2 * Definitions for the 'struct sk_buff' memory handlers.
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/cache.h>
23 #include <asm/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/dmaengine.h>
31 #include <linux/hrtimer.h>
33 /* Don't change this without changing skb_csum_unnecessary! */
34 #define CHECKSUM_NONE 0
35 #define CHECKSUM_UNNECESSARY 1
36 #define CHECKSUM_COMPLETE 2
37 #define CHECKSUM_PARTIAL 3
39 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
41 #define SKB_WITH_OVERHEAD(X) \
42 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
43 #define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
45 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
48 /* A. Checksumming of received packets by device.
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
59 * COMPLETE: the most generic way. Device supplied checksum of _all_
60 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
62 * is able to produce some skb->csum, it MUST use COMPLETE,
65 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
72 * B. Checksumming on output.
74 * NONE: skb is checksummed by protocol or csum is not required.
76 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
77 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
89 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
91 * Any questions? No questions, good. --ANK
96 struct pipe_inode_info;
98 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
104 #ifdef CONFIG_BRIDGE_NETFILTER
105 struct nf_bridge_info {
107 struct net_device *physindev;
108 struct net_device *physoutdev;
110 unsigned long data[32 / sizeof(unsigned long)];
114 struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
125 /* To allow 64K frame to be packed as single skb without frag_list */
126 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
128 typedef struct skb_frag_struct skb_frag_t;
130 struct skb_frag_struct {
136 #define HAVE_HW_TIME_STAMP
139 * struct skb_shared_hwtstamps - hardware time stamps
140 * @hwtstamp: hardware time stamp transformed into duration
141 * since arbitrary point in time
142 * @syststamp: hwtstamp transformed to system time base
144 * Software time stamps generated by ktime_get_real() are stored in
145 * skb->tstamp. The relation between the different kinds of time
146 * stamps is as follows:
148 * syststamp and tstamp can be compared against each other in
149 * arbitrary combinations. The accuracy of a
150 * syststamp/tstamp/"syststamp from other device" comparison is
151 * limited by the accuracy of the transformation into system time
152 * base. This depends on the device driver and its underlying
155 * hwtstamps can only be compared against other hwtstamps from
158 * This structure is attached to packets as part of the
159 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
161 struct skb_shared_hwtstamps {
167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
172 * @flags: all shared_tx flags
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
177 union skb_shared_tx {
186 /* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
189 struct skb_shared_info {
190 unsigned short nr_frags;
191 unsigned short gso_size;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs;
194 unsigned short gso_type;
196 union skb_shared_tx tx_flags;
197 struct sk_buff *frag_list;
198 struct skb_shared_hwtstamps hwtstamps;
201 * Warning : all fields before dataref are cleared in __alloc_skb()
205 /* Intermediate layers must ensure that destructor_arg
206 * remains valid until skb destructor */
207 void * destructor_arg;
208 /* must be last field, see pskb_expand_head() */
209 skb_frag_t frags[MAX_SKB_FRAGS];
212 /* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
223 #define SKB_DATAREF_SHIFT 16
224 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
228 SKB_FCLONE_UNAVAILABLE,
234 SKB_GSO_TCPV4 = 1 << 0,
235 SKB_GSO_UDP = 1 << 1,
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY = 1 << 2,
240 /* This indicates the tcp segment has CWR set. */
241 SKB_GSO_TCP_ECN = 1 << 3,
243 SKB_GSO_TCPV6 = 1 << 4,
245 SKB_GSO_FCOE = 1 << 5,
248 #if BITS_PER_LONG > 32
249 #define NET_SKBUFF_DATA_USES_OFFSET 1
252 #ifdef NET_SKBUFF_DATA_USES_OFFSET
253 typedef unsigned int sk_buff_data_t;
255 typedef unsigned char *sk_buff_data_t;
259 * struct sk_buff - socket buffer
260 * @next: Next buffer in list
261 * @prev: Previous buffer in list
262 * @sk: Socket we are owned by
263 * @tstamp: Time we arrived
264 * @dev: Device we arrived on/are leaving by
265 * @transport_header: Transport layer header
266 * @network_header: Network layer header
267 * @mac_header: Link layer header
268 * @_skb_refdst: destination entry (with norefcount bit)
269 * @sp: the security path, used for xfrm
270 * @cb: Control buffer. Free for use by every layer. Put private vars here
271 * @len: Length of actual data
272 * @data_len: Data length
273 * @mac_len: Length of link layer header
274 * @hdr_len: writable header length of cloned skb
275 * @csum: Checksum (must include start/offset pair)
276 * @csum_start: Offset from skb->head where checksumming should start
277 * @csum_offset: Offset from csum_start where checksum should be stored
278 * @local_df: allow local fragmentation
279 * @cloned: Head may be cloned (check refcnt to be sure)
280 * @nohdr: Payload reference only, must not modify header
281 * @pkt_type: Packet class
282 * @fclone: skbuff clone status
283 * @ip_summed: Driver fed us an IP checksum
284 * @priority: Packet queueing priority
285 * @users: User count - see {datagram,tcp}.c
286 * @protocol: Packet protocol from driver
287 * @truesize: Buffer size
288 * @head: Head of buffer
289 * @data: Data head pointer
290 * @tail: Tail pointer
292 * @destructor: Destruct function
293 * @mark: Generic packet mark
294 * @nfct: Associated connection, if any
295 * @ipvs_property: skbuff is owned by ipvs
296 * @peeked: this packet has been seen already, so stats have been
297 * done for it, don't do them again
298 * @nf_trace: netfilter packet trace flag
299 * @nfctinfo: Relationship of this skb to the connection
300 * @nfct_reasm: netfilter conntrack re-assembly pointer
301 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
302 * @skb_iif: ifindex of device we arrived on
303 * @rxhash: the packet hash computed on receive
304 * @queue_mapping: Queue mapping for multiqueue devices
305 * @tc_index: Traffic control index
306 * @tc_verd: traffic control verdict
307 * @ndisc_nodetype: router type (from link layer)
308 * @dma_cookie: a cookie to one of several possible DMA operations
309 * done by skb DMA functions
310 * @secmark: security marking
311 * @vlan_tci: vlan tag control information
315 /* These two members must be first. */
316 struct sk_buff *next;
317 struct sk_buff *prev;
322 struct net_device *dev;
325 * This is the control buffer. It is free to use for every
326 * layer. Please put your private variables there. If you
327 * want to keep them across layers you have to do a skb_clone()
328 * first. This is owned by whoever has the skb queued ATM.
330 char cb[48] __aligned(8);
332 unsigned long _skb_refdst;
348 kmemcheck_bitfield_begin(flags1);
359 kmemcheck_bitfield_end(flags1);
362 void (*destructor)(struct sk_buff *skb);
363 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
364 struct nf_conntrack *nfct;
365 struct sk_buff *nfct_reasm;
367 #ifdef CONFIG_BRIDGE_NETFILTER
368 struct nf_bridge_info *nf_bridge;
372 #ifdef CONFIG_NET_SCHED
373 __u16 tc_index; /* traffic control index */
374 #ifdef CONFIG_NET_CLS_ACT
375 __u16 tc_verd; /* traffic control verdict */
381 kmemcheck_bitfield_begin(flags2);
382 __u16 queue_mapping:16;
383 #ifdef CONFIG_IPV6_NDISC_NODETYPE
384 __u8 ndisc_nodetype:2,
387 __u8 deliver_no_wcard:1;
389 kmemcheck_bitfield_end(flags2);
393 #ifdef CONFIG_NET_DMA
394 dma_cookie_t dma_cookie;
396 #ifdef CONFIG_NETWORK_SECMARK
406 sk_buff_data_t transport_header;
407 sk_buff_data_t network_header;
408 sk_buff_data_t mac_header;
409 /* These elements must be at the end, see alloc_skb() for details. */
414 unsigned int truesize;
420 * Handling routines are only of interest to the kernel
422 #include <linux/slab.h>
424 #include <asm/system.h>
427 * skb might have a dst pointer attached, refcounted or not.
428 * _skb_refdst low order bit is set if refcount was _not_ taken
430 #define SKB_DST_NOREF 1UL
431 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
434 * skb_dst - returns skb dst_entry
437 * Returns skb dst_entry, regardless of reference taken or not.
439 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
441 /* If refdst was not refcounted, check we still are in a
442 * rcu_read_lock section
444 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
445 !rcu_read_lock_held() &&
446 !rcu_read_lock_bh_held());
447 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
451 * skb_dst_set - sets skb dst
455 * Sets skb dst, assuming a reference was taken on dst and should
456 * be released by skb_dst_drop()
458 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
460 skb->_skb_refdst = (unsigned long)dst;
464 * skb_dst_set_noref - sets skb dst, without a reference
468 * Sets skb dst, assuming a reference was not taken on dst
469 * skb_dst_drop() should not dst_release() this dst
471 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
473 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
474 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
478 * skb_dst_is_noref - Test if skb dst isnt refcounted
481 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
483 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
486 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
488 return (struct rtable *)skb_dst(skb);
491 extern void kfree_skb(struct sk_buff *skb);
492 extern void consume_skb(struct sk_buff *skb);
493 extern void __kfree_skb(struct sk_buff *skb);
494 extern struct sk_buff *__alloc_skb(unsigned int size,
495 gfp_t priority, int fclone, int node);
496 static inline struct sk_buff *alloc_skb(unsigned int size,
499 return __alloc_skb(size, priority, 0, -1);
502 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
505 return __alloc_skb(size, priority, 1, -1);
508 extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
510 extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
511 extern struct sk_buff *skb_clone(struct sk_buff *skb,
513 extern struct sk_buff *skb_copy(const struct sk_buff *skb,
515 extern struct sk_buff *pskb_copy(struct sk_buff *skb,
517 extern int pskb_expand_head(struct sk_buff *skb,
518 int nhead, int ntail,
520 extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
521 unsigned int headroom);
522 extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
523 int newheadroom, int newtailroom,
525 extern int skb_to_sgvec(struct sk_buff *skb,
526 struct scatterlist *sg, int offset,
528 extern int skb_cow_data(struct sk_buff *skb, int tailbits,
529 struct sk_buff **trailer);
530 extern int skb_pad(struct sk_buff *skb, int pad);
531 #define dev_kfree_skb(a) consume_skb(a)
533 extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
534 int getfrag(void *from, char *to, int offset,
535 int len,int odd, struct sk_buff *skb),
536 void *from, int length);
538 struct skb_seq_state {
542 __u32 stepped_offset;
543 struct sk_buff *root_skb;
544 struct sk_buff *cur_skb;
548 extern void skb_prepare_seq_read(struct sk_buff *skb,
549 unsigned int from, unsigned int to,
550 struct skb_seq_state *st);
551 extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
552 struct skb_seq_state *st);
553 extern void skb_abort_seq_read(struct skb_seq_state *st);
555 extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
556 unsigned int to, struct ts_config *config,
557 struct ts_state *state);
559 #ifdef NET_SKBUFF_DATA_USES_OFFSET
560 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
562 return skb->head + skb->end;
565 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
572 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
574 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
576 return &skb_shinfo(skb)->hwtstamps;
579 static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
581 return &skb_shinfo(skb)->tx_flags;
585 * skb_queue_empty - check if a queue is empty
588 * Returns true if the queue is empty, false otherwise.
590 static inline int skb_queue_empty(const struct sk_buff_head *list)
592 return list->next == (struct sk_buff *)list;
596 * skb_queue_is_last - check if skb is the last entry in the queue
600 * Returns true if @skb is the last buffer on the list.
602 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
603 const struct sk_buff *skb)
605 return (skb->next == (struct sk_buff *) list);
609 * skb_queue_is_first - check if skb is the first entry in the queue
613 * Returns true if @skb is the first buffer on the list.
615 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
616 const struct sk_buff *skb)
618 return (skb->prev == (struct sk_buff *) list);
622 * skb_queue_next - return the next packet in the queue
624 * @skb: current buffer
626 * Return the next packet in @list after @skb. It is only valid to
627 * call this if skb_queue_is_last() evaluates to false.
629 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
630 const struct sk_buff *skb)
632 /* This BUG_ON may seem severe, but if we just return then we
633 * are going to dereference garbage.
635 BUG_ON(skb_queue_is_last(list, skb));
640 * skb_queue_prev - return the prev packet in the queue
642 * @skb: current buffer
644 * Return the prev packet in @list before @skb. It is only valid to
645 * call this if skb_queue_is_first() evaluates to false.
647 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
648 const struct sk_buff *skb)
650 /* This BUG_ON may seem severe, but if we just return then we
651 * are going to dereference garbage.
653 BUG_ON(skb_queue_is_first(list, skb));
658 * skb_get - reference buffer
659 * @skb: buffer to reference
661 * Makes another reference to a socket buffer and returns a pointer
664 static inline struct sk_buff *skb_get(struct sk_buff *skb)
666 atomic_inc(&skb->users);
671 * If users == 1, we are the only owner and are can avoid redundant
676 * skb_cloned - is the buffer a clone
677 * @skb: buffer to check
679 * Returns true if the buffer was generated with skb_clone() and is
680 * one of multiple shared copies of the buffer. Cloned buffers are
681 * shared data so must not be written to under normal circumstances.
683 static inline int skb_cloned(const struct sk_buff *skb)
685 return skb->cloned &&
686 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
690 * skb_header_cloned - is the header a clone
691 * @skb: buffer to check
693 * Returns true if modifying the header part of the buffer requires
694 * the data to be copied.
696 static inline int skb_header_cloned(const struct sk_buff *skb)
703 dataref = atomic_read(&skb_shinfo(skb)->dataref);
704 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
709 * skb_header_release - release reference to header
710 * @skb: buffer to operate on
712 * Drop a reference to the header part of the buffer. This is done
713 * by acquiring a payload reference. You must not read from the header
714 * part of skb->data after this.
716 static inline void skb_header_release(struct sk_buff *skb)
720 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
724 * skb_shared - is the buffer shared
725 * @skb: buffer to check
727 * Returns true if more than one person has a reference to this
730 static inline int skb_shared(const struct sk_buff *skb)
732 return atomic_read(&skb->users) != 1;
736 * skb_share_check - check if buffer is shared and if so clone it
737 * @skb: buffer to check
738 * @pri: priority for memory allocation
740 * If the buffer is shared the buffer is cloned and the old copy
741 * drops a reference. A new clone with a single reference is returned.
742 * If the buffer is not shared the original buffer is returned. When
743 * being called from interrupt status or with spinlocks held pri must
746 * NULL is returned on a memory allocation failure.
748 static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
751 might_sleep_if(pri & __GFP_WAIT);
752 if (skb_shared(skb)) {
753 struct sk_buff *nskb = skb_clone(skb, pri);
761 * Copy shared buffers into a new sk_buff. We effectively do COW on
762 * packets to handle cases where we have a local reader and forward
763 * and a couple of other messy ones. The normal one is tcpdumping
764 * a packet thats being forwarded.
768 * skb_unshare - make a copy of a shared buffer
769 * @skb: buffer to check
770 * @pri: priority for memory allocation
772 * If the socket buffer is a clone then this function creates a new
773 * copy of the data, drops a reference count on the old copy and returns
774 * the new copy with the reference count at 1. If the buffer is not a clone
775 * the original buffer is returned. When called with a spinlock held or
776 * from interrupt state @pri must be %GFP_ATOMIC
778 * %NULL is returned on a memory allocation failure.
780 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
783 might_sleep_if(pri & __GFP_WAIT);
784 if (skb_cloned(skb)) {
785 struct sk_buff *nskb = skb_copy(skb, pri);
786 kfree_skb(skb); /* Free our shared copy */
793 * skb_peek - peek at the head of an &sk_buff_head
794 * @list_: list to peek at
796 * Peek an &sk_buff. Unlike most other operations you _MUST_
797 * be careful with this one. A peek leaves the buffer on the
798 * list and someone else may run off with it. You must hold
799 * the appropriate locks or have a private queue to do this.
801 * Returns %NULL for an empty list or a pointer to the head element.
802 * The reference count is not incremented and the reference is therefore
803 * volatile. Use with caution.
805 static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
807 struct sk_buff *list = ((struct sk_buff *)list_)->next;
808 if (list == (struct sk_buff *)list_)
814 * skb_peek_tail - peek at the tail of an &sk_buff_head
815 * @list_: list to peek at
817 * Peek an &sk_buff. Unlike most other operations you _MUST_
818 * be careful with this one. A peek leaves the buffer on the
819 * list and someone else may run off with it. You must hold
820 * the appropriate locks or have a private queue to do this.
822 * Returns %NULL for an empty list or a pointer to the tail element.
823 * The reference count is not incremented and the reference is therefore
824 * volatile. Use with caution.
826 static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
828 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
829 if (list == (struct sk_buff *)list_)
835 * skb_queue_len - get queue length
836 * @list_: list to measure
838 * Return the length of an &sk_buff queue.
840 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
846 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
847 * @list: queue to initialize
849 * This initializes only the list and queue length aspects of
850 * an sk_buff_head object. This allows to initialize the list
851 * aspects of an sk_buff_head without reinitializing things like
852 * the spinlock. It can also be used for on-stack sk_buff_head
853 * objects where the spinlock is known to not be used.
855 static inline void __skb_queue_head_init(struct sk_buff_head *list)
857 list->prev = list->next = (struct sk_buff *)list;
862 * This function creates a split out lock class for each invocation;
863 * this is needed for now since a whole lot of users of the skb-queue
864 * infrastructure in drivers have different locking usage (in hardirq)
865 * than the networking core (in softirq only). In the long run either the
866 * network layer or drivers should need annotation to consolidate the
867 * main types of usage into 3 classes.
869 static inline void skb_queue_head_init(struct sk_buff_head *list)
871 spin_lock_init(&list->lock);
872 __skb_queue_head_init(list);
875 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
876 struct lock_class_key *class)
878 skb_queue_head_init(list);
879 lockdep_set_class(&list->lock, class);
883 * Insert an sk_buff on a list.
885 * The "__skb_xxxx()" functions are the non-atomic ones that
886 * can only be called with interrupts disabled.
888 extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
889 static inline void __skb_insert(struct sk_buff *newsk,
890 struct sk_buff *prev, struct sk_buff *next,
891 struct sk_buff_head *list)
895 next->prev = prev->next = newsk;
899 static inline void __skb_queue_splice(const struct sk_buff_head *list,
900 struct sk_buff *prev,
901 struct sk_buff *next)
903 struct sk_buff *first = list->next;
904 struct sk_buff *last = list->prev;
914 * skb_queue_splice - join two skb lists, this is designed for stacks
915 * @list: the new list to add
916 * @head: the place to add it in the first list
918 static inline void skb_queue_splice(const struct sk_buff_head *list,
919 struct sk_buff_head *head)
921 if (!skb_queue_empty(list)) {
922 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
923 head->qlen += list->qlen;
928 * skb_queue_splice - join two skb lists and reinitialise the emptied list
929 * @list: the new list to add
930 * @head: the place to add it in the first list
932 * The list at @list is reinitialised
934 static inline void skb_queue_splice_init(struct sk_buff_head *list,
935 struct sk_buff_head *head)
937 if (!skb_queue_empty(list)) {
938 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
939 head->qlen += list->qlen;
940 __skb_queue_head_init(list);
945 * skb_queue_splice_tail - join two skb lists, each list being a queue
946 * @list: the new list to add
947 * @head: the place to add it in the first list
949 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
950 struct sk_buff_head *head)
952 if (!skb_queue_empty(list)) {
953 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
954 head->qlen += list->qlen;
959 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
960 * @list: the new list to add
961 * @head: the place to add it in the first list
963 * Each of the lists is a queue.
964 * The list at @list is reinitialised
966 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
967 struct sk_buff_head *head)
969 if (!skb_queue_empty(list)) {
970 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
971 head->qlen += list->qlen;
972 __skb_queue_head_init(list);
977 * __skb_queue_after - queue a buffer at the list head
979 * @prev: place after this buffer
980 * @newsk: buffer to queue
982 * Queue a buffer int the middle of a list. This function takes no locks
983 * and you must therefore hold required locks before calling it.
985 * A buffer cannot be placed on two lists at the same time.
987 static inline void __skb_queue_after(struct sk_buff_head *list,
988 struct sk_buff *prev,
989 struct sk_buff *newsk)
991 __skb_insert(newsk, prev, prev->next, list);
994 extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
995 struct sk_buff_head *list);
997 static inline void __skb_queue_before(struct sk_buff_head *list,
998 struct sk_buff *next,
999 struct sk_buff *newsk)
1001 __skb_insert(newsk, next->prev, next, list);
1005 * __skb_queue_head - queue a buffer at the list head
1006 * @list: list to use
1007 * @newsk: buffer to queue
1009 * Queue a buffer at the start of a list. This function takes no locks
1010 * and you must therefore hold required locks before calling it.
1012 * A buffer cannot be placed on two lists at the same time.
1014 extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1015 static inline void __skb_queue_head(struct sk_buff_head *list,
1016 struct sk_buff *newsk)
1018 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1022 * __skb_queue_tail - queue a buffer at the list tail
1023 * @list: list to use
1024 * @newsk: buffer to queue
1026 * Queue a buffer at the end of a list. This function takes no locks
1027 * and you must therefore hold required locks before calling it.
1029 * A buffer cannot be placed on two lists at the same time.
1031 extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1032 static inline void __skb_queue_tail(struct sk_buff_head *list,
1033 struct sk_buff *newsk)
1035 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1039 * remove sk_buff from list. _Must_ be called atomically, and with
1042 extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1043 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1045 struct sk_buff *next, *prev;
1050 skb->next = skb->prev = NULL;
1056 * __skb_dequeue - remove from the head of the queue
1057 * @list: list to dequeue from
1059 * Remove the head of the list. This function does not take any locks
1060 * so must be used with appropriate locks held only. The head item is
1061 * returned or %NULL if the list is empty.
1063 extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1064 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1066 struct sk_buff *skb = skb_peek(list);
1068 __skb_unlink(skb, list);
1073 * __skb_dequeue_tail - remove from the tail of the queue
1074 * @list: list to dequeue from
1076 * Remove the tail of the list. This function does not take any locks
1077 * so must be used with appropriate locks held only. The tail item is
1078 * returned or %NULL if the list is empty.
1080 extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1081 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1083 struct sk_buff *skb = skb_peek_tail(list);
1085 __skb_unlink(skb, list);
1090 static inline int skb_is_nonlinear(const struct sk_buff *skb)
1092 return skb->data_len;
1095 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1097 return skb->len - skb->data_len;
1100 static inline int skb_pagelen(const struct sk_buff *skb)
1104 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1105 len += skb_shinfo(skb)->frags[i].size;
1106 return len + skb_headlen(skb);
1109 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1110 struct page *page, int off, int size)
1112 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1115 frag->page_offset = off;
1117 skb_shinfo(skb)->nr_frags = i + 1;
1120 extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1123 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1124 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
1125 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1127 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1128 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1130 return skb->head + skb->tail;
1133 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1135 skb->tail = skb->data - skb->head;
1138 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1140 skb_reset_tail_pointer(skb);
1141 skb->tail += offset;
1143 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1144 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1149 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1151 skb->tail = skb->data;
1154 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1156 skb->tail = skb->data + offset;
1159 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1162 * Add data to an sk_buff
1164 extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1165 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1167 unsigned char *tmp = skb_tail_pointer(skb);
1168 SKB_LINEAR_ASSERT(skb);
1174 extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1175 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1182 extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1183 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1186 BUG_ON(skb->len < skb->data_len);
1187 return skb->data += len;
1190 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1192 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1195 extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1197 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1199 if (len > skb_headlen(skb) &&
1200 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1203 return skb->data += len;
1206 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1208 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1211 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1213 if (likely(len <= skb_headlen(skb)))
1215 if (unlikely(len > skb->len))
1217 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1221 * skb_headroom - bytes at buffer head
1222 * @skb: buffer to check
1224 * Return the number of bytes of free space at the head of an &sk_buff.
1226 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1228 return skb->data - skb->head;
1232 * skb_tailroom - bytes at buffer end
1233 * @skb: buffer to check
1235 * Return the number of bytes of free space at the tail of an sk_buff
1237 static inline int skb_tailroom(const struct sk_buff *skb)
1239 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1243 * skb_reserve - adjust headroom
1244 * @skb: buffer to alter
1245 * @len: bytes to move
1247 * Increase the headroom of an empty &sk_buff by reducing the tail
1248 * room. This is only allowed for an empty buffer.
1250 static inline void skb_reserve(struct sk_buff *skb, int len)
1256 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1257 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1259 return skb->head + skb->transport_header;
1262 static inline void skb_reset_transport_header(struct sk_buff *skb)
1264 skb->transport_header = skb->data - skb->head;
1267 static inline void skb_set_transport_header(struct sk_buff *skb,
1270 skb_reset_transport_header(skb);
1271 skb->transport_header += offset;
1274 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1276 return skb->head + skb->network_header;
1279 static inline void skb_reset_network_header(struct sk_buff *skb)
1281 skb->network_header = skb->data - skb->head;
1284 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1286 skb_reset_network_header(skb);
1287 skb->network_header += offset;
1290 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1292 return skb->head + skb->mac_header;
1295 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1297 return skb->mac_header != ~0U;
1300 static inline void skb_reset_mac_header(struct sk_buff *skb)
1302 skb->mac_header = skb->data - skb->head;
1305 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1307 skb_reset_mac_header(skb);
1308 skb->mac_header += offset;
1311 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1313 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1315 return skb->transport_header;
1318 static inline void skb_reset_transport_header(struct sk_buff *skb)
1320 skb->transport_header = skb->data;
1323 static inline void skb_set_transport_header(struct sk_buff *skb,
1326 skb->transport_header = skb->data + offset;
1329 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1331 return skb->network_header;
1334 static inline void skb_reset_network_header(struct sk_buff *skb)
1336 skb->network_header = skb->data;
1339 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1341 skb->network_header = skb->data + offset;
1344 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1346 return skb->mac_header;
1349 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1351 return skb->mac_header != NULL;
1354 static inline void skb_reset_mac_header(struct sk_buff *skb)
1356 skb->mac_header = skb->data;
1359 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1361 skb->mac_header = skb->data + offset;
1363 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1365 static inline int skb_transport_offset(const struct sk_buff *skb)
1367 return skb_transport_header(skb) - skb->data;
1370 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1372 return skb->transport_header - skb->network_header;
1375 static inline int skb_network_offset(const struct sk_buff *skb)
1377 return skb_network_header(skb) - skb->data;
1381 * CPUs often take a performance hit when accessing unaligned memory
1382 * locations. The actual performance hit varies, it can be small if the
1383 * hardware handles it or large if we have to take an exception and fix it
1386 * Since an ethernet header is 14 bytes network drivers often end up with
1387 * the IP header at an unaligned offset. The IP header can be aligned by
1388 * shifting the start of the packet by 2 bytes. Drivers should do this
1391 * skb_reserve(skb, NET_IP_ALIGN);
1393 * The downside to this alignment of the IP header is that the DMA is now
1394 * unaligned. On some architectures the cost of an unaligned DMA is high
1395 * and this cost outweighs the gains made by aligning the IP header.
1397 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1400 #ifndef NET_IP_ALIGN
1401 #define NET_IP_ALIGN 2
1405 * The networking layer reserves some headroom in skb data (via
1406 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1407 * the header has to grow. In the default case, if the header has to grow
1408 * 32 bytes or less we avoid the reallocation.
1410 * Unfortunately this headroom changes the DMA alignment of the resulting
1411 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1412 * on some architectures. An architecture can override this value,
1413 * perhaps setting it to a cacheline in size (since that will maintain
1414 * cacheline alignment of the DMA). It must be a power of 2.
1416 * Various parts of the networking layer expect at least 32 bytes of
1417 * headroom, you should not reduce this.
1419 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1420 * to reduce average number of cache lines per packet.
1421 * get_rps_cpus() for example only access one 64 bytes aligned block :
1422 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1425 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1428 extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1430 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1432 if (unlikely(skb->data_len)) {
1437 skb_set_tail_pointer(skb, len);
1440 extern void skb_trim(struct sk_buff *skb, unsigned int len);
1442 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1445 return ___pskb_trim(skb, len);
1446 __skb_trim(skb, len);
1450 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1452 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1456 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1457 * @skb: buffer to alter
1460 * This is identical to pskb_trim except that the caller knows that
1461 * the skb is not cloned so we should never get an error due to out-
1464 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1466 int err = pskb_trim(skb, len);
1471 * skb_orphan - orphan a buffer
1472 * @skb: buffer to orphan
1474 * If a buffer currently has an owner then we call the owner's
1475 * destructor function and make the @skb unowned. The buffer continues
1476 * to exist but is no longer charged to its former owner.
1478 static inline void skb_orphan(struct sk_buff *skb)
1480 if (skb->destructor)
1481 skb->destructor(skb);
1482 skb->destructor = NULL;
1487 * __skb_queue_purge - empty a list
1488 * @list: list to empty
1490 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1491 * the list and one reference dropped. This function does not take the
1492 * list lock and the caller must hold the relevant locks to use it.
1494 extern void skb_queue_purge(struct sk_buff_head *list);
1495 static inline void __skb_queue_purge(struct sk_buff_head *list)
1497 struct sk_buff *skb;
1498 while ((skb = __skb_dequeue(list)) != NULL)
1503 * __dev_alloc_skb - allocate an skbuff for receiving
1504 * @length: length to allocate
1505 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1507 * Allocate a new &sk_buff and assign it a usage count of one. The
1508 * buffer has unspecified headroom built in. Users should allocate
1509 * the headroom they think they need without accounting for the
1510 * built in space. The built in space is used for optimisations.
1512 * %NULL is returned if there is no free memory.
1514 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1517 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
1519 skb_reserve(skb, NET_SKB_PAD);
1523 extern struct sk_buff *dev_alloc_skb(unsigned int length);
1525 extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1526 unsigned int length, gfp_t gfp_mask);
1529 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1530 * @dev: network device to receive on
1531 * @length: length to allocate
1533 * Allocate a new &sk_buff and assign it a usage count of one. The
1534 * buffer has unspecified headroom built in. Users should allocate
1535 * the headroom they think they need without accounting for the
1536 * built in space. The built in space is used for optimisations.
1538 * %NULL is returned if there is no free memory. Although this function
1539 * allocates memory it can be called from an interrupt.
1541 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1542 unsigned int length)
1544 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1547 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1548 unsigned int length)
1550 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1552 if (NET_IP_ALIGN && skb)
1553 skb_reserve(skb, NET_IP_ALIGN);
1557 extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1560 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1561 * @dev: network device to receive on
1563 * Allocate a new page node local to the specified device.
1565 * %NULL is returned if there is no free memory.
1567 static inline struct page *netdev_alloc_page(struct net_device *dev)
1569 return __netdev_alloc_page(dev, GFP_ATOMIC);
1572 static inline void netdev_free_page(struct net_device *dev, struct page *page)
1578 * skb_clone_writable - is the header of a clone writable
1579 * @skb: buffer to check
1580 * @len: length up to which to write
1582 * Returns true if modifying the header part of the cloned buffer
1583 * does not requires the data to be copied.
1585 static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
1587 return !skb_header_cloned(skb) &&
1588 skb_headroom(skb) + len <= skb->hdr_len;
1591 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1596 if (headroom < NET_SKB_PAD)
1597 headroom = NET_SKB_PAD;
1598 if (headroom > skb_headroom(skb))
1599 delta = headroom - skb_headroom(skb);
1601 if (delta || cloned)
1602 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1608 * skb_cow - copy header of skb when it is required
1609 * @skb: buffer to cow
1610 * @headroom: needed headroom
1612 * If the skb passed lacks sufficient headroom or its data part
1613 * is shared, data is reallocated. If reallocation fails, an error
1614 * is returned and original skb is not changed.
1616 * The result is skb with writable area skb->head...skb->tail
1617 * and at least @headroom of space at head.
1619 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1621 return __skb_cow(skb, headroom, skb_cloned(skb));
1625 * skb_cow_head - skb_cow but only making the head writable
1626 * @skb: buffer to cow
1627 * @headroom: needed headroom
1629 * This function is identical to skb_cow except that we replace the
1630 * skb_cloned check by skb_header_cloned. It should be used when
1631 * you only need to push on some header and do not need to modify
1634 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1636 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1640 * skb_padto - pad an skbuff up to a minimal size
1641 * @skb: buffer to pad
1642 * @len: minimal length
1644 * Pads up a buffer to ensure the trailing bytes exist and are
1645 * blanked. If the buffer already contains sufficient data it
1646 * is untouched. Otherwise it is extended. Returns zero on
1647 * success. The skb is freed on error.
1650 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1652 unsigned int size = skb->len;
1653 if (likely(size >= len))
1655 return skb_pad(skb, len - size);
1658 static inline int skb_add_data(struct sk_buff *skb,
1659 char __user *from, int copy)
1661 const int off = skb->len;
1663 if (skb->ip_summed == CHECKSUM_NONE) {
1665 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
1668 skb->csum = csum_block_add(skb->csum, csum, off);
1671 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1674 __skb_trim(skb, off);
1678 static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1679 struct page *page, int off)
1682 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1684 return page == frag->page &&
1685 off == frag->page_offset + frag->size;
1690 static inline int __skb_linearize(struct sk_buff *skb)
1692 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1696 * skb_linearize - convert paged skb to linear one
1697 * @skb: buffer to linarize
1699 * If there is no free memory -ENOMEM is returned, otherwise zero
1700 * is returned and the old skb data released.
1702 static inline int skb_linearize(struct sk_buff *skb)
1704 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1708 * skb_linearize_cow - make sure skb is linear and writable
1709 * @skb: buffer to process
1711 * If there is no free memory -ENOMEM is returned, otherwise zero
1712 * is returned and the old skb data released.
1714 static inline int skb_linearize_cow(struct sk_buff *skb)
1716 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1717 __skb_linearize(skb) : 0;
1721 * skb_postpull_rcsum - update checksum for received skb after pull
1722 * @skb: buffer to update
1723 * @start: start of data before pull
1724 * @len: length of data pulled
1726 * After doing a pull on a received packet, you need to call this to
1727 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1728 * CHECKSUM_NONE so that it can be recomputed from scratch.
1731 static inline void skb_postpull_rcsum(struct sk_buff *skb,
1732 const void *start, unsigned int len)
1734 if (skb->ip_summed == CHECKSUM_COMPLETE)
1735 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1738 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1741 * pskb_trim_rcsum - trim received skb and update checksum
1742 * @skb: buffer to trim
1745 * This is exactly the same as pskb_trim except that it ensures the
1746 * checksum of received packets are still valid after the operation.
1749 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1751 if (likely(len >= skb->len))
1753 if (skb->ip_summed == CHECKSUM_COMPLETE)
1754 skb->ip_summed = CHECKSUM_NONE;
1755 return __pskb_trim(skb, len);
1758 #define skb_queue_walk(queue, skb) \
1759 for (skb = (queue)->next; \
1760 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1763 #define skb_queue_walk_safe(queue, skb, tmp) \
1764 for (skb = (queue)->next, tmp = skb->next; \
1765 skb != (struct sk_buff *)(queue); \
1766 skb = tmp, tmp = skb->next)
1768 #define skb_queue_walk_from(queue, skb) \
1769 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1772 #define skb_queue_walk_from_safe(queue, skb, tmp) \
1773 for (tmp = skb->next; \
1774 skb != (struct sk_buff *)(queue); \
1775 skb = tmp, tmp = skb->next)
1777 #define skb_queue_reverse_walk(queue, skb) \
1778 for (skb = (queue)->prev; \
1779 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1783 static inline bool skb_has_frags(const struct sk_buff *skb)
1785 return skb_shinfo(skb)->frag_list != NULL;
1788 static inline void skb_frag_list_init(struct sk_buff *skb)
1790 skb_shinfo(skb)->frag_list = NULL;
1793 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1795 frag->next = skb_shinfo(skb)->frag_list;
1796 skb_shinfo(skb)->frag_list = frag;
1799 #define skb_walk_frags(skb, iter) \
1800 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1802 extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1803 int *peeked, int *err);
1804 extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1805 int noblock, int *err);
1806 extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1807 struct poll_table_struct *wait);
1808 extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1809 int offset, struct iovec *to,
1811 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
1814 extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1816 const struct iovec *from,
1819 extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1821 const struct iovec *to,
1824 extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
1825 extern void skb_free_datagram_locked(struct sock *sk,
1826 struct sk_buff *skb);
1827 extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
1828 unsigned int flags);
1829 extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1830 int len, __wsum csum);
1831 extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1833 extern int skb_store_bits(struct sk_buff *skb, int offset,
1834 const void *from, int len);
1835 extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
1836 int offset, u8 *to, int len,
1838 extern int skb_splice_bits(struct sk_buff *skb,
1839 unsigned int offset,
1840 struct pipe_inode_info *pipe,
1842 unsigned int flags);
1843 extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1844 extern void skb_split(struct sk_buff *skb,
1845 struct sk_buff *skb1, const u32 len);
1846 extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1849 extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
1851 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1852 int len, void *buffer)
1854 int hlen = skb_headlen(skb);
1856 if (hlen - offset >= len)
1857 return skb->data + offset;
1859 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1865 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1867 const unsigned int len)
1869 memcpy(to, skb->data, len);
1872 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1873 const int offset, void *to,
1874 const unsigned int len)
1876 memcpy(to, skb->data + offset, len);
1879 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1881 const unsigned int len)
1883 memcpy(skb->data, from, len);
1886 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1889 const unsigned int len)
1891 memcpy(skb->data + offset, from, len);
1894 extern void skb_init(void);
1896 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1902 * skb_get_timestamp - get timestamp from a skb
1903 * @skb: skb to get stamp from
1904 * @stamp: pointer to struct timeval to store stamp in
1906 * Timestamps are stored in the skb as offsets to a base timestamp.
1907 * This function converts the offset back to a struct timeval and stores
1910 static inline void skb_get_timestamp(const struct sk_buff *skb,
1911 struct timeval *stamp)
1913 *stamp = ktime_to_timeval(skb->tstamp);
1916 static inline void skb_get_timestampns(const struct sk_buff *skb,
1917 struct timespec *stamp)
1919 *stamp = ktime_to_timespec(skb->tstamp);
1922 static inline void __net_timestamp(struct sk_buff *skb)
1924 skb->tstamp = ktime_get_real();
1927 static inline ktime_t net_timedelta(ktime_t t)
1929 return ktime_sub(ktime_get_real(), t);
1932 static inline ktime_t net_invalid_timestamp(void)
1934 return ktime_set(0, 0);
1937 extern void skb_timestamping_init(void);
1939 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1941 extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1942 extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1944 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1946 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1950 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1955 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1958 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1960 * @skb: clone of the the original outgoing packet
1961 * @hwtstamps: hardware time stamps
1964 void skb_complete_tx_timestamp(struct sk_buff *skb,
1965 struct skb_shared_hwtstamps *hwtstamps);
1968 * skb_tstamp_tx - queue clone of skb with send time stamps
1969 * @orig_skb: the original outgoing packet
1970 * @hwtstamps: hardware time stamps, may be NULL if not available
1972 * If the skb has a socket associated, then this function clones the
1973 * skb (thus sharing the actual data and optional structures), stores
1974 * the optional hardware time stamping information (if non NULL) or
1975 * generates a software time stamp (otherwise), then queues the clone
1976 * to the error queue of the socket. Errors are silently ignored.
1978 extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1979 struct skb_shared_hwtstamps *hwtstamps);
1981 static inline void sw_tx_timestamp(struct sk_buff *skb)
1983 union skb_shared_tx *shtx = skb_tx(skb);
1984 if (shtx->software && !shtx->in_progress)
1985 skb_tstamp_tx(skb, NULL);
1989 * skb_tx_timestamp() - Driver hook for transmit timestamping
1991 * Ethernet MAC Drivers should call this function in their hard_xmit()
1992 * function as soon as possible after giving the sk_buff to the MAC
1993 * hardware, but before freeing the sk_buff.
1995 * @skb: A socket buffer.
1997 static inline void skb_tx_timestamp(struct sk_buff *skb)
1999 skb_clone_tx_timestamp(skb);
2000 sw_tx_timestamp(skb);
2003 extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2004 extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
2006 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2008 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2012 * skb_checksum_complete - Calculate checksum of an entire packet
2013 * @skb: packet to process
2015 * This function calculates the checksum over the entire packet plus
2016 * the value of skb->csum. The latter can be used to supply the
2017 * checksum of a pseudo header as used by TCP/UDP. It returns the
2020 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2021 * this function can be used to verify that checksum on received
2022 * packets. In that case the function should return zero if the
2023 * checksum is correct. In particular, this function will return zero
2024 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2025 * hardware has already verified the correctness of the checksum.
2027 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2029 return skb_csum_unnecessary(skb) ?
2030 0 : __skb_checksum_complete(skb);
2033 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2034 extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
2035 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2037 if (nfct && atomic_dec_and_test(&nfct->use))
2038 nf_conntrack_destroy(nfct);
2040 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2043 atomic_inc(&nfct->use);
2045 static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2048 atomic_inc(&skb->users);
2050 static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2056 #ifdef CONFIG_BRIDGE_NETFILTER
2057 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2059 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2062 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2065 atomic_inc(&nf_bridge->use);
2067 #endif /* CONFIG_BRIDGE_NETFILTER */
2068 static inline void nf_reset(struct sk_buff *skb)
2070 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2071 nf_conntrack_put(skb->nfct);
2073 nf_conntrack_put_reasm(skb->nfct_reasm);
2074 skb->nfct_reasm = NULL;
2076 #ifdef CONFIG_BRIDGE_NETFILTER
2077 nf_bridge_put(skb->nf_bridge);
2078 skb->nf_bridge = NULL;
2082 /* Note: This doesn't put any conntrack and bridge info in dst. */
2083 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2085 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2086 dst->nfct = src->nfct;
2087 nf_conntrack_get(src->nfct);
2088 dst->nfctinfo = src->nfctinfo;
2089 dst->nfct_reasm = src->nfct_reasm;
2090 nf_conntrack_get_reasm(src->nfct_reasm);
2092 #ifdef CONFIG_BRIDGE_NETFILTER
2093 dst->nf_bridge = src->nf_bridge;
2094 nf_bridge_get(src->nf_bridge);
2098 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2100 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2101 nf_conntrack_put(dst->nfct);
2102 nf_conntrack_put_reasm(dst->nfct_reasm);
2104 #ifdef CONFIG_BRIDGE_NETFILTER
2105 nf_bridge_put(dst->nf_bridge);
2107 __nf_copy(dst, src);
2110 #ifdef CONFIG_NETWORK_SECMARK
2111 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2113 to->secmark = from->secmark;
2116 static inline void skb_init_secmark(struct sk_buff *skb)
2121 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2124 static inline void skb_init_secmark(struct sk_buff *skb)
2128 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2130 skb->queue_mapping = queue_mapping;
2133 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2135 return skb->queue_mapping;
2138 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2140 to->queue_mapping = from->queue_mapping;
2143 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2145 skb->queue_mapping = rx_queue + 1;
2148 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2150 return skb->queue_mapping - 1;
2153 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2155 return (skb->queue_mapping != 0);
2158 extern u16 skb_tx_hash(const struct net_device *dev,
2159 const struct sk_buff *skb);
2162 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2167 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2173 static inline int skb_is_gso(const struct sk_buff *skb)
2175 return skb_shinfo(skb)->gso_size;
2178 static inline int skb_is_gso_v6(const struct sk_buff *skb)
2180 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2183 extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2185 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2187 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2188 * wanted then gso_type will be set. */
2189 struct skb_shared_info *shinfo = skb_shinfo(skb);
2190 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2191 unlikely(shinfo->gso_type == 0)) {
2192 __skb_warn_lro_forwarding(skb);
2198 static inline void skb_forward_csum(struct sk_buff *skb)
2200 /* Unfortunately we don't support this one. Any brave souls? */
2201 if (skb->ip_summed == CHECKSUM_COMPLETE)
2202 skb->ip_summed = CHECKSUM_NONE;
2205 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
2206 #endif /* __KERNEL__ */
2207 #endif /* _LINUX_SKBUFF_H */