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/compiler.h>
19 #include <linux/time.h>
20 #include <linux/cache.h>
22 #include <asm/atomic.h>
23 #include <asm/types.h>
24 #include <linux/spinlock.h>
25 #include <linux/net.h>
26 #include <linux/textsearch.h>
27 #include <net/checksum.h>
28 #include <linux/rcupdate.h>
29 #include <linux/dmaengine.h>
30 #include <linux/hrtimer.h>
32 #define HAVE_ALLOC_SKB /* For the drivers to know */
33 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
35 #define CHECKSUM_NONE 0
36 #define CHECKSUM_PARTIAL 1
37 #define CHECKSUM_UNNECESSARY 2
38 #define CHECKSUM_COMPLETE 3
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_WITH_OVERHEAD(X) \
43 (((X) - sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
45 #define SKB_MAX_ORDER(X, ORDER) \
46 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
47 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
48 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50 /* A. Checksumming of received packets by device.
52 * NONE: device failed to checksum this packet.
53 * skb->csum is undefined.
55 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
56 * skb->csum is undefined.
57 * It is bad option, but, unfortunately, many of vendors do this.
58 * Apparently with secret goal to sell you new device, when you
59 * will add new protocol to your host. F.e. IPv6. 8)
61 * COMPLETE: the most generic way. Device supplied checksum of _all_
62 * the packet as seen by netif_rx in skb->csum.
63 * NOTE: Even if device supports only some protocols, but
64 * is able to produce some skb->csum, it MUST use COMPLETE,
67 * B. Checksumming on output.
69 * NONE: skb is checksummed by protocol or csum is not required.
71 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
72 * from skb->transport_header to the end and to record the checksum
73 * at skb->transport_header + skb->csum.
75 * Device must show its capabilities in dev->features, set
76 * at device setup time.
77 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
79 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
80 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
81 * TCP/UDP over IPv4. Sigh. Vendors like this
82 * way by an unknown reason. Though, see comment above
83 * about CHECKSUM_UNNECESSARY. 8)
85 * Any questions? No questions, good. --ANK
90 #ifdef CONFIG_NETFILTER
93 void (*destroy
)(struct nf_conntrack
*);
96 #ifdef CONFIG_BRIDGE_NETFILTER
97 struct nf_bridge_info
{
99 struct net_device
*physindev
;
100 struct net_device
*physoutdev
;
101 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
102 struct net_device
*netoutdev
;
105 unsigned long data
[32 / sizeof(unsigned long)];
111 struct sk_buff_head
{
112 /* These two members must be first. */
113 struct sk_buff
*next
;
114 struct sk_buff
*prev
;
122 /* To allow 64K frame to be packed as single skb without frag_list */
123 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
125 typedef struct skb_frag_struct skb_frag_t
;
127 struct skb_frag_struct
{
133 /* This data is invariant across clones and lives at
134 * the end of the header data, ie. at skb->end.
136 struct skb_shared_info
{
138 unsigned short nr_frags
;
139 unsigned short gso_size
;
140 /* Warning: this field is not always filled in (UFO)! */
141 unsigned short gso_segs
;
142 unsigned short gso_type
;
144 struct sk_buff
*frag_list
;
145 skb_frag_t frags
[MAX_SKB_FRAGS
];
148 /* We divide dataref into two halves. The higher 16 bits hold references
149 * to the payload part of skb->data. The lower 16 bits hold references to
150 * the entire skb->data. It is up to the users of the skb to agree on
151 * where the payload starts.
153 * All users must obey the rule that the skb->data reference count must be
154 * greater than or equal to the payload reference count.
156 * Holding a reference to the payload part means that the user does not
157 * care about modifications to the header part of skb->data.
159 #define SKB_DATAREF_SHIFT 16
160 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
164 SKB_FCLONE_UNAVAILABLE
,
170 SKB_GSO_TCPV4
= 1 << 0,
171 SKB_GSO_UDP
= 1 << 1,
173 /* This indicates the skb is from an untrusted source. */
174 SKB_GSO_DODGY
= 1 << 2,
176 /* This indicates the tcp segment has CWR set. */
177 SKB_GSO_TCP_ECN
= 1 << 3,
179 SKB_GSO_TCPV6
= 1 << 4,
183 * struct sk_buff - socket buffer
184 * @next: Next buffer in list
185 * @prev: Previous buffer in list
186 * @sk: Socket we are owned by
187 * @tstamp: Time we arrived
188 * @dev: Device we arrived on/are leaving by
189 * @iif: ifindex of device we arrived on
190 * @h: Transport layer header
191 * @network_header: Network layer header
192 * @mac_header: Link layer header
193 * @dst: destination entry
194 * @sp: the security path, used for xfrm
195 * @cb: Control buffer. Free for use by every layer. Put private vars here
196 * @len: Length of actual data
197 * @data_len: Data length
198 * @mac_len: Length of link layer header
200 * @local_df: allow local fragmentation
201 * @cloned: Head may be cloned (check refcnt to be sure)
202 * @nohdr: Payload reference only, must not modify header
203 * @pkt_type: Packet class
204 * @fclone: skbuff clone status
205 * @ip_summed: Driver fed us an IP checksum
206 * @priority: Packet queueing priority
207 * @users: User count - see {datagram,tcp}.c
208 * @protocol: Packet protocol from driver
209 * @truesize: Buffer size
210 * @head: Head of buffer
211 * @data: Data head pointer
212 * @tail: Tail pointer
214 * @destructor: Destruct function
215 * @mark: Generic packet mark
216 * @nfct: Associated connection, if any
217 * @ipvs_property: skbuff is owned by ipvs
218 * @nfctinfo: Relationship of this skb to the connection
219 * @nfct_reasm: netfilter conntrack re-assembly pointer
220 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
221 * @tc_index: Traffic control index
222 * @tc_verd: traffic control verdict
223 * @dma_cookie: a cookie to one of several possible DMA operations
224 * done by skb DMA functions
225 * @secmark: security marking
229 /* These two members must be first. */
230 struct sk_buff
*next
;
231 struct sk_buff
*prev
;
235 struct net_device
*dev
;
237 /* 4 byte hole on 64 bit*/
239 unsigned char *transport_header
;
240 unsigned char *network_header
;
241 unsigned char *mac_header
;
242 struct dst_entry
*dst
;
246 * This is the control buffer. It is free to use for every
247 * layer. Please put your private variables there. If you
248 * want to keep them across layers you have to do a skb_clone()
249 * first. This is owned by whoever has the skb queued ATM.
271 void (*destructor
)(struct sk_buff
*skb
);
272 #ifdef CONFIG_NETFILTER
273 struct nf_conntrack
*nfct
;
274 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
275 struct sk_buff
*nfct_reasm
;
277 #ifdef CONFIG_BRIDGE_NETFILTER
278 struct nf_bridge_info
*nf_bridge
;
280 #endif /* CONFIG_NETFILTER */
281 #ifdef CONFIG_NET_SCHED
282 __u16 tc_index
; /* traffic control index */
283 #ifdef CONFIG_NET_CLS_ACT
284 __u16 tc_verd
; /* traffic control verdict */
287 #ifdef CONFIG_NET_DMA
288 dma_cookie_t dma_cookie
;
290 #ifdef CONFIG_NETWORK_SECMARK
296 /* These elements must be at the end, see alloc_skb() for details. */
297 unsigned int truesize
;
307 * Handling routines are only of interest to the kernel
309 #include <linux/slab.h>
311 #include <asm/system.h>
313 extern void kfree_skb(struct sk_buff
*skb
);
314 extern void __kfree_skb(struct sk_buff
*skb
);
315 extern struct sk_buff
*__alloc_skb(unsigned int size
,
316 gfp_t priority
, int fclone
, int node
);
317 static inline struct sk_buff
*alloc_skb(unsigned int size
,
320 return __alloc_skb(size
, priority
, 0, -1);
323 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
326 return __alloc_skb(size
, priority
, 1, -1);
329 extern void kfree_skbmem(struct sk_buff
*skb
);
330 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
332 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
334 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
336 extern int pskb_expand_head(struct sk_buff
*skb
,
337 int nhead
, int ntail
,
339 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
340 unsigned int headroom
);
341 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
342 int newheadroom
, int newtailroom
,
344 extern int skb_pad(struct sk_buff
*skb
, int pad
);
345 #define dev_kfree_skb(a) kfree_skb(a)
346 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
348 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
350 extern void skb_truesize_bug(struct sk_buff
*skb
);
352 static inline void skb_truesize_check(struct sk_buff
*skb
)
354 if (unlikely((int)skb
->truesize
< sizeof(struct sk_buff
) + skb
->len
))
355 skb_truesize_bug(skb
);
358 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
359 int getfrag(void *from
, char *to
, int offset
,
360 int len
,int odd
, struct sk_buff
*skb
),
361 void *from
, int length
);
368 __u32 stepped_offset
;
369 struct sk_buff
*root_skb
;
370 struct sk_buff
*cur_skb
;
374 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
375 unsigned int from
, unsigned int to
,
376 struct skb_seq_state
*st
);
377 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
378 struct skb_seq_state
*st
);
379 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
381 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
382 unsigned int to
, struct ts_config
*config
,
383 struct ts_state
*state
);
386 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
389 * skb_queue_empty - check if a queue is empty
392 * Returns true if the queue is empty, false otherwise.
394 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
396 return list
->next
== (struct sk_buff
*)list
;
400 * skb_get - reference buffer
401 * @skb: buffer to reference
403 * Makes another reference to a socket buffer and returns a pointer
406 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
408 atomic_inc(&skb
->users
);
413 * If users == 1, we are the only owner and are can avoid redundant
418 * skb_cloned - is the buffer a clone
419 * @skb: buffer to check
421 * Returns true if the buffer was generated with skb_clone() and is
422 * one of multiple shared copies of the buffer. Cloned buffers are
423 * shared data so must not be written to under normal circumstances.
425 static inline int skb_cloned(const struct sk_buff
*skb
)
427 return skb
->cloned
&&
428 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
432 * skb_header_cloned - is the header a clone
433 * @skb: buffer to check
435 * Returns true if modifying the header part of the buffer requires
436 * the data to be copied.
438 static inline int skb_header_cloned(const struct sk_buff
*skb
)
445 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
446 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
451 * skb_header_release - release reference to header
452 * @skb: buffer to operate on
454 * Drop a reference to the header part of the buffer. This is done
455 * by acquiring a payload reference. You must not read from the header
456 * part of skb->data after this.
458 static inline void skb_header_release(struct sk_buff
*skb
)
462 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
466 * skb_shared - is the buffer shared
467 * @skb: buffer to check
469 * Returns true if more than one person has a reference to this
472 static inline int skb_shared(const struct sk_buff
*skb
)
474 return atomic_read(&skb
->users
) != 1;
478 * skb_share_check - check if buffer is shared and if so clone it
479 * @skb: buffer to check
480 * @pri: priority for memory allocation
482 * If the buffer is shared the buffer is cloned and the old copy
483 * drops a reference. A new clone with a single reference is returned.
484 * If the buffer is not shared the original buffer is returned. When
485 * being called from interrupt status or with spinlocks held pri must
488 * NULL is returned on a memory allocation failure.
490 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
493 might_sleep_if(pri
& __GFP_WAIT
);
494 if (skb_shared(skb
)) {
495 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
503 * Copy shared buffers into a new sk_buff. We effectively do COW on
504 * packets to handle cases where we have a local reader and forward
505 * and a couple of other messy ones. The normal one is tcpdumping
506 * a packet thats being forwarded.
510 * skb_unshare - make a copy of a shared buffer
511 * @skb: buffer to check
512 * @pri: priority for memory allocation
514 * If the socket buffer is a clone then this function creates a new
515 * copy of the data, drops a reference count on the old copy and returns
516 * the new copy with the reference count at 1. If the buffer is not a clone
517 * the original buffer is returned. When called with a spinlock held or
518 * from interrupt state @pri must be %GFP_ATOMIC
520 * %NULL is returned on a memory allocation failure.
522 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
525 might_sleep_if(pri
& __GFP_WAIT
);
526 if (skb_cloned(skb
)) {
527 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
528 kfree_skb(skb
); /* Free our shared copy */
536 * @list_: list to peek at
538 * Peek an &sk_buff. Unlike most other operations you _MUST_
539 * be careful with this one. A peek leaves the buffer on the
540 * list and someone else may run off with it. You must hold
541 * the appropriate locks or have a private queue to do this.
543 * Returns %NULL for an empty list or a pointer to the head element.
544 * The reference count is not incremented and the reference is therefore
545 * volatile. Use with caution.
547 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
549 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
550 if (list
== (struct sk_buff
*)list_
)
557 * @list_: list to peek at
559 * Peek an &sk_buff. Unlike most other operations you _MUST_
560 * be careful with this one. A peek leaves the buffer on the
561 * list and someone else may run off with it. You must hold
562 * the appropriate locks or have a private queue to do this.
564 * Returns %NULL for an empty list or a pointer to the tail element.
565 * The reference count is not incremented and the reference is therefore
566 * volatile. Use with caution.
568 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
570 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
571 if (list
== (struct sk_buff
*)list_
)
577 * skb_queue_len - get queue length
578 * @list_: list to measure
580 * Return the length of an &sk_buff queue.
582 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
588 * This function creates a split out lock class for each invocation;
589 * this is needed for now since a whole lot of users of the skb-queue
590 * infrastructure in drivers have different locking usage (in hardirq)
591 * than the networking core (in softirq only). In the long run either the
592 * network layer or drivers should need annotation to consolidate the
593 * main types of usage into 3 classes.
595 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
597 spin_lock_init(&list
->lock
);
598 list
->prev
= list
->next
= (struct sk_buff
*)list
;
602 static inline void skb_queue_head_init_class(struct sk_buff_head
*list
,
603 struct lock_class_key
*class)
605 skb_queue_head_init(list
);
606 lockdep_set_class(&list
->lock
, class);
610 * Insert an sk_buff at the start of a list.
612 * The "__skb_xxxx()" functions are the non-atomic ones that
613 * can only be called with interrupts disabled.
617 * __skb_queue_after - queue a buffer at the list head
619 * @prev: place after this buffer
620 * @newsk: buffer to queue
622 * Queue a buffer int the middle of a list. This function takes no locks
623 * and you must therefore hold required locks before calling it.
625 * A buffer cannot be placed on two lists at the same time.
627 static inline void __skb_queue_after(struct sk_buff_head
*list
,
628 struct sk_buff
*prev
,
629 struct sk_buff
*newsk
)
631 struct sk_buff
*next
;
637 next
->prev
= prev
->next
= newsk
;
641 * __skb_queue_head - queue a buffer at the list head
643 * @newsk: buffer to queue
645 * Queue a buffer at the start of a list. This function takes no locks
646 * and you must therefore hold required locks before calling it.
648 * A buffer cannot be placed on two lists at the same time.
650 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
651 static inline void __skb_queue_head(struct sk_buff_head
*list
,
652 struct sk_buff
*newsk
)
654 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
658 * __skb_queue_tail - queue a buffer at the list tail
660 * @newsk: buffer to queue
662 * Queue a buffer at the end of a list. This function takes no locks
663 * and you must therefore hold required locks before calling it.
665 * A buffer cannot be placed on two lists at the same time.
667 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
668 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
669 struct sk_buff
*newsk
)
671 struct sk_buff
*prev
, *next
;
674 next
= (struct sk_buff
*)list
;
678 next
->prev
= prev
->next
= newsk
;
683 * __skb_dequeue - remove from the head of the queue
684 * @list: list to dequeue from
686 * Remove the head of the list. This function does not take any locks
687 * so must be used with appropriate locks held only. The head item is
688 * returned or %NULL if the list is empty.
690 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
691 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
693 struct sk_buff
*next
, *prev
, *result
;
695 prev
= (struct sk_buff
*) list
;
704 result
->next
= result
->prev
= NULL
;
711 * Insert a packet on a list.
713 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
714 static inline void __skb_insert(struct sk_buff
*newsk
,
715 struct sk_buff
*prev
, struct sk_buff
*next
,
716 struct sk_buff_head
*list
)
720 next
->prev
= prev
->next
= newsk
;
725 * Place a packet after a given packet in a list.
727 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
728 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
730 __skb_insert(newsk
, old
, old
->next
, list
);
734 * remove sk_buff from list. _Must_ be called atomically, and with
737 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
738 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
740 struct sk_buff
*next
, *prev
;
745 skb
->next
= skb
->prev
= NULL
;
751 /* XXX: more streamlined implementation */
754 * __skb_dequeue_tail - remove from the tail of the queue
755 * @list: list to dequeue from
757 * Remove the tail of the list. This function does not take any locks
758 * so must be used with appropriate locks held only. The tail item is
759 * returned or %NULL if the list is empty.
761 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
762 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
764 struct sk_buff
*skb
= skb_peek_tail(list
);
766 __skb_unlink(skb
, list
);
771 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
773 return skb
->data_len
;
776 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
778 return skb
->len
- skb
->data_len
;
781 static inline int skb_pagelen(const struct sk_buff
*skb
)
785 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
786 len
+= skb_shinfo(skb
)->frags
[i
].size
;
787 return len
+ skb_headlen(skb
);
790 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
791 struct page
*page
, int off
, int size
)
793 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
796 frag
->page_offset
= off
;
798 skb_shinfo(skb
)->nr_frags
= i
+ 1;
801 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
802 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
803 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
806 * Add data to an sk_buff
808 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
810 unsigned char *tmp
= skb
->tail
;
811 SKB_LINEAR_ASSERT(skb
);
818 * skb_put - add data to a buffer
819 * @skb: buffer to use
820 * @len: amount of data to add
822 * This function extends the used data area of the buffer. If this would
823 * exceed the total buffer size the kernel will panic. A pointer to the
824 * first byte of the extra data is returned.
826 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
828 unsigned char *tmp
= skb
->tail
;
829 SKB_LINEAR_ASSERT(skb
);
832 if (unlikely(skb
->tail
>skb
->end
))
833 skb_over_panic(skb
, len
, current_text_addr());
837 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
845 * skb_push - add data to the start of a buffer
846 * @skb: buffer to use
847 * @len: amount of data to add
849 * This function extends the used data area of the buffer at the buffer
850 * start. If this would exceed the total buffer headroom the kernel will
851 * panic. A pointer to the first byte of the extra data is returned.
853 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
857 if (unlikely(skb
->data
<skb
->head
))
858 skb_under_panic(skb
, len
, current_text_addr());
862 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
865 BUG_ON(skb
->len
< skb
->data_len
);
866 return skb
->data
+= len
;
870 * skb_pull - remove data from the start of a buffer
871 * @skb: buffer to use
872 * @len: amount of data to remove
874 * This function removes data from the start of a buffer, returning
875 * the memory to the headroom. A pointer to the next data in the buffer
876 * is returned. Once the data has been pulled future pushes will overwrite
879 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
881 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
884 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
886 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
888 if (len
> skb_headlen(skb
) &&
889 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
892 return skb
->data
+= len
;
895 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
897 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
900 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
902 if (likely(len
<= skb_headlen(skb
)))
904 if (unlikely(len
> skb
->len
))
906 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
910 * skb_headroom - bytes at buffer head
911 * @skb: buffer to check
913 * Return the number of bytes of free space at the head of an &sk_buff.
915 static inline int skb_headroom(const struct sk_buff
*skb
)
917 return skb
->data
- skb
->head
;
921 * skb_tailroom - bytes at buffer end
922 * @skb: buffer to check
924 * Return the number of bytes of free space at the tail of an sk_buff
926 static inline int skb_tailroom(const struct sk_buff
*skb
)
928 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
932 * skb_reserve - adjust headroom
933 * @skb: buffer to alter
934 * @len: bytes to move
936 * Increase the headroom of an empty &sk_buff by reducing the tail
937 * room. This is only allowed for an empty buffer.
939 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
945 static inline unsigned char *skb_transport_header(const struct sk_buff
*skb
)
947 return skb
->transport_header
;
950 static inline void skb_reset_transport_header(struct sk_buff
*skb
)
952 skb
->transport_header
= skb
->data
;
955 static inline void skb_set_transport_header(struct sk_buff
*skb
,
958 skb
->transport_header
= skb
->data
+ offset
;
961 static inline int skb_transport_offset(const struct sk_buff
*skb
)
963 return skb
->transport_header
- skb
->data
;
966 static inline unsigned char *skb_network_header(const struct sk_buff
*skb
)
968 return skb
->network_header
;
971 static inline void skb_reset_network_header(struct sk_buff
*skb
)
973 skb
->network_header
= skb
->data
;
976 static inline void skb_set_network_header(struct sk_buff
*skb
, const int offset
)
978 skb
->network_header
= skb
->data
+ offset
;
981 static inline int skb_network_offset(const struct sk_buff
*skb
)
983 return skb
->network_header
- skb
->data
;
986 static inline u32
skb_network_header_len(const struct sk_buff
*skb
)
988 return skb
->transport_header
- skb
->network_header
;
991 static inline unsigned char *skb_mac_header(const struct sk_buff
*skb
)
993 return skb
->mac_header
;
996 static inline int skb_mac_header_was_set(const struct sk_buff
*skb
)
998 return skb
->mac_header
!= NULL
;
1001 static inline void skb_reset_mac_header(struct sk_buff
*skb
)
1003 skb
->mac_header
= skb
->data
;
1006 static inline void skb_set_mac_header(struct sk_buff
*skb
, const int offset
)
1008 skb
->mac_header
= skb
->data
+ offset
;
1012 * CPUs often take a performance hit when accessing unaligned memory
1013 * locations. The actual performance hit varies, it can be small if the
1014 * hardware handles it or large if we have to take an exception and fix it
1017 * Since an ethernet header is 14 bytes network drivers often end up with
1018 * the IP header at an unaligned offset. The IP header can be aligned by
1019 * shifting the start of the packet by 2 bytes. Drivers should do this
1022 * skb_reserve(NET_IP_ALIGN);
1024 * The downside to this alignment of the IP header is that the DMA is now
1025 * unaligned. On some architectures the cost of an unaligned DMA is high
1026 * and this cost outweighs the gains made by aligning the IP header.
1028 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1031 #ifndef NET_IP_ALIGN
1032 #define NET_IP_ALIGN 2
1036 * The networking layer reserves some headroom in skb data (via
1037 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1038 * the header has to grow. In the default case, if the header has to grow
1039 * 16 bytes or less we avoid the reallocation.
1041 * Unfortunately this headroom changes the DMA alignment of the resulting
1042 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1043 * on some architectures. An architecture can override this value,
1044 * perhaps setting it to a cacheline in size (since that will maintain
1045 * cacheline alignment of the DMA). It must be a power of 2.
1047 * Various parts of the networking layer expect at least 16 bytes of
1048 * headroom, you should not reduce this.
1051 #define NET_SKB_PAD 16
1054 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
1056 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
1058 if (unlikely(skb
->data_len
)) {
1063 skb
->tail
= skb
->data
+ len
;
1067 * skb_trim - remove end from a buffer
1068 * @skb: buffer to alter
1071 * Cut the length of a buffer down by removing data from the tail. If
1072 * the buffer is already under the length specified it is not modified.
1073 * The skb must be linear.
1075 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
1078 __skb_trim(skb
, len
);
1082 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1085 return ___pskb_trim(skb
, len
);
1086 __skb_trim(skb
, len
);
1090 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1092 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1096 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1097 * @skb: buffer to alter
1100 * This is identical to pskb_trim except that the caller knows that
1101 * the skb is not cloned so we should never get an error due to out-
1104 static inline void pskb_trim_unique(struct sk_buff
*skb
, unsigned int len
)
1106 int err
= pskb_trim(skb
, len
);
1111 * skb_orphan - orphan a buffer
1112 * @skb: buffer to orphan
1114 * If a buffer currently has an owner then we call the owner's
1115 * destructor function and make the @skb unowned. The buffer continues
1116 * to exist but is no longer charged to its former owner.
1118 static inline void skb_orphan(struct sk_buff
*skb
)
1120 if (skb
->destructor
)
1121 skb
->destructor(skb
);
1122 skb
->destructor
= NULL
;
1127 * __skb_queue_purge - empty a list
1128 * @list: list to empty
1130 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1131 * the list and one reference dropped. This function does not take the
1132 * list lock and the caller must hold the relevant locks to use it.
1134 extern void skb_queue_purge(struct sk_buff_head
*list
);
1135 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1137 struct sk_buff
*skb
;
1138 while ((skb
= __skb_dequeue(list
)) != NULL
)
1143 * __dev_alloc_skb - allocate an skbuff for receiving
1144 * @length: length to allocate
1145 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1147 * Allocate a new &sk_buff and assign it a usage count of one. The
1148 * buffer has unspecified headroom built in. Users should allocate
1149 * the headroom they think they need without accounting for the
1150 * built in space. The built in space is used for optimisations.
1152 * %NULL is returned if there is no free memory.
1154 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1157 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1159 skb_reserve(skb
, NET_SKB_PAD
);
1164 * dev_alloc_skb - allocate an skbuff for receiving
1165 * @length: length to allocate
1167 * Allocate a new &sk_buff and assign it a usage count of one. The
1168 * buffer has unspecified headroom built in. Users should allocate
1169 * the headroom they think they need without accounting for the
1170 * built in space. The built in space is used for optimisations.
1172 * %NULL is returned if there is no free memory. Although this function
1173 * allocates memory it can be called from an interrupt.
1175 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1177 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1180 extern struct sk_buff
*__netdev_alloc_skb(struct net_device
*dev
,
1181 unsigned int length
, gfp_t gfp_mask
);
1184 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1185 * @dev: network device to receive on
1186 * @length: length to allocate
1188 * Allocate a new &sk_buff and assign it a usage count of one. The
1189 * buffer has unspecified headroom built in. Users should allocate
1190 * the headroom they think they need without accounting for the
1191 * built in space. The built in space is used for optimisations.
1193 * %NULL is returned if there is no free memory. Although this function
1194 * allocates memory it can be called from an interrupt.
1196 static inline struct sk_buff
*netdev_alloc_skb(struct net_device
*dev
,
1197 unsigned int length
)
1199 return __netdev_alloc_skb(dev
, length
, GFP_ATOMIC
);
1203 * skb_cow - copy header of skb when it is required
1204 * @skb: buffer to cow
1205 * @headroom: needed headroom
1207 * If the skb passed lacks sufficient headroom or its data part
1208 * is shared, data is reallocated. If reallocation fails, an error
1209 * is returned and original skb is not changed.
1211 * The result is skb with writable area skb->head...skb->tail
1212 * and at least @headroom of space at head.
1214 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1216 int delta
= (headroom
> NET_SKB_PAD
? headroom
: NET_SKB_PAD
) -
1222 if (delta
|| skb_cloned(skb
))
1223 return pskb_expand_head(skb
, (delta
+ (NET_SKB_PAD
-1)) &
1224 ~(NET_SKB_PAD
-1), 0, GFP_ATOMIC
);
1229 * skb_padto - pad an skbuff up to a minimal size
1230 * @skb: buffer to pad
1231 * @len: minimal length
1233 * Pads up a buffer to ensure the trailing bytes exist and are
1234 * blanked. If the buffer already contains sufficient data it
1235 * is untouched. Otherwise it is extended. Returns zero on
1236 * success. The skb is freed on error.
1239 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1241 unsigned int size
= skb
->len
;
1242 if (likely(size
>= len
))
1244 return skb_pad(skb
, len
-size
);
1247 static inline int skb_add_data(struct sk_buff
*skb
,
1248 char __user
*from
, int copy
)
1250 const int off
= skb
->len
;
1252 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1254 __wsum csum
= csum_and_copy_from_user(from
, skb_put(skb
, copy
),
1257 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1260 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1263 __skb_trim(skb
, off
);
1267 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1268 struct page
*page
, int off
)
1271 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1273 return page
== frag
->page
&&
1274 off
== frag
->page_offset
+ frag
->size
;
1279 static inline int __skb_linearize(struct sk_buff
*skb
)
1281 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1285 * skb_linearize - convert paged skb to linear one
1286 * @skb: buffer to linarize
1288 * If there is no free memory -ENOMEM is returned, otherwise zero
1289 * is returned and the old skb data released.
1291 static inline int skb_linearize(struct sk_buff
*skb
)
1293 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1297 * skb_linearize_cow - make sure skb is linear and writable
1298 * @skb: buffer to process
1300 * If there is no free memory -ENOMEM is returned, otherwise zero
1301 * is returned and the old skb data released.
1303 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1305 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1306 __skb_linearize(skb
) : 0;
1310 * skb_postpull_rcsum - update checksum for received skb after pull
1311 * @skb: buffer to update
1312 * @start: start of data before pull
1313 * @len: length of data pulled
1315 * After doing a pull on a received packet, you need to call this to
1316 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1317 * CHECKSUM_NONE so that it can be recomputed from scratch.
1320 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1321 const void *start
, unsigned int len
)
1323 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1324 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1327 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1330 * pskb_trim_rcsum - trim received skb and update checksum
1331 * @skb: buffer to trim
1334 * This is exactly the same as pskb_trim except that it ensures the
1335 * checksum of received packets are still valid after the operation.
1338 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1340 if (likely(len
>= skb
->len
))
1342 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
1343 skb
->ip_summed
= CHECKSUM_NONE
;
1344 return __pskb_trim(skb
, len
);
1347 #define skb_queue_walk(queue, skb) \
1348 for (skb = (queue)->next; \
1349 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1352 #define skb_queue_reverse_walk(queue, skb) \
1353 for (skb = (queue)->prev; \
1354 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1358 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1359 int noblock
, int *err
);
1360 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1361 struct poll_table_struct
*wait
);
1362 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1363 int offset
, struct iovec
*to
,
1365 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1368 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1369 extern void skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1370 unsigned int flags
);
1371 extern __wsum
skb_checksum(const struct sk_buff
*skb
, int offset
,
1372 int len
, __wsum csum
);
1373 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1375 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1376 void *from
, int len
);
1377 extern __wsum
skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1378 int offset
, u8
*to
, int len
,
1380 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1381 extern void skb_split(struct sk_buff
*skb
,
1382 struct sk_buff
*skb1
, const u32 len
);
1384 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, int features
);
1386 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1387 int len
, void *buffer
)
1389 int hlen
= skb_headlen(skb
);
1391 if (hlen
- offset
>= len
)
1392 return skb
->data
+ offset
;
1394 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1400 extern void skb_init(void);
1401 extern void skb_add_mtu(int mtu
);
1404 * skb_get_timestamp - get timestamp from a skb
1405 * @skb: skb to get stamp from
1406 * @stamp: pointer to struct timeval to store stamp in
1408 * Timestamps are stored in the skb as offsets to a base timestamp.
1409 * This function converts the offset back to a struct timeval and stores
1412 static inline void skb_get_timestamp(const struct sk_buff
*skb
, struct timeval
*stamp
)
1414 *stamp
= ktime_to_timeval(skb
->tstamp
);
1417 static inline void __net_timestamp(struct sk_buff
*skb
)
1419 skb
->tstamp
= ktime_get_real();
1423 extern __sum16
__skb_checksum_complete_head(struct sk_buff
*skb
, int len
);
1424 extern __sum16
__skb_checksum_complete(struct sk_buff
*skb
);
1427 * skb_checksum_complete - Calculate checksum of an entire packet
1428 * @skb: packet to process
1430 * This function calculates the checksum over the entire packet plus
1431 * the value of skb->csum. The latter can be used to supply the
1432 * checksum of a pseudo header as used by TCP/UDP. It returns the
1435 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1436 * this function can be used to verify that checksum on received
1437 * packets. In that case the function should return zero if the
1438 * checksum is correct. In particular, this function will return zero
1439 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1440 * hardware has already verified the correctness of the checksum.
1442 static inline unsigned int skb_checksum_complete(struct sk_buff
*skb
)
1444 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
1445 __skb_checksum_complete(skb
);
1448 #ifdef CONFIG_NETFILTER
1449 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1451 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1452 nfct
->destroy(nfct
);
1454 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1457 atomic_inc(&nfct
->use
);
1459 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1460 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
1463 atomic_inc(&skb
->users
);
1465 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
1471 #ifdef CONFIG_BRIDGE_NETFILTER
1472 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1474 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1477 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1480 atomic_inc(&nf_bridge
->use
);
1482 #endif /* CONFIG_BRIDGE_NETFILTER */
1483 static inline void nf_reset(struct sk_buff
*skb
)
1485 nf_conntrack_put(skb
->nfct
);
1487 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1488 nf_conntrack_put_reasm(skb
->nfct_reasm
);
1489 skb
->nfct_reasm
= NULL
;
1491 #ifdef CONFIG_BRIDGE_NETFILTER
1492 nf_bridge_put(skb
->nf_bridge
);
1493 skb
->nf_bridge
= NULL
;
1497 /* Note: This doesn't put any conntrack and bridge info in dst. */
1498 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
1500 dst
->nfct
= src
->nfct
;
1501 nf_conntrack_get(src
->nfct
);
1502 dst
->nfctinfo
= src
->nfctinfo
;
1503 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1504 dst
->nfct_reasm
= src
->nfct_reasm
;
1505 nf_conntrack_get_reasm(src
->nfct_reasm
);
1507 #ifdef CONFIG_BRIDGE_NETFILTER
1508 dst
->nf_bridge
= src
->nf_bridge
;
1509 nf_bridge_get(src
->nf_bridge
);
1513 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
)
1515 nf_conntrack_put(dst
->nfct
);
1516 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1517 nf_conntrack_put_reasm(dst
->nfct_reasm
);
1519 #ifdef CONFIG_BRIDGE_NETFILTER
1520 nf_bridge_put(dst
->nf_bridge
);
1522 __nf_copy(dst
, src
);
1525 #else /* CONFIG_NETFILTER */
1526 static inline void nf_reset(struct sk_buff
*skb
) {}
1527 static inline void __nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
) {}
1528 static inline void nf_copy(struct sk_buff
*dst
, const struct sk_buff
*src
) {}
1529 #endif /* CONFIG_NETFILTER */
1531 #ifdef CONFIG_NETWORK_SECMARK
1532 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1534 to
->secmark
= from
->secmark
;
1537 static inline void skb_init_secmark(struct sk_buff
*skb
)
1542 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1545 static inline void skb_init_secmark(struct sk_buff
*skb
)
1549 static inline int skb_is_gso(const struct sk_buff
*skb
)
1551 return skb_shinfo(skb
)->gso_size
;
1554 #endif /* __KERNEL__ */
1555 #endif /* _LINUX_SKBUFF_H */