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>
26 #include <linux/highmem.h>
27 #include <linux/poll.h>
28 #include <linux/net.h>
29 #include <linux/textsearch.h>
30 #include <net/checksum.h>
31 #include <linux/dmaengine.h>
33 #define HAVE_ALLOC_SKB /* For the drivers to know */
34 #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
36 #define CHECKSUM_NONE 0
38 #define CHECKSUM_UNNECESSARY 2
40 #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
41 ~(SMP_CACHE_BYTES - 1))
42 #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
43 sizeof(struct skb_shared_info)) & \
44 ~(SMP_CACHE_BYTES - 1))
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 * HW: 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 HW,
65 * B. Checksumming on output.
67 * NONE: skb is checksummed by protocol or csum is not required.
69 * HW: device is required to csum packet as seen by hard_start_xmit
70 * from skb->h.raw to the end and to record the checksum
71 * at skb->h.raw+skb->csum.
73 * Device must show its capabilities in dev->features, set
74 * at device setup time.
75 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
77 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
78 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
79 * TCP/UDP over IPv4. Sigh. Vendors like this
80 * way by an unknown reason. Though, see comment above
81 * about CHECKSUM_UNNECESSARY. 8)
83 * Any questions? No questions, good. --ANK
88 #ifdef CONFIG_NETFILTER
91 void (*destroy
)(struct nf_conntrack
*);
94 #ifdef CONFIG_BRIDGE_NETFILTER
95 struct nf_bridge_info
{
97 struct net_device
*physindev
;
98 struct net_device
*physoutdev
;
99 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
100 struct net_device
*netoutdev
;
103 unsigned long data
[32 / sizeof(unsigned long)];
109 struct sk_buff_head
{
110 /* These two members must be first. */
111 struct sk_buff
*next
;
112 struct sk_buff
*prev
;
120 /* To allow 64K frame to be packed as single skb without frag_list */
121 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
123 typedef struct skb_frag_struct skb_frag_t
;
125 struct skb_frag_struct
{
131 /* This data is invariant across clones and lives at
132 * the end of the header data, ie. at skb->end.
134 struct skb_shared_info
{
136 unsigned short nr_frags
;
137 unsigned short gso_size
;
138 /* Warning: this field is not always filled in (UFO)! */
139 unsigned short gso_segs
;
140 unsigned short gso_type
;
141 unsigned int ip6_frag_id
;
142 struct sk_buff
*frag_list
;
143 skb_frag_t frags
[MAX_SKB_FRAGS
];
146 /* We divide dataref into two halves. The higher 16 bits hold references
147 * to the payload part of skb->data. The lower 16 bits hold references to
148 * the entire skb->data. It is up to the users of the skb to agree on
149 * where the payload starts.
151 * All users must obey the rule that the skb->data reference count must be
152 * greater than or equal to the payload reference count.
154 * Holding a reference to the payload part means that the user does not
155 * care about modifications to the header part of skb->data.
157 #define SKB_DATAREF_SHIFT 16
158 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
167 SKB_FCLONE_UNAVAILABLE
,
173 SKB_GSO_TCPV4
= 1 << 0,
174 SKB_GSO_UDPV4
= 1 << 1,
178 * struct sk_buff - socket buffer
179 * @next: Next buffer in list
180 * @prev: Previous buffer in list
181 * @sk: Socket we are owned by
182 * @tstamp: Time we arrived
183 * @dev: Device we arrived on/are leaving by
184 * @input_dev: Device we arrived on
185 * @h: Transport layer header
186 * @nh: Network layer header
187 * @mac: Link layer header
188 * @dst: destination entry
189 * @sp: the security path, used for xfrm
190 * @cb: Control buffer. Free for use by every layer. Put private vars here
191 * @len: Length of actual data
192 * @data_len: Data length
193 * @mac_len: Length of link layer header
195 * @local_df: allow local fragmentation
196 * @cloned: Head may be cloned (check refcnt to be sure)
197 * @nohdr: Payload reference only, must not modify header
198 * @pkt_type: Packet class
199 * @fclone: skbuff clone status
200 * @ip_summed: Driver fed us an IP checksum
201 * @priority: Packet queueing priority
202 * @users: User count - see {datagram,tcp}.c
203 * @protocol: Packet protocol from driver
204 * @truesize: Buffer size
205 * @head: Head of buffer
206 * @data: Data head pointer
207 * @tail: Tail pointer
209 * @destructor: Destruct function
210 * @nfmark: Can be used for communication between hooks
211 * @nfct: Associated connection, if any
212 * @ipvs_property: skbuff is owned by ipvs
213 * @nfctinfo: Relationship of this skb to the connection
214 * @nfct_reasm: netfilter conntrack re-assembly pointer
215 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
216 * @tc_index: Traffic control index
217 * @tc_verd: traffic control verdict
218 * @secmark: security marking
222 /* These two members must be first. */
223 struct sk_buff
*next
;
224 struct sk_buff
*prev
;
227 struct skb_timeval tstamp
;
228 struct net_device
*dev
;
229 struct net_device
*input_dev
;
234 struct icmphdr
*icmph
;
235 struct igmphdr
*igmph
;
237 struct ipv6hdr
*ipv6h
;
243 struct ipv6hdr
*ipv6h
;
252 struct dst_entry
*dst
;
256 * This is the control buffer. It is free to use for every
257 * layer. Please put your private variables there. If you
258 * want to keep them across layers you have to do a skb_clone()
259 * first. This is owned by whoever has the skb queued ATM.
278 void (*destructor
)(struct sk_buff
*skb
);
279 #ifdef CONFIG_NETFILTER
280 struct nf_conntrack
*nfct
;
281 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
282 struct sk_buff
*nfct_reasm
;
284 #ifdef CONFIG_BRIDGE_NETFILTER
285 struct nf_bridge_info
*nf_bridge
;
288 #endif /* CONFIG_NETFILTER */
289 #ifdef CONFIG_NET_SCHED
290 __u16 tc_index
; /* traffic control index */
291 #ifdef CONFIG_NET_CLS_ACT
292 __u16 tc_verd
; /* traffic control verdict */
295 #ifdef CONFIG_NET_DMA
296 dma_cookie_t dma_cookie
;
298 #ifdef CONFIG_NETWORK_SECMARK
303 /* These elements must be at the end, see alloc_skb() for details. */
304 unsigned int truesize
;
314 * Handling routines are only of interest to the kernel
316 #include <linux/slab.h>
318 #include <asm/system.h>
320 extern void kfree_skb(struct sk_buff
*skb
);
321 extern void __kfree_skb(struct sk_buff
*skb
);
322 extern struct sk_buff
*__alloc_skb(unsigned int size
,
323 gfp_t priority
, int fclone
);
324 static inline struct sk_buff
*alloc_skb(unsigned int size
,
327 return __alloc_skb(size
, priority
, 0);
330 static inline struct sk_buff
*alloc_skb_fclone(unsigned int size
,
333 return __alloc_skb(size
, priority
, 1);
336 extern struct sk_buff
*alloc_skb_from_cache(kmem_cache_t
*cp
,
339 extern void kfree_skbmem(struct sk_buff
*skb
);
340 extern struct sk_buff
*skb_clone(struct sk_buff
*skb
,
342 extern struct sk_buff
*skb_copy(const struct sk_buff
*skb
,
344 extern struct sk_buff
*pskb_copy(struct sk_buff
*skb
,
346 extern int pskb_expand_head(struct sk_buff
*skb
,
347 int nhead
, int ntail
,
349 extern struct sk_buff
*skb_realloc_headroom(struct sk_buff
*skb
,
350 unsigned int headroom
);
351 extern struct sk_buff
*skb_copy_expand(const struct sk_buff
*skb
,
352 int newheadroom
, int newtailroom
,
354 extern int skb_pad(struct sk_buff
*skb
, int pad
);
355 #define dev_kfree_skb(a) kfree_skb(a)
356 extern void skb_over_panic(struct sk_buff
*skb
, int len
,
358 extern void skb_under_panic(struct sk_buff
*skb
, int len
,
360 extern void skb_truesize_bug(struct sk_buff
*skb
);
362 static inline void skb_truesize_check(struct sk_buff
*skb
)
364 if (unlikely((int)skb
->truesize
< sizeof(struct sk_buff
) + skb
->len
))
365 skb_truesize_bug(skb
);
368 extern int skb_append_datato_frags(struct sock
*sk
, struct sk_buff
*skb
,
369 int getfrag(void *from
, char *to
, int offset
,
370 int len
,int odd
, struct sk_buff
*skb
),
371 void *from
, int length
);
378 __u32 stepped_offset
;
379 struct sk_buff
*root_skb
;
380 struct sk_buff
*cur_skb
;
384 extern void skb_prepare_seq_read(struct sk_buff
*skb
,
385 unsigned int from
, unsigned int to
,
386 struct skb_seq_state
*st
);
387 extern unsigned int skb_seq_read(unsigned int consumed
, const u8
**data
,
388 struct skb_seq_state
*st
);
389 extern void skb_abort_seq_read(struct skb_seq_state
*st
);
391 extern unsigned int skb_find_text(struct sk_buff
*skb
, unsigned int from
,
392 unsigned int to
, struct ts_config
*config
,
393 struct ts_state
*state
);
396 #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
399 * skb_queue_empty - check if a queue is empty
402 * Returns true if the queue is empty, false otherwise.
404 static inline int skb_queue_empty(const struct sk_buff_head
*list
)
406 return list
->next
== (struct sk_buff
*)list
;
410 * skb_get - reference buffer
411 * @skb: buffer to reference
413 * Makes another reference to a socket buffer and returns a pointer
416 static inline struct sk_buff
*skb_get(struct sk_buff
*skb
)
418 atomic_inc(&skb
->users
);
423 * If users == 1, we are the only owner and are can avoid redundant
428 * skb_cloned - is the buffer a clone
429 * @skb: buffer to check
431 * Returns true if the buffer was generated with skb_clone() and is
432 * one of multiple shared copies of the buffer. Cloned buffers are
433 * shared data so must not be written to under normal circumstances.
435 static inline int skb_cloned(const struct sk_buff
*skb
)
437 return skb
->cloned
&&
438 (atomic_read(&skb_shinfo(skb
)->dataref
) & SKB_DATAREF_MASK
) != 1;
442 * skb_header_cloned - is the header a clone
443 * @skb: buffer to check
445 * Returns true if modifying the header part of the buffer requires
446 * the data to be copied.
448 static inline int skb_header_cloned(const struct sk_buff
*skb
)
455 dataref
= atomic_read(&skb_shinfo(skb
)->dataref
);
456 dataref
= (dataref
& SKB_DATAREF_MASK
) - (dataref
>> SKB_DATAREF_SHIFT
);
461 * skb_header_release - release reference to header
462 * @skb: buffer to operate on
464 * Drop a reference to the header part of the buffer. This is done
465 * by acquiring a payload reference. You must not read from the header
466 * part of skb->data after this.
468 static inline void skb_header_release(struct sk_buff
*skb
)
472 atomic_add(1 << SKB_DATAREF_SHIFT
, &skb_shinfo(skb
)->dataref
);
476 * skb_shared - is the buffer shared
477 * @skb: buffer to check
479 * Returns true if more than one person has a reference to this
482 static inline int skb_shared(const struct sk_buff
*skb
)
484 return atomic_read(&skb
->users
) != 1;
488 * skb_share_check - check if buffer is shared and if so clone it
489 * @skb: buffer to check
490 * @pri: priority for memory allocation
492 * If the buffer is shared the buffer is cloned and the old copy
493 * drops a reference. A new clone with a single reference is returned.
494 * If the buffer is not shared the original buffer is returned. When
495 * being called from interrupt status or with spinlocks held pri must
498 * NULL is returned on a memory allocation failure.
500 static inline struct sk_buff
*skb_share_check(struct sk_buff
*skb
,
503 might_sleep_if(pri
& __GFP_WAIT
);
504 if (skb_shared(skb
)) {
505 struct sk_buff
*nskb
= skb_clone(skb
, pri
);
513 * Copy shared buffers into a new sk_buff. We effectively do COW on
514 * packets to handle cases where we have a local reader and forward
515 * and a couple of other messy ones. The normal one is tcpdumping
516 * a packet thats being forwarded.
520 * skb_unshare - make a copy of a shared buffer
521 * @skb: buffer to check
522 * @pri: priority for memory allocation
524 * If the socket buffer is a clone then this function creates a new
525 * copy of the data, drops a reference count on the old copy and returns
526 * the new copy with the reference count at 1. If the buffer is not a clone
527 * the original buffer is returned. When called with a spinlock held or
528 * from interrupt state @pri must be %GFP_ATOMIC
530 * %NULL is returned on a memory allocation failure.
532 static inline struct sk_buff
*skb_unshare(struct sk_buff
*skb
,
535 might_sleep_if(pri
& __GFP_WAIT
);
536 if (skb_cloned(skb
)) {
537 struct sk_buff
*nskb
= skb_copy(skb
, pri
);
538 kfree_skb(skb
); /* Free our shared copy */
546 * @list_: list to peek at
548 * Peek an &sk_buff. Unlike most other operations you _MUST_
549 * be careful with this one. A peek leaves the buffer on the
550 * list and someone else may run off with it. You must hold
551 * the appropriate locks or have a private queue to do this.
553 * Returns %NULL for an empty list or a pointer to the head element.
554 * The reference count is not incremented and the reference is therefore
555 * volatile. Use with caution.
557 static inline struct sk_buff
*skb_peek(struct sk_buff_head
*list_
)
559 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->next
;
560 if (list
== (struct sk_buff
*)list_
)
567 * @list_: list to peek at
569 * Peek an &sk_buff. Unlike most other operations you _MUST_
570 * be careful with this one. A peek leaves the buffer on the
571 * list and someone else may run off with it. You must hold
572 * the appropriate locks or have a private queue to do this.
574 * Returns %NULL for an empty list or a pointer to the tail element.
575 * The reference count is not incremented and the reference is therefore
576 * volatile. Use with caution.
578 static inline struct sk_buff
*skb_peek_tail(struct sk_buff_head
*list_
)
580 struct sk_buff
*list
= ((struct sk_buff
*)list_
)->prev
;
581 if (list
== (struct sk_buff
*)list_
)
587 * skb_queue_len - get queue length
588 * @list_: list to measure
590 * Return the length of an &sk_buff queue.
592 static inline __u32
skb_queue_len(const struct sk_buff_head
*list_
)
597 static inline void skb_queue_head_init(struct sk_buff_head
*list
)
599 spin_lock_init(&list
->lock
);
600 list
->prev
= list
->next
= (struct sk_buff
*)list
;
605 * Insert an sk_buff at the start of a list.
607 * The "__skb_xxxx()" functions are the non-atomic ones that
608 * can only be called with interrupts disabled.
612 * __skb_queue_after - queue a buffer at the list head
614 * @prev: place after this buffer
615 * @newsk: buffer to queue
617 * Queue a buffer int the middle of a list. This function takes no locks
618 * and you must therefore hold required locks before calling it.
620 * A buffer cannot be placed on two lists at the same time.
622 static inline void __skb_queue_after(struct sk_buff_head
*list
,
623 struct sk_buff
*prev
,
624 struct sk_buff
*newsk
)
626 struct sk_buff
*next
;
632 next
->prev
= prev
->next
= newsk
;
636 * __skb_queue_head - queue a buffer at the list head
638 * @newsk: buffer to queue
640 * Queue a buffer at the start of a list. This function takes no locks
641 * and you must therefore hold required locks before calling it.
643 * A buffer cannot be placed on two lists at the same time.
645 extern void skb_queue_head(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
646 static inline void __skb_queue_head(struct sk_buff_head
*list
,
647 struct sk_buff
*newsk
)
649 __skb_queue_after(list
, (struct sk_buff
*)list
, newsk
);
653 * __skb_queue_tail - queue a buffer at the list tail
655 * @newsk: buffer to queue
657 * Queue a buffer at the end of a list. This function takes no locks
658 * and you must therefore hold required locks before calling it.
660 * A buffer cannot be placed on two lists at the same time.
662 extern void skb_queue_tail(struct sk_buff_head
*list
, struct sk_buff
*newsk
);
663 static inline void __skb_queue_tail(struct sk_buff_head
*list
,
664 struct sk_buff
*newsk
)
666 struct sk_buff
*prev
, *next
;
669 next
= (struct sk_buff
*)list
;
673 next
->prev
= prev
->next
= newsk
;
678 * __skb_dequeue - remove from the head of the queue
679 * @list: list to dequeue from
681 * Remove the head of the list. This function does not take any locks
682 * so must be used with appropriate locks held only. The head item is
683 * returned or %NULL if the list is empty.
685 extern struct sk_buff
*skb_dequeue(struct sk_buff_head
*list
);
686 static inline struct sk_buff
*__skb_dequeue(struct sk_buff_head
*list
)
688 struct sk_buff
*next
, *prev
, *result
;
690 prev
= (struct sk_buff
*) list
;
699 result
->next
= result
->prev
= NULL
;
706 * Insert a packet on a list.
708 extern void skb_insert(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
709 static inline void __skb_insert(struct sk_buff
*newsk
,
710 struct sk_buff
*prev
, struct sk_buff
*next
,
711 struct sk_buff_head
*list
)
715 next
->prev
= prev
->next
= newsk
;
720 * Place a packet after a given packet in a list.
722 extern void skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
);
723 static inline void __skb_append(struct sk_buff
*old
, struct sk_buff
*newsk
, struct sk_buff_head
*list
)
725 __skb_insert(newsk
, old
, old
->next
, list
);
729 * remove sk_buff from list. _Must_ be called atomically, and with
732 extern void skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
);
733 static inline void __skb_unlink(struct sk_buff
*skb
, struct sk_buff_head
*list
)
735 struct sk_buff
*next
, *prev
;
740 skb
->next
= skb
->prev
= NULL
;
746 /* XXX: more streamlined implementation */
749 * __skb_dequeue_tail - remove from the tail of the queue
750 * @list: list to dequeue from
752 * Remove the tail of the list. This function does not take any locks
753 * so must be used with appropriate locks held only. The tail item is
754 * returned or %NULL if the list is empty.
756 extern struct sk_buff
*skb_dequeue_tail(struct sk_buff_head
*list
);
757 static inline struct sk_buff
*__skb_dequeue_tail(struct sk_buff_head
*list
)
759 struct sk_buff
*skb
= skb_peek_tail(list
);
761 __skb_unlink(skb
, list
);
766 static inline int skb_is_nonlinear(const struct sk_buff
*skb
)
768 return skb
->data_len
;
771 static inline unsigned int skb_headlen(const struct sk_buff
*skb
)
773 return skb
->len
- skb
->data_len
;
776 static inline int skb_pagelen(const struct sk_buff
*skb
)
780 for (i
= (int)skb_shinfo(skb
)->nr_frags
- 1; i
>= 0; i
--)
781 len
+= skb_shinfo(skb
)->frags
[i
].size
;
782 return len
+ skb_headlen(skb
);
785 static inline void skb_fill_page_desc(struct sk_buff
*skb
, int i
,
786 struct page
*page
, int off
, int size
)
788 skb_frag_t
*frag
= &skb_shinfo(skb
)->frags
[i
];
791 frag
->page_offset
= off
;
793 skb_shinfo(skb
)->nr_frags
= i
+ 1;
796 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
797 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
798 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
801 * Add data to an sk_buff
803 static inline unsigned char *__skb_put(struct sk_buff
*skb
, unsigned int len
)
805 unsigned char *tmp
= skb
->tail
;
806 SKB_LINEAR_ASSERT(skb
);
813 * skb_put - add data to a buffer
814 * @skb: buffer to use
815 * @len: amount of data to add
817 * This function extends the used data area of the buffer. If this would
818 * exceed the total buffer size the kernel will panic. A pointer to the
819 * first byte of the extra data is returned.
821 static inline unsigned char *skb_put(struct sk_buff
*skb
, unsigned int len
)
823 unsigned char *tmp
= skb
->tail
;
824 SKB_LINEAR_ASSERT(skb
);
827 if (unlikely(skb
->tail
>skb
->end
))
828 skb_over_panic(skb
, len
, current_text_addr());
832 static inline unsigned char *__skb_push(struct sk_buff
*skb
, unsigned int len
)
840 * skb_push - add data to the start of a buffer
841 * @skb: buffer to use
842 * @len: amount of data to add
844 * This function extends the used data area of the buffer at the buffer
845 * start. If this would exceed the total buffer headroom the kernel will
846 * panic. A pointer to the first byte of the extra data is returned.
848 static inline unsigned char *skb_push(struct sk_buff
*skb
, unsigned int len
)
852 if (unlikely(skb
->data
<skb
->head
))
853 skb_under_panic(skb
, len
, current_text_addr());
857 static inline unsigned char *__skb_pull(struct sk_buff
*skb
, unsigned int len
)
860 BUG_ON(skb
->len
< skb
->data_len
);
861 return skb
->data
+= len
;
865 * skb_pull - remove data from the start of a buffer
866 * @skb: buffer to use
867 * @len: amount of data to remove
869 * This function removes data from the start of a buffer, returning
870 * the memory to the headroom. A pointer to the next data in the buffer
871 * is returned. Once the data has been pulled future pushes will overwrite
874 static inline unsigned char *skb_pull(struct sk_buff
*skb
, unsigned int len
)
876 return unlikely(len
> skb
->len
) ? NULL
: __skb_pull(skb
, len
);
879 extern unsigned char *__pskb_pull_tail(struct sk_buff
*skb
, int delta
);
881 static inline unsigned char *__pskb_pull(struct sk_buff
*skb
, unsigned int len
)
883 if (len
> skb_headlen(skb
) &&
884 !__pskb_pull_tail(skb
, len
-skb_headlen(skb
)))
887 return skb
->data
+= len
;
890 static inline unsigned char *pskb_pull(struct sk_buff
*skb
, unsigned int len
)
892 return unlikely(len
> skb
->len
) ? NULL
: __pskb_pull(skb
, len
);
895 static inline int pskb_may_pull(struct sk_buff
*skb
, unsigned int len
)
897 if (likely(len
<= skb_headlen(skb
)))
899 if (unlikely(len
> skb
->len
))
901 return __pskb_pull_tail(skb
, len
-skb_headlen(skb
)) != NULL
;
905 * skb_headroom - bytes at buffer head
906 * @skb: buffer to check
908 * Return the number of bytes of free space at the head of an &sk_buff.
910 static inline int skb_headroom(const struct sk_buff
*skb
)
912 return skb
->data
- skb
->head
;
916 * skb_tailroom - bytes at buffer end
917 * @skb: buffer to check
919 * Return the number of bytes of free space at the tail of an sk_buff
921 static inline int skb_tailroom(const struct sk_buff
*skb
)
923 return skb_is_nonlinear(skb
) ? 0 : skb
->end
- skb
->tail
;
927 * skb_reserve - adjust headroom
928 * @skb: buffer to alter
929 * @len: bytes to move
931 * Increase the headroom of an empty &sk_buff by reducing the tail
932 * room. This is only allowed for an empty buffer.
934 static inline void skb_reserve(struct sk_buff
*skb
, int len
)
941 * CPUs often take a performance hit when accessing unaligned memory
942 * locations. The actual performance hit varies, it can be small if the
943 * hardware handles it or large if we have to take an exception and fix it
946 * Since an ethernet header is 14 bytes network drivers often end up with
947 * the IP header at an unaligned offset. The IP header can be aligned by
948 * shifting the start of the packet by 2 bytes. Drivers should do this
951 * skb_reserve(NET_IP_ALIGN);
953 * The downside to this alignment of the IP header is that the DMA is now
954 * unaligned. On some architectures the cost of an unaligned DMA is high
955 * and this cost outweighs the gains made by aligning the IP header.
957 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
961 #define NET_IP_ALIGN 2
965 * The networking layer reserves some headroom in skb data (via
966 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
967 * the header has to grow. In the default case, if the header has to grow
968 * 16 bytes or less we avoid the reallocation.
970 * Unfortunately this headroom changes the DMA alignment of the resulting
971 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
972 * on some architectures. An architecture can override this value,
973 * perhaps setting it to a cacheline in size (since that will maintain
974 * cacheline alignment of the DMA). It must be a power of 2.
976 * Various parts of the networking layer expect at least 16 bytes of
977 * headroom, you should not reduce this.
980 #define NET_SKB_PAD 16
983 extern int ___pskb_trim(struct sk_buff
*skb
, unsigned int len
);
985 static inline void __skb_trim(struct sk_buff
*skb
, unsigned int len
)
987 if (unlikely(skb
->data_len
)) {
992 skb
->tail
= skb
->data
+ len
;
996 * skb_trim - remove end from a buffer
997 * @skb: buffer to alter
1000 * Cut the length of a buffer down by removing data from the tail. If
1001 * the buffer is already under the length specified it is not modified.
1002 * The skb must be linear.
1004 static inline void skb_trim(struct sk_buff
*skb
, unsigned int len
)
1007 __skb_trim(skb
, len
);
1011 static inline int __pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1014 return ___pskb_trim(skb
, len
);
1015 __skb_trim(skb
, len
);
1019 static inline int pskb_trim(struct sk_buff
*skb
, unsigned int len
)
1021 return (len
< skb
->len
) ? __pskb_trim(skb
, len
) : 0;
1025 * skb_orphan - orphan a buffer
1026 * @skb: buffer to orphan
1028 * If a buffer currently has an owner then we call the owner's
1029 * destructor function and make the @skb unowned. The buffer continues
1030 * to exist but is no longer charged to its former owner.
1032 static inline void skb_orphan(struct sk_buff
*skb
)
1034 if (skb
->destructor
)
1035 skb
->destructor(skb
);
1036 skb
->destructor
= NULL
;
1041 * __skb_queue_purge - empty a list
1042 * @list: list to empty
1044 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1045 * the list and one reference dropped. This function does not take the
1046 * list lock and the caller must hold the relevant locks to use it.
1048 extern void skb_queue_purge(struct sk_buff_head
*list
);
1049 static inline void __skb_queue_purge(struct sk_buff_head
*list
)
1051 struct sk_buff
*skb
;
1052 while ((skb
= __skb_dequeue(list
)) != NULL
)
1056 #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
1058 * __dev_alloc_skb - allocate an skbuff for sending
1059 * @length: length to allocate
1060 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1062 * Allocate a new &sk_buff and assign it a usage count of one. The
1063 * buffer has unspecified headroom built in. Users should allocate
1064 * the headroom they think they need without accounting for the
1065 * built in space. The built in space is used for optimisations.
1067 * %NULL is returned in there is no free memory.
1069 static inline struct sk_buff
*__dev_alloc_skb(unsigned int length
,
1072 struct sk_buff
*skb
= alloc_skb(length
+ NET_SKB_PAD
, gfp_mask
);
1074 skb_reserve(skb
, NET_SKB_PAD
);
1078 extern struct sk_buff
*__dev_alloc_skb(unsigned int length
, int gfp_mask
);
1082 * dev_alloc_skb - allocate an skbuff for sending
1083 * @length: length to allocate
1085 * Allocate a new &sk_buff and assign it a usage count of one. The
1086 * buffer has unspecified headroom built in. Users should allocate
1087 * the headroom they think they need without accounting for the
1088 * built in space. The built in space is used for optimisations.
1090 * %NULL is returned in there is no free memory. Although this function
1091 * allocates memory it can be called from an interrupt.
1093 static inline struct sk_buff
*dev_alloc_skb(unsigned int length
)
1095 return __dev_alloc_skb(length
, GFP_ATOMIC
);
1099 * skb_cow - copy header of skb when it is required
1100 * @skb: buffer to cow
1101 * @headroom: needed headroom
1103 * If the skb passed lacks sufficient headroom or its data part
1104 * is shared, data is reallocated. If reallocation fails, an error
1105 * is returned and original skb is not changed.
1107 * The result is skb with writable area skb->head...skb->tail
1108 * and at least @headroom of space at head.
1110 static inline int skb_cow(struct sk_buff
*skb
, unsigned int headroom
)
1112 int delta
= (headroom
> NET_SKB_PAD
? headroom
: NET_SKB_PAD
) -
1118 if (delta
|| skb_cloned(skb
))
1119 return pskb_expand_head(skb
, (delta
+ (NET_SKB_PAD
-1)) &
1120 ~(NET_SKB_PAD
-1), 0, GFP_ATOMIC
);
1125 * skb_padto - pad an skbuff up to a minimal size
1126 * @skb: buffer to pad
1127 * @len: minimal length
1129 * Pads up a buffer to ensure the trailing bytes exist and are
1130 * blanked. If the buffer already contains sufficient data it
1131 * is untouched. Otherwise it is extended. Returns zero on
1132 * success. The skb is freed on error.
1135 static inline int skb_padto(struct sk_buff
*skb
, unsigned int len
)
1137 unsigned int size
= skb
->len
;
1138 if (likely(size
>= len
))
1140 return skb_pad(skb
, len
-size
);
1143 static inline int skb_add_data(struct sk_buff
*skb
,
1144 char __user
*from
, int copy
)
1146 const int off
= skb
->len
;
1148 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1150 unsigned int csum
= csum_and_copy_from_user(from
,
1154 skb
->csum
= csum_block_add(skb
->csum
, csum
, off
);
1157 } else if (!copy_from_user(skb_put(skb
, copy
), from
, copy
))
1160 __skb_trim(skb
, off
);
1164 static inline int skb_can_coalesce(struct sk_buff
*skb
, int i
,
1165 struct page
*page
, int off
)
1168 struct skb_frag_struct
*frag
= &skb_shinfo(skb
)->frags
[i
- 1];
1170 return page
== frag
->page
&&
1171 off
== frag
->page_offset
+ frag
->size
;
1176 static inline int __skb_linearize(struct sk_buff
*skb
)
1178 return __pskb_pull_tail(skb
, skb
->data_len
) ? 0 : -ENOMEM
;
1182 * skb_linearize - convert paged skb to linear one
1183 * @skb: buffer to linarize
1185 * If there is no free memory -ENOMEM is returned, otherwise zero
1186 * is returned and the old skb data released.
1188 static inline int skb_linearize(struct sk_buff
*skb
)
1190 return skb_is_nonlinear(skb
) ? __skb_linearize(skb
) : 0;
1194 * skb_linearize_cow - make sure skb is linear and writable
1195 * @skb: buffer to process
1197 * If there is no free memory -ENOMEM is returned, otherwise zero
1198 * is returned and the old skb data released.
1200 static inline int skb_linearize_cow(struct sk_buff
*skb
)
1202 return skb_is_nonlinear(skb
) || skb_cloned(skb
) ?
1203 __skb_linearize(skb
) : 0;
1207 * skb_postpull_rcsum - update checksum for received skb after pull
1208 * @skb: buffer to update
1209 * @start: start of data before pull
1210 * @len: length of data pulled
1212 * After doing a pull on a received packet, you need to call this to
1213 * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
1214 * so that it can be recomputed from scratch.
1217 static inline void skb_postpull_rcsum(struct sk_buff
*skb
,
1218 const void *start
, unsigned int len
)
1220 if (skb
->ip_summed
== CHECKSUM_HW
)
1221 skb
->csum
= csum_sub(skb
->csum
, csum_partial(start
, len
, 0));
1224 unsigned char *skb_pull_rcsum(struct sk_buff
*skb
, unsigned int len
);
1227 * pskb_trim_rcsum - trim received skb and update checksum
1228 * @skb: buffer to trim
1231 * This is exactly the same as pskb_trim except that it ensures the
1232 * checksum of received packets are still valid after the operation.
1235 static inline int pskb_trim_rcsum(struct sk_buff
*skb
, unsigned int len
)
1237 if (likely(len
>= skb
->len
))
1239 if (skb
->ip_summed
== CHECKSUM_HW
)
1240 skb
->ip_summed
= CHECKSUM_NONE
;
1241 return __pskb_trim(skb
, len
);
1244 static inline void *kmap_skb_frag(const skb_frag_t
*frag
)
1246 #ifdef CONFIG_HIGHMEM
1251 return kmap_atomic(frag
->page
, KM_SKB_DATA_SOFTIRQ
);
1254 static inline void kunmap_skb_frag(void *vaddr
)
1256 kunmap_atomic(vaddr
, KM_SKB_DATA_SOFTIRQ
);
1257 #ifdef CONFIG_HIGHMEM
1262 #define skb_queue_walk(queue, skb) \
1263 for (skb = (queue)->next; \
1264 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1267 #define skb_queue_reverse_walk(queue, skb) \
1268 for (skb = (queue)->prev; \
1269 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1273 extern struct sk_buff
*skb_recv_datagram(struct sock
*sk
, unsigned flags
,
1274 int noblock
, int *err
);
1275 extern unsigned int datagram_poll(struct file
*file
, struct socket
*sock
,
1276 struct poll_table_struct
*wait
);
1277 extern int skb_copy_datagram_iovec(const struct sk_buff
*from
,
1278 int offset
, struct iovec
*to
,
1280 extern int skb_copy_and_csum_datagram_iovec(struct sk_buff
*skb
,
1283 extern void skb_free_datagram(struct sock
*sk
, struct sk_buff
*skb
);
1284 extern void skb_kill_datagram(struct sock
*sk
, struct sk_buff
*skb
,
1285 unsigned int flags
);
1286 extern unsigned int skb_checksum(const struct sk_buff
*skb
, int offset
,
1287 int len
, unsigned int csum
);
1288 extern int skb_copy_bits(const struct sk_buff
*skb
, int offset
,
1290 extern int skb_store_bits(const struct sk_buff
*skb
, int offset
,
1291 void *from
, int len
);
1292 extern unsigned int skb_copy_and_csum_bits(const struct sk_buff
*skb
,
1293 int offset
, u8
*to
, int len
,
1295 extern void skb_copy_and_csum_dev(const struct sk_buff
*skb
, u8
*to
);
1296 extern void skb_split(struct sk_buff
*skb
,
1297 struct sk_buff
*skb1
, const u32 len
);
1299 extern void skb_release_data(struct sk_buff
*skb
);
1300 extern struct sk_buff
*skb_segment(struct sk_buff
*skb
, int sg
);
1302 static inline void *skb_header_pointer(const struct sk_buff
*skb
, int offset
,
1303 int len
, void *buffer
)
1305 int hlen
= skb_headlen(skb
);
1307 if (hlen
- offset
>= len
)
1308 return skb
->data
+ offset
;
1310 if (skb_copy_bits(skb
, offset
, buffer
, len
) < 0)
1316 extern void skb_init(void);
1317 extern void skb_add_mtu(int mtu
);
1320 * skb_get_timestamp - get timestamp from a skb
1321 * @skb: skb to get stamp from
1322 * @stamp: pointer to struct timeval to store stamp in
1324 * Timestamps are stored in the skb as offsets to a base timestamp.
1325 * This function converts the offset back to a struct timeval and stores
1328 static inline void skb_get_timestamp(const struct sk_buff
*skb
, struct timeval
*stamp
)
1330 stamp
->tv_sec
= skb
->tstamp
.off_sec
;
1331 stamp
->tv_usec
= skb
->tstamp
.off_usec
;
1335 * skb_set_timestamp - set timestamp of a skb
1336 * @skb: skb to set stamp of
1337 * @stamp: pointer to struct timeval to get stamp from
1339 * Timestamps are stored in the skb as offsets to a base timestamp.
1340 * This function converts a struct timeval to an offset and stores
1343 static inline void skb_set_timestamp(struct sk_buff
*skb
, const struct timeval
*stamp
)
1345 skb
->tstamp
.off_sec
= stamp
->tv_sec
;
1346 skb
->tstamp
.off_usec
= stamp
->tv_usec
;
1349 extern void __net_timestamp(struct sk_buff
*skb
);
1351 extern unsigned int __skb_checksum_complete(struct sk_buff
*skb
);
1354 * skb_checksum_complete - Calculate checksum of an entire packet
1355 * @skb: packet to process
1357 * This function calculates the checksum over the entire packet plus
1358 * the value of skb->csum. The latter can be used to supply the
1359 * checksum of a pseudo header as used by TCP/UDP. It returns the
1362 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1363 * this function can be used to verify that checksum on received
1364 * packets. In that case the function should return zero if the
1365 * checksum is correct. In particular, this function will return zero
1366 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1367 * hardware has already verified the correctness of the checksum.
1369 static inline unsigned int skb_checksum_complete(struct sk_buff
*skb
)
1371 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
1372 __skb_checksum_complete(skb
);
1375 #ifdef CONFIG_NETFILTER
1376 static inline void nf_conntrack_put(struct nf_conntrack
*nfct
)
1378 if (nfct
&& atomic_dec_and_test(&nfct
->use
))
1379 nfct
->destroy(nfct
);
1381 static inline void nf_conntrack_get(struct nf_conntrack
*nfct
)
1384 atomic_inc(&nfct
->use
);
1386 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1387 static inline void nf_conntrack_get_reasm(struct sk_buff
*skb
)
1390 atomic_inc(&skb
->users
);
1392 static inline void nf_conntrack_put_reasm(struct sk_buff
*skb
)
1398 #ifdef CONFIG_BRIDGE_NETFILTER
1399 static inline void nf_bridge_put(struct nf_bridge_info
*nf_bridge
)
1401 if (nf_bridge
&& atomic_dec_and_test(&nf_bridge
->use
))
1404 static inline void nf_bridge_get(struct nf_bridge_info
*nf_bridge
)
1407 atomic_inc(&nf_bridge
->use
);
1409 #endif /* CONFIG_BRIDGE_NETFILTER */
1410 static inline void nf_reset(struct sk_buff
*skb
)
1412 nf_conntrack_put(skb
->nfct
);
1414 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1415 nf_conntrack_put_reasm(skb
->nfct_reasm
);
1416 skb
->nfct_reasm
= NULL
;
1418 #ifdef CONFIG_BRIDGE_NETFILTER
1419 nf_bridge_put(skb
->nf_bridge
);
1420 skb
->nf_bridge
= NULL
;
1424 #else /* CONFIG_NETFILTER */
1425 static inline void nf_reset(struct sk_buff
*skb
) {}
1426 #endif /* CONFIG_NETFILTER */
1428 #ifdef CONFIG_NETWORK_SECMARK
1429 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1431 to
->secmark
= from
->secmark
;
1434 static inline void skb_init_secmark(struct sk_buff
*skb
)
1439 static inline void skb_copy_secmark(struct sk_buff
*to
, const struct sk_buff
*from
)
1442 static inline void skb_init_secmark(struct sk_buff
*skb
)
1446 #endif /* __KERNEL__ */
1447 #endif /* _LINUX_SKBUFF_H */