];
757 char optionrom_version[64];
758 char firmware_version[64];
760 char model_description[256];
764 #define MAX_PHYS_ITEM_ID_LEN 32
766 /* This structure holds a unique identifier to identify some
767 * physical item (port for example) used by a netdevice.
769 struct netdev_phys_item_id {
770 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 unsigned char id_len;
774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 struct netdev_phys_item_id *b)
777 return a->id_len == b->id_len &&
778 memcmp(a->id, b->id, a->id_len) == 0;
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 struct sk_buff *skb);
784 /* These structures hold the attributes of qdisc and classifiers
785 * that are being passed to the netdevice through the setup_tc op.
795 struct tc_cls_u32_offload;
797 struct tc_to_netdev {
801 struct tc_cls_u32_offload *cls_u32;
802 struct tc_cls_flower_offload *cls_flower;
803 struct tc_cls_matchall_offload *cls_mall;
804 struct tc_cls_bpf_offload *cls_bpf;
808 /* These structures hold the attributes of xdp state that are being passed
809 * to the netdevice through the xdp op.
811 enum xdp_netdev_command {
812 /* Set or clear a bpf program used in the earliest stages of packet
813 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
814 * is responsible for calling bpf_prog_put on any old progs that are
815 * stored. In case of error, the callee need not release the new prog
816 * reference, but on success it takes ownership and must bpf_prog_put
817 * when it is no longer used.
820 /* Check if a bpf program is set on the device. The callee should
821 * return true if a program is currently attached and running.
827 enum xdp_netdev_command command;
830 struct bpf_prog *prog;
837 * This structure defines the management hooks for network devices.
838 * The following hooks can be defined; unless noted otherwise, they are
839 * optional and can be filled with a null pointer.
841 * int (*ndo_init)(struct net_device *dev);
842 * This function is called once when a network device is registered.
843 * The network device can use this for any late stage initialization
844 * or semantic validation. It can fail with an error code which will
845 * be propagated back to register_netdev.
847 * void (*ndo_uninit)(struct net_device *dev);
848 * This function is called when device is unregistered or when registration
849 * fails. It is not called if init fails.
851 * int (*ndo_open)(struct net_device *dev);
852 * This function is called when a network device transitions to the up
855 * int (*ndo_stop)(struct net_device *dev);
856 * This function is called when a network device transitions to the down
859 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
860 * struct net_device *dev);
861 * Called when a packet needs to be transmitted.
862 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
863 * the queue before that can happen; it's for obsolete devices and weird
864 * corner cases, but the stack really does a non-trivial amount
865 * of useless work if you return NETDEV_TX_BUSY.
866 * Required; cannot be NULL.
868 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
869 * struct net_device *dev
870 * netdev_features_t features);
871 * Called by core transmit path to determine if device is capable of
872 * performing offload operations on a given packet. This is to give
873 * the device an opportunity to implement any restrictions that cannot
874 * be otherwise expressed by feature flags. The check is called with
875 * the set of features that the stack has calculated and it returns
876 * those the driver believes to be appropriate.
878 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
879 * void *accel_priv, select_queue_fallback_t fallback);
880 * Called to decide which queue to use when device supports multiple
883 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
884 * This function is called to allow device receiver to make
885 * changes to configuration when multicast or promiscuous is enabled.
887 * void (*ndo_set_rx_mode)(struct net_device *dev);
888 * This function is called device changes address list filtering.
889 * If driver handles unicast address filtering, it should set
890 * IFF_UNICAST_FLT in its priv_flags.
892 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
893 * This function is called when the Media Access Control address
894 * needs to be changed. If this interface is not defined, the
895 * MAC address can not be changed.
897 * int (*ndo_validate_addr)(struct net_device *dev);
898 * Test if Media Access Control address is valid for the device.
900 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
901 * Called when a user requests an ioctl which can't be handled by
902 * the generic interface code. If not defined ioctls return
903 * not supported error code.
905 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
906 * Used to set network devices bus interface parameters. This interface
907 * is retained for legacy reasons; new devices should use the bus
908 * interface (PCI) for low level management.
910 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
911 * Called when a user wants to change the Maximum Transfer Unit
912 * of a device. If not defined, any request to change MTU will
913 * will return an error.
915 * void (*ndo_tx_timeout)(struct net_device *dev);
916 * Callback used when the transmitter has not made any progress
917 * for dev->watchdog ticks.
919 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
920 * struct rtnl_link_stats64 *storage);
921 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
922 * Called when a user wants to get the network device usage
923 * statistics. Drivers must do one of the following:
924 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
925 * rtnl_link_stats64 structure passed by the caller.
926 * 2. Define @ndo_get_stats to update a net_device_stats structure
927 * (which should normally be dev->stats) and return a pointer to
928 * it. The structure may be changed asynchronously only if each
929 * field is written atomically.
930 * 3. Update dev->stats asynchronously and atomically, and define
933 * bool (*ndo_has_offload_stats)(int attr_id)
934 * Return true if this device supports offload stats of this attr_id.
936 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
938 * Get statistics for offload operations by attr_id. Write it into the
941 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
942 * If device supports VLAN filtering this function is called when a
943 * VLAN id is registered.
945 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
946 * If device supports VLAN filtering this function is called when a
947 * VLAN id is unregistered.
949 * void (*ndo_poll_controller)(struct net_device *dev);
951 * SR-IOV management functions.
952 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
953 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
954 * u8 qos, __be16 proto);
955 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
957 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
958 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
959 * int (*ndo_get_vf_config)(struct net_device *dev,
960 * int vf, struct ifla_vf_info *ivf);
961 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
962 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
963 * struct nlattr *port[]);
965 * Enable or disable the VF ability to query its RSS Redirection Table and
966 * Hash Key. This is needed since on some devices VF share this information
967 * with PF and querying it may introduce a theoretical security risk.
968 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
969 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
970 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
971 * Called to setup 'tc' number of traffic classes in the net device. This
972 * is always called from the stack with the rtnl lock held and netif tx
973 * queues stopped. This allows the netdevice to perform queue management
976 * Fiber Channel over Ethernet (FCoE) offload functions.
977 * int (*ndo_fcoe_enable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to start using LLD for FCoE
979 * so the underlying device can perform whatever needed configuration or
980 * initialization to support acceleration of FCoE traffic.
982 * int (*ndo_fcoe_disable)(struct net_device *dev);
983 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
984 * so the underlying device can perform whatever needed clean-ups to
985 * stop supporting acceleration of FCoE traffic.
987 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
988 * struct scatterlist *sgl, unsigned int sgc);
989 * Called when the FCoE Initiator wants to initialize an I/O that
990 * is a possible candidate for Direct Data Placement (DDP). The LLD can
991 * perform necessary setup and returns 1 to indicate the device is set up
992 * successfully to perform DDP on this I/O, otherwise this returns 0.
994 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
995 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
996 * indicated by the FC exchange id 'xid', so the underlying device can
997 * clean up and reuse resources for later DDP requests.
999 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1000 * struct scatterlist *sgl, unsigned int sgc);
1001 * Called when the FCoE Target wants to initialize an I/O that
1002 * is a possible candidate for Direct Data Placement (DDP). The LLD can
1003 * perform necessary setup and returns 1 to indicate the device is set up
1004 * successfully to perform DDP on this I/O, otherwise this returns 0.
1006 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1007 * struct netdev_fcoe_hbainfo *hbainfo);
1008 * Called when the FCoE Protocol stack wants information on the underlying
1009 * device. This information is utilized by the FCoE protocol stack to
1010 * register attributes with Fiber Channel management service as per the
1011 * FC-GS Fabric Device Management Information(FDMI) specification.
1013 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1014 * Called when the underlying device wants to override default World Wide
1015 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1016 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1017 * protocol stack to use.
1020 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1021 * u16 rxq_index, u32 flow_id);
1022 * Set hardware filter for RFS. rxq_index is the target queue index;
1023 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1024 * Return the filter ID on success, or a negative error code.
1026 * Slave management functions (for bridge, bonding, etc).
1027 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1028 * Called to make another netdev an underling.
1030 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1031 * Called to release previously enslaved netdev.
1033 * Feature/offload setting functions.
1034 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1035 * netdev_features_t features);
1036 * Adjusts the requested feature flags according to device-specific
1037 * constraints, and returns the resulting flags. Must not modify
1040 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1041 * Called to update device configuration to new features. Passed
1042 * feature set might be less than what was returned by ndo_fix_features()).
1043 * Must return >0 or -errno if it changed dev->features itself.
1045 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1046 * struct net_device *dev,
1047 * const unsigned char *addr, u16 vid, u16 flags)
1048 * Adds an FDB entry to dev for addr.
1049 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1050 * struct net_device *dev,
1051 * const unsigned char *addr, u16 vid)
1052 * Deletes the FDB entry from dev coresponding to addr.
1053 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1054 * struct net_device *dev, struct net_device *filter_dev,
1056 * Used to add FDB entries to dump requests. Implementers should add
1057 * entries to skb and update idx with the number of entries.
1059 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1061 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1062 * struct net_device *dev, u32 filter_mask,
1064 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1067 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1068 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1069 * which do not represent real hardware may define this to allow their
1070 * userspace components to manage their virtual carrier state. Devices
1071 * that determine carrier state from physical hardware properties (eg
1072 * network cables) or protocol-dependent mechanisms (eg
1073 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1075 * int (*ndo_get_phys_port_id)(struct net_device *dev,
1076 * struct netdev_phys_item_id *ppid);
1077 * Called to get ID of physical port of this device. If driver does
1078 * not implement this, it is assumed that the hw is not able to have
1079 * multiple net devices on single physical port.
1081 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1082 * struct udp_tunnel_info *ti);
1083 * Called by UDP tunnel to notify a driver about the UDP port and socket
1084 * address family that a UDP tunnel is listnening to. It is called only
1085 * when a new port starts listening. The operation is protected by the
1088 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1089 * struct udp_tunnel_info *ti);
1090 * Called by UDP tunnel to notify the driver about a UDP port and socket
1091 * address family that the UDP tunnel is not listening to anymore. The
1092 * operation is protected by the RTNL.
1094 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1095 * struct net_device *dev)
1096 * Called by upper layer devices to accelerate switching or other
1097 * station functionality into hardware. 'pdev is the lowerdev
1098 * to use for the offload and 'dev' is the net device that will
1099 * back the offload. Returns a pointer to the private structure
1100 * the upper layer will maintain.
1101 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1102 * Called by upper layer device to delete the station created
1103 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1104 * the station and priv is the structure returned by the add
1106 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1107 * struct net_device *dev,
1109 * Callback to use for xmit over the accelerated station. This
1110 * is used in place of ndo_start_xmit on accelerated net
1112 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1113 * int queue_index, u32 maxrate);
1114 * Called when a user wants to set a max-rate limitation of specific
1116 * int (*ndo_get_iflink)(const struct net_device *dev);
1117 * Called to get the iflink value of this device.
1118 * void (*ndo_change_proto_down)(struct net_device *dev,
1120 * This function is used to pass protocol port error state information
1121 * to the switch driver. The switch driver can react to the proto_down
1122 * by doing a phys down on the associated switch port.
1123 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1124 * This function is used to get egress tunnel information for given skb.
1125 * This is useful for retrieving outer tunnel header parameters while
1127 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1128 * This function is used to specify the headroom that the skb must
1129 * consider when allocation skb during packet reception. Setting
1130 * appropriate rx headroom value allows avoiding skb head copy on
1131 * forward. Setting a negative value resets the rx headroom to the
1133 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1134 * This function is used to set or query state related to XDP on the
1135 * netdevice. See definition of enum xdp_netdev_command for details.
1138 struct net_device_ops {
1139 int (*ndo_init)(struct net_device *dev);
1140 void (*ndo_uninit)(struct net_device *dev);
1141 int (*ndo_open)(struct net_device *dev);
1142 int (*ndo_stop)(struct net_device *dev);
1143 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1144 struct net_device *dev);
1145 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1146 struct net_device *dev,
1147 netdev_features_t features);
1148 u16 (*ndo_select_queue)(struct net_device *dev,
1149 struct sk_buff *skb,
1151 select_queue_fallback_t fallback);
1152 void (*ndo_change_rx_flags)(struct net_device *dev,
1154 void (*ndo_set_rx_mode)(struct net_device *dev);
1155 int (*ndo_set_mac_address)(struct net_device *dev,
1157 int (*ndo_validate_addr)(struct net_device *dev);
1158 int (*ndo_do_ioctl)(struct net_device *dev,
1159 struct ifreq *ifr, int cmd);
1160 int (*ndo_set_config)(struct net_device *dev,
1162 int (*ndo_change_mtu)(struct net_device *dev,
1164 int (*ndo_neigh_setup)(struct net_device *dev,
1165 struct neigh_parms *);
1166 void (*ndo_tx_timeout) (struct net_device *dev);
1168 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1169 struct rtnl_link_stats64 *storage);
1170 bool (*ndo_has_offload_stats)(int attr_id);
1171 int (*ndo_get_offload_stats)(int attr_id,
1172 const struct net_device *dev,
1174 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1177 __be16 proto, u16 vid);
1178 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1179 __be16 proto, u16 vid);
1180 #ifdef CONFIG_NET_POLL_CONTROLLER
1181 void (*ndo_poll_controller)(struct net_device *dev);
1182 int (*ndo_netpoll_setup)(struct net_device *dev,
1183 struct netpoll_info *info);
1184 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1186 #ifdef CONFIG_NET_RX_BUSY_POLL
1187 int (*ndo_busy_poll)(struct napi_struct *dev);
1189 int (*ndo_set_vf_mac)(struct net_device *dev,
1190 int queue, u8 *mac);
1191 int (*ndo_set_vf_vlan)(struct net_device *dev,
1192 int queue, u16 vlan,
1193 u8 qos, __be16 proto);
1194 int (*ndo_set_vf_rate)(struct net_device *dev,
1195 int vf, int min_tx_rate,
1197 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1198 int vf, bool setting);
1199 int (*ndo_set_vf_trust)(struct net_device *dev,
1200 int vf, bool setting);
1201 int (*ndo_get_vf_config)(struct net_device *dev,
1203 struct ifla_vf_info *ivf);
1204 int (*ndo_set_vf_link_state)(struct net_device *dev,
1205 int vf, int link_state);
1206 int (*ndo_get_vf_stats)(struct net_device *dev,
1208 struct ifla_vf_stats
1210 int (*ndo_set_vf_port)(struct net_device *dev,
1212 struct nlattr *port[]);
1213 int (*ndo_get_vf_port)(struct net_device *dev,
1214 int vf, struct sk_buff *skb);
1215 int (*ndo_set_vf_guid)(struct net_device *dev,
1218 int (*ndo_set_vf_rss_query_en)(
1219 struct net_device *dev,
1220 int vf, bool setting);
1221 int (*ndo_setup_tc)(struct net_device *dev,
1224 struct tc_to_netdev *tc);
1225 #if IS_ENABLED(CONFIG_FCOE)
1226 int (*ndo_fcoe_enable)(struct net_device *dev);
1227 int (*ndo_fcoe_disable)(struct net_device *dev);
1228 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1230 struct scatterlist *sgl,
1232 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1234 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1236 struct scatterlist *sgl,
1238 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1239 struct netdev_fcoe_hbainfo *hbainfo);
1242 #if IS_ENABLED(CONFIG_LIBFCOE)
1243 #define NETDEV_FCOE_WWNN 0
1244 #define NETDEV_FCOE_WWPN 1
1245 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1246 u64 *wwn, int type);
1249 #ifdef CONFIG_RFS_ACCEL
1250 int (*ndo_rx_flow_steer)(struct net_device *dev,
1251 const struct sk_buff *skb,
1255 int (*ndo_add_slave)(struct net_device *dev,
1256 struct net_device *slave_dev);
1257 int (*ndo_del_slave)(struct net_device *dev,
1258 struct net_device *slave_dev);
1259 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1260 netdev_features_t features);
1261 int (*ndo_set_features)(struct net_device *dev,
1262 netdev_features_t features);
1263 int (*ndo_neigh_construct)(struct net_device *dev,
1264 struct neighbour *n);
1265 void (*ndo_neigh_destroy)(struct net_device *dev,
1266 struct neighbour *n);
1268 int (*ndo_fdb_add)(struct ndmsg *ndm,
1269 struct nlattr *tb[],
1270 struct net_device *dev,
1271 const unsigned char *addr,
1274 int (*ndo_fdb_del)(struct ndmsg *ndm,
1275 struct nlattr *tb[],
1276 struct net_device *dev,
1277 const unsigned char *addr,
1279 int (*ndo_fdb_dump)(struct sk_buff *skb,
1280 struct netlink_callback *cb,
1281 struct net_device *dev,
1282 struct net_device *filter_dev,
1285 int (*ndo_bridge_setlink)(struct net_device *dev,
1286 struct nlmsghdr *nlh,
1288 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1290 struct net_device *dev,
1293 int (*ndo_bridge_dellink)(struct net_device *dev,
1294 struct nlmsghdr *nlh,
1296 int (*ndo_change_carrier)(struct net_device *dev,
1298 int (*ndo_get_phys_port_id)(struct net_device *dev,
1299 struct netdev_phys_item_id *ppid);
1300 int (*ndo_get_phys_port_name)(struct net_device *dev,
1301 char *name, size_t len);
1302 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1303 struct udp_tunnel_info *ti);
1304 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1305 struct udp_tunnel_info *ti);
1306 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1307 struct net_device *dev);
1308 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1311 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1312 struct net_device *dev,
1314 int (*ndo_get_lock_subclass)(struct net_device *dev);
1315 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1318 int (*ndo_get_iflink)(const struct net_device *dev);
1319 int (*ndo_change_proto_down)(struct net_device *dev,
1321 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1322 struct sk_buff *skb);
1323 void (*ndo_set_rx_headroom)(struct net_device *dev,
1324 int needed_headroom);
1325 int (*ndo_xdp)(struct net_device *dev,
1326 struct netdev_xdp *xdp);
1330 * enum net_device_priv_flags - &struct net_device priv_flags
1332 * These are the &struct net_device, they are only set internally
1333 * by drivers and used in the kernel. These flags are invisible to
1334 * userspace; this means that the order of these flags can change
1335 * during any kernel release.
1337 * You should have a pretty good reason to be extending these flags.
1339 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1340 * @IFF_EBRIDGE: Ethernet bridging device
1341 * @IFF_BONDING: bonding master or slave
1342 * @IFF_ISATAP: ISATAP interface (RFC4214)
1343 * @IFF_WAN_HDLC: WAN HDLC device
1344 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1346 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1347 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1348 * @IFF_MACVLAN_PORT: device used as macvlan port
1349 * @IFF_BRIDGE_PORT: device used as bridge port
1350 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1351 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1352 * @IFF_UNICAST_FLT: Supports unicast filtering
1353 * @IFF_TEAM_PORT: device used as team port
1354 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1355 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1356 * change when it's running
1357 * @IFF_MACVLAN: Macvlan device
1358 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1359 * underlying stacked devices
1360 * @IFF_IPVLAN_MASTER: IPvlan master device
1361 * @IFF_IPVLAN_SLAVE: IPvlan slave device
1362 * @IFF_L3MDEV_MASTER: device is an L3 master device
1363 * @IFF_NO_QUEUE: device can run without qdisc attached
1364 * @IFF_OPENVSWITCH: device is a Open vSwitch master
1365 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1366 * @IFF_TEAM: device is a team device
1367 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1368 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1369 * entity (i.e. the master device for bridged veth)
1370 * @IFF_MACSEC: device is a MACsec device
1371 * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1373 enum netdev_priv_flags {
1374 IFF_802_1Q_VLAN = 1<<0,
1378 IFF_WAN_HDLC = 1<<4,
1379 IFF_XMIT_DST_RELEASE = 1<<5,
1380 IFF_DONT_BRIDGE = 1<<6,
1381 IFF_DISABLE_NETPOLL = 1<<7,
1382 IFF_MACVLAN_PORT = 1<<8,
1383 IFF_BRIDGE_PORT = 1<<9,
1384 IFF_OVS_DATAPATH = 1<<10,
1385 IFF_TX_SKB_SHARING = 1<<11,
1386 IFF_UNICAST_FLT = 1<<12,
1387 IFF_TEAM_PORT = 1<<13,
1388 IFF_SUPP_NOFCS = 1<<14,
1389 IFF_LIVE_ADDR_CHANGE = 1<<15,
1390 IFF_MACVLAN = 1<<16,
1391 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1392 IFF_IPVLAN_MASTER = 1<<18,
1393 IFF_IPVLAN_SLAVE = 1<<19,
1394 IFF_L3MDEV_MASTER = 1<<20,
1395 IFF_NO_QUEUE = 1<<21,
1396 IFF_OPENVSWITCH = 1<<22,
1397 IFF_L3MDEV_SLAVE = 1<<23,
1399 IFF_RXFH_CONFIGURED = 1<<25,
1400 IFF_PHONY_HEADROOM = 1<<26,
1402 IFF_L3MDEV_RX_HANDLER = 1<<28,
1405 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1406 #define IFF_EBRIDGE IFF_EBRIDGE
1407 #define IFF_BONDING IFF_BONDING
1408 #define IFF_ISATAP IFF_ISATAP
1409 #define IFF_WAN_HDLC IFF_WAN_HDLC
1410 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1411 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1412 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1413 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1414 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1415 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1416 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1417 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1418 #define IFF_TEAM_PORT IFF_TEAM_PORT
1419 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1420 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1421 #define IFF_MACVLAN IFF_MACVLAN
1422 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1423 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1424 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1425 #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
1426 #define IFF_NO_QUEUE IFF_NO_QUEUE
1427 #define IFF_OPENVSWITCH IFF_OPENVSWITCH
1428 #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
1429 #define IFF_TEAM IFF_TEAM
1430 #define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
1431 #define IFF_MACSEC IFF_MACSEC
1432 #define IFF_L3MDEV_RX_HANDLER IFF_L3MDEV_RX_HANDLER
1435 * struct net_device - The DEVICE structure.
1436 * Actually, this whole structure is a big mistake. It mixes I/O
1437 * data with strictly "high-level" data, and it has to know about
1438 * almost every data structure used in the INET module.
1440 * @name: This is the first field of the "visible" part of this structure
1441 * (i.e. as seen by users in the "Space.c" file). It is the name
1444 * @name_hlist: Device name hash chain, please keep it close to name[]
1445 * @ifalias: SNMP alias
1446 * @mem_end: Shared memory end
1447 * @mem_start: Shared memory start
1448 * @base_addr: Device I/O address
1449 * @irq: Device IRQ number
1451 * @carrier_changes: Stats to monitor carrier on<->off transitions
1453 * @state: Generic network queuing layer state, see netdev_state_t
1454 * @dev_list: The global list of network devices
1455 * @napi_list: List entry used for polling NAPI devices
1456 * @unreg_list: List entry when we are unregistering the
1457 * device; see the function unregister_netdev
1458 * @close_list: List entry used when we are closing the device
1459 * @ptype_all: Device-specific packet handlers for all protocols
1460 * @ptype_specific: Device-specific, protocol-specific packet handlers
1462 * @adj_list: Directly linked devices, like slaves for bonding
1463 * @all_adj_list: All linked devices, *including* neighbours
1464 * @features: Currently active device features
1465 * @hw_features: User-changeable features
1467 * @wanted_features: User-requested features
1468 * @vlan_features: Mask of features inheritable by VLAN devices
1470 * @hw_enc_features: Mask of features inherited by encapsulating devices
1471 * This field indicates what encapsulation
1472 * offloads the hardware is capable of doing,
1473 * and drivers will need to set them appropriately.
1475 * @mpls_features: Mask of features inheritable by MPLS
1477 * @ifindex: interface index
1478 * @group: The group the device belongs to
1480 * @stats: Statistics struct, which was left as a legacy, use
1481 * rtnl_link_stats64 instead
1483 * @rx_dropped: Dropped packets by core network,
1484 * do not use this in drivers
1485 * @tx_dropped: Dropped packets by core network,
1486 * do not use this in drivers
1487 * @rx_nohandler: nohandler dropped packets by core network on
1488 * inactive devices, do not use this in drivers
1490 * @wireless_handlers: List of functions to handle Wireless Extensions,
1492 * see <net/iw_handler.h> for details.
1493 * @wireless_data: Instance data managed by the core of wireless extensions
1495 * @netdev_ops: Includes several pointers to callbacks,
1496 * if one wants to override the ndo_*() functions
1497 * @ethtool_ops: Management operations
1498 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1499 * discovery handling. Necessary for e.g. 6LoWPAN.
1500 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1501 * of Layer 2 headers.
1503 * @flags: Interface flags (a la BSD)
1504 * @priv_flags: Like 'flags' but invisible to userspace,
1505 * see if.h for the definitions
1506 * @gflags: Global flags ( kept as legacy )
1507 * @padded: How much padding added by alloc_netdev()
1508 * @operstate: RFC2863 operstate
1509 * @link_mode: Mapping policy to operstate
1510 * @if_port: Selectable AUI, TP, ...
1512 * @mtu: Interface MTU value
1513 * @type: Interface hardware type
1514 * @hard_header_len: Maximum hardware header length.
1515 * @min_header_len: Minimum hardware header length
1517 * @needed_headroom: Extra headroom the hardware may need, but not in all
1518 * cases can this be guaranteed
1519 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1520 * cases can this be guaranteed. Some cases also use
1521 * LL_MAX_HEADER instead to allocate the skb
1523 * interface address info:
1525 * @perm_addr: Permanent hw address
1526 * @addr_assign_type: Hw address assignment type
1527 * @addr_len: Hardware address length
1528 * @neigh_priv_len: Used in neigh_alloc()
1529 * @dev_id: Used to differentiate devices that share
1530 * the same link layer address
1531 * @dev_port: Used to differentiate devices that share
1533 * @addr_list_lock: XXX: need comments on this one
1534 * @uc_promisc: Counter that indicates promiscuous mode
1535 * has been enabled due to the need to listen to
1536 * additional unicast addresses in a device that
1537 * does not implement ndo_set_rx_mode()
1538 * @uc: unicast mac addresses
1539 * @mc: multicast mac addresses
1540 * @dev_addrs: list of device hw addresses
1541 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1542 * @promiscuity: Number of times the NIC is told to work in
1543 * promiscuous mode; if it becomes 0 the NIC will
1544 * exit promiscuous mode
1545 * @allmulti: Counter, enables or disables allmulticast mode
1547 * @vlan_info: VLAN info
1548 * @dsa_ptr: dsa specific data
1549 * @tipc_ptr: TIPC specific data
1550 * @atalk_ptr: AppleTalk link
1551 * @ip_ptr: IPv4 specific data
1552 * @dn_ptr: DECnet specific data
1553 * @ip6_ptr: IPv6 specific data
1554 * @ax25_ptr: AX.25 specific data
1555 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1557 * @last_rx: Time of last Rx
1558 * @dev_addr: Hw address (before bcast,
1559 * because most packets are unicast)
1561 * @_rx: Array of RX queues
1562 * @num_rx_queues: Number of RX queues
1563 * allocated at register_netdev() time
1564 * @real_num_rx_queues: Number of RX queues currently active in device
1566 * @rx_handler: handler for received packets
1567 * @rx_handler_data: XXX: need comments on this one
1568 * @ingress_queue: XXX: need comments on this one
1569 * @broadcast: hw bcast address
1571 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1572 * indexed by RX queue number. Assigned by driver.
1573 * This must only be set if the ndo_rx_flow_steer
1574 * operation is defined
1575 * @index_hlist: Device index hash chain
1577 * @_tx: Array of TX queues
1578 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1579 * @real_num_tx_queues: Number of TX queues currently active in device
1580 * @qdisc: Root qdisc from userspace point of view
1581 * @tx_queue_len: Max frames per queue allowed
1582 * @tx_global_lock: XXX: need comments on this one
1584 * @xps_maps: XXX: need comments on this one
1586 * @watchdog_timeo: Represents the timeout that is used by
1587 * the watchdog (see dev_watchdog())
1588 * @watchdog_timer: List of timers
1590 * @pcpu_refcnt: Number of references to this device
1591 * @todo_list: Delayed register/unregister
1592 * @link_watch_list: XXX: need comments on this one
1594 * @reg_state: Register/unregister state machine
1595 * @dismantle: Device is going to be freed
1596 * @rtnl_link_state: This enum represents the phases of creating
1599 * @destructor: Called from unregister,
1600 * can be used to call free_netdev
1601 * @npinfo: XXX: need comments on this one
1602 * @nd_net: Network namespace this network device is inside
1604 * @ml_priv: Mid-layer private
1605 * @lstats: Loopback statistics
1606 * @tstats: Tunnel statistics
1607 * @dstats: Dummy statistics
1608 * @vstats: Virtual ethernet statistics
1613 * @dev: Class/net/name entry
1614 * @sysfs_groups: Space for optional device, statistics and wireless
1617 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1618 * @rtnl_link_ops: Rtnl_link_ops
1620 * @gso_max_size: Maximum size of generic segmentation offload
1621 * @gso_max_segs: Maximum number of segments that can be passed to the
1624 * @dcbnl_ops: Data Center Bridging netlink ops
1625 * @num_tc: Number of traffic classes in the net device
1626 * @tc_to_txq: XXX: need comments on this one
1627 * @prio_tc_map: XXX: need comments on this one
1629 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1631 * @priomap: XXX: need comments on this one
1632 * @phydev: Physical device may attach itself
1633 * for hardware timestamping
1635 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1636 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1638 * @proto_down: protocol port state information can be sent to the
1639 * switch driver and used to set the phys state of the
1642 * FIXME: cleanup struct net_device such that network protocol info
1647 char name[IFNAMSIZ];
1648 struct hlist_node name_hlist;
1651 * I/O specific fields
1652 * FIXME: Merge these and struct ifmap into one
1654 unsigned long mem_end;
1655 unsigned long mem_start;
1656 unsigned long base_addr;
1659 atomic_t carrier_changes;
1662 * Some hardware also needs these fields (state,dev_list,
1663 * napi_list,unreg_list,close_list) but they are not
1664 * part of the usual set specified in Space.c.
1667 unsigned long state;
1669 struct list_head dev_list;
1670 struct list_head napi_list;
1671 struct list_head unreg_list;
1672 struct list_head close_list;
1673 struct list_head ptype_all;
1674 struct list_head ptype_specific;
1677 struct list_head upper;
1678 struct list_head lower;
1682 struct list_head upper;
1683 struct list_head lower;
1686 netdev_features_t features;
1687 netdev_features_t hw_features;
1688 netdev_features_t wanted_features;
1689 netdev_features_t vlan_features;
1690 netdev_features_t hw_enc_features;
1691 netdev_features_t mpls_features;
1692 netdev_features_t gso_partial_features;
1697 struct net_device_stats stats;
1699 atomic_long_t rx_dropped;
1700 atomic_long_t tx_dropped;
1701 atomic_long_t rx_nohandler;
1703 #ifdef CONFIG_WIRELESS_EXT
1704 const struct iw_handler_def *wireless_handlers;
1705 struct iw_public_data *wireless_data;
1707 const struct net_device_ops *netdev_ops;
1708 const struct ethtool_ops *ethtool_ops;
1709 #ifdef CONFIG_NET_SWITCHDEV
1710 const struct switchdev_ops *switchdev_ops;
1712 #ifdef CONFIG_NET_L3_MASTER_DEV
1713 const struct l3mdev_ops *l3mdev_ops;
1715 #if IS_ENABLED(CONFIG_IPV6)
1716 const struct ndisc_ops *ndisc_ops;
1719 const struct header_ops *header_ops;
1722 unsigned int priv_flags;
1724 unsigned short gflags;
1725 unsigned short padded;
1727 unsigned char operstate;
1728 unsigned char link_mode;
1730 unsigned char if_port;
1733 /* Note : dev->mtu is often read without holding a lock.
1734 * Writers usually hold RTNL.
1735 * It is recommended to use READ_ONCE() to annotate the reads,
1736 * and to use WRITE_ONCE() to annotate the writes.
1739 unsigned short type;
1740 unsigned short hard_header_len;
1741 unsigned short min_header_len;
1743 unsigned short needed_headroom;
1744 unsigned short needed_tailroom;
1746 /* Interface address info. */
1747 unsigned char perm_addr[MAX_ADDR_LEN];
1748 unsigned char addr_assign_type;
1749 unsigned char addr_len;
1750 unsigned short neigh_priv_len;
1751 unsigned short dev_id;
1752 unsigned short dev_port;
1753 spinlock_t addr_list_lock;
1754 unsigned char name_assign_type;
1756 struct netdev_hw_addr_list uc;
1757 struct netdev_hw_addr_list mc;
1758 struct netdev_hw_addr_list dev_addrs;
1761 struct kset *queues_kset;
1763 unsigned int promiscuity;
1764 unsigned int allmulti;
1767 /* Protocol-specific pointers */
1769 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1770 struct vlan_info __rcu *vlan_info;
1772 #if IS_ENABLED(CONFIG_NET_DSA)
1773 struct dsa_switch_tree *dsa_ptr;
1775 #if IS_ENABLED(CONFIG_TIPC)
1776 struct tipc_bearer __rcu *tipc_ptr;
1779 struct in_device __rcu *ip_ptr;
1780 struct dn_dev __rcu *dn_ptr;
1781 struct inet6_dev __rcu *ip6_ptr;
1783 struct wireless_dev *ieee80211_ptr;
1784 struct wpan_dev *ieee802154_ptr;
1785 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1786 struct mpls_dev __rcu *mpls_ptr;
1790 * Cache lines mostly used on receive path (including eth_type_trans())
1792 unsigned long last_rx;
1794 /* Interface address info used in eth_type_trans() */
1795 unsigned char *dev_addr;
1798 struct netdev_rx_queue *_rx;
1800 unsigned int num_rx_queues;
1801 unsigned int real_num_rx_queues;
1804 unsigned long gro_flush_timeout;
1805 rx_handler_func_t __rcu *rx_handler;
1806 void __rcu *rx_handler_data;
1808 #ifdef CONFIG_NET_CLS_ACT
1809 struct tcf_proto __rcu *ingress_cl_list;
1811 struct netdev_queue __rcu *ingress_queue;
1812 #ifdef CONFIG_NETFILTER_INGRESS
1813 struct nf_hook_entry __rcu *nf_hooks_ingress;
1816 unsigned char broadcast[MAX_ADDR_LEN];
1817 #ifdef CONFIG_RFS_ACCEL
1818 struct cpu_rmap *rx_cpu_rmap;
1820 struct hlist_node index_hlist;
1823 * Cache lines mostly used on transmit path
1825 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1826 unsigned int num_tx_queues;
1827 unsigned int real_num_tx_queues;
1828 struct Qdisc *qdisc;
1829 #ifdef CONFIG_NET_SCHED
1830 DECLARE_HASHTABLE (qdisc_hash, 4);
1832 unsigned long tx_queue_len;
1833 spinlock_t tx_global_lock;
1837 struct xps_dev_maps __rcu *xps_maps;
1839 #ifdef CONFIG_NET_CLS_ACT
1840 struct tcf_proto __rcu *egress_cl_list;
1843 /* These may be needed for future network-power-down code. */
1844 struct timer_list watchdog_timer;
1846 int __percpu *pcpu_refcnt;
1847 struct list_head todo_list;
1849 struct list_head link_watch_list;
1851 enum { NETREG_UNINITIALIZED=0,
1852 NETREG_REGISTERED, /* completed register_netdevice */
1853 NETREG_UNREGISTERING, /* called unregister_netdevice */
1854 NETREG_UNREGISTERED, /* completed unregister todo */
1855 NETREG_RELEASED, /* called free_netdev */
1856 NETREG_DUMMY, /* dummy device for NAPI poll */
1862 RTNL_LINK_INITIALIZED,
1863 RTNL_LINK_INITIALIZING,
1864 } rtnl_link_state:16;
1866 void (*destructor)(struct net_device *dev);
1868 #ifdef CONFIG_NETPOLL
1869 struct netpoll_info __rcu *npinfo;
1872 possible_net_t nd_net;
1874 /* mid-layer private */
1877 struct pcpu_lstats __percpu *lstats;
1878 struct pcpu_sw_netstats __percpu *tstats;
1879 struct pcpu_dstats __percpu *dstats;
1880 struct pcpu_vstats __percpu *vstats;
1883 struct garp_port __rcu *garp_port;
1884 struct mrp_port __rcu *mrp_port;
1887 const struct attribute_group *sysfs_groups[4];
1888 const struct attribute_group *sysfs_rx_queue_group;
1890 const struct rtnl_link_ops *rtnl_link_ops;
1892 /* for setting kernel sock attribute on TCP connection setup */
1893 #define GSO_MAX_SIZE 65536
1894 unsigned int gso_max_size;
1895 #define GSO_MAX_SEGS 65535
1899 const struct dcbnl_rtnl_ops *dcbnl_ops;
1902 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1903 u8 prio_tc_map[TC_BITMASK + 1];
1905 #if IS_ENABLED(CONFIG_FCOE)
1906 unsigned int fcoe_ddp_xid;
1908 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1909 struct netprio_map __rcu *priomap;
1911 struct phy_device *phydev;
1912 struct lock_class_key *qdisc_tx_busylock;
1913 struct lock_class_key *qdisc_running_key;
1916 #define to_net_dev(d) container_of(d, struct net_device, dev)
1918 #define NETDEV_ALIGN 32
1921 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1923 return dev->prio_tc_map[prio & TC_BITMASK];
1927 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1929 if (tc >= dev->num_tc)
1932 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1937 void netdev_reset_tc(struct net_device *dev)
1940 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1941 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1945 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1947 if (tc >= dev->num_tc)
1950 dev->tc_to_txq[tc].count = count;
1951 dev->tc_to_txq[tc].offset = offset;
1956 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1958 if (num_tc > TC_MAX_QUEUE)
1961 dev->num_tc = num_tc;
1966 int netdev_get_num_tc(struct net_device *dev)
1972 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1975 return &dev->_tx[index];
1978 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1979 const struct sk_buff *skb)
1981 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1984 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1985 void (*f)(struct net_device *,
1986 struct netdev_queue *,
1992 for (i = 0; i < dev->num_tx_queues; i++)
1993 f(dev, &dev->_tx[i], arg);
1996 #define netdev_lockdep_set_classes(dev) \
1998 static struct lock_class_key qdisc_tx_busylock_key; \
1999 static struct lock_class_key qdisc_running_key; \
2000 static struct lock_class_key qdisc_xmit_lock_key; \
2001 static struct lock_class_key dev_addr_list_lock_key; \
2004 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
2005 (dev)->qdisc_running_key = &qdisc_running_key; \
2006 lockdep_set_class(&(dev)->addr_list_lock, \
2007 &dev_addr_list_lock_key); \
2008 for (i = 0; i < (dev)->num_tx_queues; i++) \
2009 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
2010 &qdisc_xmit_lock_key); \
2013 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2014 struct sk_buff *skb,
2017 /* returns the headroom that the master device needs to take in account
2018 * when forwarding to this dev
2020 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2022 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2025 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2027 if (dev->netdev_ops->ndo_set_rx_headroom)
2028 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2031 /* set the device rx headroom to the dev's default */
2032 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2034 netdev_set_rx_headroom(dev, -1);
2038 * Net namespace inlines
2041 struct net *dev_net(const struct net_device *dev)
2043 return read_pnet(&dev->nd_net);
2047 void dev_net_set(struct net_device *dev, struct net *net)
2049 write_pnet(&dev->nd_net, net);
2052 static inline bool netdev_uses_dsa(struct net_device *dev)
2054 #if IS_ENABLED(CONFIG_NET_DSA)
2055 if (dev->dsa_ptr != NULL)
2056 return dsa_uses_tagged_protocol(dev->dsa_ptr);
2062 * netdev_priv - access network device private data
2063 * @dev: network device
2065 * Get network device private data
2067 static inline void *netdev_priv(const struct net_device *dev)
2069 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2072 /* Set the sysfs physical device reference for the network logical device
2073 * if set prior to registration will cause a symlink during initialization.
2075 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
2077 /* Set the sysfs device type for the network logical device to allow
2078 * fine-grained identification of different network device types. For
2079 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2081 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2083 /* Default NAPI poll() weight
2084 * Device drivers are strongly advised to not use bigger value
2086 #define NAPI_POLL_WEIGHT 64
2089 * netif_napi_add - initialize a NAPI context
2090 * @dev: network device
2091 * @napi: NAPI context
2092 * @poll: polling function
2093 * @weight: default weight
2095 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2096 * *any* of the other NAPI-related functions.
2098 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2099 int (*poll)(struct napi_struct *, int), int weight);
2102 * netif_tx_napi_add - initialize a NAPI context
2103 * @dev: network device
2104 * @napi: NAPI context
2105 * @poll: polling function
2106 * @weight: default weight
2108 * This variant of netif_napi_add() should be used from drivers using NAPI
2109 * to exclusively poll a TX queue.
2110 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2112 static inline void netif_tx_napi_add(struct net_device *dev,
2113 struct napi_struct *napi,
2114 int (*poll)(struct napi_struct *, int),
2117 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2118 netif_napi_add(dev, napi, poll, weight);
2122 * netif_napi_del - remove a NAPI context
2123 * @napi: NAPI context
2125 * netif_napi_del() removes a NAPI context from the network device NAPI list
2127 void netif_napi_del(struct napi_struct *napi);
2129 struct napi_gro_cb {
2130 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2133 /* Length of frag0. */
2134 unsigned int frag0_len;
2136 /* This indicates where we are processing relative to skb->data. */
2139 /* This is non-zero if the packet cannot be merged with the new skb. */
2142 /* Save the IP ID here and check when we get to the transport layer */
2145 /* Number of segments aggregated. */
2148 /* Start offset for remote checksum offload */
2149 u16 gro_remcsum_start;
2151 /* jiffies when first packet was created/queued */
2154 /* Used in ipv6_gro_receive() and foo-over-udp */
2157 /* This is non-zero if the packet may be of the same flow. */
2160 /* Used in tunnel GRO receive */
2163 /* GRO checksum is valid */
2166 /* Number of checksums via CHECKSUM_UNNECESSARY */
2171 #define NAPI_GRO_FREE 1
2172 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2174 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2177 /* Used in GRE, set in fou/gue_gro_receive */
2180 /* Used to determine if flush_id can be ignored */
2183 /* Number of gro_receive callbacks this packet already went through */
2184 u8 recursion_counter:4;
2188 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2191 /* used in skb_gro_receive() slow path */
2192 struct sk_buff *last;
2195 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2197 #define GRO_RECURSION_LIMIT 15
2198 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2200 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2203 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2204 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2205 struct sk_buff **head,
2206 struct sk_buff *skb)
2208 if (unlikely(gro_recursion_inc_test(skb))) {
2209 NAPI_GRO_CB(skb)->flush |= 1;
2213 return cb(head, skb);
2216 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2218 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2220 struct sk_buff **head,
2221 struct sk_buff *skb)
2223 if (unlikely(gro_recursion_inc_test(skb))) {
2224 NAPI_GRO_CB(skb)->flush |= 1;
2228 return cb(sk, head, skb);
2231 struct packet_type {
2232 __be16 type; /* This is really htons(ether_type). */
2233 struct net_device *dev; /* NULL is wildcarded here */
2234 int (*func) (struct sk_buff *,
2235 struct net_device *,
2236 struct packet_type *,
2237 struct net_device *);
2238 bool (*id_match)(struct packet_type *ptype,
2240 struct net *af_packet_net;
2241 void *af_packet_priv;
2242 struct list_head list;
2245 struct offload_callbacks {
2246 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2247 netdev_features_t features);
2248 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2249 struct sk_buff *skb);
2250 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2253 struct packet_offload {
2254 __be16 type; /* This is really htons(ether_type). */
2256 struct offload_callbacks callbacks;
2257 struct list_head list;
2260 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2261 struct pcpu_sw_netstats {
2266 struct u64_stats_sync syncp;
2269 #define __netdev_alloc_pcpu_stats(type, gfp) \
2271 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2274 for_each_possible_cpu(__cpu) { \
2275 typeof(type) *stat; \
2276 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2277 u64_stats_init(&stat->syncp); \
2283 #define netdev_alloc_pcpu_stats(type) \
2284 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2286 enum netdev_lag_tx_type {
2287 NETDEV_LAG_TX_TYPE_UNKNOWN,
2288 NETDEV_LAG_TX_TYPE_RANDOM,
2289 NETDEV_LAG_TX_TYPE_BROADCAST,
2290 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2291 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2292 NETDEV_LAG_TX_TYPE_HASH,
2295 struct netdev_lag_upper_info {
2296 enum netdev_lag_tx_type tx_type;
2299 struct netdev_lag_lower_state_info {
2304 #include <linux/notifier.h>
2306 /* netdevice notifier chain. Please remember to update the rtnetlink
2307 * notification exclusion list in rtnetlink_event() when adding new
2310 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2311 #define NETDEV_DOWN 0x0002
2312 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2313 detected a hardware crash and restarted
2314 - we can use this eg to kick tcp sessions
2316 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2317 #define NETDEV_REGISTER 0x0005
2318 #define NETDEV_UNREGISTER 0x0006
2319 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2320 #define NETDEV_CHANGEADDR 0x0008
2321 #define NETDEV_GOING_DOWN 0x0009
2322 #define NETDEV_CHANGENAME 0x000A
2323 #define NETDEV_FEAT_CHANGE 0x000B
2324 #define NETDEV_BONDING_FAILOVER 0x000C
2325 #define NETDEV_PRE_UP 0x000D
2326 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2327 #define NETDEV_POST_TYPE_CHANGE 0x000F
2328 #define NETDEV_POST_INIT 0x0010
2329 #define NETDEV_UNREGISTER_FINAL 0x0011
2330 #define NETDEV_RELEASE 0x0012
2331 #define NETDEV_NOTIFY_PEERS 0x0013
2332 #define NETDEV_JOIN 0x0014
2333 #define NETDEV_CHANGEUPPER 0x0015
2334 #define NETDEV_RESEND_IGMP 0x0016
2335 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2336 #define NETDEV_CHANGEINFODATA 0x0018
2337 #define NETDEV_BONDING_INFO 0x0019
2338 #define NETDEV_PRECHANGEUPPER 0x001A
2339 #define NETDEV_CHANGELOWERSTATE 0x001B
2340 #define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
2341 #define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
2343 int register_netdevice_notifier(struct notifier_block *nb);
2344 int unregister_netdevice_notifier(struct notifier_block *nb);
2346 struct netdev_notifier_info {
2347 struct net_device *dev;
2350 struct netdev_notifier_info_ext {
2351 struct netdev_notifier_info info; /* must be first */
2357 struct netdev_notifier_change_info {
2358 struct netdev_notifier_info info; /* must be first */
2359 unsigned int flags_changed;
2362 struct netdev_notifier_changeupper_info {
2363 struct netdev_notifier_info info; /* must be first */
2364 struct net_device *upper_dev; /* new upper dev */
2365 bool master; /* is upper dev master */
2366 bool linking; /* is the notification for link or unlink */
2367 void *upper_info; /* upper dev info */
2370 struct netdev_notifier_changelowerstate_info {
2371 struct netdev_notifier_info info; /* must be first */
2372 void *lower_state_info; /* is lower dev state */
2375 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2376 struct net_device *dev)
2381 static inline struct net_device *
2382 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2387 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2390 extern rwlock_t dev_base_lock; /* Device list lock */
2392 #define for_each_netdev(net, d) \
2393 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2394 #define for_each_netdev_reverse(net, d) \
2395 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2396 #define for_each_netdev_rcu(net, d) \
2397 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2398 #define for_each_netdev_safe(net, d, n) \
2399 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2400 #define for_each_netdev_continue(net, d) \
2401 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2402 #define for_each_netdev_continue_rcu(net, d) \
2403 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2404 #define for_each_netdev_in_bond_rcu(bond, slave) \
2405 for_each_netdev_rcu(&init_net, slave) \
2406 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2407 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2409 static inline struct net_device *next_net_device(struct net_device *dev)
2411 struct list_head *lh;
2415 lh = dev->dev_list.next;
2416 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2419 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2421 struct list_head *lh;
2425 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2426 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2429 static inline struct net_device *first_net_device(struct net *net)
2431 return list_empty(&net->dev_base_head) ? NULL :
2432 net_device_entry(net->dev_base_head.next);
2435 static inline struct net_device *first_net_device_rcu(struct net *net)
2437 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2439 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2442 int netdev_boot_setup_check(struct net_device *dev);
2443 unsigned long netdev_boot_base(const char *prefix, int unit);
2444 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2445 const char *hwaddr);
2446 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2447 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2448 void dev_add_pack(struct packet_type *pt);
2449 void dev_remove_pack(struct packet_type *pt);
2450 void __dev_remove_pack(struct packet_type *pt);
2451 void dev_add_offload(struct packet_offload *po);
2452 void dev_remove_offload(struct packet_offload *po);
2454 int dev_get_iflink(const struct net_device *dev);
2455 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2456 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2457 unsigned short mask);
2458 struct net_device *dev_get_by_name(struct net *net, const char *name);
2459 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2460 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2461 int dev_alloc_name(struct net_device *dev, const char *name);
2462 int dev_open(struct net_device *dev);
2463 int dev_close(struct net_device *dev);
2464 int dev_close_many(struct list_head *head, bool unlink);
2465 void dev_disable_lro(struct net_device *dev);
2466 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2467 int dev_queue_xmit(struct sk_buff *skb);
2468 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2469 int register_netdevice(struct net_device *dev);
2470 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2471 void unregister_netdevice_many(struct list_head *head);
2472 static inline void unregister_netdevice(struct net_device *dev)
2474 unregister_netdevice_queue(dev, NULL);
2477 int netdev_refcnt_read(const struct net_device *dev);
2478 void free_netdev(struct net_device *dev);
2479 void netdev_freemem(struct net_device *dev);
2480 void synchronize_net(void);
2481 int init_dummy_netdev(struct net_device *dev);
2483 DECLARE_PER_CPU(int, xmit_recursion);
2484 #define XMIT_RECURSION_LIMIT 8
2486 static inline int dev_recursion_level(void)
2488 return this_cpu_read(xmit_recursion);
2491 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2492 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2493 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2494 int netdev_get_name(struct net *net, char *name, int ifindex);
2495 int dev_restart(struct net_device *dev);
2496 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2498 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2500 return NAPI_GRO_CB(skb)->data_offset;
2503 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2505 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2508 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2510 NAPI_GRO_CB(skb)->data_offset += len;
2513 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2514 unsigned int offset)
2516 return NAPI_GRO_CB(skb)->frag0 + offset;
2519 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2521 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2524 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2526 NAPI_GRO_CB(skb)->frag0 = NULL;
2527 NAPI_GRO_CB(skb)->frag0_len = 0;
2530 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2531 unsigned int offset)
2533 if (!pskb_may_pull(skb, hlen))
2536 skb_gro_frag0_invalidate(skb);
2537 return skb->data + offset;
2540 static inline void *skb_gro_network_header(struct sk_buff *skb)
2542 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2543 skb_network_offset(skb);
2546 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2547 const void *start, unsigned int len)
2549 if (NAPI_GRO_CB(skb)->csum_valid)
2550 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2551 csum_partial(start, len, 0));
2554 /* GRO checksum functions. These are logical equivalents of the normal
2555 * checksum functions (in skbuff.h) except that they operate on the GRO
2556 * offsets and fields in sk_buff.
2559 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2561 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2563 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2566 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2570 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2571 skb_checksum_start_offset(skb) <
2572 skb_gro_offset(skb)) &&
2573 !skb_at_gro_remcsum_start(skb) &&
2574 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2575 (!zero_okay || check));
2578 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2581 if (NAPI_GRO_CB(skb)->csum_valid &&
2582 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2585 NAPI_GRO_CB(skb)->csum = psum;
2587 return __skb_gro_checksum_complete(skb);
2590 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2592 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2593 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2594 NAPI_GRO_CB(skb)->csum_cnt--;
2596 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2597 * verified a new top level checksum or an encapsulated one
2598 * during GRO. This saves work if we fallback to normal path.
2600 __skb_incr_checksum_unnecessary(skb);
2604 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2607 __sum16 __ret = 0; \
2608 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2609 __ret = __skb_gro_checksum_validate_complete(skb, \
2610 compute_pseudo(skb, proto)); \
2612 __skb_mark_checksum_bad(skb); \
2614 skb_gro_incr_csum_unnecessary(skb); \
2618 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2619 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2621 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2623 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2625 #define skb_gro_checksum_simple_validate(skb) \
2626 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2628 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2630 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2631 !NAPI_GRO_CB(skb)->csum_valid);
2634 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2635 __sum16 check, __wsum pseudo)
2637 NAPI_GRO_CB(skb)->csum = ~pseudo;
2638 NAPI_GRO_CB(skb)->csum_valid = 1;
2641 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2643 if (__skb_gro_checksum_convert_check(skb)) \
2644 __skb_gro_checksum_convert(skb, check, \
2645 compute_pseudo(skb, proto)); \
2648 struct gro_remcsum {
2653 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2659 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2660 unsigned int off, size_t hdrlen,
2661 int start, int offset,
2662 struct gro_remcsum *grc,
2666 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2668 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2671 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2675 ptr = skb_gro_header_fast(skb, off);
2676 if (skb_gro_header_hard(skb, off + plen)) {
2677 ptr = skb_gro_header_slow(skb, off + plen, off);
2682 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2685 /* Adjust skb->csum since we changed the packet */
2686 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2688 grc->offset = off + hdrlen + offset;
2694 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2695 struct gro_remcsum *grc)
2698 size_t plen = grc->offset + sizeof(u16);
2703 ptr = skb_gro_header_fast(skb, grc->offset);
2704 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2705 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2710 remcsum_unadjust((__sum16 *)ptr, grc->delta);
2713 struct skb_csum_offl_spec {
2727 bool __skb_csum_offload_chk(struct sk_buff *skb,
2728 const struct skb_csum_offl_spec *spec,
2729 bool *csum_encapped,
2732 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2733 const struct skb_csum_offl_spec *spec,
2734 bool *csum_encapped,
2737 if (skb->ip_summed != CHECKSUM_PARTIAL)
2740 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2743 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2744 const struct skb_csum_offl_spec *spec)
2748 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2751 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2753 static const struct skb_csum_offl_spec csum_offl_spec = {
2755 .ip_options_okay = 1,
2762 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2765 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2767 static const struct skb_csum_offl_spec csum_offl_spec = {
2769 .ip_options_okay = 1,
2775 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2778 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2779 unsigned short type,
2780 const void *daddr, const void *saddr,
2783 if (!dev->header_ops || !dev->header_ops->create)
2786 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2789 static inline int dev_parse_header(const struct sk_buff *skb,
2790 unsigned char *haddr)
2792 const struct net_device *dev = skb->dev;
2794 if (!dev->header_ops || !dev->header_ops->parse)
2796 return dev->header_ops->parse(skb, haddr);
2799 /* ll_header must have at least hard_header_len allocated */
2800 static inline bool dev_validate_header(const struct net_device *dev,
2801 char *ll_header, int len)
2803 if (likely(len >= dev->hard_header_len))
2805 if (len < dev->min_header_len)
2808 if (capable(CAP_SYS_RAWIO)) {
2809 memset(ll_header + len, 0, dev->hard_header_len - len);
2813 if (dev->header_ops && dev->header_ops->validate)
2814 return dev->header_ops->validate(ll_header, len);
2819 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2820 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2821 static inline int unregister_gifconf(unsigned int family)
2823 return register_gifconf(family, NULL);
2826 #ifdef CONFIG_NET_FLOW_LIMIT
2827 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2828 struct sd_flow_limit {
2830 unsigned int num_buckets;
2831 unsigned int history_head;
2832 u16 history[FLOW_LIMIT_HISTORY];
2836 extern int netdev_flow_limit_table_len;
2837 #endif /* CONFIG_NET_FLOW_LIMIT */
2840 * Incoming packets are placed on per-CPU queues
2842 struct softnet_data {
2843 struct list_head poll_list;
2844 struct sk_buff_head process_queue;
2847 unsigned int processed;
2848 unsigned int time_squeeze;
2849 unsigned int received_rps;
2851 struct softnet_data *rps_ipi_list;
2853 #ifdef CONFIG_NET_FLOW_LIMIT
2854 struct sd_flow_limit __rcu *flow_limit;
2856 struct Qdisc *output_queue;
2857 struct Qdisc **output_queue_tailp;
2858 struct sk_buff *completion_queue;
2861 /* input_queue_head should be written by cpu owning this struct,
2862 * and only read by other cpus. Worth using a cache line.
2864 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2866 /* Elements below can be accessed between CPUs for RPS/RFS */
2867 struct call_single_data csd ____cacheline_aligned_in_smp;
2868 struct softnet_data *rps_ipi_next;
2870 unsigned int input_queue_tail;
2872 unsigned int dropped;
2873 struct sk_buff_head input_pkt_queue;
2874 struct napi_struct backlog;
2878 static inline void input_queue_head_incr(struct softnet_data *sd)
2881 sd->input_queue_head++;
2885 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2886 unsigned int *qtail)
2889 *qtail = ++sd->input_queue_tail;
2893 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2895 void __netif_schedule(struct Qdisc *q);
2896 void netif_schedule_queue(struct netdev_queue *txq);
2898 static inline void netif_tx_schedule_all(struct net_device *dev)
2902 for (i = 0; i < dev->num_tx_queues; i++)
2903 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2906 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2908 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2912 * netif_start_queue - allow transmit
2913 * @dev: network device
2915 * Allow upper layers to call the device hard_start_xmit routine.
2917 static inline void netif_start_queue(struct net_device *dev)
2919 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2922 static inline void netif_tx_start_all_queues(struct net_device *dev)
2926 for (i = 0; i < dev->num_tx_queues; i++) {
2927 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2928 netif_tx_start_queue(txq);
2932 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2935 * netif_wake_queue - restart transmit
2936 * @dev: network device
2938 * Allow upper layers to call the device hard_start_xmit routine.
2939 * Used for flow control when transmit resources are available.
2941 static inline void netif_wake_queue(struct net_device *dev)
2943 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2946 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2950 for (i = 0; i < dev->num_tx_queues; i++) {
2951 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2952 netif_tx_wake_queue(txq);
2956 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2958 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2962 * netif_stop_queue - stop transmitted packets
2963 * @dev: network device
2965 * Stop upper layers calling the device hard_start_xmit routine.
2966 * Used for flow control when transmit resources are unavailable.
2968 static inline void netif_stop_queue(struct net_device *dev)
2970 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2973 void netif_tx_stop_all_queues(struct net_device *dev);
2975 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2977 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2981 * netif_queue_stopped - test if transmit queue is flowblocked
2982 * @dev: network device
2984 * Test if transmit queue on device is currently unable to send.
2986 static inline bool netif_queue_stopped(const struct net_device *dev)
2988 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2991 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2993 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2997 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2999 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3003 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3005 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3009 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3010 * @dev_queue: pointer to transmit queue
3012 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3013 * to give appropriate hint to the CPU.
3015 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3018 prefetchw(&dev_queue->dql.num_queued);
3023 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3024 * @dev_queue: pointer to transmit queue
3026 * BQL enabled drivers might use this helper in their TX completion path,
3027 * to give appropriate hint to the CPU.
3029 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3032 prefetchw(&dev_queue->dql.limit);
3036 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3040 dql_queued(&dev_queue->dql, bytes);
3042 if (likely(dql_avail(&dev_queue->dql) >= 0))
3045 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3048 * The XOFF flag must be set before checking the dql_avail below,
3049 * because in netdev_tx_completed_queue we update the dql_completed
3050 * before checking the XOFF flag.
3054 /* check again in case another CPU has just made room avail */
3055 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3056 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3061 * netdev_sent_queue - report the number of bytes queued to hardware
3062 * @dev: network device
3063 * @bytes: number of bytes queued to the hardware device queue
3065 * Report the number of bytes queued for sending/completion to the network
3066 * device hardware queue. @bytes should be a good approximation and should
3067 * exactly match netdev_completed_queue() @bytes
3069 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3071 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3074 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3075 unsigned int pkts, unsigned int bytes)
3078 if (unlikely(!bytes))
3081 dql_completed(&dev_queue->dql, bytes);
3084 * Without the memory barrier there is a small possiblity that
3085 * netdev_tx_sent_queue will miss the update and cause the queue to
3086 * be stopped forever
3090 if (dql_avail(&dev_queue->dql) < 0)
3093 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3094 netif_schedule_queue(dev_queue);
3099 * netdev_completed_queue - report bytes and packets completed by device
3100 * @dev: network device
3101 * @pkts: actual number of packets sent over the medium
3102 * @bytes: actual number of bytes sent over the medium
3104 * Report the number of bytes and packets transmitted by the network device
3105 * hardware queue over the physical medium, @bytes must exactly match the
3106 * @bytes amount passed to netdev_sent_queue()
3108 static inline void netdev_completed_queue(struct net_device *dev,
3109 unsigned int pkts, unsigned int bytes)
3111 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3114 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3117 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3123 * netdev_reset_queue - reset the packets and bytes count of a network device
3124 * @dev_queue: network device
3126 * Reset the bytes and packet count of a network device and clear the
3127 * software flow control OFF bit for this network device
3129 static inline void netdev_reset_queue(struct net_device *dev_queue)
3131 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3135 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3136 * @dev: network device
3137 * @queue_index: given tx queue index
3139 * Returns 0 if given tx queue index >= number of device tx queues,
3140 * otherwise returns the originally passed tx queue index.
3142 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3144 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3145 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3146 dev->name, queue_index,
3147 dev->real_num_tx_queues);
3155 * netif_running - test if up
3156 * @dev: network device
3158 * Test if the device has been brought up.
3160 static inline bool netif_running(const struct net_device *dev)
3162 return test_bit(__LINK_STATE_START, &dev->state);
3166 * Routines to manage the subqueues on a device. We only need start,
3167 * stop, and a check if it's stopped. All other device management is
3168 * done at the overall netdevice level.
3169 * Also test the device if we're multiqueue.
3173 * netif_start_subqueue - allow sending packets on subqueue
3174 * @dev: network device
3175 * @queue_index: sub queue index
3177 * Start individual transmit queue of a device with multiple transmit queues.
3179 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3181 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3183 netif_tx_start_queue(txq);
3187 * netif_stop_subqueue - stop sending packets on subqueue
3188 * @dev: network device
3189 * @queue_index: sub queue index
3191 * Stop individual transmit queue of a device with multiple transmit queues.
3193 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3195 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3196 netif_tx_stop_queue(txq);
3200 * netif_subqueue_stopped - test status of subqueue
3201 * @dev: network device
3202 * @queue_index: sub queue index
3204 * Check individual transmit queue of a device with multiple transmit queues.
3206 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3209 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3211 return netif_tx_queue_stopped(txq);
3214 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3215 struct sk_buff *skb)
3217 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3220 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3223 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3226 static inline int netif_set_xps_queue(struct net_device *dev,
3227 const struct cpumask *mask,
3234 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3235 unsigned int num_tx_queues);
3238 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3239 * as a distribution range limit for the returned value.
3241 static inline u16 skb_tx_hash(const struct net_device *dev,
3242 struct sk_buff *skb)
3244 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3248 * netif_is_multiqueue - test if device has multiple transmit queues
3249 * @dev: network device
3251 * Check if device has multiple transmit queues
3253 static inline bool netif_is_multiqueue(const struct net_device *dev)
3255 return dev->num_tx_queues > 1;
3258 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3261 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3263 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3271 static inline unsigned int get_netdev_rx_queue_index(
3272 struct netdev_rx_queue *queue)
3274 struct net_device *dev = queue->dev;
3275 int index = queue - dev->_rx;
3277 BUG_ON(index >= dev->num_rx_queues);
3282 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
3283 int netif_get_num_default_rss_queues(void);
3285 enum skb_free_reason {
3286 SKB_REASON_CONSUMED,
3290 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3291 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3294 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3295 * interrupt context or with hardware interrupts being disabled.
3296 * (in_irq() || irqs_disabled())
3298 * We provide four helpers that can be used in following contexts :
3300 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3301 * replacing kfree_skb(skb)
3303 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3304 * Typically used in place of consume_skb(skb) in TX completion path
3306 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3307 * replacing kfree_skb(skb)
3309 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3310 * and consumed a packet. Used in place of consume_skb(skb)
3312 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3314 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3317 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3319 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3322 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3324 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3327 static inline void dev_consume_skb_any(struct sk_buff *skb)
3329 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3332 int netif_rx(struct sk_buff *skb);
3333 int netif_rx_ni(struct sk_buff *skb);
3334 int netif_receive_skb(struct sk_buff *skb);
3335 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3336 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3337 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3338 gro_result_t napi_gro_frags(struct napi_struct *napi);
3339 struct packet_offload *gro_find_receive_by_type(__be16 type);
3340 struct packet_offload *gro_find_complete_by_type(__be16 type);
3342 static inline void napi_free_frags(struct napi_struct *napi)
3344 kfree_skb(napi->skb);
3348 bool netdev_is_rx_handler_busy(struct net_device *dev);
3349 int netdev_rx_handler_register(struct net_device *dev,
3350 rx_handler_func_t *rx_handler,
3351 void *rx_handler_data);
3352 void netdev_rx_handler_unregister(struct net_device *dev);
3354 bool dev_valid_name(const char *name);
3355 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3356 int dev_ethtool(struct net *net, struct ifreq *);
3357 unsigned int dev_get_flags(const struct net_device *);
3358 int __dev_change_flags(struct net_device *, unsigned int flags);
3359 int dev_change_flags(struct net_device *, unsigned int);
3360 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3361 unsigned int gchanges);
3362 int dev_change_name(struct net_device *, const char *);
3363 int dev_set_alias(struct net_device *, const char *, size_t);
3364 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3365 int dev_set_mtu(struct net_device *, int);
3366 void dev_set_group(struct net_device *, int);
3367 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3368 int dev_change_carrier(struct net_device *, bool new_carrier);
3369 int dev_get_phys_port_id(struct net_device *dev,
3370 struct netdev_phys_item_id *ppid);
3371 int dev_get_phys_port_name(struct net_device *dev,
3372 char *name, size_t len);
3373 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3374 int dev_change_xdp_fd(struct net_device *dev, int fd);
3375 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3376 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3377 struct netdev_queue *txq, int *ret);
3378 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3379 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3380 bool is_skb_forwardable(const struct net_device *dev,
3381 const struct sk_buff *skb);
3383 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3384 struct sk_buff *skb)
3386 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3387 unlikely(!is_skb_forwardable(dev, skb))) {
3388 atomic_long_inc(&dev->rx_dropped);
3393 skb_scrub_packet(skb, true);
3398 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3400 extern int netdev_budget;
3402 /* Called by rtnetlink.c:rtnl_unlock() */
3403 void netdev_run_todo(void);
3406 * dev_put - release reference to device
3407 * @dev: network device
3409 * Release reference to device to allow it to be freed.
3411 static inline void dev_put(struct net_device *dev)
3414 this_cpu_dec(*dev->pcpu_refcnt);
3418 * dev_hold - get reference to device
3419 * @dev: network device
3421 * Hold reference to device to keep it from being freed.
3423 static inline void dev_hold(struct net_device *dev)
3426 this_cpu_inc(*dev->pcpu_refcnt);
3429 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3430 * and _off may be called from IRQ context, but it is caller
3431 * who is responsible for serialization of these calls.
3433 * The name carrier is inappropriate, these functions should really be
3434 * called netif_lowerlayer_*() because they represent the state of any
3435 * kind of lower layer not just hardware media.
3438 void linkwatch_init_dev(struct net_device *dev);
3439 void linkwatch_fire_event(struct net_device *dev);
3440 void linkwatch_forget_dev(struct net_device *dev);
3443 * netif_carrier_ok - test if carrier present
3444 * @dev: network device
3446 * Check if carrier is present on device
3448 static inline bool netif_carrier_ok(const struct net_device *dev)
3450 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3453 unsigned long dev_trans_start(struct net_device *dev);
3455 void __netdev_watchdog_up(struct net_device *dev);
3457 void netif_carrier_on(struct net_device *dev);
3459 void netif_carrier_off(struct net_device *dev);
3462 * netif_dormant_on - mark device as dormant.
3463 * @dev: network device
3465 * Mark device as dormant (as per RFC2863).
3467 * The dormant state indicates that the relevant interface is not
3468 * actually in a condition to pass packets (i.e., it is not 'up') but is
3469 * in a "pending" state, waiting for some external event. For "on-
3470 * demand" interfaces, this new state identifies the situation where the
3471 * interface is waiting for events to place it in the up state.
3473 static inline void netif_dormant_on(struct net_device *dev)
3475 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3476 linkwatch_fire_event(dev);
3480 * netif_dormant_off - set device as not dormant.
3481 * @dev: network device
3483 * Device is not in dormant state.
3485 static inline void netif_dormant_off(struct net_device *dev)
3487 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3488 linkwatch_fire_event(dev);
3492 * netif_dormant - test if carrier present
3493 * @dev: network device
3495 * Check if carrier is present on device
3497 static inline bool netif_dormant(const struct net_device *dev)
3499 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3504 * netif_oper_up - test if device is operational
3505 * @dev: network device
3507 * Check if carrier is operational
3509 static inline bool netif_oper_up(const struct net_device *dev)
3511 return (dev->operstate == IF_OPER_UP ||
3512 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3516 * netif_device_present - is device available or removed
3517 * @dev: network device
3519 * Check if device has not been removed from system.
3521 static inline bool netif_device_present(struct net_device *dev)
3523 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3526 void netif_device_detach(struct net_device *dev);
3528 void netif_device_attach(struct net_device *dev);
3531 * Network interface message level settings
3535 NETIF_MSG_DRV = 0x0001,
3536 NETIF_MSG_PROBE = 0x0002,
3537 NETIF_MSG_LINK = 0x0004,
3538 NETIF_MSG_TIMER = 0x0008,
3539 NETIF_MSG_IFDOWN = 0x0010,
3540 NETIF_MSG_IFUP = 0x0020,
3541 NETIF_MSG_RX_ERR = 0x0040,
3542 NETIF_MSG_TX_ERR = 0x0080,
3543 NETIF_MSG_TX_QUEUED = 0x0100,
3544 NETIF_MSG_INTR = 0x0200,
3545 NETIF_MSG_TX_DONE = 0x0400,
3546 NETIF_MSG_RX_STATUS = 0x0800,
3547 NETIF_MSG_PKTDATA = 0x1000,
3548 NETIF_MSG_HW = 0x2000,
3549 NETIF_MSG_WOL = 0x4000,
3552 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3553 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3554 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3555 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3556 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3557 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3558 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3559 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3560 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3561 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3562 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3563 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3564 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3565 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3566 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3568 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3571 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3572 return default_msg_enable_bits;
3573 if (debug_value == 0) /* no output */
3575 /* set low N bits */
3576 return (1U << debug_value) - 1;
3579 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3581 spin_lock(&txq->_xmit_lock);
3582 txq->xmit_lock_owner = cpu;
3585 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3587 spin_lock_bh(&txq->_xmit_lock);
3588 txq->xmit_lock_owner = smp_processor_id();
3591 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3593 bool ok = spin_trylock(&txq->_xmit_lock);
3595 txq->xmit_lock_owner = smp_processor_id();
3599 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3601 txq->xmit_lock_owner = -1;
3602 spin_unlock(&txq->_xmit_lock);
3605 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3607 txq->xmit_lock_owner = -1;
3608 spin_unlock_bh(&txq->_xmit_lock);
3611 static inline void txq_trans_update(struct netdev_queue *txq)
3613 if (txq->xmit_lock_owner != -1)
3614 txq->trans_start = jiffies;
3617 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3618 static inline void netif_trans_update(struct net_device *dev)
3620 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3622 if (txq->trans_start != jiffies)
3623 txq->trans_start = jiffies;
3627 * netif_tx_lock - grab network device transmit lock
3628 * @dev: network device
3630 * Get network device transmit lock
3632 static inline void netif_tx_lock(struct net_device *dev)
3637 spin_lock(&dev->tx_global_lock);
3638 cpu = smp_processor_id();
3639 for (i = 0; i < dev->num_tx_queues; i++) {
3640 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3642 /* We are the only thread of execution doing a
3643 * freeze, but we have to grab the _xmit_lock in
3644 * order to synchronize with threads which are in
3645 * the ->hard_start_xmit() handler and already
3646 * checked the frozen bit.
3648 __netif_tx_lock(txq, cpu);
3649 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3650 __netif_tx_unlock(txq);
3654 static inline void netif_tx_lock_bh(struct net_device *dev)
3660 static inline void netif_tx_unlock(struct net_device *dev)
3664 for (i = 0; i < dev->num_tx_queues; i++) {
3665 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3667 /* No need to grab the _xmit_lock here. If the
3668 * queue is not stopped for another reason, we
3671 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3672 netif_schedule_queue(txq);
3674 spin_unlock(&dev->tx_global_lock);
3677 static inline void netif_tx_unlock_bh(struct net_device *dev)
3679 netif_tx_unlock(dev);
3683 #define HARD_TX_LOCK(dev, txq, cpu) { \
3684 if ((dev->features & NETIF_F_LLTX) == 0) { \
3685 __netif_tx_lock(txq, cpu); \
3689 #define HARD_TX_TRYLOCK(dev, txq) \
3690 (((dev->features & NETIF_F_LLTX) == 0) ? \
3691 __netif_tx_trylock(txq) : \
3694 #define HARD_TX_UNLOCK(dev, txq) { \
3695 if ((dev->features & NETIF_F_LLTX) == 0) { \
3696 __netif_tx_unlock(txq); \
3700 static inline void netif_tx_disable(struct net_device *dev)
3706 cpu = smp_processor_id();
3707 spin_lock(&dev->tx_global_lock);
3708 for (i = 0; i < dev->num_tx_queues; i++) {
3709 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3711 __netif_tx_lock(txq, cpu);
3712 netif_tx_stop_queue(txq);
3713 __netif_tx_unlock(txq);
3715 spin_unlock(&dev->tx_global_lock);
3719 static inline void netif_addr_lock(struct net_device *dev)
3721 spin_lock(&dev->addr_list_lock);
3724 static inline void netif_addr_lock_nested(struct net_device *dev)
3726 int subclass = SINGLE_DEPTH_NESTING;
3728 if (dev->netdev_ops->ndo_get_lock_subclass)
3729 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3731 spin_lock_nested(&dev->addr_list_lock, subclass);
3734 static inline void netif_addr_lock_bh(struct net_device *dev)
3736 spin_lock_bh(&dev->addr_list_lock);
3739 static inline void netif_addr_unlock(struct net_device *dev)
3741 spin_unlock(&dev->addr_list_lock);
3744 static inline void netif_addr_unlock_bh(struct net_device *dev)
3746 spin_unlock_bh(&dev->addr_list_lock);
3750 * dev_addrs walker. Should be used only for read access. Call with
3751 * rcu_read_lock held.
3753 #define for_each_dev_addr(dev, ha) \
3754 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3756 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3758 void ether_setup(struct net_device *dev);
3760 /* Support for loadable net-drivers */
3761 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3762 unsigned char name_assign_type,
3763 void (*setup)(struct net_device *),
3764 unsigned int txqs, unsigned int rxqs);
3765 int dev_get_valid_name(struct net *net, struct net_device *dev,
3768 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3769 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3771 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3772 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3775 int register_netdev(struct net_device *dev);
3776 void unregister_netdev(struct net_device *dev);
3778 /* General hardware address lists handling functions */
3779 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3780 struct netdev_hw_addr_list *from_list, int addr_len);
3781 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3782 struct netdev_hw_addr_list *from_list, int addr_len);
3783 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3784 struct net_device *dev,
3785 int (*sync)(struct net_device *, const unsigned char *),
3786 int (*unsync)(struct net_device *,
3787 const unsigned char *));
3788 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3789 struct net_device *dev,
3790 int (*unsync)(struct net_device *,
3791 const unsigned char *));
3792 void __hw_addr_init(struct netdev_hw_addr_list *list);
3794 /* Functions used for device addresses handling */
3795 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3796 unsigned char addr_type);
3797 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3798 unsigned char addr_type);
3799 void dev_addr_flush(struct net_device *dev);
3800 int dev_addr_init(struct net_device *dev);
3802 /* Functions used for unicast addresses handling */
3803 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3804 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3805 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3806 int dev_uc_sync(struct net_device *to, struct net_device *from);
3807 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3808 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3809 void dev_uc_flush(struct net_device *dev);
3810 void dev_uc_init(struct net_device *dev);
3813 * __dev_uc_sync - Synchonize device's unicast list
3814 * @dev: device to sync
3815 * @sync: function to call if address should be added
3816 * @unsync: function to call if address should be removed
3818 * Add newly added addresses to the interface, and release
3819 * addresses that have been deleted.
3821 static inline int __dev_uc_sync(struct net_device *dev,
3822 int (*sync)(struct net_device *,
3823 const unsigned char *),
3824 int (*unsync)(struct net_device *,
3825 const unsigned char *))
3827 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3831 * __dev_uc_unsync - Remove synchronized addresses from device
3832 * @dev: device to sync
3833 * @unsync: function to call if address should be removed
3835 * Remove all addresses that were added to the device by dev_uc_sync().
3837 static inline void __dev_uc_unsync(struct net_device *dev,
3838 int (*unsync)(struct net_device *,
3839 const unsigned char *))
3841 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3844 /* Functions used for multicast addresses handling */
3845 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3846 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3847 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3848 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3849 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3850 int dev_mc_sync(struct net_device *to, struct net_device *from);
3851 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3852 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3853 void dev_mc_flush(struct net_device *dev);
3854 void dev_mc_init(struct net_device *dev);
3857 * __dev_mc_sync - Synchonize device's multicast list
3858 * @dev: device to sync
3859 * @sync: function to call if address should be added
3860 * @unsync: function to call if address should be removed
3862 * Add newly added addresses to the interface, and release
3863 * addresses that have been deleted.
3865 static inline int __dev_mc_sync(struct net_device *dev,
3866 int (*sync)(struct net_device *,
3867 const unsigned char *),
3868 int (*unsync)(struct net_device *,
3869 const unsigned char *))
3871 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3875 * __dev_mc_unsync - Remove synchronized addresses from device
3876 * @dev: device to sync
3877 * @unsync: function to call if address should be removed
3879 * Remove all addresses that were added to the device by dev_mc_sync().
3881 static inline void __dev_mc_unsync(struct net_device *dev,
3882 int (*unsync)(struct net_device *,
3883 const unsigned char *))
3885 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3888 /* Functions used for secondary unicast and multicast support */
3889 void dev_set_rx_mode(struct net_device *dev);
3890 void __dev_set_rx_mode(struct net_device *dev);
3891 int dev_set_promiscuity(struct net_device *dev, int inc);
3892 int dev_set_allmulti(struct net_device *dev, int inc);
3893 void netdev_state_change(struct net_device *dev);
3894 void netdev_notify_peers(struct net_device *dev);
3895 void netdev_features_change(struct net_device *dev);
3896 /* Load a device via the kmod */
3897 void dev_load(struct net *net, const char *name);
3898 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3899 struct rtnl_link_stats64 *storage);
3900 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3901 const struct net_device_stats *netdev_stats);
3903 extern int netdev_max_backlog;
3904 extern int netdev_tstamp_prequeue;
3905 extern int weight_p;
3907 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3908 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3909 struct list_head **iter);
3910 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3911 struct list_head **iter);
3913 /* iterate through upper list, must be called under RCU read lock */
3914 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3915 for (iter = &(dev)->adj_list.upper, \
3916 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3918 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3920 /* iterate through upper list, must be called under RCU read lock */
3921 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3922 for (iter = &(dev)->all_adj_list.upper, \
3923 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3925 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3927 bool netdev_has_any_upper_dev(struct net_device *dev);
3929 void *netdev_lower_get_next_private(struct net_device *dev,
3930 struct list_head **iter);
3931 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3932 struct list_head **iter);
3934 #define netdev_for_each_lower_private(dev, priv, iter) \
3935 for (iter = (dev)->adj_list.lower.next, \
3936 priv = netdev_lower_get_next_private(dev, &(iter)); \
3938 priv = netdev_lower_get_next_private(dev, &(iter)))
3940 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3941 for (iter = &(dev)->adj_list.lower, \
3942 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3944 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3946 void *netdev_lower_get_next(struct net_device *dev,
3947 struct list_head **iter);
3949 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3950 for (iter = (dev)->adj_list.lower.next, \
3951 ldev = netdev_lower_get_next(dev, &(iter)); \
3953 ldev = netdev_lower_get_next(dev, &(iter)))
3955 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3956 struct list_head **iter);
3957 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3958 struct list_head **iter);
3960 #define netdev_for_each_all_lower_dev(dev, ldev, iter) \
3961 for (iter = (dev)->all_adj_list.lower.next, \
3962 ldev = netdev_all_lower_get_next(dev, &(iter)); \
3964 ldev = netdev_all_lower_get_next(dev, &(iter)))
3966 #define netdev_for_each_all_lower_dev_rcu(dev, ldev, iter) \
3967 for (iter = &(dev)->all_adj_list.lower, \
3968 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)); \
3970 ldev = netdev_all_lower_get_next_rcu(dev, &(iter)))
3972 void *netdev_adjacent_get_private(struct list_head *adj_list);
3973 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3974 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3975 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3976 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3977 int netdev_master_upper_dev_link(struct net_device *dev,
3978 struct net_device *upper_dev,
3979 void *upper_priv, void *upper_info);
3980 void netdev_upper_dev_unlink(struct net_device *dev,
3981 struct net_device *upper_dev);
3982 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3983 void *netdev_lower_dev_get_private(struct net_device *dev,
3984 struct net_device *lower_dev);
3985 void netdev_lower_state_changed(struct net_device *lower_dev,
3986 void *lower_state_info);
3987 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3988 struct neighbour *n);
3989 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3990 struct neighbour *n);
3992 /* RSS keys are 40 or 52 bytes long */
3993 #define NETDEV_RSS_KEY_LEN 52
3994 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3995 void netdev_rss_key_fill(void *buffer, size_t len);
3997 int dev_get_nest_level(struct net_device *dev);
3998 int skb_checksum_help(struct sk_buff *skb);
3999 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4000 netdev_features_t features, bool tx_path);
4001 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4002 netdev_features_t features);
4004 struct netdev_bonding_info {
4009 struct netdev_notifier_bonding_info {
4010 struct netdev_notifier_info info; /* must be first */
4011 struct netdev_bonding_info bonding_info;
4014 void netdev_bonding_info_change(struct net_device *dev,
4015 struct netdev_bonding_info *bonding_info);
4018 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4020 return __skb_gso_segment(skb, features, true);
4022 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4024 static inline bool can_checksum_protocol(netdev_features_t features,
4027 if (protocol == htons(ETH_P_FCOE))
4028 return !!(features & NETIF_F_FCOE_CRC);
4030 /* Assume this is an IP checksum (not SCTP CRC) */
4032 if (features & NETIF_F_HW_CSUM) {
4033 /* Can checksum everything */
4038 case htons(ETH_P_IP):
4039 return !!(features & NETIF_F_IP_CSUM);
4040 case htons(ETH_P_IPV6):
4041 return !!(features & NETIF_F_IPV6_CSUM);
4047 /* Map an ethertype into IP protocol if possible */
4048 static inline int eproto_to_ipproto(int eproto)
4051 case htons(ETH_P_IP):
4053 case htons(ETH_P_IPV6):
4054 return IPPROTO_IPV6;
4061 void netdev_rx_csum_fault(struct net_device *dev);
4063 static inline void netdev_rx_csum_fault(struct net_device *dev)
4067 /* rx skb timestamps */
4068 void net_enable_timestamp(void);
4069 void net_disable_timestamp(void);
4071 #ifdef CONFIG_PROC_FS
4072 int __init dev_proc_init(void);
4074 #define dev_proc_init() 0
4077 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4078 struct sk_buff *skb, struct net_device *dev,
4081 skb->xmit_more = more ? 1 : 0;
4082 return ops->ndo_start_xmit(skb, dev);
4085 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4086 struct netdev_queue *txq, bool more)
4088 const struct net_device_ops *ops = dev->netdev_ops;
4091 rc = __netdev_start_xmit(ops, skb, dev, more);
4092 if (rc == NETDEV_TX_OK)
4093 txq_trans_update(txq);
4098 int netdev_class_create_file_ns(struct class_attribute *class_attr,
4100 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4103 static inline int netdev_class_create_file(struct class_attribute *class_attr)
4105 return netdev_class_create_file_ns(class_attr, NULL);
4108 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4110 netdev_class_remove_file_ns(class_attr, NULL);
4113 extern struct kobj_ns_type_operations net_ns_type_operations;
4115 const char *netdev_drivername(const struct net_device *dev);
4117 void linkwatch_run_queue(void);
4119 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4120 netdev_features_t f2)
4122 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4123 if (f1 & NETIF_F_HW_CSUM)
4124 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4126 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4132 static inline netdev_features_t netdev_get_wanted_features(
4133 struct net_device *dev)
4135 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4137 netdev_features_t netdev_increment_features(netdev_features_t all,
4138 netdev_features_t one, netdev_features_t mask);
4140 /* Allow TSO being used on stacked device :
4141 * Performing the GSO segmentation before last device
4142 * is a performance improvement.
4144 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4145 netdev_features_t mask)
4147 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4150 int __netdev_update_features(struct net_device *dev);
4151 void netdev_update_features(struct net_device *dev);
4152 void netdev_change_features(struct net_device *dev);
4154 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4155 struct net_device *dev);
4157 netdev_features_t passthru_features_check(struct sk_buff *skb,
4158 struct net_device *dev,
4159 netdev_features_t features);
4160 netdev_features_t netif_skb_features(struct sk_buff *skb);
4162 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4164 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4166 /* check flags correspondence */
4167 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4168 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4169 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4170 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4171 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4172 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4173 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4174 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4175 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4176 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4177 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4178 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4179 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4180 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4181 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4182 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4184 return (features & feature) == feature;
4187 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4189 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4190 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4193 static inline bool netif_needs_gso(struct sk_buff *skb,
4194 netdev_features_t features)
4196 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4197 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4198 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4201 static inline void netif_set_gso_max_size(struct net_device *dev,
4204 dev->gso_max_size = size;
4207 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4208 int pulled_hlen, u16 mac_offset,
4211 skb->protocol = protocol;
4212 skb->encapsulation = 1;
4213 skb_push(skb, pulled_hlen);
4214 skb_reset_transport_header(skb);
4215 skb->mac_header = mac_offset;
4216 skb->network_header = skb->mac_header + mac_len;
4217 skb->mac_len = mac_len;
4220 static inline bool netif_is_macsec(const struct net_device *dev)
4222 return dev->priv_flags & IFF_MACSEC;
4225 static inline bool netif_is_macvlan(const struct net_device *dev)
4227 return dev->priv_flags & IFF_MACVLAN;
4230 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4232 return dev->priv_flags & IFF_MACVLAN_PORT;
4235 static inline bool netif_is_ipvlan(const struct net_device *dev)
4237 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4240 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4242 return dev->priv_flags & IFF_IPVLAN_MASTER;
4245 static inline bool netif_is_bond_master(const struct net_device *dev)
4247 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4250 static inline bool netif_is_bond_slave(const struct net_device *dev)
4252 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4255 static inline bool netif_supports_nofcs(struct net_device *dev)
4257 return dev->priv_flags & IFF_SUPP_NOFCS;
4260 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4262 return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4265 static inline bool netif_is_l3_master(const struct net_device *dev)
4267 return dev->priv_flags & IFF_L3MDEV_MASTER;
4270 static inline bool netif_is_l3_slave(const struct net_device *dev)
4272 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4275 static inline bool netif_is_bridge_master(const struct net_device *dev)
4277 return dev->priv_flags & IFF_EBRIDGE;
4280 static inline bool netif_is_bridge_port(const struct net_device *dev)
4282 return dev->priv_flags & IFF_BRIDGE_PORT;
4285 static inline bool netif_is_ovs_master(const struct net_device *dev)
4287 return dev->priv_flags & IFF_OPENVSWITCH;
4290 static inline bool netif_is_team_master(const struct net_device *dev)
4292 return dev->priv_flags & IFF_TEAM;
4295 static inline bool netif_is_team_port(const struct net_device *dev)
4297 return dev->priv_flags & IFF_TEAM_PORT;
4300 static inline bool netif_is_lag_master(const struct net_device *dev)
4302 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4305 static inline bool netif_is_lag_port(const struct net_device *dev)
4307 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4310 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4312 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4315 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4316 static inline void netif_keep_dst(struct net_device *dev)
4318 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4321 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4322 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4324 /* TODO: reserve and use an additional IFF bit, if we get more users */
4325 return dev->priv_flags & IFF_MACSEC;
4328 extern struct pernet_operations __net_initdata loopback_net_ops;
4330 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4332 /* netdev_printk helpers, similar to dev_printk */
4334 static inline const char *netdev_name(const struct net_device *dev)
4336 if (!dev->name[0] || strchr(dev->name, '%'))
4337 return "(unnamed net_device)";
4341 static inline const char *netdev_reg_state(const struct net_device *dev)
4343 switch (dev->reg_state) {
4344 case NETREG_UNINITIALIZED: return " (uninitialized)";
4345 case NETREG_REGISTERED: return "";
4346 case NETREG_UNREGISTERING: return " (unregistering)";
4347 case NETREG_UNREGISTERED: return " (unregistered)";
4348 case NETREG_RELEASED: return " (released)";
4349 case NETREG_DUMMY: return " (dummy)";
4352 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4353 return " (unknown)";
4357 void netdev_printk(const char *level, const struct net_device *dev,
4358 const char *format, ...);
4360 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4362 void netdev_alert(const struct net_device *dev, const char *format, ...);
4364 void netdev_crit(const struct net_device *dev, const char *format, ...);
4366 void netdev_err(const struct net_device *dev, const char *format, ...);
4368 void netdev_warn(const struct net_device *dev, const char *format, ...);
4370 void netdev_notice(const struct net_device *dev, const char *format, ...);
4372 void netdev_info(const struct net_device *dev, const char *format, ...);
4374 #define MODULE_ALIAS_NETDEV(device) \
4375 MODULE_ALIAS("netdev-" device)
4377 #if defined(CONFIG_DYNAMIC_DEBUG)
4378 #define netdev_dbg(__dev, format, args...) \
4380 dynamic_netdev_dbg(__dev, format, ##args); \
4382 #elif defined(DEBUG)
4383 #define netdev_dbg(__dev, format, args...) \
4384 netdev_printk(KERN_DEBUG, __dev, format, ##args)
4386 #define netdev_dbg(__dev, format, args...) \
4389 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4393 #if defined(VERBOSE_DEBUG)
4394 #define netdev_vdbg netdev_dbg
4397 #define netdev_vdbg(dev, format, args...) \
4400 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4406 * netdev_WARN() acts like dev_printk(), but with the key difference
4407 * of using a WARN/WARN_ON to get the message out, including the
4408 * file/line information and a backtrace.
4410 #define netdev_WARN(dev, format, args...) \
4411 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4412 netdev_reg_state(dev), ##args)
4414 /* netif printk helpers, similar to netdev_printk */
4416 #define netif_printk(priv, type, level, dev, fmt, args...) \
4418 if (netif_msg_##type(priv)) \
4419 netdev_printk(level, (dev), fmt, ##args); \
4422 #define netif_level(level, priv, type, dev, fmt, args...) \
4424 if (netif_msg_##type(priv)) \
4425 netdev_##level(dev, fmt, ##args); \
4428 #define netif_emerg(priv, type, dev, fmt, args...) \
4429 netif_level(emerg, priv, type, dev, fmt, ##args)
4430 #define netif_alert(priv, type, dev, fmt, args...) \
4431 netif_level(alert, priv, type, dev, fmt, ##args)
4432 #define netif_crit(priv, type, dev, fmt, args...) \
4433 netif_level(crit, priv, type, dev, fmt, ##args)
4434 #define netif_err(priv, type, dev, fmt, args...) \
4435 netif_level(err, priv, type, dev, fmt, ##args)
4436 #define netif_warn(priv, type, dev, fmt, args...) \
4437 netif_level(warn, priv, type, dev, fmt, ##args)
4438 #define netif_notice(priv, type, dev, fmt, args...) \
4439 netif_level(notice, priv, type, dev, fmt, ##args)
4440 #define netif_info(priv, type, dev, fmt, args...) \
4441 netif_level(info, priv, type, dev, fmt, ##args)
4443 #if defined(CONFIG_DYNAMIC_DEBUG)
4444 #define netif_dbg(priv, type, netdev, format, args...) \
4446 if (netif_msg_##type(priv)) \
4447 dynamic_netdev_dbg(netdev, format, ##args); \
4449 #elif defined(DEBUG)
4450 #define netif_dbg(priv, type, dev, format, args...) \
4451 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4453 #define netif_dbg(priv, type, dev, format, args...) \
4456 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4461 #if defined(VERBOSE_DEBUG)
4462 #define netif_vdbg netif_dbg
4464 #define netif_vdbg(priv, type, dev, format, args...) \
4467 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4473 * The list of packet types we will receive (as opposed to discard)
4474 * and the routines to invoke.
4476 * Why 16. Because with 16 the only overlap we get on a hash of the
4477 * low nibble of the protocol value is RARP/SNAP/X.25.
4479 * NOTE: That is no longer true with the addition of VLAN tags. Not
4480 * sure which should go first, but I bet it won't make much
4481 * difference if we are running VLANs. The good news is that
4482 * this protocol won't be in the list unless compiled in, so
4483 * the average user (w/out VLANs) will not be adversely affected.
4499 #define PTYPE_HASH_SIZE (16)
4500 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4502 #endif /* _LINUX_NETDEVICE_H */